FOOD AND DRUG ADMINISTRATION
NINETY-SIXTH MEETING OF THE
CARDIOVASCULAR AND RENAL DRUG ADVISORY COMMITTEE
Friday, April 12, 2002
8777 Georgia Avenue
Silver Spring, Maryland
JEFFREY BORER, M.D., Chairman
Director, Division of Pathophysiology
Weill Medical College
525 East 68th Street, Room F467
New York, New York 10021
JAYNE E. PETERSON, R.PH., J.D., Acting Executive Secretary
Advisors and Consultants Staff
Center for Drug Evaluation and Research
Food and Drug Administration
5630 Fishers Lane, Room 1093
Rockville, Maryland 20857
PAUL ARMSTRONG, M.D.
Professor, Department of Medicine
University of Alberta
251 Medical Science Building
Edmonton, Alberta, Canada T6C37
BLASE A. CARABELLO, M.D.
Professor of Medicine
Houston Veterans Affairs Medical Center
Medical Service (111)
2002 Holcombe Boulevard
Houston, Texas 77030
SUSANNA L. CUNNINGHAM, PH.D.
Professor, Department of Biobehavioral Nursing
School of Nursing, Box 357266
Seattle, Washington 98195-7266
THOMAS FLEMING, PH.D.
Professor and Chair
Department of Biostatistics
University of Washington
Seattle, Washington 98195-7232
COMMITTEE MEMBERS: (Continued)
ALAN T. HIRSCH, M.D.
Associate Professor of Medicine and Radiology
Minnesota Vascular Diseases Center
University of Minnesota Medical School
420 Delaware Street, S.E., Box 508 FUMC
Minneapolis, Minnesota 55455
JOANN LINDENFELD, M.D.
Professor of Medicine
Division of Cardiology
University of Colorado Health Science Center
4200 East Ninth Avenue, B-130
Denver, Colorado 80262
BEVERLY H. LORELL, M.D.
Professor of Medicine
Beth Israel Deaconess Medical Center
330 Brookline Avenue
Boston, Massachusetts 02215
STEVEN NISSEN, M.D., F.A.C.C.
Vice Chairman, Department of Cardiology
Professor of Medicine
Ohio State University
The Cleveland Clinic Foundation
9500 Euclid Avenue, F15
Cleveland, Ohio 44195
SPECIAL GOVERNMENT EMPLOYEES:
ANDREW S. BREM, M.D.
Professor in Pediatrics
Brown University School of Medicine
Director, Division of Pediatric Nephrology
Rhode Island Hospital
593 Eddy Street
Providence, Rhode Island 02902
SPECIAL GOVERNMENT EMPLOYEES: (Continued)
JEFFREY B. KOPP, M.D.
Kidney Disease Section, Metabolic Diseases Branch
National Institute of Diabetes, Digestive
and Kidney Diseases
National Institutes of Health
Building 10, Room 3N116
Bethesda, Maryland 20892-1268
FOOD AND DRUG ADMINISTRATION STAFF:
JAMES HUNG, PH.D.
ROBERT TEMPLE, M.D.
DOUGLAS THROCKMORTON, M.D.
RAY BAIN, PH.D.
BARRY BRENNER, M.D.
MICHAEL C. ELIA, PH.D.
JONATHAN FOX, M.D.
BONNIE GOLDMAN, M.D.
STEVEN HAFFNER, M.D.
WILLIAM KEANE, M.D.
MARVIN KONSTAM, M.D.
PETER KOWEY, M.D.
SHAHNAZ SHAHINFAR, M.D.
SCOTT ZEGER, M.D.
C O N T E N T S
NDA 20-386/S028, Cozaar (losartan potassium)
Merck and Company, Inc.
Proposed Indication: For the Treatment of
Type II Diabetic Patients with Nephropathy
* * *
AGENDA ITEM PAGE
CONFLICT OF INTEREST STATEMENT
By Dr. Jayne Peterson 7
By Dr. Michael C. Elia 9
Background, Rationale and Results
of the RENAAL Study
By Dr. Shahnaz Shahinfar 14
Review of the Evidence and Conclusions
By Dr. William Keane 117
OPEN PUBLIC HEARING 204
COMMITTEE DISCUSSION OF QUESTIONS PRESENTED 205
P R O C E E D I N G S
DR. BORER: We will call the meeting to order. This is the 96th meeting of the Cardiovascular and Renal Drugs Advisory Committee.
The agenda consists of consideration of one NDA for losartan potassium for the indication of treatment of diabetic patients with type II diabetes with nephropathy.
We have no applicants for public discussion this morning.
I need to announce that Dr. Michael Artman, who made every effort to be here, was unable to get here because of a last minute problem, so the committee will be without Dr. Artman today.
I also want to announce prophylactically that I have been importuned to provide a break at 9:45. We didn't do that the last time, and so we will. So, don't worry.
Now, we will begin then with the conflict of interest statement. Dr. Peterson, do we have one?
DR. PETERSON: Yes. Before we go on I just want to caution the committee that if you are going to speak you need to push the center button to turn the mikes on.
DR. BORER: Yes, I'm sorry. I should have mentioned as well that as for the committee, as well as for anybody among the presenters or among the audience, if you are recognized, we do need you to state your name before you start talking so that the transcription can be as accurate as possible. Thank you.
DR. PETERSON: I'll read the conflict of interest statement now. The following announcement addresses the issue of conflict of issue with regard to this meeting and is made a part of this record to preclude even the appearance of such at this meeting.
Based on the submitted agenda for the meeting and all the financial interests reported by the committee participants, it has been determined that all interests in firms regulated by the Center for Drug Evaluation and Research, which have been reported by the participants present, present no potential for an appearance of a conflict of interest at this meeting with the following exceptions.
Dr. Susanna Cunningham has been granted waivers under 18 U.S.C. 208(b)(3) and under 21 U.S.C. 355(n)(4), an amendment of Section 505 of the Food and Drug Administration Act, for ownership of stock in the sponsor valued between $25,001 and $50,000.
Dr. Andrew Brem has been granted waivers under 18 U.S.C. 208(b)(3) and under 21 U.S.C. 355(n)(4), an amendment of Section 505(a) of the Food and Drug Administration Modernization Act, for ownership of stock in the sponsor valued between $5,001 and $25,000.
Dr. Alan Hirsch has been granted a waiver under 18 U.S.C. 208(b)(3) for his participation on a sponsor's bureau supported by firms that have a financial interest in the product at issue and the competing product. He receives less than $10,001 per year.
Dr. Steven Nissen has been granted a waiver under 18 U.S.C. 208(b)(3) for his consulting with the sponsor on unrelated matters. He received less than $10,001 over the past year.
A copy of these waiver statements may be obtained by submitting a written request to the agency's Freedom of Information Office, room 12A-30 of the Parklawn Building.
In the event that the discussions involve any other products and firms not already on the agenda, for which an FDA participant has a financial interest, the participants are aware of the need to exclude themselves from such involvement, and their exclusion will be noted for the record.
With respect to all other participants, we ask in the interest of fairness that they address any current or previous financial involvement with any firm whose products they may wish to comment upon. Thank you.
DR. BORER: Thank you. We will begin with the sponsor's presentation. Dr. Elia.
DR. ELIA: Good morning Mr. Chairman, members of the advisory committee, FDA, ladies and gentlemen. My name is Michael Elia from the Department of Regulatory Affairs at Merck Research Laboratories.
I would like to thank the Advisory Committee and the FDA for the opportunity to present Merck's data on the efficacy and safety of losartan in providing renal protection in type II diabetic patients with nephropathy.
I am going to provide an introduction today to our presentation that will focus on the results of Merck's renal outcome study RENAAL, which stands for reduction in endpoints in non-insulin dependent diabetes mellitus with the angiotensin II receptor antagonist losartan.
RENAAL is a multicenter international, double-blind, randomized, placebo-controlled trial to evaluate the protective effects of losartan in patients with type II diabetes and proteinuria.
Just to remind you, as we stated in our briefing document, we will use the words "nephropathy" and "proteinuria" interchangeably today.
Prior to initiating RENAAL, no convincing long-term renal protection outcomes data were available in type II diabetic patients with nephropathy.
To address this, RENAAL was designed as a long term renal protection outcome study. In RENAAL the primary endpoint was a time to event analysis of a the composite of doubling of serum creatinine, end-stage renal disease, defined as the need for chronic dialysis or transplantation, or death.
As we will show you today, this study provides persuasive evidence that losartan delays the progression of renal disease in type II diabetic patients with nephropathy.
Complete data on the hard endpoints of end-stage renal disease and death were collected on all randomized patients. No patients were lost to follow-up. The results of RENAAL confirm that the safety and tolerability profile of losartan is these patients is consistent with that in the currently approved U.S. product circular for Cozaar.
Furthermore, RENAAL demonstrated that the benefit of losartan treatment in these patients exceeds that attributable to blood pressure reduction alone. We believe that these results merit modification of our product label, with the following new indication. Under the heading of renal protection in type II diabetic patients with proteinuria, Cozaar is indicated to delay the progression of renal disease, as measured by a reduction in the combined incidence of doubling of serum creatinine, end-stage renal disease, or death.
It's important to note that at present there are no drugs approved in the United States for renal protection in type II diabetic patients with proteinuria.
While we believe that the RENAAL study provides compelling evidence to support our proposed indication, it is a single study. In an earlier meeting this year, this committee and the FDA discussed the evidence needed to support a new claim for renal protection in type II diabetic patients with proteinuria. In considering the questions before the committee today, it's useful to review the evidentiary standard needed to support a new claim for an approved drug. This is especially important when the basis for the proposed claim rests largely on the results of a single study such as RENAAL.
Although the FDA has relied on data from a single study to support a new indication, their views on the matter had not been fully delineated prior to the issuance of a guidance document.
In 1998 the FDA issued a guidance document entitled "Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products", that gives the agency its current thinking on the approval of new claims based on data from a single trial.
When regulatory decisions are based primarily on a single study it is important to determine the level of confidence one can have in the results of that study. Are the data scientifically convincing or simply due to chance? FDA's guidance document provides several points to consider in evaluating whether one has sufficient confidence in the results of a single study to support a new effectiveness claim.
As you will see you a few minutes in Dr. Shahinfar's presentation, the RENAAL study satisfies several of the key features of a single study that can be used to support an effectiveness claim. For example, RENAAL is a large, multicenter study conducted at 250 clinical sites in 28 countries, and it provides persuasive evidence that losartan delays the progression of renal disease in type II diabetic patients with proteinuria.
The consistent and significant benefits of losartan across multiple endpoints and multiple subgroups in the RENAAL study promote confidence in its findings and provide strong support for our proposed indication.
Furthermore, as described in the advisory committee's background package, several separate smaller clinical studies and preclinical studies have demonstrated the beneficial effect of losartan therapy on renal endpoints. These latter studies support the biological plausibility of renal protective effects of losartan observed in RENAAL.
In summary, we will show you today that the efficacy and safety of losartan in patients with type II diabetes and proteinuria support our proposed indication.
Finally, the agenda for today's Merck presentation is as follows. After discussing the natural history of type II diabetes and current treatments, Dr. Shahnaz Shahinfar from Merck's Department of Cardiovascular Clinical Research will describe the efficacy and safety results from the RENAAL study. Next, Dr. William Keane, Vice President of Clinical Development in Merck's U.S. Human Health Division, will summarize the evidence that supports and confirms the RENAAL results, and end our presentation with our overall conclusions.
The advisory committee members have previously received a briefing document from Merck that provides more detailed information than time allows us to present here this morning.
In addition, Merck has brought several consultants to the meeting. These experts are available to facilitate the Advisory Committee's discussions and deliberations.
Here today are Dr. Barry Brenner from the Harvard Medical School, who served as Chair of the RENAAL Steering Committee; Dr. Steven Haffner from the University of Texas, who chaired the RENAAL Endpoint Adjudication Committee; Dr. Carl Erik Mogensen from Aarhus Kommunehospital in Denmark, who chaired the Data and Safety Monitoring Committee; Dr. Peter Kowey from Jefferson Medical College, who was a member of the Data and Safety Monitoring Committee for RENAAL; Dr. Marvin Konstam from the Tufts New England Medical Center: and Dr. Scott Zeger from the Johns Hopkins University.
I would now like to turn the podium over to Dr. Shahinfar.
DR. SHAHINFAR: Good morning. I'm Shahnaz Shahinfar of Cardiovascular Clinical Research of Merck Research Laboratory. I was the clinical monitor for the RENAAL study, which I have the pleasure of presenting to you.
Today I will be reviewing the background and rationale for the RENAAL study, followed by a presentation of demographics, efficacy results, and safety results. Dr. William Keane will conclude our presentation with a review of the evidence and our conclusion regarding the renal protective effect of losartan in type II diabetes.
Diabetes mellitus is a major public health issue worldwide. The most common type of diabetes is type II diabetes, which is the focus of our discussion today. Up to 40 percent of type II diabetic patients develop kidney disease. End-stage renal disease is a devastating complication of diabetes mellitus. Studies have shown that the incidence of end-stage renal disease is increasing worldwide. In the United States diabetic nephropathy is the leading cause of end-stage renal disease.
It should be noted that end-stage renal disease is an irreversible condition, and dialysis is a life-support system to prevent death in these patients. Even with dialysis, up to 40 percent of diabetic patients die within 2 years of its initiation, a mortality rate comparable to that of end-stage heart failure.
No treatment has shown conclusively to delay end-stage renal disease in type II diabetic patients with nephropathy. It is extremely important to identify a therapeutic intervention for this unmet medical need.
In the search for a therapeutic intervention to reduce end-stage renal disease, it should be remembered that diabetic nephropathy is primarily a glomerular disease. The exact mechanism of progression of diabetic nephropathies are known, but many factors, both hemodynamic and nonhemodynamic, may contribute to the development of this glomerular injury. Angiotensin II has been hypothesized to play a role in the progressive nature of this glomerular injury.
The proposed mechanisms by which angiotensin II is involved in diabetic glomerular injury are shown pictorially on this slide. The glomerulus is the site of the original injury in diabetic nephropathy. As a result of the original insult to the glomerulus and the loss of the nephrons, there is a remarkable adaptation within the kidney to compensate for this nephron loss. The remaining nephrons increase the workload through a state of hyperfiltration in order to maintain overall glomerular filtration rate.
Angiotensin II plays an important role in this adaption by increasing resistance in post-glomerular vessel or efferent arteriole. In addition, in diabetic patients, resistant in preglomerular vessel or afferent arteriole is reduced. This combination of increased efferent arteriolar resistance, mediated by angiotensin II, and decreased efferent arteriolar resistance results in an increase in intraglomerular pressure or glomerular hypertension.
This glomerular hypertension, while favorable in the short term, has long-term detrimental effects on the nephron. This increasing glomerular pressure initiates an up-regulation of a series of nonhemodynamic factors. These factors include increased permeability of the filtering membrane, which may result in proteinuria and activation of fibrotic and inflammatory processes, in which angiotensin II also plays a role. The end result is glomerulosclerosis and death of the nephron. This cycle of nephron death continues until all nephrons are lost, and that's when end-stage renal disease occurs.
Since angiotensin II appears to be play an central role in many of the hemodynamic and nonhemodynamic mechanisms of the progression of diabetic nephropathy, it has been hypothesized that the blockade of angiotensin II will provide renal protection in this disease.
This hypothesis has been tested repeatedly with both losartan and ACE inhibitors in animal models of diabetic nephropathy. There is no perfect experimental model for type II diabetic nephropathy. However, the streptozotocin-induced diabetic rats have been used for this purpose. It has been demonstrated that the blockade of angiotensin II in this model is associated with a reduction in glomerulosclerosis and proteinuria.
Furthermore, in an animal model of non-diabetic renal disease, blockade of angiotensin II reduced proteinuria and glomerulosclerosis. In the same studies, other antihypertensive agents did not confirm these renal protective effects. These findings support the biological plausibility of the effect of angiotensin II blockade in renal protection.
This hypothesis was tested clinically with captopril in type I diabetic patients with nephropathy, approximately a decade ago. In this study, in 409 type I diabetic patients with proteinuria and retinopathy, with a mean age of approximately 35, captopril significantly reduced end-stage renal disease or death. However, until now, conclusive clinical data on end-stage renal disease have not been available in patients with type II diabetic nephropathy, which is the most common type of diabetes. It is important to note that type I and type II diabetic patients represent two different populations, and it is difficult to extrapolate data from type I to type II and vice versa.
Patients with type II diabetes are typically older, obese, and have insulin resistance, advanced atherosclerosis, and long-standing hypertension. Many of these patients are hypertensive even before the onset of nephropathy. The kidney of a patient with type II diabetes, in addition to the glomerular injury of diabetic nephropathy, also has other morphological changes, which may be a reflection of long-standing hypertension, older age and macrovascular disease. Only a portion of this disease burden may be susceptible to the blockade of the renin-angiotensin system.
In evaluating a therapeutic intervention, especially the magnitude of benefit, these differences between type I and type II diabetes should be taken into consideration.
In the absence of definitive data, current therapeutic approaches among clinicians for renal protection in type II diabetic nephropathy have focused on metabolic control and blood pressure control. Valuable amounts of data are available for each strategy. However, there is no conclusive clinical evidence that these therapeutic approaches are associated with a reduction in end-stage renal disease in type II diabetic patients with nephropathy.
Prior to the initiation of RENAAL, the question remained. In patients with type II diabetes and nephropathy does angiotensin II blockade with losartan offer renal protection? The RENAAL study was designed to answer that question.
RENAAL, reduction of endpoints in non-insulin dependent diabetes with the AII antagonist losartan, was a multicenter, multinational, double-blind, randomized, placebo-controlled study to evaluate the renal protective effects of losartan in patients with type II diabetes and nephropathy.
There were three oversight committees in RENAAL: the Steering Committee, the Data and Safety Monitoring Committee, and the Endpoint Adjudication Committee.
The Steering Committee, chaired by Dr. Barry Brenner, was blinded to the study results and oversaw the overall conduct of the study.
The Data and Safety Monitoring Committee was chaired by Dr. Carl Erik Mogensen. This committee was unblinded to the results of RENAAL, and oversaw the safety of the patients in the study.
An Endpoint Adjudication Committee, chaired by Dr. Steven Haffner, included two cardiologists, three nephrologists and one endocrinologist, who were blinded to the results, and adjudicated all the primary renal and secondary cardiovascular endpoints in RENAAL.
Merck functioned as the coordinating and data management center, with national and regional coordinators. There were 250 centers from 28 countries worldwide.
The primary hypothesis in RENAAL was, in type II diabetic patients with nephropathy, losartan compared to placebo would increase the time to the first event of the composite endpoint of doubling of serum creatinine, which represented more than 50 percent loss of renal function, end-stage renal disease, defined as need for chronic dialysis or transplantation, or death, defined as all-cause mortality.
These components of the primary endpoints were selected based on the natural course of nephropathy in type II diabetic patients. Serum creatinine progressively rises, leading to a doubling of this biochemical marker, which represents approximately 50 percent loss of renal function, followed eventually by the clinically irreversible condition of end-stage renal disease.
In patients who reach end-stage renal disease, dialysis or transplantation is necessary to sustain life. Thus, death and dialysis are competing events. Despite dialysis, mortality remains high in these patients. Of course, death from any cause can occur at any time.
There were three secondary hypotheses in RENAAL. The first two were renal hypotheses. Losartan would reduce the rate of progression of renal disease as measured by the slope of reciprocal of serum creatinine compared to placebo. Second, losartan would reduce proteinuria, compared to placebo, during the course of this study.
Another secondary hypothesis was a cardiovascular hypothesis. It should be noted that RENAAL was specifically designed as a renal protection study. However, since cardiovascular events are common in these type II diabetic patients, we made cardiovascular morbidity and mortality a secondary hypothesis and adjudicated all cardiovascular events.
The cardiovascular hypothesis was that in type II diabetic patients with nephropathy losartan, compared to placebo, would increase the time to first event of the composite endpoint of cardiovascular morbidity and mortality. This cardiovascular composite endpoint included cardiovascular death, myocardial infarction, stroke, first hospitalization for heart failure, first hospitalization for angina and revascularization, both coronary and peripheral revascularization.
The major inclusion and exclusion criteria in RENAAL are noted on this slide. Since RENAAL was a renal protection study, we enriched our population with patients who had risk of progression of renal disease.
Important inclusion criteria were that these patients were required to have type II diabetes and had to be between the ages of 31 and 70 years, with proteinuria defined as an albumin-to-creatinine ratio of more than 300 milligrams per gram on a first morning void, which demonstrate macroalbuminuria, or greater than 500 milligrams protein in a 24-hour urine. Serum creatinine was required to be between 1.3 and 3 milligrams per deciliter.
Important exclusion criteria included patients with known non-diabetic renal disease, such as patients with renal artery stenosis and polycystic kidney were excluded. Patients with uncontrolled diabetes, defined as hemoglobin A1C greater than 12 percent, were excluded. Patients with a history of myocardial infarction or CABG within a month, stroke or PTCA within 6 months, and TIA within a year of randomization, were excluded. Patients with a history of heart failure were excluded. A heart failure exclusion criterion was added shortly after the initiation of the study.
The RENAAL study design is shown on the next few slides.
RENAAL was a double-blind, randomized, placebo-controlled, multicenter study. Because of the known effect of proteinuria on the progression of renal disease in type I diabetics and in non-diabetic patients with renal disease, patients were stratified, based on an urinary albumin-to-creatinine ratio, less than 2,000 or greater than or equal to 2,000 milligrams per gram.
During the follow-up, clinic and laboratory evaluations were performed every 3 months. We planned for 1-year enrollment and 5 years maximum follow-up.
This is a schematic diagram of the study design. Qualified patients with type II diabetes were screened by urine protein dipstick, and were placed in a run-in period for 6 weeks. During this period, prior antihypertensive therapy was maintained, except for angiotensin converting enzyme inhibitors or angiotensin II receptor antagonists, which were stopped 6 weeks prior to randomization, and were replaced by other antihypertensive drugs if needed.
To qualify patients at baseline, they first were stratified, based on urinary albumin-to-creatinine ratio, less than 2,000 or greater than 2,000 milligrams per gram. Within each stratum, patients were randomized to receive either 50 milligrams losartan or matching placebo once daily, on the background of the run-in period antihypertensive therapy. The goal was a trough blood pressure of systolic less than 140 millimeters of mercury and diastolic less than 90, which was the WHO-recommended guideline for diabetics at the time of initiation of the study.
If this goal blood pressure was not achieved within 4 weeks, the study drug could be increased to 100 milligrams losartan once daily or matching placebo. If the goal blood pressure was still not achieved, other open-label antihypertensives, except angiotensin II antagonists or ACE inhibitors, could be added, or the dose of the existing drug could be adjusted.
Comprehensive patient follow-up was an important feature of RENAAL, to ensure that doubling of serum creatinine, end-stage renal disease, and death events were collected in all patients.
Importantly, in RENAAL patients were required to remain on a study therapy, regardless of nonfatal events until the completion of the study. For example, if patients doubled their serum creatinine, they were to remain on therapy until the end of the study unless the patient died. If patients experienced end-stage renal disease, they were to remain on a study drug until the end of the study, unless the patient died.
We collected end-stage renal disease and death data in all patients, regardless of doubling of serum creatinine events. This permitted the independent assessment of treatment on the clinical endpoints of end-stage renal disease in all patients. For all patients, including those who discontinued the study drug, clinic visits were to continue every 3 months to capture renal and cardiovascular endpoint information.
After discontinuation of the study drug, if clinic visits were not feasible, telephone follow-up was done to capture end-stage renal disease and death information. Whereas doubling of serum creatinine and cardiovascular morbidity information were not captured in telephone follow-up, data on end-stage renal disease and death were collected for all patients randomized.
On February 10, 2001, the Steering Committee, while blinded to the study results, voted unanimously to end RENAAL prior to its planned termination date of March 2002, because of concerns of continuing the placebo group without the blockade of the renin-angiotensin system. This decision was based on increasing evidence that ACE inhibitors may be effective in reducing cardiovascular events in patients with cardiovascular risk factors. These data were from the Heart Outcome Prevention Evaluation study, HOPE, and a subpopulation of patients with renal impairment, which was reported by Mann, et al. in Annals of Internal Medicine, April 2001.
I would now like to discuss the RENAAL demographics and efficacy results. In discussing the efficacy results, I will first present the information under primary and secondary renal data, followed by the secondary cardiovascular data.
This slide summarizes the patient disposition in RENAAL. In RENAAL, 3,893 patients were screened, 1,513 patients were randomized, 751 patients were allocated to losartan, 762 patients were allocated to placebo.
The breakdown of patients who completed on the study drug and the number of patients who discontinued the study drug after reaching a primary event or prior to reaching a primary event are shown in these boxes. There are no patients lost to follow-up in RENAAL. Outcome data on end-stage renal disease and death are available for all patients randomized.
The next 4 slides provide the baseline demographic data in RENAAL.
Patients were equally distributed with respect to baseline demographics between losartan and placebo arms of the study. With respect to gender, age, blood pressure and body mass index, the two treatment groups were comparable.
The two treatment groups were also comparable with respect to race and region. The RENAAL study achieved an excellent representation from diverse ethnic groups. Among the patients in RENAAL, about 17 percent were Asian, 15 percent were Black, 49 percent were Caucasian, and 18 percent were Hispanic.
The treatment groups were comparable with respect to past medical history at baseline. As anticipated, the majority of patients in RENAAL were hypertensive at the beginning of the study and were treated with antihypertensive drugs. At baseline, the number of patients with cardiovascular history of angina, myocardial infarction, and stroke was relatively low in our study.
This slide provides selected mean laboratory values at baseline in RENAAL. The two treatment groups were comparable with respect to mean serum creatinine, serum potassium, hemoglobin, and hemoglobin A1C. I will discuss the differences baseline proteinuria later in my presentation.
As a reminder, the primary hypothesis of RENAAL was that losartan compared to placebo would increase the time to the first event of the composite endpoint of doubling of serum creatinine, end-stage renal disease, or death. The analytical approach to the primary composite endpoint was the time to each patient's first event. In the next two slides I will illustrate how patients contributed to this time-to-event analysis.
In the left column, we have 5 hypothetical patients, A through E. In the next 3 columns, we show three endpoints that contributed to the primary composite endpoint. As you can see, patient A doubled his serum creatinine, developed end-stage renal disease, and died. Compare this patient to patient D, who died without doubling of serum creatinine or developing end-stage renal disease.
On this slide we have circled the endpoint that occurred first for each patient. Each of these values were used as the first event in the analysis of the primary composite endpoint. Thus, for patient A, doubling of serum creatinine was the first event captured, and this event was used in the analysis of the primary composite endpoint.
For patient D, death was the first event, and this event was used in the analysis of the primary composite endpoint. Note that also 4 of the 5 patients in this example died. Only the death of patient D would contribute to the primary composite endpoint.
The next slide will illustrate the results of our analysis on the primary composite endpoint.
DR. BORER: Dr. Shahinfar, can you just stop for a moment? Before you actually show the outcome data, I would like a clarification that sounds like you won't get to later.
In the book that you sent us, it suggests that there was a distinct difference in the average years of patient follow-up. It says here on your page 12 that you had an average of 173 patient-years of follow-up in the losartan group and 234 years of patient follow-up in the placebo group. That is clinic follow-up. And with regard to telephone follow-up, there was the same discrepancy.
Now, in the FDA analysis, however -- and this was looked at in a different way -- the conclusion was that exposure to drug in the placebo group and the losartan group was approximately equivalent.
Before you actually go on and present the data, I'd like to understand why this apparent discrepancy exists, or is not a discrepancy, or do I not understand these data correctly?
DR. SHAHINFAR: Thank you very much for your question. The numbers that you are referring to in the background refer to the patients who were discontinued and they went to some kind of follow-up.
Basically, after patients were discontinued from the study, we tried to keep them in the clinic regardless of whether they were on drug or were not on drug. So, in those patients, some of them couldn't come back for subsequent clinic visits, but we followed these patients in telephone follow-up. Does that answer your question?
DR. BORER: Yes, it does indeed. What we are seeing here is that people stayed on losartan longer. That's why there were presumably fewer patient years of follow-up in the losartan group. I understand. Thank you.
DR. SHAHINFAR: The next slide will illustrate the results of our analysis on the primary composite endpoint. In all the slides that you will see today, losartan is depicted in yellow and placebo in white. Unless otherwise noted, all efficacy analyses are based on intention to treat.
The primary composite endpoint results of RENAAL are demonstrated in this Kaplan-Meier curve. In this slide, the y axis is the percentage of patients with events. The x axis is the duration of follow-up. The number below the x axis represents the number of patients at risk at each time point.
As this slide illustrates, RENAAL began with 762 patients in the placebo group and 751 patients in the losartan group. By 36 months in this study, year 3, there still remained 296 patients at risk in the placebo group, and 300 patients in the losartan group. As anticipated, by month 48 a relatively small number of patients at risk is available in each treatment group.
The risk reduction for losartan was calculated using the Cox proportional hazard regression model. The results demonstrate that losartan significantly reduced the risk of the primary composite endpoint of time to first event of doubling of serum creatinine, end-stage renal disease, or death by 16.1 percent; p equals .022.
In RENAAL, there were prespecified analyses of the irreversible clinical endpoints, collectively referred to as end-stage renal disease, death, and a composite of end-stage renal disease or death. Since patients were followed after the occurrence of a nonfatal primary endpoint, many patients experienced multiple clinical endpoints. In these cases the key principles applied were that a patient counted as having had an endpoint in all relevant analyses and that a patient counted only once in any analysis.
This slide summarizes the analytical approach taken for the irreversible clinical endpoints. For end-stage renal disease, the analytical approach included all patients who reached end-stage renal disease, regardless of whether doubling of serum creatinine occurred first.
For death, the analytical approach included all patients who died, regardless of whether doubling of serum creatinine or end-stage renal disease occurred first.
For end-stage renal disease or death, the analytical approach included patients who reached a first event of end-stage renal disease or death, regardless of whether doubling of serum creatinine occurred first.
These analyses were performed on the entire patient cohort, because all patients were followed for the occurrence of end-stage renal disease and death for the entire study.
The next slide illustrates how these events would be captured for separate analysis of ESRD, death, and the composite of ESRD or death.
This slide illustrates the same 5 hypothetical patients that we showed you earlier. For each patient, the first occurring primary event is circled. Now, let me show you how these events in these patients would be captured in each of the analyses of the irreversible clinical endpoints of ESRD, death, and the composite of ESRD or death.
The three columns on the right-hand side of the slide illustrate the analysis of end-stage renal disease, death, and the composite of ESRD or death, and the contribution of each patient's event to each analysis.
Patient A contributes to all three analyses, whereas patient D contributes to the analysis of death and to the analysis of ESRD or death.
Please note that although a patient can experience each of the three endpoints of doubling of serum creatinine, end-stage renal disease, and death, a patient is only counted once in each analysis.
For the prespecified analysis of end-stage renal disease, defined as need for chronic dialysis or transplantation, the y axis shows the percentage of patients with events. The x axis demonstrates duration of follow-up.
Losartan treatment reduced the risk of end-stage renal disease by 28.6 percent; p equals .002. As you see in this Kaplan-Meier curve, the effect of therapy is observed about 18 months after the initiation of losartan, as demonstrated by the separation of the curve at this time point.
This is the first time that a therapeutic intervention has shown a beneficial effect on end-stage renal disease in type II diabetic patients with nephropathy.
For the prespecified analysis of all-cause mortality, there was no significant difference in this endpoint between losartan and placebo; p equals .884.
End-stage renal disease and death are competing events in type II diabetic patients with proteinuria because these patients may die before reaching end-stage renal disease or die as a result of requiring but not receiving dialysis, which is a life support therapy in these patients. Therefore, we evaluated the effect of losartan on the risk of experiencing either of these two endpoints. The prespecified analysis of the composite endpoint of end-stage renal disease or death demonstrated that losartan significantly reduced the risk of this outcome by 19.9 percent; p equals .009.
To summarize the results of the RENAAL primary hypothesis and prespecified analysis of irreversible clinical endpoints, this chart displays the percent risk reduction and its 95 percent confidence interval.
Note that the horizontal scale is a logarithmic scale from plus 50 percent on the left, corresponding to a reduction risk due to losartan, to minus 50 percent on the right, corresponding to an increase in risk with losartan. The confidence intervals are symmetric on this logarithmic scale. Losartan significantly reduced the risk of the primary composite endpoint and the irreversible clinical endpoint of end-stage renal disease and end-stage renal disease or death.
Before the initiation of RENAAL, we recognized the overall importance of proteinuria on the progression of renal disease, but the precise nature of this relationship in type II diabetes was unknown. Therefore, RENAAL patients were stratified at baseline, based on a level of proteinuria less than 2,000 or greater than 2,000 albumin-to-creatinine ratio.
As a result of the stratification, within each stratum an equal number of patients were randomized to each treatment group. However, there was an imbalance in the distribution of baseline proteinuria within the higher stratum of greater than 2,000. Specifically, more patients with baseline proteinuria above 4,000, the highest level of proteinuria, were randomized to the losartan arm compared to placebo. I will now illustrate these two points in each of the next two slides.
This slide demonstrates the patient randomized to losartan or placebo in each stratum. The x axis demonstrates categories baseline proteinuria and the y axis is the percentage of patients in each category. The two strata are noted: the lower stratum below 2,000 on the left, and the higher stratum above 2,000 on the right.
As you see from this slide, stratification ensured that overall there was an equal number of patients on losartan and placebo in each stratum. But such a stratification does not eliminate the possibility of an imbalance of distribution of patients within each stratum. As I will show you next, that occurred in RENAAL.
As demonstrated on this slide, there is a statistically significant imbalance in the distribution of patients between losartan and placebo with respect to baseline proteinuria within the higher stratum. This imbalance is especially pronounced in the category of greater than 4,000 patients where we have more losartan than placebo patients. I will show you later that these patients are at the highest risk of progression of renal disease, and that 80 percent of these patients had a primary endpoint during the study.
In RENAAL, we demonstrated that in type II diabetic patients there is a strong relationship between baseline proteinuria and the risk of the primary composite endpoint as shown on this slide. In this analysis we pooled the losartan and placebo groups. On the x axis, different levels of baseline proteinuria are listed. On the y axis is the hazards ratio for the primary composite endpoint, which is the hazard rate for each level of proteinuria relative to 300 milligrams albumin-to-creatinine ratio, which was the entry criterion for proteinuria in our study. The larger the hazard ratio represents the higher risk for primary event. As it is shown, the risk for primary events increases substantially as baseline proteinuria increased.
For example, in a patient at baseline proteinuria of 2,000 albumin-to-creatinine ratio, the hazard rate for the primary outcome is approximately 3 times higher than a patient with 300 milligrams albumin-to-creatinine ratio. In a patient with 4,000 albumin-to-creatinine ratio at baseline, the hazard rate is about 8 times higher than a patient with 300 milligrams albumin-to-creatinine ratio.
Because of the important role of proteinuria and the risk of the progression of renal disease, and since prespecified primary analysis did not adjust for imbalances within strata, it was reasonable to adjust for imbalances in baseline proteinuria for RENAAL. The results of this analysis are shown in the next two slides.
Using the baseline proteinuria as a continuous covariate, in the Cox proportional hazard regression model, the risk reduction with losartan on the primary composite endpoint increases from 16.1 percent to 22.2 percent; p equals .001.
Adjustment for baseline proteinuria also results in an increase in treatment effect on irreversible clinical endpoints of end-stage renal disease and a composite of end-stage renal disease or death. This slide demonstrates the risk reduction with losartan for the primary composite endpoint, end-stage renal disease, death, and end-stage renal disease or death. The solid line with a circle represent the prespecified analysis, and the dotted line with the square represents the adjusted value, using baseline proteinuria as a continuous covariate. As you can see, the dotted line with the square moves toward the left of the 0 line in favor of losartan.
DR. BORER: Before you go on, Dr. Shahinfar, this is Jeff Borer for the microphone for the tape there. Those adjustments are very interesting, and intuitively they seem reasonable and I don't want to suggest that there isn't an effect here and that the proteinuria isn't important. But I have some concern about accepting adjustments and numbers based on adjustments per se because my understanding is that to make an adjustment you have to assume that you know the relationship quantitatively between the variable for which you are adjusting and the dependent variable, the outcome variable. I don't think we know that. So, you're assuming a model.
I am going to ask Tom Fleming and the FDA statistician, if we have him or her here, to comment on that. But, what I would like to see, if you have the data -- and you may not have them this minute, so you can pull them out -- is an analysis of absolute numbers for the outcome in people above 4,000, understanding that there's an imbalance in favor of more on losartan, and for 2,000 to 4,000, which are the patients where there was an imbalance in favor of placebo, still above your prespecified cut point.
I would like to know that, and, Tom, I would like to have some comment from you about the validity of this kind of adjustment, if you would.
DR. FLEMING: I would be happy to look at the results as you have asked for them, and then I can comment. You have asked, since in particular what we are seeing is a breakdown of imbalances in those above 4,000, it would be interesting to see the results in that stratum above 4,000, and then in the complement.
DR. BORER: I was also asking about the generic issue of making imbalances based on models when we don't really know the relation between the variables that we are adjusting for.
DR. FLEMING: Okay, I'll go ahead a comment now then.
The approach that is being used is a very standard approach to address any potential confounding that can exist, and a confounder arises, as you know, when you have a very predictive variable that is imbalanced between the two arms.
What we are seeing here is a recognition that baseline proteinuria in advance is obviously a very important predicted variable. The structure that was imposed was to balance for those below 2,000 and above 2,000.
What we're seeing here -- and I would like to probe a little bit later, at exactly how this functional form was derived -- I think it is in slide 54, which is certainly very informative and relevant. Taking that as the truth, what it is showing is that there's a striking monotonic trend toward increasing risk of the outcome as you increase that baseline proteinuria level, such that there could readily be emerging imbalances in the cohort above 2,000 because there was no structure imposed beyond what randomization does itself to assure a balance in those people who were above 2,000.
What emerged in the data was an excess of 92 versus 71 people with values above 4,000. So, there was a confounding that emerged in spite of the structure imposed at randomization, and that confounding certainly has the potential of biasing the results.
If you use a Cox regression or a stratified log rank analysis, you get a very appropriate and legitimate adjustment for that balance. You don't have to assume you know the functional form. If it's a highly predictive covariate and it's imbalanced, we don't have to assume we know that to do a Cox regression analysis or a stratified log rank.
The problem that we run into in interpreting this, though, is post hoc analyses that are adjusting for imbalances that can occur even in a random way in a randomized stratified trial require some careful interpretation because one could do these analyses in numerous ways; i.e., you could form innumerable different types of covariate adjustments. And I'd like to come back to this discussion later on. That's the issue that I think we have to be a bit cautious about how we interpret.
But it is certainly very appropriate in a supportive analysis to look for what might be profound or very substantial evidence of confounding and in a supportive analysis to look to see whether that strengthens or weakens our sense of association.
DR. BORER: Thank you.
DR. TEMPLE: So, I think the first thing Jeffrey asked was where did you get those numbers. How did you decide how to make the adjustment? And, your answer, if I understand it was, they looked at the relationship between proteinuria in both the untreated and treated people and found that relationship in these data. The risk in there is that you can find it even if it's not true as a matter of chance sometimes. So, you have to judge its degree of plausibility. If it were extremely plausible, of course it would have been prospectively defined. This comes up all the time. People are smarter after the fact than before.
The other point I guess is that looking at the two groups, people over 4,000 and under 4,000 is sort of the poor man nonstatistician covariate analysis, and it's very helpful for people who don't quite understand what covariate analyses are, like me. So, that's always very helpful.
DR. FLEMING: In this discussion, I think if you could put slide 54 up as well, which I think is your slide that showed what I call the functional form. This is very important.
For a covariate to be a confounder you need to have an imbalance. Okay, we've seen that, 92 against 71. It has to be imbalanced in a way that matters, i.e. those people, i.e., the 92/71, or the people who are at 4,000 or to the right, and those people are at much higher risk than people who were to the left. Two questions.
I'm always a little skeptical about a curve that is so smooth. That probably is what nature really is, but our data usually has much more noise. So, how did this function form? Was it derived from this study or other data? That's question 1.
Question 2 is, is this the relationship of baseline proteinuria with a triple endpoint, and if so, I might speculate the relationship with what I care more about, which is the double endpoint. End-stage renal disease/death would be even a more striking gradient.
Can you answer those two questions?
DR. BAIN: Yes. Ray Bain, clinical biostatistics, Merck Research Lab.
You're correct. This slide here was based on the pooled groups from the RENAAL study, the 15/13 patients. You are also right that this particular analysis, this hazard ratio of increasing baseline proteinuria, relative to a patient who has 300 milligrams per gram, is based on our primary triple endpoint, doubling, death, or dialysis.
Now, Dr. Temple is also correct that sometimes if we were more intelligent we would have probably introduced this correction way back when we were designing the protocol. It turns out that we were half intelligent. We did recognize the importance of baseline protein and therefore prestratified. Now, unfortunately we only prestratified by less than 2 and greater than 2, but we recognize that it was a very important baseline risk factor.
In addition to that -- and we can go into this later -- we also prespecified in our data analysis plan to do a risk score analysis, which looked at a number of different baseline risk factors. It turns out 1 of the risk factors that we included as potentially being a risk factor for our triple endpoint was, again, baseline proteinuria. We can go into that later if you like.
Dr. Borer, did we answer your question?
DR. BORER: Reasonably.
DR. TEMPLE: Do you have that curve for the renal endpoint alone?
DR. BAIN: Yes, we do.
DR. LINDENFELD: Jeff, may I ask?
DR. BORER: Yes.
DR. LINDENFELD: I wonder if we could back to the last slide. Can you just show us that same data on slide 54 when the endpoint is end-stage renal disease or death?
The reason I ask this -- and everyone can help me with this -- is I think all of us who see these patients recognize as the proteinuria goes up and up the diuretics go up and up. And while in the past we have considered a doubling of creatinine a real endpoint -- that is, it reflects loss of renal function -- when we start to add those diuretics, we have all seen the creatinine double. We withdraw diuretics and it goes back down. So, I wonder if this same endpoint holds when we don't use the doubling of creatinine.
DR. BAIN: We do not have that particular graph, but we could probably get it.
DR. FLEMING: The second question that I was intending to ask was that I wanted to see that same curve not for the triple endpoint, but what I call the double endpoint. I think of the double endpoint as end-stage renal disease/death, where my speculation is the gradient or the slope I would expect would be even greater, but I'd like to see that.
DR. BAIN: Okay, we'll work on that.
But, as you can see here, this is again one of these curves where now on the x axis here is hazard ratio, so the .5 is that 50 percent reduction. You can see the doubling of serum creatinine if you look at that line. I'm sorry. The primary composite outcome is at the top there. At the very top, the solid line is our primary outcome, the .022. The next line below that is what Dr. Shahinfar already discussed. When you adjust for baseline proteinuria, it moves the effect to the left and you get a p value of .001 that she showed.
Now, we haven't shown you that curve for baseline proteinuria as a prediction of ESRD or death, but you can see the same type of effect here when you look at the ESRD. We haven't gotten there yet I think, but you'll see that endpoint of ESRD or death, and then you'll also see that when you adjust again for continuous baseline proteinuria it moves the treatment effect to the left.
DR. FLEMING: Let me just clarify a little bit. Put that slide back up.
I guess to provide the most precise statement, where I'd expect the biggest gradient is end-stage renal disease as a single endpoint, because certainly what we are seeing is an effect on end-stage renal disease much more so than death, doubling in serum creatinine for reasons we can discuss later. It doesn't strike me as an endpoint that should be as strongly affected by baseline proteinuria as end-stage renal disease itself. In fact, if we look at the adjustment here, the adjustment is in fact visually the greatest for end-stage renal disease.
So, my speculation is if we show that gradient curve that you were showing on slide 54, it would be very interesting to show what the relationship is of baseline proteinuria not just with the triple endpoint, but with the double endpoint and with end-stage renal disease itself.
DR. BAIN: Okay, so your request is to look at slide 54 for the single endpoint of ESRD and then the double endpoint of ESRD or death. Will do.
DR. TEMPLE: And end-stage renal disease or doubling, just for me.
DR. BAIN: And end-stage renal disease or doubling. Okay.
DR. SHAHINFAR: In RENAAL, we prespecified several sensitivity analyses for the primary composite endpoint, using hemoglobin A1C and mean arterial pressure as time varying covariates. We also performed analyses of baseline subgroups. Hemoglobin A1C, as mentioned in your background, was comparable between the two treatment groups. In the next several slides, I will discuss the sensitivity analysis for mean arterial pressure and baseline subgroups.
Blood pressure control was an important treatment goal in the RENAAL study. Blood pressure was aggressively treated in both treatment groups in order to get to equal blood pressure levels between losartan and placebo, a trough systolic less than 140 and diastolic less than 90 millimeters of mercury.
The goal blood pressure was achieved by titrating losartan or matching placebo from 50 to 100 milligrams first and then titrating other open-label antihypertensives from different classes, with the exception of ACE inhibitors and AII receptor antagonists.
The next slide shows concomitant antihypertensive drugs used during the study. As you see from this slide, many drugs from different classes of antihypertensives had to be used in order to control blood pressure in these patients.
It should be noted that in addition to study drug, patients in both treatment groups took an average of 3-and-a-half antihypertensive drugs from different classes. The use of each class of antihypertensive agents was comparable between the losartan and the placebo group.
This slide demonstrates the mean systolic and diastolic blood pressure in each treatment group during the study. The y axis is blood pressure in millimeters of mercury, and the x axis is duration of follow-up. Overall, the mean systolic and diastolic blood pressure were reduced throughout the study. In the losartan group blood pressure declined from 152 systolic and 82 diastolic at baseline to 140 systolic and 75 millimeters of mercury diastolic at the study end, while in the placebo group, blood pressure declined from 153 systolic and 82 millimeters of mercury diastolic to 142 systolic and 75 diastolic at the study end.
Since both systolic and diastolic blood pressure were aggressively treated, the best approach would be evaluation of mean arterial pressure, which reflects both.
This slide demonstrates by percentile the distribution of mean arterial pressure. The y axis is the mean arterial pressure in millimeters of mercury, and the x axis is duration of follow-up. The line in the middle of each box represents the 50th percentile of mean arterial pressure. The bottom of the box is the 25th percentile, and the top of the box is the 75th percentile of mean arterial pressure. The whiskers represent the 5th and 95th percentile of mean arterial pressure in all patients.
As you can see, we were successful in reducing blood pressure in these patients. Mean arterial pressure decreased in all patients throughout the study. Overall, except for the first year of the study, the mean arterial pressure was comparable between the losartan and the placebo group. On the average there was a 2 millimeter of mercury difference in mean arterial pressure, with the losartan group having lower blood pressure. This difference, although small, did achieve a statistical significance.
Adjusting for differences in mean arterial pressure, using a predefined analysis of a time varying covariate, demonstrated that the losartan effect on the primary composite endpoint and on the irreversible clinical endpoints of end-stage renal disease, death, and a composite of end-stage renal disease or death, were minimally affected by this adjustment.
As is demonstrated on this slide, the dotted lines with squares shows the treatment effect after adjustment for mean arterial pressure, using mean arterial pressure as a time varying covariate.
This supports the conclusion that the renal protective effects of losartan on the primary composite endpoint and on irreversible clinical endpoints of end-stage renal disease, and end-stage renal disease or death, is over and above the antihypertensive effect of losartan.
DR. BORER: Excuse me, Dr. Shahinfar, two things. First, because I always keep my word, it's 30 seconds short of 9:45, and we've been importuned to have a break at 9:45 for about 10 minutes, which we will. This seems like a reasonable break point.
At this point I want to just take my 30 seconds to ask you about these data, because to me these are particularly important in judging the independent effect of losartan for this indication.
Here I am even more concerned about the relation of the model to the conclusion, because, in fact, the adjustment makes the drug appear to be somewhat less effective. I don't care about the p values and all because it's hard for me to interpret those in this setting when we're talking about adjustments.
But, did you do anything with blood pressure that was analogous to what you did with protein? For example, you found 4,000, after the fact, that there was an imbalance above 4,000. Wasn't there an imbalance above 160 or 180 in favor of placebo, and is that where most of the events lay in the placebo group? I don't know.
The other question I would ask is why, and we don't need a prolonged answer here, and maybe one of your consultants wants to get up and talk about this, because they know more than I do. You looked at mean pressure and you said that really was the best integrator, and I would wonder about that.
As data emerged it appears that systolic pressure and pulse pressure, particularly in the population that you studied, might be more important, and mean blood pressure would tend to minimize the effect of the widening pulse pressure that you see when you look at the difference in the systolic and diastolic blood pressure curves on drug and on placebo. There is a wider pulse pressure on average throughout the course of the study in the patients on placebo than on losartan. You tend to lose that when you define a mean blood pressure.
So, I would like some quick answers to those questions, and if one of them is going to be long, we'll wait until after the break. But first, how about the cut points for blood pressure?
DR. SHAHINFAR: We didn't do exactly what we did for proteinuria. It's not exactly the same picture. It was to see where are the outliers.
One important factor is that we saw events in all different levels of blood pressure control. But the important finding is that the better the blood pressure is controlled, so even in the placebo group, when blood pressure is controlled absolutely great, we have better treatment effect with losartan, which basically proved that this is not just blood pressure.
The answer to your next question, I think is very important, and I agree with you. In looking at the pulse pressure and systolic and diastolic in these patients, we look at all of those factors. Mean arterial pressure was predefined. What I presented today is the predefined analysis and the correction that we did. Did I answer your question?
DR. BORER: Pretty much. But it looks like we have an additional answer here.
DR. KEANE: Thank you. I'm Bill Keane. I'm the last presenter this morning, but maybe I could just quickly come in here and talk to the issue before we do break. I'm Vice President for Clinical Development in U.S. Human Health.
Let me give you some numbers, and if we want to discuss these more -- we did look at the prespecified, as Dr. Shahinfar has indicated, the mean arterial pressure -- and you've already seen that data.
We've also looked at obviously each of the components, the systolic, the diastolic, as well as the pulse pressure, to see the influence on our risk reduction. In essence, the risk reduction all was approximately 16 percent. For the mean arterial pressure, it was 15 percent. For the diastolic pressure alone, it was a 17 percent risk reduction. For the systolic blood pressure alone, there was a 13 percent risk reduction, and for the pulse pressure, there was a 13 percent risk reduction. So, this basically I think answers the issue that you are raising.
I think that this is an important point because overall, as Dr. Shahinfar has said, we did aggressively lower blood pressure in the trial. We came from about 150 over 80 at the start. We brought it down to about 140 over about 75 at the end. Blood pressure changed, and these were our adjustments as we looked at them.
DR. BORER: We may want to come back to this, but I wonder if anybody, before the break, has any other comments they want to make on the committee about this issue. Otherwise, we can come back to it in later discussion. Tom?
DR. FLEMING: Just very briefly, if that slide can go back up. It confirms what your sense was, Jeff. Basically the imbalance seems to be more with systolic than with diastolic, and so when you adjust for systolic rather than the mean you do see the bigger effect of the confounding.
DR. BORER: Okay. Thank you very much. We may come back to this later.
Let's take a short break now and reconvene at 10 o'clock.
DR. BORER: Okay. Dr. Bain.
DR. BAIN: Dr. Borer, your original question was, was there a difference in baseline blood pressure? Actually you were interested in some of the upper distributions. We looked at the distribution of blood pressure at baseline and there was no difference in the distribution, if you look at the percentiles over time.
Now, this morning already we've talked about two different adjustments, and what I want to do is make sure that everybody understands the two different adjustments we've talked about so far.
The first one we introduced was the adjustment because we observed this difference in baseline proteinuria. As we showed, there were higher levels of proteinuria in our upper stratum, the greater than 2 grams. The rate was higher in the losartan group. Because of that baseline imbalance we showed you, when we adjusted the treatment effect, the treatment effect got stronger because of the losartan group having the higher baseline protein. So, that's the first one we talked about.
Then we introduced another adjustment when Dr. Shahinfar was talking about the difference that we observed in mean arterial pressure over the course of the trial. Okay. So, now we're not talking about baseline anymore. Now we're talking about adjusting for that difference in mean arterial pressure during the study. And what we showed there was, when you adjust for the mean arterial pressure as a time varying covariate, is it had minimal impact on that same treatment affect that we observed on our primary composite endpoint. Okay?
DR. BORER: Okay. Thank you.
DR. KOPP: Could I ask a blood pressure question? You have blood pressures out to 48 months. Were patients censored for collection of that data when they went on dialysis?
DR. SHAHINFAR: No. All patients were included in the analysis, so that included all of them.
DR. BAIN: Let me clarify that. If you are talking about the blood pressures that we were using to adjust for the primary composite endpoint, in that adjustment we were only adjusting up to the time they had the event. We continued to collect blood pressures after that, but the adjustment is only in effect up to the time of the event.
DR. BORER: Dr. Hirsch?
DR. HIRSCH: Thank you. Alan Hirsch.
I want to talk about a little bit of physiology and the pre hoc hypothesis for the adjustments. As a non-nephrologist, for me, I am still a little bewildered by what I would have anticipated the adjustment would have accomplished. In other words, in a patient with greater than 4 grams of urinary protein excretion, was it anticipated that losartan would have a greater impact or a lesser impact on the primary outcome, based on the time to the first event? In other words, if we adjust for those patients who are sicker, would we have anticipated a lesser or a greater effect of the drug?
DR. BAIN: I think the question was, in patients that entered -- we're back to baseline proteinuria -- with relatively high levels of protein, we were adjusting for that imbalance, because that's mainly where the imbalance was. It turns out that, as Dr. Shahinfar has already indicated, in those patients that have high baseline protein -- like, say, greater than 4 milligrams per gram -- 80 percent of them had an event. And when you adjust for that baseline difference, that's what's having the effect to push the treatment effect to be larger than what we observe when we don't adjust for that imbalance in the high levels of baseline protein.
DR. BORER: Does that satisfy you, Alan?
DR. HIRSCH: Well, I'll come back to this later then. I'm not entirely satisfied.
DR. BORER: Before you start the next section, we've been talking about nephropathy and proteinuria, and Dr. Elia said that the terms are going to be used interchangeably. I would just like to hear from the nephrology members of the panel about that equivalence and how we should interpret these data based on definitions and what have you. Dr. Brem?
DR. BREM: Andrew Brem.
I was a little bit troubled I guess by the opening remarks, stating something to the effect that proteinuria will equal nephropathy. Proteinuria is a sign of the nephropathy perhaps, and biopsy is the gold standard to establish what the nephropathy is and to make that diagnosis.
I certainly realize that one is not going to biopsy 1,500 patients, so that's not a possibility, but I think it's important to recognize that we are using proteinuria as a surrogate for the nephropathy, not as the nephropathy itself.
As such, is it well established enough to be used as a surrogate? You're doing your statistics and presenting outcomes based on it as a surrogate. Have we established well enough that higher the proteinuria, the worse the outcome? I think probably much of the discussion that we're having about the statistics right now sort of centers on that question.
DR. GOLDMAN: Yes. That's a great question. Bonnie Goldman, Regulatory Affairs.
I'd asked that we try to come back because we do have data, if the group would like to see it, especially within this study, about essentially what you're asking about the surrogacy. But it literally would take quite a while to go through during the course of this.
The bottom line is, in this study, in fact, there's a very good relationship between the proteinuria and the outcomes. Okay?
DR. BORER: At the end of the day, we are not looking at proteinuria as an endpoint, just as an admission point, so that may mitigate against any concern about that.
DR. FLEMING: I think it just might be worth a quick comment, that we've been thinking and talking about proteinuria now in at least three different ways, as a predictor, as an effect modifier, and as a surrogate. Those are in fact completely separate concepts.
I am persuaded that proteinuria is a predictor. It is related to the risk of the outcome. Whether it's the triple endpoint, the double endpoint, or the single endpoint, the higher the level of baseline proteinuria the higher the risk of these outcomes we're looking at. So, it is a predictor.
Is it an effect a modifier, which might be related to Alan's comment. I'm not sure, but as an effect modifier, the question is, is the treatment effect different in people above 4,000, versus in people 2,000 to 4,000, versus less. I don't know the answer to that, but that's an entirely separate question than whether we should adjust for baseline proteinuria if there is an imbalance because it is a potential confounder.
The third issue is, is it is a true surrogate? That's the question I ask when I wonder is looking at treatment effect in change in proteinuria levels an adequate measure of whether I'm truly achieving clinical benefit. I would agree that that is entirely unestablished, that proteinuria levels would be an adequate replacement endpoint.
DR. BORER: Here we're not doing that, fortunately.
DR. TEMPLE: Tom's second question would be answered by the display that you initially requested, and that I asked for also, which is to look at what the effect of treatment is in the greater than 4,000 and less than 4,000, which is still your poor man's covariate adjustment for people who don't understand covariate adjustments very well. It is very informative. We still hope to see that.
DR. BORER: We poor men probably will see that after lunch, and poor women, if there are any of you among us.
DR. SHAHINFAR: I think this is where we left off.
Next I would like to show you the analysis of primary composite endpoint across predefined baseline subgroups.
This slide is intended to convey by visual impact the consistent benefit of losartan across a variety of subgroups. The primary composite endpoint was explored in 18 predefined subgroups, which are shown on this slide. For each subgroup, the point estimates on the graph are all on the left side of the zero line in favor of losartan. Overall, there was no significant interaction between losartan treatment and these predefined subgroups, except for region.
When testing as many as 18 subgroups, it's not unexpected that by chance alone one subgroup would have a significant interaction. While it appears that the effect of losartan varies among regions, note than none of the point estimates are to the right of the 0 line.
As noted in the FDA review, there is not a significant regional heterogeneity.
DR. BORER: Can we just stop you there for one second? I think that one issue that may come up and that I don't understand so well may have to do with definition.
Dr. Armstrong, did you want to say something about that?
DR. ARMSTRONG: As a non-nephrologist, I would appreciate some discussion about the definition of end-stage renal disease, which is the clinically robust, meaningful component of the double endpoint, which we are relying on. I guess the issue is how systematic is this definition applied across centers and countries, since there were many in your study? To what extent did it relate to some of these other markers such as proteinuria and creatinine?
For example, was there heterogeneity across centers or countries as it relates to the application of this to a non-nephrologist's subjective decision about transplantation or dialysis, which is obviously modulated by several factors. So, I would really appreciate some enhanced understanding around this issue.
DR. SHAHINFAR: That's a very important question, and the criteria for the definition of end-stage renal disease was a requirement for dialysis or transplantation. As you mentioned, there are some countries where there is less transplantation, and there are more transplantations in other countries.
But if the patient required dialysis or transplantation, and the investigator identified that, that this patient needs to be dialyzed, and if the patient refused to be dialyzed, or dialysis was not available, which can happen in some countries, then the Adjudication Committee used the criteria, estimated GFR. The GFR had to be below 10 ml per minute, corrected for body surface area, in order to adjudicate that patient as an end-stage renal disease patient. So, this way we have information on end-stage renal disease in all patient randomized.
DR. ARMSTRONG: May I, Mr. Chairman, just pursue this? So, could you then reassure me about the homogeneity versus the heterogeneity as it relates to the application of this across the centers and countries, and the extent to which the definition was aligned with, as I understand it -- we've just had a discussion about proteinuria for example -- could you comment about alignment of that measurement with this endpoint?
DR. SHAHINFAR: We used a central lab for measurement for all proteinuria, so it was done uniformly across all countries in the world. With respect to the definition of end-stage renal disease, there are very few patients who either required and did not get dialysis or refused dialysis.
DR. HAFFNER: I'm Steve Haffner. I was chair of the Events Committee. It turns out that very few people actually fit into this category. I think there were 16 subjects. 10 were in one group and 6 were in the other; 10 in placebo and 6 in losartan. I couldn't remember which group it was. It turns out that it was a very small percentage of the people who were identified as end-stage renal disease. We had to go to the Walser formula to characterize them. Of those, I think most were refusals, and not the absence of dialysis as a practical issue.
DR. BORER: Looking at it the other way, was there a substantial number of people who were put on dialysis with GFRs that were greater than 10 milligrams per hour or cc per hour, whatever it is, adjusted for body surface area? No.
DR. SHAHINFAR: We did not systematically measure estimated GFR. When the physician made a distinction that this patient needed to be dialyzed, based on the condition of the patient and requirement for dialysis, and if that patient was dialyzed, that was captured as dialysis.
DR. BORER: So, there might be some non-homogeneity across regions in that regard or no?
DR. HAFFNER: We can look at that but we have not looked at that at this point. Once they went on dialysis, we didn't calculate Walser's.
There's another issue -- there were some people who got end-stage renal disease who were not classified as doubling. That can happen for two reasons. One is that serum creatinine didn't actually double, or two, in order to be called doubling, they had to get two measures of serum creatinine, and some people were put on dialysis before the second confirmation of it.
We actually had long discussions that if there was doubling after dialysis -- you'd figure it would probably double before dialysis -- but we excluded that from the doubling. So, we took what I think is a fairly rigorous point about doubling, and the second measurement also had been done by the central lab. So, there were a few people who didn't make that, but we could do that analysis. But we have not yet by region.
DR. BORER: Dr. Nissen?
DR. NISSEN: Could you put that slide up on the various subgroups please? I need some clarification because the slide is at variance with the briefing document we received from the FDA.
According to our briefing document, the hazard ratio for the Europeans is actually 1.05, so it actually goes in the wrong direction, and you show it in the favoring losartan group.
I guess what I want clarification about is that it looks to me, from the FDA briefing document, that virtually the entire benefit for the whole cohort comes from the group from Asia. That is, if you look at the rest of the cohort, there are virtually no effects. The 20 percent of patients who were from Asia completely drove the endpoint in the trial. Is that correct or incorrect?
DR. SHAHINFAR: You're referring to the region?
DR. NISSEN: Yes.
DR. KEANE: Again, I'm Bill Keane. Maybe I can take a shot at that at this point in time. I'm in clinical development. I think we were going to get into some of these questions later on, so we can answer it now or we can postpone it to later on.
I think one of the important points that Dr. Shahinfar has already mentioned to you is that in fact there was by region a slight interaction that achieved a level of about .047 that was seen in the Asian group. I think as we looked at this with the FDA and in greater detail, actually the FDA -- and I think that's in your briefing document itself -- did not really underscore that this was an important interaction.
Now, part of the reason for that is the way we defined regions, first of all. The region was a geographic region that included not only the Pacific Rim but also included the Middle East as well, specifically Israel. There were, from a regional perspective, just inherent reasons in our definitions that maybe this might have been shown as a difference basically, with the point estimates slightly to the left for Asia.
DR. NISSEN: No, but your point estimate for Europe favors losartan, and in the FDA briefing document the point estimate favors placebo. So, in Europe this doesn't gibe with the FDA's data that we got.
DR. BAIN: Ray Bain, Clinical Biostatistics, Merck Research Labs.
If you look at the Merck briefing document, they are all in the right direction. I assume possibly that the FDA was doing a different baseline adjustment. I have to look at their briefing document to figure out exactly what the slight differences were.
DR. NISSEN: For the committee, I am looking at page 22 of the FDA's briefing document, and the data don't agree with the data up there. So, I think we really do need to clarify this from the agency as to what are the correct numbers here?
DR. BAIN: We also need to look at page 63 of the Merck briefing document because that's what this slide is based on. In the Merck briefing document, we're adjusting for baseline protein as a stratum, but not region, since region is the variable that we're looking at. So, if you look at page 63 in the Merck --
DR. NISSEN: I now understand the difference. So, what you did was you took the post hoc adjustment for protein and added it to this.
DR. BAIN: No. We took our prespecified adjustment, which is the categories of baseline protein. That was our prespecified analysis. In that analysis those numbers on page 63, although they're in hazard ratios, so you just take 1 minus that times 100, you'll get everything on the left. For example, in Europe it's .943, which is a risk reduction of roughly 5 percent.
DR. NISSEN: Tom Fleming, maybe you can help me understand this. I don't understand why the two don't agree here.
DR. FLEMING: I think we need to sort this out, as we have a chance maybe over the break, to get a clear -- my sense is it has to do with the difference in either -- and I don't think it's this exactly -- who's in which region or what covariates are used in these two analyses for adjustments. I would suggest maybe at the break we can get that ironed out.
DR. NISSEN: All right. Maybe not now, but I do want to come back to this question later on whether the entire treatment effect for the entire study was driven by the 20 percent of patients that were in Asia.
DR. FLEMING: Just on that point, Steve, in either of these two analyses the same general, at least qualitative picture is that Asia carries the greatest part of the signal in either of the analyses.
DR. NISSEN: The overwhelming majority of the signal.
DR. BORER: Dr. Konstam.
DR. KONSTAM: Hi. I'm Marv Konstam from Tufts New England Medical Center. I am here as a paid consultant to Merck.
You know, I just wanted to comment from my own experience and perspective about subgroup analyses, and I think the panel is all aware about how treacherous this can be of identifying one subgroup.
I just want to point the panel to the FDA medical reviewer analysis of this, because I personally found it enormously instructive. Specifically, in the medical review, page 24, figure 7, it breaks it down by individual countries. What's nice about it is that it shows you the impact of size of the particular subgroups on the relevance of the relative risk that comes out.
You can break these down all sorts of different ways, but if you look at this analysis, the individual countries really are all over the map. Actually, it's interesting that the big Asian contribution comes from Israel, by the way, so you can think about that.
But, I just want to point that out. I think to me this was a very instructive way of looking at it, and sort of not focusing too much on this Asian business is at least my perspective on it.
DR. BORER: Dr. Kopp, you had a comment. Before you do, I just want to point out, Dr. Konstam, I never understood you to be such a humorist. In fact, the countries are all over the map.
DR. KOPP: I wanted to follow-up on a response to Dr. Armstrong's point. I think we had a lot of discussion about how to adjudicate people at the low end of the GFR spectrum who weren't placed on dialysis. I think it is worth thinking a little bit about the patients with presumably GFRs between 10 and 15 or 10 and 16, whatever, who were started on dialysis, and admit that there is undoubtedly both heterogeneity across countries and within countries about whether a particular patient gets started on dialysis on the typical day, and think about what the indications would be. They would certainly, for most nephrologists, include fatigue, anorexia, falling albumin, among the key criteria that would be looked at, that I think would be unrelated that I can think of to the therapy in question.
But there are two that at least might play a role that could potentially could be related, and one would be potassium levels, difficulty maintaining potassium, admittedly not too much of a problem with chronic renal failure patients. The other would be difficult to maintain congestive heart failure, a patient for whom diuresis was becoming increasingly difficult. Each of those at least -- and here I'm speculating -- run in opposite directions. That is, a losartan patient might be more likely to have hyperkalemia than a placebo treated patient, but less likely to have difficult to manage congestive heart failure.
Having said that, there is no way we can really be sure, and I guess the major protection is this idea that it was all placebo-controlled, and so the decision was being made by somebody without advance knowledge about how the treatment was based.
But at least maybe you'd comment upon the possibility of how some of those factors that are potentially related to losartan might play into the decision to start dialysis.
DR. BORER: You can hold that until you have a chance to think about and come back to it later if you like.
Dr. Lindenfeld, did you want to make a comment?
DR. LINDENFELD: Just a quick question. When we come back to this issue of region -- and I think we fully understand the problem with subgroups -- I wondered if we could see by these regions the numbers of patients that had a doubling of creatinine versus end-stage renal disease. If we could divide those up by regions and just see if that was different by region.
DR. SHAHINFAR: We will show that.
DR. BORER: While you're doing that, I think there will be time to respond specifically to Dr. Kopp's question. I know you're going to show us a cardiovascular endpoint analysis, where the heart failure issue will be death with and a safety analysis where the hyperkalemia issue and its relation to sudden death will be dealt with. So, let's bookmark that so you can specifically respond to the question when you get to it.
DR. SHAHINFAR: There were two prespecified secondary hypotheses to support the renal protective effect of losartan. The first was that losartan, compared to placebo, would reduce the rate of progression of renal disease as measured by the slope of reciprocal of serum creatinine.
This is an important measure that is used by clinical nephrologists to predict the time to dialysis for the patient. Note that the literature on nephropathy indicates that the inverse of serum creatinine value for an individual tends to fall in a linear fashion over time. The slope of this line is an indication of the speed of progression of renal disease. The more negative the slope, the faster the progression. This analysis takes into consideration all patients, not just those patients who have reached a renal endpoint.
Another secondary endpoint hypothesis in RENAAL was that losartan would reduce proteinuria during the course of the study, compared to placebo. The next two slides demonstrate the effect of losartan on these secondary hypotheses.
For the analysis of the progression of renal disease, we calculated the slope for each patient, and present here in this chart the median slope for two treatment groups. As I referred before, the more negative the slope, the faster is the progression.
This slide illustrates the rate of loss of renal function as measured by the slope of reciprocal of serum creatinine. On the y axis is the change in the slope of 1 over sCr. There was a significant reduction in the rate of loss of renal function by 18 percent with losartan with p equals .01.
Another secondary hypothesis was the effect of losartan on proteinuria. Reduction of proteinuria has been considered an important therapeutic target for treatment of diabetic nephropathy among clinical nephrologists.
On this slide the y axis demonstrates the mean percent change based on geometric mean for urinary protein excretion. The x axis is the duration of follow-up.
Losartan was associated with a reduction in urinary protein excretion of 25 percent within 3 months, and this reduction was maintained throughout the study. There was an overall 34 percent reduction in proteinuria compared to placebo, with p less than .001.
The significant treatment effect of losartan on proteinuria was unchanged after adjustment for blood pressure at each time point, supporting the conclusion that the antiproteinuric effect of losartan is over and above its blood pressure lowering effect.
In summary, in type II diabetic patients with proteinuria, losartan is renal protective by delaying the onset of the primary composite endpoint of doubling of serum creatinine, end-stage renal disease, or death. In the entire cohort losartan reduced the risk of end-stage renal disease by 28.6 percent.
Losartan reduces the rate of decline of renal function as measured by the slope of reciprocal of serum creatinine.
Losartan reduces proteinuria, and has a beneficial effect on the primary composite endpoint and proteinuria beyond its beneficial effect on blood pressure.
I would now like to turn to a discussion of cardiovascular secondary hypotheses. We designed RENAAL as a renal protection study, but recognizing the importance of cardiovascular events in type II diabetic patients, we made a cardiovascular morbidity and mortality a prespecified secondary hypothesis, and adjudicated cardiovascular outcomes.
The cardiovascular hypothesis in RENAAL was based on the effect of losartan on cardiovascular morbidity and mortality. We hypothesized that losartan compared to placebo would increase the time to the first event of the cardiovascular morbidity and mortality which was a composite of cardiovascular death, myocardial infarction, stroke, first hospitalization for heart failure, first hospitalization for angina, and revascularization. The next three slides summarize cardiovascular secondary endpoint data in RENAAL.
This Kaplan-Meier curve demonstrates the result of the secondary composite endpoint of cardiovascular morbidity and mortality. Although numerically lower on losartan, there was no statistically significant effect of losartan compared to placebo on the composite endpoint of cardiovascular morbidity and mortality.
This plot demonstrates the effect of losartan on cardiovascular morbidity and mortality and the individual component of this composite endpoint. The risk reduction and 95 percent confidence interval are presented for the composite endpoint and each of the components of this composite endpoint: cardiovascular death, myocardial infarction, stroke, first hospitalization for heart failure, first hospitalization for angina, and revascularization, both coronary and peripheral revascularization.
As you see, the point estimates are all distributed around the 0 line for different components, but overall there is no significant difference in the composite endpoint of cardiovascular morbidity and mortality.
During the January 17th advisory committee meeting for irbesartan, this committee expressed interest in the effect of irbesartan treatment on the composite of renal and cardiovascular endpoints in the irbesartan diabetic nephropathy trial. In light of this we performed a post hoc analysis, time-to-event analysis, of the composite of irreversible clinical endpoints of end-stage renal disease, myocardial infarction, stroke, or death in the RENAAL study.
As is demonstrated in this Kaplan-Meier curve, losartan significantly reduced this combined endpoint by 21.2 percent; p equals .003. This analysis demonstrates that the renal protective benefits of losartan do not come at the expense of increased risk of cardiovascular endpoints, and in fact demonstrates the benefit of losartan in this population.
As mentioned earlier, we designed RENAAL as a renal protection study, rather than a cardiovascular study. Therefore, we enriched the population with patients at high risk for progression of renal disease.
In summary, in RENAAL there was no statistical significant difference in cardiovascular morbidity and mortality.
Post hoc analysis of end-stage renal disease, myocardial infarction, stroke, and death indicate that renal protective benefits of losartan in RENAAL did not come at the expense of increased risk of cardiovascular events, and therefore supports the overall benefits of therapy in these patients.
DR. BORER: Excuse me one second.
Dr. Kopp, do you want to come back to your issue here. We have just seen the total cardiovascular endpoints. The sponsor hasn't broken out the heart failure, which they did in their briefing document. They did actually have one slide that did. Did you want to restate your question, or are you satisfied with what you've gotten here?
DR. KOPP: I don't think it directly addresses the issue. Is it possible that when a clinician is sitting with a patient with a creatinine of 5.5 and trying to decide does this person needs dialysis, would the presence of heart failure or the presence of hyperkalemia have tilted one way or the other that decision. But I think we may get that later from Dr. Haffner's group.
DR. HAFFNER: I don't think we're going to get really good data on this. We did not ask people, when they went to end-stage renal disease, to answer a special questionnaire on why they went to end-stage renal disease. Was it dialysis, hyperkalemia, nausea and vomiting?
We do have some data on prior events for CHF. We have events on potassiums. But it would be linking data and this isn't all entered. My guess is eventually we could probably figure some of this out, but because they weren't done at the time of dialysis, I'm not sure it's going to be really good data, to tell you the truth.
What we clearly could do is we can calculate by regions for instance the Walser to see if creatinine clearance differed by region. We could clearly do that. I am not sure we can do that today. We can clearly do that by whether they were on losartan or in the placebo group, so we can see whether this is systematic bias by area for people who enter end-stage renal disease.
But I don't think we're going be able to get a detailed thing about hyperkalemia versus failure. We can look at the events prior to it. We haven't done that because that doesn't enter into the composite. But I am really not so sure, as an epidemiologist who works on clinical trials, how worthwhile that data really is.
DR. BORER: Dr. Lorell.
DR. LORELL: Yes, I wonder if you can provide us with data, in regard to the all-clause mortality, as to a breakdown of cause of death, including what percent of deaths were cardiovascular deaths? I think it is a potentially important issue for this committee to discuss for a couple of reasons.
One is that this trial was, in fact, shaped and stopped early because of ethical safety concerns of the committee regarding the growing evidence of cardioprotective effects of ACE inhibitors on multiple cardiac events. That letter that was sent to the investigators is in our brochure.
So, it raises the question, in regard to your last slide, as to the real issue, a real issue that was also addressed by the Safety and Monitoring Board, not whether the renal protective effects of losartan were at the expense of increased risk, but were at the expense of the absence of cardioprotection from a competitive therapy. That was an issue that clearly was a concern for the Safety and Monitoring Board.
So, I think it would be of interest for this committee to know the breakdown of all-cause mortality, if you have that data.
DR. SHAHINFAR: Can I first clarify one question you raise? You mentioned the Safety and Monitoring Committee. In fact, the study was not stopped by the Safety Committee. It was stopped by the blinded Steering Committee, who didn't have access to unblinded information. So, that decision was clearly unrelated to internal information from RENAAL. That was the decision of the Steering Committee.
Now, with respect to the causes of death and all-cause mortality, would you like me to present that now, or do you want me to present the safety and we can present all the information?
DR. LORELL: Why don't you include it whenever you think it would be fit. That would be fine.
DR. SHAHINFAR: If you agree, I'll finish the presentation of safety. Thank you.
Now I would like to present to you the RENAAL safety results. Overall, the safety profile of losartan in this study was consistent with that listed in the U.S. prescribing information for losartan. In this population of patients with type II diabetes and underlying kidney disease, we expected to see many clinical or laboratory adverse experiences. In fact, this is demonstrated in high event rates of clinical adverse experiences in both treatment groups.
On this slide, on the y axis we present the percentage of patients with clinical adverse experiences. For clarity the percentage of patients are also shown on the top of each bar. Up to 95 percent of patients in each treatment group had clinical adverse experiences during the study. There was a slightly higher number of drug-related clinical adverse experiences in the losartan group. There were more patients on placebo who had discontinued for clinical adverse experiences. The number of patients who died because of clinical adverse experiences was comparable between losartan and placebo group.
With respect to laboratory adverse experiences, we see a similar pattern. It is important to note that the laboratory adverse experiences were those that were reported by the investigators, not based on a predefined laboratory value. The number of patients with drug-related laboratory adverse experiences was higher in the losartan group. This was mostly attributed to a higher number of patients with hyperkalemia. No patient died because of a laboratory adverse experience.
Because of the population that we studied in RENAAL, we predefined six adverse experiences of interest, and performed a prespecified analysis. This table provides the results of our analysis on the six predefined adverse experiences that are listed on the left column.
The reason for selecting the six adverse experiences was that acute renal failure has been reported by ACE inhibitors and AII receptor antagonists. Hyperkalemia and hypokalemia were adverse experiences of interest since we were studying diabetic patients with underlying kidney disease. Anemia has been reported with ACE inhibitors and angiotensin II antagonists. It is listed in the prescribing information for losartan. It is also common in patients with advanced renal disease. Since we studied diabetic patients, hyperglycemia and hypoglycemia were also adverse experiences of interest.
Except for hyperkalemia with losartan, and hypokalemia in the placebo arm, there was no significant difference in any of the predefined adverse experiences between losartan and placebo. Hyperkalemia is not unexpected in patients with type II diabetes and underlying kidney disease, especially when a drug that blocks the renin-angiotensin-aldosterone system, such as losartan is used.
In contrast, the higher incidence of hypokalemia in the placebo arm is probably related to the high use of diuretics. Hypokalemia is important in patients with chronic renal disease. These patients have low bicarbonate, and therefore low serum potassium levels are a true reflection of low intracellular levels of potassium.
This slide demonstrates the distribution of serum potassium by percentile in the two treatment groups during the study. The y axis is the level of serum potassium in milliequivalents per liter. The x axis is duration of follow-up. The line in the middle of each box represents the 50th percentile of serum potassium. The bottom of the box is the 25th percentile, and the top of the box is the 75th percentile of serum potassium. The whiskers represent 5th and 95th percentile of serum potassium at each time point.
Overall, mean serum potassium was significantly higher in patients who were treated with losartan at each time point. However, the mean differences did not exceed .3 milliequivalent per liter. 95 percent of all patients' serum potassium values were below 6 and above 3.5 milliequivalents per liter at each time.
This table shows the number of patients who had serum potassium less than or equal to 3.5 and equal or greater than 6.0 milliequivalents in each treatment group at any time during the study. As you can see on this slide, there were more patients in the placebo group with potassium less than 3.5 and more patients in the losartan group with potassium greater than 6.0 milliequivalents per liter.
There were more patients with adverse experiences of hyperkalemia in the losartan group. The number of patients with serious adverse experiences of hyperkalemia was small, but higher in the losartan group. Relatively a small number of patients had to be discontinued for hyperkalemia in each treatment group, indicating that hyperkalemia was clinically manageable in these patients. No deaths were attributed to hyperkalemia during therapy.
DR. BORER: Dr. Shahinfar, can you just go back to that slide, the last one? That's it. I don't want to overstate the case here. You've shown us total mortality. You're going to show us the breakdown of cause of death, and that's of interest, and overall the difference in deaths was so small that I think that this may not be a major issue.
But you say that no death is due to the adverse event or has been attributed -- that was what you said -- attributed to the adverse event of hyperkalemia, and I'm sure that's true. I think the only way you could, with certainty, make such an attribution is, the last time you saw the patient alive, the potassium was high.
But it's interesting that nominally there is an excess of sudden deaths in the losartan group from your briefing document. Not a tremendous number, but an excess percentage of sudden deaths nominally. It is not statistically significant.
I have to wonder whether -- and this may relate more to labeling if the drug is ultimately approved for this indication -- whether this may not have been related somehow to electrolyte imbalance that just wasn't picked up because deaths occur when they occur, not necessarily right after the last lab value was done.
Can you just comment on that?
DR. SHAHINFAR: What I would like to do, basically in response to all-cause death and cardiovascular death, and for sudden death that you just mentioned, I would like to ask Dr. Jonathan Fox, who is the cardiologist at Merck, to respond to this question.
DR. FOX: Thank you, Dr. Shahinfar. My name is Jonathan Fox. I am a cardiologist with cardiovascular clinical research, Merck Research Labs. I will try to address the question. If I could have slide 1228, please.
Just to remind members of the committee, you have already seen these data. These are the data described in the cardiovascular composite endpoint and the components, and I believe you're most interested, as Dr. Lorell already pointed out, the cardiovascular deaths. What I'm going to try to do over the next few slides is to walk you through all of the adjudicated causes of death, and to focus on those that were adjudicated as cardiovascular causes.
This table shows you a breakdown of the adjudicated causes of death by the Endpoint Committee on the entire intention to treat population. Those causes were adjudicated into the following categories: fatal myocardial infarction, known non-cardiovascular cause. In other words, that included categories of non-cardiovascular death, for which there were sufficient data or other clinical information to allow the Endpoint Committee to come to that conclusion.
There was a category of "not determined". There were other cardiac causes, and those could include arrhythmia, for example; other vascular causes, including hemorrhagic death, progressive heart failure, and the category of sudden cardiac death, which Dr. Borer has already pointed out. There was a numerical imbalance in favor of placebo 45 deaths compared to 30 in the placebo arm, for a total of 158 deaths in the losartan arm and 155 deaths in the placebo arm, which contributes to the overall Kaplan-Meier curve you saw earlier, which were almost entirely superimposed.
Next, please, 1231. This is a subset from that same table, so the totals are different. These are Endpoint Committee adjudicated causes of cardiovascular death in the entire intention-to-treat population.
So, again, the categories are exactly the same as you saw in the previous slide, except that the non-cardiovascular causes have been omitted. The upshot of this slide is in the last category of "sudden cardiac death", where, again, you see 45 deaths in losartan and 30 in placebo. But in terms of the percentages, those have now been boosted by the removal of the non-cardiovascular deaths, so that the sudden cardiac deaths comprised 50 percent of those adjudicated causes of cardiovascular death in the trial, compared to 38 percent.
1232 please. Now, I think it's important at this point to point out what the definition was of sudden cardiac death that was used by the Endpoint Committee to adjudicate these cases of patients who died. If there were insufficient clinical information or other supporting data for the Endpoint Committee to categorize patients' death events as one of the other categories of cardiovascular death, for example, fatal myocardial infarction, those patients were adjudicated as sudden death, regardless of what the actual mechanism might have been. There just was not sufficient information.
So, the definition of sudden cardiac death in a way becomes a default category when there is a lack of specific information. So, that included any death occurring without warning, any death occurring within 24 hours of new symptoms, and any unwitnessed death at home.
Let me just pause there and ask if that satisfies Dr. Lorell and Dr. Borer, if that answers the question.
DR. BORER: Dr. Lorell.
DR. LORELL: Thank you. That's very helpful. Just to be very clear to me, because I didn't see either of those two tables in our briefing books, could you just restate, of the total deaths in the losartan and placebo groups, what percent of total deaths in each group was cardiovascular?
DR. FOX: I believe we can get that information from 1230. There were 90 cardiovascular deaths in the losartan arm out of a total of 158. There were 79 cardiovascular deaths out of a total of 155 in placebo.
DR. LORELL: Thank you.
DR. ARMSTRONG: I found this safety presentation very helpful. Thank you. As an entree to my question, I just want to recap my understanding and focus on hyperkalemia and the relationship between hyperkalemia and the endpoint, i.e., transplantation or dialysis. As I understand it, 1 out of 4 losartan patients and 1 out of 8 placebo patients had hyperkalemia, and 1 out of 10 losartan patients had hyperkalemia in excess of 6, and 1 out of 20 placebo patients had hyperkalemia in excess of 6.
As I understand it, also, this hyperkalemia was perceived to be clinically manageable, and none of this hyperkalemia was perceived to have related to death, notwithstanding the chairman's caveat.
My question is, if you look at hyperkalemia, all comers, what proportion of those patients went on to dialysis or to transplantation in the context, as Dr. Kopp pointed out, of that being a meaningful indication for that endpoint, and what proportion of the losartan versus the placebo hyperkalemia led to the development of that endpoint?
DR. SHAHINFAR: We can get that information for you. We haven't looked at the relationship between hyperkalemia and initiation of dialysis, if I understand your question correctly. We can look into that and get back to you.
DR. KEANE: I think one has to recognize that our values of potassium that were determined were determined on a regular interval, and actually dialysis, as you're rightly pointing out, is a therapeutic intervention that may occur. So, our ability to measure potassium immediately prior to the initiation of dialysis is really not within the design of the trial. So, we really don't have that kind of information those kinds of patients that are actually starting dialysis, to define whether that was truly an indication for initiation of dialysis. I don't think we're really going to be able to adequately define with precision and clinical reliability the exact answer to your question.
DR. BORER: Dr. Kopp?
DR. KOPP: Just to clarify -- if a nephrologist had a potassium come back of 6, he could call the study and say, tell us if the patient is on losartan or not, and he would have the option of stopping the study drug -- well, I guess at that point they would come off the study in either case. But that would be an alternative to putting the patient on dialysis.
DR. SHAHINFAR: They managed patients as they would manage these patients in their clinical practices. Only 6 patients were unblinded throughout the study, so we did not unblind every single hyperkalemia in the study. It was not actually the reason for unblinding. Is that the question you're asking?
DR. KOPP: Is that because the physician never contacted the study, or they weren't allowed to ask that question?
DR. SHAHINFAR: No, they were allowed to ask any question, but they never asked that question. In fact, the reason for unblinding those 6 patients was not by the investigator; it was by a cardiologist. The patient had heart failure, MI, and they asked for the unblinding so they could make a clinical decision in that patient.
Hyperkalemia was overall managed. We had guidelines in the protocol of how to handle hyperkalemia, and people were used to handling these patients with hyperkalemia and underlying kidney disease. So, they used potassium lowering agents in both treatment groups.
DR. BORER: Just for clarification, in the FDA briefing document there is a table that presents all the patients who were unblinded, so we have that information.
Let me just remind the committee or ask the committee, because there will undoubtedly be some discussion as we move forward through the summary here, if you want to say something and you press the button on your microphone, the light comes on and I catch it immediately. We've taken care of 99 percent of the problem that we've had by shifting the table so you don't have to break your neck to see the slides. But what that does is cause everybody to focus on the slides instead of on the hands being raised. So, if you press the button that will make it easier to recognize whoever wants to talk.
DR. FOX: Jonathan Fox, Merck Research Labs.
Mr. Chairman, if I could just expand just a little bit on the death issue. I think this will also perhaps anticipate the intersection between the concerns of several of the committee members, in terms of the relationship between hyperkalemia and sudden death, which I believe you raised yourself earlier. If I could have 1236 please.
This slide is a table of some selected laboratory test results that were associated with patients who died in the trial, specifically patients who were categorized in that adjudicated category of sudden cardiac death that I explained a moment ago.
The first column is the last central lab measure that was obtained prior to death, and I'll tell you something about that time interval in a moment. It is broken down between the losartan arm and the placebo arm.
We have included some laboratory measures that I think are relevant in this patient population who is quite ill in the trial. It would be the serum potassiums and the serum bicarbs, blood glucose, in this diabetic patient population, the serum creatinine, and the urine protein.
The conclusion I would like to leave you with from this slide is that in terms of overall means, for what that's worth, the potassiums were in the normal range at that last central lab visit. The bicarbs were near normal. Blood glucose was reflective of this diabetic population, but not indicative of hypoglycemia. Serum creatinines indicated that these patients were well along the way in their nephropathy, and that's also reflected in the means urine protein.
Could I have 1237? This table shows you the results of the deaths of any cause of death in patients, and their potassium results if they had a potassium result less than or equal to 3.5 milliequivalents per liter at any point during the study.
So, in the losartan arm there were 70 patients who fell into that category and 90 in the placebo arm. Those are all patients who had that measurement. There were 12 patients in that category who actually died during the trial and 11 in placebo. 10 of the 12 in losartan were categorized as 1 or more causes of cardiovascular death, and 3 of those 10 as sudden cardiac death. 2 of the 11 patients who died in the placebo arm were cardiovascular deaths, a 1 of those was a sudden cardiac death.
Also on this next slide, these are investigated causes of death in those same patients who had any measure of potassium less than or equal to 3.5 at any point during the study.
So, I want to emphasize that, independent of what the adjudication was of the Endpoint Committee, again, they relied on a strict set of criteria that had to be satisfied by a sufficient body of information for them to adjudicate a particular death to one of those categories that I showed you earlier. So, if there is insufficient evidence available to them, many of those patients were categorized as sudden death.
So, independent of that adjudication, this table shows you clinical information that was obtained from the CRF narratives entered by the investigator. As you can see, there is quite a broad collection of different diagnoses without any particular pattern that I can discern for you in this table.
Now, this takes a look at the same kind of an examination of the information that we have, but looking at patients who had a measurement of potassium greater than or equal to 6 milliequivalents per liter at any time during the study. There were 123 patients in the losartan arm and 61 in the placebo. Of those patients, of the 123 in losartan, 23 patients died, 7 of those had a cardiovascular death, and 4 of those were adjudicated as sudden cardiac death. 61 patients in placebo had the results of a potassium greater than 6. 13 of those patients died. 6 of those had a cardiovascular death, and 2 were categorized as sudden cardiac death.
This is a similar table to what you saw a few slides ago. The investigator reported causes of death in patients with a potassium result greater than or equal to 6 at any time during the study who were adjudicated as cardiac death. Again, it contains quite a broad range of diagnoses without any particular pattern.
I promised you that I was going to show you some information that related the measurement of serum potassium to the time interval between when that last laboratory measure was obtained and when the patient actually died.
What this plot shows you on the ordinate is the last potassium value that was obtained that went to the central laboratory, and on the abscissa, the time in days to the death event from that last measurement. As you would expect, there is a scatter of those data, both in terms of the values of potassium on the y axis, and a scatter with respect to time on the x axis.
The legend is shown underneath the title of the slide: in yellow, losartan; in white, placebo. The circles are patients who were still on study therapy at the time they died. The triangles are those were off therapy at the time they died. In fact, there was a group of patients you might consider outliers and just in anticipation of the question as to what happened to those 2 patients who had values at or near 6 or above at the time they died and close in time to the time they died.
This is a listing of some clinical characteristics of those 2 patients. The first patient, allocation number 4141, was a patient who had already achieved end-stage renal disease and was on dialysis. This laboratory result was obtained approximately 6 days prior to death. It was obtained immediately prior to the dialysis session.
Concomitant medications the patient was taking are listed there; and concurrent condition, as I mentioned, the patient already had end-stage renal disease.
The other patient, allocation 4455, had the laboratory measure taken approximately 4 days prior to death, was on the medications you see listed there. Of note, the patient had chronic acidosis as revealed by the serum bicarb measurement that was obtained at the same time.
Is that helpful to the committee?
DR. BORER: It is to me. Let me ask you one more question before we get to the summary. You know, I am a cardiologist. I don't know much about endocrinology, but you have a lot of fire power sitting there.
It was interesting to me that on losartan and on placebo, although the frequency of hyper- and hypoglycemia was not statistically significantly different in the two groups, there was a different direction in the trend of the events; that is, hypoglycemia up in one, and hyperglycemia up in the other group. Although statistical significance wasn't reached, and in fact, even the nominal criterion for a statistical trend wasn't reached -- the p values were both about .12 and .15, -- still, the fact that they were directionally opposite was interesting to me. Is there any reason why losartan should affect blood glucose differently than placebo?
DR. HAFFNER: This is interesting. I'm not sure it's going to be resolved, but I'll tell you what I believe. I was one of the PIs in the diabetes prevention program. This is an area we're interested in.
First of all, the data you've seen is among deaths. So, it's not truly representative of the overall population.
The hemoglobin A1Cs during the trial did not differ, and that suggests to me there may not have been a major effect.
On the other hand, there's some very interesting literature with both ACEs and now with ARBs, that maybe these agents could prevent type II diabetes.
Now, the mechanism is, again, kind of interesting. There is one report of losartan improving insulin sensitivity, but it's a relatively small study, and it clearly needs to replicated.
So, I think the safe answer would be that the data within this trial does not provide strong evidence for a glycemic effect. There is a related trial we haven't talked about, which did sort of quasi show an effect relative to atenolol, but it could have been the atenolol or it could have been the losartan. And there are certain equivalent data with captopril. I think it's not quite resolved at this point.
DR. SHAHINFAR: Just answering, glycemic control was comparable, and glucose levels were comparable between the two treatment groups throughout this study.
DR. LINDENFELD: JoAnn Lindenfeld. I'd like to just ask a couple of questions before you finish.
Patients were withdrawn from ACE inhibitors, about half the patients in this study as I understand it. There was a 6-week period between withdrawal and randomization. Am I correct there?
DR. SHAHINFAR: Yes.
DR. LINDENFELD: Can you assure me that those groups were equal in the randomization -- in other words, patients who were withdrawn from ACE inhibitors were equal in both groups?
DR. SHAHINFAR: Yes. There was an equal number of patients in each treatment group.
DR. LINDENFELD: For my interest, and just a quick answer if you have it -- if not I don't desperately need it -- is there a difference in the results of the study in patients who were previously on ACE inhibitors?
DR. SHAHINFAR: There is no interaction between losartan treatment and use of ACE inhibitors on the primary outcome of the study.
DR. LINDENFELD: I don't mean the use of ACE inhibitors during the study.
DR. SHAHINFAR: At baseline, that's right. And that's in the background that we provided.
DR. LINDENFELD: Okay. For clarification, when we come back to discuses cardiovascular endpoints, I know there were a number of exclusions here with recent MI, et cetera. Can you give me a percentage of patients who had had a cardiovascular endpoint at baseline prior to randomization? In other words, what percentage of patients in this trial had had some previous cardiovascular endpoint? I know MI was excluded within 6 months, but how many patients had had a distant cardiovascular endpoint of some type? Again, I'm just trying to get at the risk of this group.
DR. SHAHINFAR: This is the history for myocardial infarction, angina, stroke.
DR. LINDENFELD: So, less than 1 percent had had a previous cardiovascular event?
DR. SHAHINFAR: Yes, that's right.
DR. LINDENFELD: Okay. That's 12 percent. I'm sorry. Right.
DR. CARABELLO: What was the mean time to difference in endpoint? The time difference. In other words, the losartan delayed the endpoint. What was the difference in time between when patients reached the endpoint?
DR. SHAHINFAR: Our analysis was a time-to-event analysis. In the study, during the 3-and-a-half years period of follow-up, we were able to demonstrate that for every 16 patients treated, 1 ESRD was delayed basically or was prevented.
DR. BORER: But Blase is asking a different question, which is of interest, although I understand that it's difficult to draw firm inferences since some people didn't ever have an endpoint. Among those who had an endpoint, what was the average time to endpoint in the placebo group versus the average time to endpoint in the losartan group? Do you have that? I think there is something in the FDA briefing document about it. Do you have those numbers?
DR. KEANE: That's actually a relatively difficult number to calculate as you would imagine. I think as Dr. Shahinfar has already indicated, what was easier to identify and to calculate is really the number needed to treat, which was 16 patients needed to be treated to prevent 1 case of end-stage renal disease.
We were also able to demonstrate that we reduced overall end-stage renal disease days by some 32 percent in the losartan group. So, that gives you an estimation of the impact.
Considering the fact that the diabetic patient today represents almost 40 percent of end-stage renal disease, and we have already projected by the end of this decade that it's going to up to 50 percent, that's a substantial impact I think from a public health perspective.
DR. CARABELLO: I'm still not clear. We have X number of patients that reached an endpoint on losartan, and X number of patients that reached an endpoint on placebo, and each patient reached that endpoint in a certain number of days from when he or she began the study. My question is, what's the difference in days or months between those two endpoints?
DR. KEANE: I think we can go ahead and get that analysis for you in a little bit more specific way. I don't have the exact answer right now beyond what I've given you, but we'll provide that in a moment or two.
DR. CARABELLO: Thank you.
DR. FLEMING: Yes, I think it's in the statistical review and evaluation hand out. I think what you're looking for is on page 8, which is best obtained with a Kaplan-Meier estimate of time-to-event distribution. Is that what you're looking for? The difference in the median times to events by Kaplan-Meier estimates?
It is on page 8 of the statistical review and evaluation, and the numbers here are 1,303 days versus 1,373 days. It does raise another methodologic issue that I'd like to pursue, but I'd be happy to delay that. I don't need to discuss it yet.
DR. BORER: Dr. Kopp?
DR. KOPP: I had a question on the same point, and maybe, Tom, this is what you're going to get at. As I understand, that was the primary proposed method of analysis, the time to event. The slide you showed this morning talked about time to event, but then you actually gave risk reduction of 16 percent and p .022.
I notice in this table on page 8, that it also gives p .022. So, again, it may be my statistical ignorance, but both are ways to look at time to event? Am I being clear?
DR. BAIN: I'm not sure I understand the question.
DR. KOPP: I'm confused about when you started out saying this was going to be a time-to-event analysis, and then the slide that kind of captured the key data was actually risk reduction. I had the same question you did: what was the actual data? Now I see it on page 8 of the FDA analysis with the same p value.
DR. BORER: I think we may be getting into sort of a semantic issue.
DR. FLEMING: This is the slide to show. There really are two separate analyses that are being done. One is where do these two curves cross the 50 percent line, which is basically a difference in median time to event. And another very mainstream analysis -- it's really a separate parameter looking at treatment effect -- is to say suppose there's an underlying failure rate over time on the control arm, and suppose the intervention arm alters that by a multiplicative constant, the hazard ratio. What is that reduction in relative risk? And that's what the hazard ratio is, and that's a 16.1 percent relative reduction in the failure rate. So, those are related, but those are, nevertheless, separate measures of treatment effect.
DR. KOPP: So, is it by chance that they're both .022, or is that inherent in the way these statistical tests are done?
DR. FLEMING: The p of .022 I'm sure relates only to this analysis of hazard ratio estimates in the confidence interval. We rarely test that difference in medians as the basis for statistical p values. The p values that you see are going to be based on the proportional hazards model and the estimate of the relative risk, the confidence interval, and the corresponding significance from that.
DR. BORER: Dr. Temple.
DR. TEMPLE: Tom, help us a little more. It's common in representing the results of a study to give the total number of events, because you can work with that. You know what that means. Whereas, Kaplan-Meier curves are sort of something that doesn't have a number attached to them. You just have to look at them.
But you could apply a hazard ratio risk reduction statement either to the total number of events, sort of independent of time -- some studies do that -- or you can do it this way, which as I understand it, is the risk reduction or hazard ratio of the likelihood of getting an event in a given time or over the entire course of the study. The hazard ratio is a time-related function as usually presented in these analyses.
DR. FLEMING: There are many different measures that we could use to assess the nature of treatment effect and the strength of evidence that it's been an established effect.
The most traditional or most common analysis in a time-to-event setting is to say we're going to make no assumptions on the nature of the failure rate on the control arm. We're only going to make an assumption that if treatment affects that failure rate, it does so in a multiplicative fashion. It reduces the failure rate by some multiplicative constant, that is constant over time. That's what proportional hazards mean. If treatment reduces the failure rate early in time by 16 percent, we're going to assume it's reducing it late in time by 16 percent. What is that percent?
That is the fundamental common analysis. It's called the log rank test, the Cox regression. Those are all based on that same fundamental principle. Essentially almost all estimates that we would see, which is the relative risk or the reduction, is based on that hazard ratio estimate, with confidence intervals and p values.
Now, there are other analyses that we could do. You can look at total numbers of events. I like to look at that as well as just a descriptor. You could look at the differences in rates at a given time, so you might say is what I really care about is, is there a difference in these events at 2 years? In which case, I would look at the difference in Kaplan-Meier estimates standardized by Greenwood variance estimates. That's a different analysis.
Another analysis is to look at the medians. It's a good descriptor, although I tend to favor the proportional hazards analysis because the medians is only telling you how the curves separate at one place, whereas the proportional hazards analysis is looking at a weighted average of what the nature of the effect is on the outcome failure risk over the entire duration of the curve.
DR. BORER: Still, Blase is raising a point that we really haven't discussed that I think is very important. If I am understanding correctly, the time-to-event analyses that have been done show us that there is a significant effect by this analysis and that the magnitude of the effect -- that is, the failure rate -- is a reduction of 16 percent. Blase is saying, what does that mean to the patient? You know, 73 days more of dialysis? 73 less days?
DR. CARABELLO: Yes. As sort of the dull normal on the committee --
DR. CARABELLO: -- that was really the intent. If I start a group of patients on this medicine, by how many days will I delay an endpoint? Specifically, that's the question I'm asking.
DR. FLEMING: If these curves were exponential, and they are not -- but if they were -- then a 16 percent reduction in the failure rate would translate into a 16 percent extension in the average duration of time to when the event occurs.
DR. BORER: Good. Well, we've got that one settled for the moment. Ray, did you have anything else you wanted?
DR. BAIN: Just as Tom mentioned, the counts are in your background document by the two treatment groups. Our analysis is that 16 percent is based on the Cox regression hazard, with the prespecified adjustment for baseline proteinuria stratum and region.
DR. FLEMING: There is one methodologic issue here that I'd like to pursue at least briefly. It doesn't trouble me greatly, and I'll preface my comments by saying that the reason it doesn't trouble me greatly is because -- at least as I noted back in January -- I'm much more persuaded by end-stage renal disease/death as a dual endpoint than the primary specified doubling in creatinine/end-stage renal disease/death as a triple endpoint. I am even more persuaded that's the case here because of my understanding of the way follow-up was done, and I'd like to at least clarify to make sure my understanding is correct.
I am absolutely delighted to hear, by my understanding, that all patients were followed until the outcomes for end-stage renal disease and death. So, we have complete information on the endpoint that I am referring to as my preferred endpoint, that dual endpoint.
But there were a number of people who had a termination of their serum creatinine assessments prior to the time that any of those endpoints, i.e., the triple endpoints, prior to the time that either a doubling of end-stage renal disease or death occurred. So, they hadn't had any of those three components as yet.
Those people were in fact, though, followed for subsequent assessments of end-stage renal disease and death. So, the analyses that we get for the endpoint I care most about are fully valid and are perfectly fine, in spite of the concern I am about to raise.
My understanding is for the triple endpoint, though, in those cases -- and if I understand there were 130 of those people on losartan and 137 on placebo -- 267 -- so about 17.6 percent of our cohort were people who did have cessation of assessments of serum creatinine before a triple endpoint occurred and were subsequently followed for the other elements. They are included in your primary analysis.
In essence, in those 267 people as you are continuing to follow them, we're really underestimating the subsequent rates at which the triple endpoint occurs, because we're only using the double endpoint assessments in those people. Now, of course that's happening in both treatment arms. So, both treatment arms are being underestimated.
But then when questions come up from my colleagues about how do we interpret the extension in time or delay in the triple endpoint, I can't interpret that anymore, because I haven't followed people for the triple endpoint. I followed some for the triple endpoint, but 267 people were followed over a substantial duration only for the double endpoint.
Of course, I don't care that much about the triple endpoint. I like the double endpoint. So, this is no problem for me on the double endpoint.
DR. FLEMING: But it makes me even more worried about interpreting your data on the triple endpoint. Am I off target here or is my interpretation correct?
DR. BAIN: Your interpretation is correct. What we're focusing on right now are those patients who discontinued their therapy, their blinded therapy, prior to one of those triple endpoint events. Once those patients discontinue their primary therapy, their follow-up could be classified into one of three groups.
Either they continue their quarterly measurements, which was by protocol, and we continue to get serum creatinines, a potentially documented doubling, which they get counted in the intention to treat analysis. That's one bucket.
The other bucket is, as soon as they discontinue their therapy, they go into what we classified as telephone follow-up, where during telephone follow-up we're not collecting serum creatinines.
The third bucket is patients discontinue their meds. They go into clinic follow-up for a period of time, and then they go into telephone follow-up.
With those three kinds of different classes, it's difficult to really give you a number as to how much is in telephone, et cetera, but we attempted to. What we did was look at the total follow-up time after discontinuation of their study therapy and determined patient years of follow-up from that time of discontinuation of therapy. Then we also calculated what percentage of that time was in clinic follow-up versus what percentage of that time was in telephone follow-up. We estimated that, and 60 percent of the follow-up post discontinuation of the study therapy is 60 percent. 60 percent are followed in clinic that time.
So, that gives you a general idea of how we are following those patients post discontinuation of the therapy. Obviously we continued to follow them for the hard endpoints.
Now, when we go ahead and do that triple endpoint intention to treat analysis, what we are actually doing is what I would consider to be a conservative analysis, because we're not documenting doubling in either group, so we're assuming that any effect that losartan would have would be the same in the losartan and the placebo group after discontinuation. So, our triple endpoint, even though we had 40 percent of the patients in telephone follow-up, actually the analysis is holding up mainly because it's conservative.
DR. FLEMING: Just a brief addition to what you were saying. The only conclusion I am confident to make here is, in these 40 percent of the person years of follow-up time after discontinuation, where we no longer continue to follow the triple endpoint -- we're only following the double endpoint -- if you really care about the triple endpoint -- we're underestimating in both arms the time to the occurrence of the triple endpoint.
Whether that leads to a conservative or anti-conservative estimate of treatment effect entirely depends on in these people who are discontinued for follow-up of serum creatinine doubling times, in those people who are discontinued, would the doubling time have occurred more rapidly in the placebo versus the losartan arm? I have no clue about that.
So, it leaves me uncertain about how to interpret the data, to an extent at least, if I really care about the triple endpoint, because, again, in 40 percent of the person-years of follow-up after discontinuation -- and after treatment discontinuation I do care to continue to follow people -- I might surmise that I would have less treatment effect after treatment discontinuation. I don't know. That's only a guess. But I do care, if I really cared about the triple endpoint, to know what time is to triple endpoint in all patients.
My conclusion in this is that there are 267 people, balanced by treatment arm, in whom I am underestimating the time to the triple endpoint in both arms. I have no clue if it's more so in one than the other, and it's just another reason that from my perspective the double endpoint, which occurs in this setting not that much later than the triple endpoint, and is clinically a much harder endpoint, and is free of all of these problems, is an endpoint that I particularly would focus on more so than the triple endpoint.
DR. BAIN: Correct. Let me make one other point. When we are talking about these patient-years of follow-up and we're talking about the triple endpoint, if you actually calculate the patient-years of follow-up in clinic over the whole time, not just focusing on those patients who discontinued their therapy, the clinic follow-up is over 90 percent. It's only when you get into that subset of patients that it drops to 40 percent.
DR. BORER: Dr. Lorell.
DR. LORELL: Thank you.
In follow-up of Dr. Fleming's comments, there were two queries that I had. One is, in this difficult group of the 40 percent patient-years who have the issues of discontinuing treatment and problems of telephone contact, do you know what percentage of those patients in each of the intention-to-treat groups actually got put on an ACE inhibitor by their clinical nephrologist? Is that known?
DR. SHAHINFAR: We know for the entire cohort those patients who were discontinued from the study, and they were placed on ACE inhibitors or AII receptor antagonists. I have to remind you that this collection of the concomitant medication was extremely difficult after a patient was discontinued from the study.
DR. LORELL: I understand that.
DR. SHAHINFAR: So, it's going to be either underestimation or -- the data that we have -- can I have that?
DR. LORELL: Perhaps you can get that to us in a little bit.
DR. SHAHINFAR: This is basically those patients who discontinued the study drug, and we have information on them that they went to either ACE inhibitors or angiotensin II receptor antagonist therapy.
DR. LORELL: That's very helpful.
The second question I had that relates to Dr. Carabello's efforts to understand the clinical impact. In our briefing supplement that we were provided with in table 1-A, there is a comment about the median time to reach the triple event, but we're not provided with data, I don't believe, that might address Dr. Carabello's concern and mine too, as to what the median time is in the losartan and placebo group of reaching end-stage renal disease or reaching the double endpoint. Maybe we could be provided with that.
DR. BAIN: Ray Bain, Merck Research Labs.
Now, Tom kind of caveated around doing this but we did calculate the number. If I understand the question, it's of the patients who reached a primary composite endpoint event, which there are 327 in the losartan group and 359 in the placebo group, if you actually calculate the median number of days to the event, it's 652 days in the placebo group and 724 in the losartan group.
So, there is a delay, but that's a subset of the patients. You're just looking at those patients who actually had the event, ignoring all the other patients, which, if you do the median analysis and go across the median and drop it down to your time line, you'll get estimates of differences in time.
DR. BORER: That's for the double endpoint?
DR. BAIN: That was the triple endpoint.
DR. LORELL: That's different from the table.
DR. BORER: Yes. Maybe you can pursue this just a little bit.
DR. LORELL: In the table 1A that we were provided, the number is actually quite different, but I would also be interested in following up Tom Fleming's comment that I think for many of us the important issue is end-stage renal disease/death -- what the median time was for reaching that double endpoint. Do you have that data?
DR. KEANE: Again, are you asking the median time for doubling, or are you asking the median time for end-stage renal disease?
DR. LORELL: No. We're provided in table 1A with the median time to reach the triple endpoint. In follow-up of Dr. Fleming's comment and Dr. Carabello's, can we know what the median time is to reach end-stage renal disease and the median time to reach end-stage renal disease or death.
DR. KEANE: I think we can go back and calculate that. We just don't have that immediately on our data set here, so we'll find that out and get back to you.
DR. LORELL: Thank you, sir.
DR. KOPP: Could you explain the discrepancy that your last question just brought up between the median time to event in our table of about 1,300 days for the two groups, and the median time of about, I think you said, about 650 and 704? This is in both cases the median time to the composite endpoint, triple.
DR. FLEMING: I think I might be able to, while you're bringing somebody up. The distinction between the two -- the Kaplan-Meier is the 1,300, whereas I think the analysis of the 700 was if you just take the people with the event, what's the median time at which that event occurred. And that's always going to be a lot less. That was 700.
DR. BAIN: Yes.
DR. FLEMING: And I would caution against ever looking at that because it's very misleading.
What I really care about is what's the Kaplan-Meier? What's the percent free of having an event? If I have two treatment arms, and let's say in one treatment arm 10 percent of the people have an event at a year, and in the other treatment arm, 10 percent have an event at a year, and another 10 percent have an event at 2 years, the latter is a worse scenario. And yet the median time to event in the latter case is a year and a half, whereas it's a year in the first arm.
So, I don't want to look at, given you've had an event, what's the time to event. I want to look at the entire cohort and see what's the percent free of the event and what's the time at which that crosses the median. That's the 1,300 type number
DR. SHAHINFAR: In summary, in type II diabetic patients with proteinuria, there were no unusual or unexpected adverse experiences beyond those already noted in the U.S. prescribing information for losartan. Losartan had a higher incidence of hyperkalemia, and lower incidence of hypokalemia compared to placebo. Losartan was generally well tolerated.
I would now like to introduce Dr. William Keane, Vice President of Clinical Development, Merck U.S. Human Health. Dr. Keane will provide the review of evidence and conclusions regarding the renal protective effect of losartan in type II diabetes. Thank you.
DR. KEANE: Thank you, Dr. Shahinfar.
Members of the committee, representatives of the FDA, other invited guests, ladies and gentlemen. My name is Bill Keane, and I'm delighted to have the opportunity to appear before the advisory committee again in my new capacity.
As many of you know, I have recently joined Merck as Vice President of Clinical Development in U.S. Human Health after 28 years in an academic practice of nephrology.
Most recently I was professor and chairman of the Department of Medicine at Hennepin County Medical Center at the University of Minnesota in Minneapolis. I was intimately involved with the RENAAL study, both as a member of the Steering Committee, as well as a primary investigator, and participated in the design, the conduct, the oversight, the analysis and publication of this important study.
In addition, as President of the National Kidney Foundation, which is a patient-focused organization, I can assure you that we have recognized that the treatment of the type II diabetic patient with proteinuria is an unmet medical need.
I would like to provide you with my summary that emphasizes the salient features of this study that make it a compelling data set supporting our proposed indication.
The strength of the evidence provided by the RENAAL study is sufficient to support our conclusions, based on a number of critical features, including the robust study design, the results that are clinically important, as well as statistically significant, and the presence of internal consistencies across multiple endpoints and multiple subgroups. Moreover, as you've seen, after adjusting for differences in relevant patient characteristics, particularly the chance imbalance that was observed in baseline proteinuria, an even more dramatic impact can be seen.
First, let me re-emphasize that this was a large, multinational study conducted in 28 countries at 250 clinical sites. It included a diverse study population, with demographics of patients with type II diabetes here in the United States. There were no patients lost to follow-up in terms of ascertaining their clinical status with respect to end-stage renal disease and death. The key renal endpoints and cardiovascular endpoints were independently adjudicated. Furthermore, there were no patients missing from our intention-to-treat analysis for the primary composite endpoints.
I would now like to review the main study results in order to emphasize the internal consistencies and reliability across the multiple endpoints and subgroups of the study.
RENAAL provides persuasive evidence that losartan delays the progression of kidney disease in type II diabetic patients with proteinuria, as demonstrated by the following results.
The primary composite endpoint of the time to the first event of doubling of the serum creatinine, end-stage renal disease or death, which revealed a risk reduction of 16 percent, with a p value of .02.
The clinical endpoint of end-stage renal disease, which revealed a very robust risk reduction of nearly 29 percent. Again, I'll emphasize in this particular clinical component, there were 147 events of end-stage renal disease in the losartan group and 194 end-stage renal events that we observed in the placebo group.
The clinical endpoint of end-stage renal disease or death, which also showed a significant risk reduction of nearly 20 percent.
Our prespecified subgroup analyses, which included those performed for age, gender, and a variety of other categories, demonstrated that losartan provided renal protection across a wide range of patient subgroups.
Dr. Shahinfar has already reviewed with you that we stratified patients by baseline proteinuria, either those below 2,000 milligrams of albumin per gram of creatinine, or above 2,000 milligrams of albumin per gram of creatinine.
Despite this, we observed an imbalance in the distribution of baseline proteinuria values between the treatment groups, particularly within the higher proteinuria stratum. That is, in those patients that had a urine albumin to creatinine ratio of greater than 2,000 milligrams per gram of creatinine.
Given that baseline proteinuria was a strong predictor of the risk of subsequent renal events for patients in both treatment groups, and the risk dramatically rises for patients with high levels of baseline proteinuria, it seemed particularly appropriate to adjust for this observed imbalance.
As you can see in the right-hand column, when we adjusted for baseline proteinuria as a continuous covariate, these post hoc analyses showed that the risk reduction for the primary composite endpoint of doubling of serum creatinine, end-stage renal disease or death improved now to some 22 percent, with a p value of .001.
The risk reduction for end-stage renal disease improved to nearly 37 percent, and the risk reduction for end-stage renal disease or death improved to greater than 25 percent.
Thus, these analyses reinforce the conclusions of our primary analyses.
It is critical to emphasize the importance of effective blood pressure control in patients with type II diabetes and proteinuria. It is well recognized that patients with type II diabetes and nephropathy either have hypertension at the time they begin to develop nephropathy or become hypertensive by the time they reach one of the clinical endpoints defined in our RENAAL study.
In the RENAAL study aggressive treatment of blood pressure was specified in the protocol, and as Dr. Shahinfar has shared with you today, the majority of patients were well controlled in both groups. The differences in blood pressure between the treatment arms were small, and our prespecified analysis for the small 2 millimeter differences in mean arterial pressure did not substantially alter the treatment effect on the various renal endpoints. This supports the overall conclusion that the benefits of losartan treatment in this population are not attributable to blood pressure control alone.
There were additional secondary renal endpoints in the study. These also provide consistent and persuasive evidence of the treatment benefit of losartan in this patient population. Compared to placebo, losartan showed a significant 18 percent reduction in the rate of progressive loss of kidney function, as estimated by the reciprocal of the serum creatinine. In addition, we found that, compared to placebo, losartan also had a significant 34 percent reduction in proteinuria.
Although RENAAL was designed specifically as a renal protection study, because of the recognized increased risk of cardiovascular events in the RENAAL patient population, we prespecified the composite endpoint of cardiovascular morbidity and mortality as an important endpoint. This composite endpoint included all-cause cardiovascular death, myocardial infarction, strokes, time to first hospitalization for heart failure, and time to first hospitalization for angina, and all revascularization procedures, both within the coronary vasculature as well as in the peripheral vasculature.
There was no treatment effect demonstrated for the cardiovascular composite endpoint overall. An examination of the individual components revealed the expected fluctuations -- some positive, some negative -- that led to the nonsignificant p value for the cardiovascular composite endpoint.
While Dr. Shahinfar has already reviewed these findings in detail and we have discussed them also in detail, I simply really wish to emphasize that the benefits of losartan treatment on our renal endpoints did not come at the expense of an increased risk of cardiovascular events in the losartan arm. This is probably best illustrated by our post hoc analysis on the bottom portion of this slide which demonstrated the combined endpoint of end-stage renal disease, myocardial infarction, stroke, or death, and this showed a 21 percent risk reduction with losartan.
In conclusion, for patients with type II diabetes and proteinuria, the specter of inexorable progression to end-stage renal disease requiring renal replacement therapy is a frightening prospect. Those of us that have cared for these patients recognize the enormity of disease burden that is borne by these individuals. In my view, the sum evidence of our analysis for the RENAAL data that we have shown you today demonstrate robust, reliable, and clinically relevant evidence that is applicable to a broad population of type II diabetic patients.
In addition, our data confirm that the safety profile of losartan in this population is consistent with the U.S. prescribing information for losartan.
Finally, in patients with type II diabetes and proteinuria, the treatment effects of losartan across multiple endpoints in the RENAAL study provide persuasive evidence for the renal protective effects of losartan. Indeed, we estimated that over a 3-and-a-half year time period for every 16 patients treated, 1 case of end-stage renal disease will be prevented. RENAAL provides new data that support the role of the renin-angiotensin system blockade for renal protection in diabetes and kidney disease.
Thank you very much for your time and attention.
Mr. Chairman, I was going to just take the liberty, because there were a number of questions that had come up during the presentation that we thought we could provide some answers to.
DR. BORER: That would be helpful, but I think on your behalf really we have some more questions from the committee that may require that you gather some more data, and it might be nice for you to have the break to do that. So, let me take the last few minutes before the lunch break to go to some of these questions. You may be able to answer them now, you may not. If you can't, then you'll have a few minutes to get the information together and we'll go back over all the questions after lunch.
Dr. Nissen, you had some concerns?
DR. NISSEN: Obviously, we're all very cautious about subgroup analyses, but there were, I think, a lot of questions that I had about the subgroups in the study. I would hope maybe somebody could put of table 14 from the Merck briefing document. That's on page 63 for those of you on the panel that want to look at it. I assume you have a slide with that because I think there are some issues that it brings up about who in fact benefits from this therapy. Is that possible?
DR. KEANE: I don't have a slide of that directly, but we'll get maybe an overhead.
DR. NISSEN: All right. Well, let me just prime you then with the questions. When I look at this, what I see is, just in terms of counting events -- I understand what we're looking at is time to event, but just counting events -- there were 78 in the losartan group and 80 in the placebo group, virtually an identical number, in Latin America. For Europe, there were 58 in the losartan group and 51 in the placebo group. There were more events in the losartan group in Europe. In North America, there were 142 in losartan and 150 in placebo. So, if you look at the Asia group, the 250 patients in Asia provided virtually all of the endpoint differences in the trial. I mean, it is very, very striking.
And I recognize the hazards of subgroup analysis, but it's so striking that I went and I looked at the racial background, which is also in table 14, and you see exactly the same thing: 40 percent event rate among African American or black; Hispanic, 55 and 54; white, 40.5 and 43; and then a huge difference in the 250 Asians. So, it does speak to the heterogeneity of the response and the robustness of the findings.
It looks to me, when I look at these subgroups, that it's all driven by those 250 Asian patients, and I need to understand that better to understand whether this is, in fact, an effect of this drug that's applicable in the patients that come in my clinic and that we see with the disorder.
Similarly, if you look at the creatinine data, there's also this very striking disparity where, in fact, there's a higher event rate in the losartan group for those with a creatinine of less than 2, which is by the way the vast majority of patients, and for the somewhat smaller subgroup, about a third of patients where the creatinine is elevated, there's a very striking difference.
So, again, not necessarily now, but I really need to understand this very striking heterogeneity, which I recognize does not rise to the level of statistical significance with the exception of region, which does reach statistical significance. But in terms of the point estimates, it's really a striking difference, and I really need to understand that better.
DR. LINDENFELD: Just to emphasize that, I wonder if maybe you could show us the creatinine and proteinuria in the Asian region group compared to the others.
DR. KEANE: If you have time now, I can sort of begin to address some of these issues because they are obviously important and we obviously recognize that region, and the equivalence of region and race I think we need to recognize is different because region is a geographic area that is quite larger than what actually the ethnicity or racial backgrounds would be.
DR. NISSEN: But it appears in both columns.
DR. KEANE: Let me then, for purposes of what I think is really sort of more consistent, look at race at this point in time. We can always go back to region per se, but I think the racial differences are important for us to understand because I believe that's applicable then regionally when we start looking at that.
So, let me first reiterate your point, that this was a subgroup analysis, and we in fact had 18 subgroups, and that in and of itself creates a set of problems.
The second thing that I would like to actually address is the fact that there are differences between races and between regions. Specifically there are differences between our regions or races. Let me have slide 624.
As you recall, when we looked at proteinuria and showed you the risk that proteinuria has in terms of predicting subsequent events, as you increase your degree of proteinuria, the amount of hazard or the risk for a primary event and also the primary events that occur in the highest group of proteinuria patients dramatically increase. It's in many ways like the old adage that we have in clinical medicine, that is, 10 percent of the people have 80 percent of the events. So, this appears to be in the area of proteinuria as well.
And this was an important analysis going into the trial that we recognized. In terms of recognizing, the risk was part of a larger component of things in terms of risk that we actually prespecified. It's called the risk score, and I'd be more than happy to get into that in a moment or two. But we had prespecified our analyses based upon risks that we thought were important in the patients with type II diabetes.
We also recognized that proteinuria was one of the most important drivers of the overall risk, so that we focused on proteinuria because it's a clinically relevant measure. It's simple to do.
And already, as you see within this particular slide, there are ethnic/racial differences within this important risk factor -- this progression promoter, as it's called -- between Asians, the black patients, the Hispanics, and the whites. In fact, the Asian population had one of the highest rates or highest levels of urine protein excretion rate, so pushing them, if you will, up the slope of risk for a renal event.
Now, within the different groups, as Dr. Shahinfar has already alluded to, there were imbalances in terms of stratification to losartan versus placebo that also occurred within the trial. So, this is one big issue.
The second big issue that I think is important for us to recognize --
DR. FLEMING: Before we leave this big issue --
DR. KEANE: Yes.
DR. FLEMING: Slide 624. Just to interpret, this is at a certain level relevant. What we're seeing is that the Hispanic population, the Asian population had higher baseline proteinuria. That would lead me to think that they should have had a higher rate of events.
So, if we go back to page 63, table 14, that explains the relationship of the events within the placebo arm by country. It explains why the Hispanics and the Asians in the placebo arm had the higher rate. That explains that region would be a predictor of rate of outcome.
It tells us nothing, though, about why region should be an effect modifier, so it doesn't answer Steve's question unless you want to go one step further, unless you want to make the statement that not only is proteinuria a predictor, but it's also an effect modifier so that we would expect higher rates in Asia and Hispanics and we expect a bigger effect. But those don't logically follow to me because they are different concepts.
DR. BORER: Not only that, we didn't see it.
DR. TEMPLE: Yes, you did. If you look at the results by various underlying factors, almost the entire effect is in the people with proteinuria over 2 grams. You did see that before. It would be equally true for 4 grams I'm sure.
DR. NISSEN: Yes, but Bob, the 277 patients in the Hispanic group who have the highest proteinuria, there's no treatment effect. So, this doesn't make any sense to me. One of the high groups, the Asians, has all the benefit and a group that's even higher with proteinuria has actually a hazard ratio that's worse in the losartan group. So, this doesn't explain anything.
DR. TEMPLE: Well, you have to see how these break down by region as well. This is not broken down by region because some people in each region are both. But you did actually see -- I thought it was fairly striking -- that if you look at the hazard ratios, essentially all of the effect was in the people who were above 2 grams, which could have something to do with that.
DR. KEANE: Can I go on? Because I think that's an important set of issues here. But I do have, beyond this, some specific, I think, explanations that we need to see to interpret actually what we're talking about right now.
May I have the next slide? This really then underscores the second feature that is present that is different amongst Asians or different, I should say, amongst ethnic groups, racial groups, as well different amongst the different regions.
But here displayed are really the racial events. As you can see, there are different levels of discontinuation that are occurring prior to the first event, so prior to our primary event. Within the Asian group, there are highly compliant groups of individuals while we have in caucasian and Hispanics somewhat less compliance or utilization of the drug. Now, this has an impact, and let me show you the next slide please, and that should be 627.
If one takes our primary endpoint and just does the evaluation -- this is the primary composite endpoint indicated in the dark lines here, the solid lines with the circle point estimates -- as you look at those point estimates and confidence intervals, and then you adjust for those patients who are on treatment who actually took the drug, and then you adjust for those patients who were on treatment that had a change in urine protein excretion rate, as you can see, the point estimates for the Asians, the blacks, Hispanics, and whites gradually evolve to the left-hand side of the line of 0, supporting at least the importance of both proteinuria, as well as being on therapy as modifying this primary composite endpoint across all races.
This is, again, supported by our overall look at there was no interaction between races and treatment arms. So, I think this is supportive and consistent data that shows that we can actually benefit these renal outcomes in all of our patients of these different ethnic and racial backgrounds.
DR. BORER: Rather than go on to any other issues -- I think these were important slides, and you may have some more that you will want to show us after the break. Having been stung by the criticism of lack of break last week and since it is now exactly 12 o'clock and it says on the schedule that you can go to lunch at 12 o'clock, including the FDA people, 12 o'clock it is and lunchtime it is, and we'll get back together at 1:00.
(Whereupon, at 12:00 p.m., the committee was recessed, to reconvene at 1:00 p.m., this same day.)
DR. BORER: Okay. It's 1 o'clock and 30 seconds, so we're a little late starting here.
At this point there are several unanswered questions and, as yet, unasked questions from the committee that we'd like to raise, and I want to say for all of us to the sponsor that, especially since we're going to move into a period where the questions may become more intense and the discussion may become more intense, all of us appreciate the efficiency and completeness with which the sponsor has presented the data. I know you'll have additional information for us, and we appreciate the superb consultants you brought along to help answer the questions. So, with that having been said, let's move right along here.
Dr. Nissen, you were in the middle of your questions and then we have several more that I marked down to go back to from other committee members, and there may be some additional ones as well. Steve, why don't you start?
DR. NISSEN: Thanks very much, Jeff.
I want to continue with this regional and ethnic imbalance and see if I can explore this a little bit more. But actually before I do, we really do have a problem here with the FDA data and the company data not agreeing. Tom, I particularly think I need your help with this.
If you look at the table on region, you see that there are 58 events in your table in the losartan group and 51 in the placebo group, and the hazard ratio is .943. Now, I assume that that's the case because, even though there are more events, the time to events would presumably have been much shorter in the placebo group.
But the same number of events in the FDA analysis on page 22 of the briefing document gives a completely different hazard ratio. It's 1.05. That's a pretty big difference, and I don't get it, Tom. Or to the company, I don't understand why they don't agree.
DR. BAIN: Yes. This was one of the housecleaning activities that I was going to come back to.
We went back and looked at the FDA background document versus the Merck background document. Let me point out that in the Merck background document, as prespecified in our protocol and data analysis plan, whenever we do a subgroup analysis, we also adjust for the baseline proteinuria category or stratum. And when we do that in our analysis, which is the prespecified analysis, we get our results.
We went back and confirmed that if you eliminate that baseline proteinuria as a covariate in the analysis, you get the FDA results. So, theirs is completely unadjusted. Ours is adjusted for what we randomized on, which was a prestratification factor, which was baseline proteinuria less than 2/greater than 2.
DR. NISSEN: That helps me a lot.
Again, we can argue about how robust the effect is, but clearly those 250 patients that were in Asia, that are either defined racially as Asians or geographically as Asians, provided almost all the treatment effect in the study. That's obviously an extremely striking finding. So, in exploring your data, I was looking for an explanation for that.
I'd like to ask you if you could show by region the rate of discontinuation of medication. I have it from the FDA briefing document on page 16 for the panel members. What is very interesting here is that there is a huge imbalance between placebo and losartan in drug discontinuation in the Asian group but not in the other groups. So, I think potentially this is a smoking gun that explains why there's such a big difference here.
If you look at this, fully 45 percent of the placebo discontinued and 26 percent of the losartan group discontinued; whereas, in the other regions, there's very little difference in discontinuation rate. I'd sure like an explanation for that because, again, it's a very striking finding. I think it could explain all of this because if a patient discontinues, they obviously get less intense follow-up, and I'm worried that those discontinued patients had their blood pressures going out of control and that's why we see this. So, many more of the patients in the placebo arm were discontinued in Asia than in the treatment arm.
Can you help me with this?
DR. KEANE: Maybe I can. Let me just jump back a little bit to sort of set the stage. Again, these are all a result of our subgroup analyses that were done. I want to emphasize two points.
One is the fact that these patients did have differences in proteinuria, and I did want to emphasize the point that we have differences in discontinuations, as you've already alluded to.
Now, probably the most important point, in terms of us now looking at this, will be to look at what, in fact, happened to end-stage renal disease and death because that's where we have captured 100 percent of all of the events in all of the patients. So, this is very, very, I think, powerful data for us to look at.
I can show you I think both by region, as well as by race -- and recognizing we have these differences in terms of the geography, as well as compared to ethnicity. So, let me just show slide 629, if I may.
I think as you can see here, this is for end-stage renal disease or death, which I think we have been discussing this morning as, in fact, being the hard outcome that is of particular relevance in any kind of renal protection trial. I think across the board, as you can see in the solid line, which is our primary analysis, and then as we look at on-treatment, that is the patients who sustained therapy throughout the trial, as well as then those patients who were on treatment who had an adjustment for proteinuria, again this being an important risk for progression of kidney disease, our point estimates for the different ethnic groups, again, move slightly to the left of the 0 line, again favoring losartan. Now, this is by race.
Let me just show you then by region, and I think that's slide 850.
DR. NISSEN: You already showed this to us I think before lunch, didn't you? This is the same slide?
DR. KEANE: No, no.
This is ESRD. This is 829. This is ESRD or death. Again, I think the same events are evident in here, with these adjustments that I've just gone through, on-treatment, on-treatment, and adjusted.
Now, if we go by region, again by Asia, Europe, Latin America, and North America, again we believe that there is a shift of the point estimate more or less to the left, again favoring losartan in all the different regions.
So, I think that at least gives us confidence in these exploratory analyses that in fact the direction that we're going is correct. It's applicable across multiple groups, multiple ethnic groups, and in fact across this region that we prespecified as one of our prespecified subanalyses.
DR. BORER: Dr. Brem?
DR. FLEMING: Can we comment on this before we take off?
DR. BORER: We can, sure. Is this about this particular issue?
DR. FLEMING: Yes.
DR. KEANE: Yes. I was just turning to my colleague now because I think part of Dr. Nissen's question had been related to what happens if you do the analysis with or without different countries or different regions, and I was going to ask Dr. Bain to actually address that specifically because we have done that.
DR. NISSEN: I do want to see it for the unadjusted because I really want to see the raw data. I want to see the data without all that proteinuria adjustment and all that. I just want to understand it with the raw data.
DR. BAIN: In that primary analysis up there, the only adjustment was for what we prespecified in the data analysis plan as being our primary. It's just the adjustment for baseline proteinuria strata, two strata, and the region.
Now, what Dr. Keane alluded to was the issue that Dr. Nissen brought up about, gee, all the effect seems to be going away if, for example, you looked in the regions other than Asia. We were, obviously, interested in exploring that issue also. Unfortunately, I'm going to have to put an overhead slide up here, which is going to be pretty difficult to read, but let me do that.
What we're doing here is we're looking for influential either regions or countries within a region, and I'll explain this. I actually have hard copies for the committee.
Now, what I'm focusing on there is the composite endpoint of ESRD or death by the intention-to-treat analysis. And if you remember from the main presentation, the risk reduction for this particular outcome, ESRD or death by intention-to-treat, was 20 percent overall.
Now, if you look within a region -- and those regions are highlighted with brackets around them -- Asia, Europe, Latin America, North America -- what this table shows you is when you eliminate that country or region that's on the left-hand column, if you go across to the third column, it has a risk reduction. That tells you what the resulting risk reduction is. So, for example, if you were to eliminate the region Asia from the analysis, the risk reduction goes from 20 percent down to 13 percent. If you eliminate Europe, 21.9; Latin America, 21.6; North America, 23.3. Now, that's by region.
Now, what has been alluded to earlier is we prespecified region and the countries that went into that region, but that's just geographic. But now, look at the countries within Asia. The countries are Hong Kong, Israel, Japan, Malaysia, Singapore. What this shows is if you eliminate them one at a time -- for example, if you eliminate Hong Kong, the risk reduction goes from 20 percent to 16.5. If you eliminate Israel, it goes to 16.8. So, you see within Asia, it ranges from 16 to 21.
As a matter of fact, if you do that one-country elimination at a time -- without eliminating any countries it was a 20 percent reduction. And now if you eliminate one country at a time, it ranges anywhere from 16.5 to 23 is the number I'm picking out, which is roughly around that 20 percent. So, what we're doing here is an analysis to try to figure out if there are influential countries.
DR. NISSEN: This is for the double endpoint.
DR. BAIN: This is ESRD or death.
DR. NISSEN: Do you have it for the triple endpoint? Because that was your primary efficacy parameter.
DR. BAIN: Now, this one, the overall result was 16 percent when we did the triple endpoint, 16 percent reduction. Again, when you eliminate a whole region, it has a major impact: 7.1 percent in Asia, which is a pretty big impact. But if you actually go in and eliminate one country at a time, now we're talking about a range from 13 to 23 in a risk reduction.
DR. NISSEN: Right, but what happens when those 257 patients from Asia are eliminated out of the 1,500, the statistical significance is lost completely. Now the confidence intervals overlap no benefit.
DR. BAIN: I would agree that eliminating a whole region is going to have an impact here, but what we're interested in looking at, we're looking at the countries within a region. So, when you look at eliminating one country at a time, within Asia it's 13 to 16.
DR. TEMPLE: But to be fair, it's not eliminating a whole region that's the problem. It's eliminating Asia. If you eliminate the other regions, nothing happens at all. In fact, the whole thing gets better. If you eliminate North America, and all of a sudden you're up to 22.4 percent, so that's a big improvement. So, it's that Asia drives the triple endpoint.
DR. ZEGER: My name is Scott Zeger.
DR. TEMPLE: What that means is hard to say, but it's true.
DR. NISSEN: I'm not passing judgment on what we do with that information, but I guess what I wanted to drill down on is the fact that if you take the 257 patients out of the 1,500 in Asia, the study loses any statistical significance for the triple endpoint.
DR. FLEMING: Could I maybe add a little bit to this? There have at least maybe three issues put on the table as we've been trying to explore and understand this better.
One is that these analyses are being done by the stratification factor for baseline proteinuria, and I would strongly agree that's the right thing to do. As your primary analysis, you ought to be at analysis accounting for structure you imposed at randomization. So, what the sponsor is doing here in this stratification seems appropriate.
The explanation, given a couple slides back on what happens when you look at the on-study or per-protocol treatment, I'm less persuaded by because that's not preserving the randomization. And to defend the sponsor, they weren't putting that forward as their primary analysis either, but I don't see that as being particularly persuasive.
The third issue, though, is they are looking at this issue as it relates not only to the triple endpoint but also the double endpoint. Again, it may be my perspective coming through again that I believe that double endpoint is, A, the most clinically relevant and, B, the most interpretable because I'm not exactly sure what impact it has when we stop following some 267 people in the course of the study for changes in doubling in creatinine times.
As a result, if either you look at this slide or you go back to the previous slide that shows these results of the impact of taking out region, looking at this -- actually in fact maybe what I'm really looking at -- well, this is fine. I wanted to actually look at the overhead slide as well, but this is fine.
Essentially what we're seeing, when we take out Asia, is a reduction in the estimate here that would have been 19.9 percent down to 13 percent, and for the triple endpoint, it was a reduction from 16.1 down to 7. So, taking out Asia has a bigger relative effect here in the robustness of the triple endpoint, taking it from 16 down to 7; whereas for the double endpoint, it takes it down from 19.9 to 13.1, and it's almost significant.
And in fact, to be fair to the data, if we eye the data and we find the worst region or the region that accounts for most of the effect, and we pull that out and say, gee, does the remainder still achieve statistical significance, there are two problems with that. One is there's less data, so you weren't necessarily powered to show significance, and secondly, it's data dredging in the reverse direction where you're taking out the best results and hoping that it stands.
So, Steve, your issues are relevant. I think it's very important to do the probing that you're doing so that we can get a sense because we're going to be asked later on, does a single study stand, and we're going to be asked to address this particular kind of issue. Is there robustness when you look at issues such as subgroups?
My own sense about this, though, is I find it reinforcing that when I look at the double endpoint, the amount of the effect that's being accounted for by Asia to me seems to be less. I'm more confident that these data seem robust through this issue on the double endpoint than on the triple endpoint.
DR. NISSEN: Although you would have to agree, Tom, that the primary prespecified efficacy parameter was the triple endpoint. So, I would tend to focus on what was prespecified up front, and what happens, when you take out the Asian population, is it's no longer significant.
DR. BORER: But to be entirely fair, the prespecified endpoint didn't include a subanalysis by region.
Did you have a --
DR. ZEGER: I could never have made the point as eloquently as Dr. Fleming.
DR. KEANE: I wonder if we could just continue along this vein because I think there have been some other issues that came up, Mr. Chairman, that I think are important for us to answer to help further clarify some of the relationships between proteinuria and some of the endpoints that we looked at.
Specifically, I think the question this morning of end-stage renal disease and whether or not proteinuria was a useful parameter or risk to actually predict end-stage renal disease in this whole setting, and I think that would be an important component to our discussion, if I may.
DR. BORER: Before you do, Dr. Brem had a question earlier. Go ahead.
DR. BREM: I wanted to ask whether two issues might have impacted your information that you've presented.
One is in the stratification, you had patients who received 50 milligrams and another group of people who got 100 milligrams. And is there a dose dependency in this, and did one group perhaps in Asia get 50 milligrams or 100 milligrams and others get 50? Was there a difference by dose?
And then the other question I had -- and perhaps you may not be able to answer this -- is these patients were older and could certainly have acquired other forms of preexistent renal disease on top of their diabetes. And since we do know that IgA nephropathy tends to occur in significant amounts in Asian populations, could these patients have two processes that would perhaps both benefit from treatment with an ACE inhibitor or an ARB?
DR. KEANE: Let me respond to the initial question first in terms of actually the amount of drug that was used. Over 70 percent of our patients actually were utilizing 100 milligrams of losartan per day. In fact, the way we utilized losartan was to titrate it up based upon blood pressure. Really, that's the fairest way to look at this. If we were going to do something, did we see a dose dependency, in terms of any effect, I think we would have had to look at a totally different design for our clinical trial. So, we haven't really addressed that within the RENAAL trial as a specific endpoint.
The second issue is that the diagnoses that were made here in terms of the clinical diagnosis of type II diabetes and nephropathy, one would have to recognize that this was a clinical diagnosis. It was made in clinic. In fact, it was made by experienced clinical investigators, mind you, and experienced nephrologists and endocrinologists used to dealing with this patient population. So, while we did not, and were not able to by the design of the study, ascertain whether or not there was something else going on, all of these patients had clinical characteristics, as you already saw today, that were consistent with the diagnosis of type II diabetes, as best we're able to do in a clinical trial of this nature.
So, if that helps to answer your questions.
DR. BREM: Yes.
The second point I just raise out of curiosity, not that you'd have the answer.
The first point, however, the dose dependence. Is there any information that you've been able to put together that looked, at least, at whether 50 milligrams in that 30 percent of patients had an effect? Was there any effect at all?
DR. KEANE: I think the fairest answer in that was that we have not looked and we didn't plan to look in terms of whether or not there was any dose dependency in terms of the effect. We were again titrating towards blood pressure and achieving the blood pressure that we have already talked about this afternoon.
DR. ARMSTRONG: I was going to wait, but since the issue of dose has been raised, Mr. Chairman, first a comment and then a question.
I'm persuaded that in the Asian subgroup that it wasn't so much the discontinuation of placebo, which is comparable, but as has been said by the sponsor, compliance. In fact, they complied with losartan far better than any other subgroup. So, if there is a plausibility associated with this, I think it probably is driven by that.
But do I understand correctly that the sponsor is not in a position to address the issue of dose response and efficacy? If that's correct, fine. Are they in a position to address the issue of safety? Because it took about 6 months to get the 70 percent of the population up to 100 milligrams and although we've seen the Kaplan-Meier curves on efficacy, we've seen the safety data not presented according to its time dependency. And it would be of interest and I think relevance to the label and to advice to practitioners for me to understand, in relationship to the final ascertainment, about the issues of safety and dose and time course.
DR. KEANE: Again, I don't think we have a specifically evaluated dose and time and safety related issues. What I can say clearly is that overall throughout the entire conduct of the study, the adverse events that we saw, both clinical and laboratory, were relatively small. Dr. Shahinfar has already presented that information to you today.
Again, I'll go back to my answer previously. We really didn't look at any kind of dose dependency with regards to any of the responses that we saw in these patients.
Finally, I'll make the point that you have to remember that all of these patients, in large part, have progressive loss of kidney function, and at any time interval throughout their course, they're going to have different clinical manifestations of their disease. Their blood pressure may be more difficult to control. They may have differences in levels of proteinuria when you look at the disease itself. So, I think that should answer, I believe, your questions in terms of differences in the drug utilization.
DR. ARMSTRONG: With respect, Mr. Chairman, the notion of safety and issues that were relatively small, as I understood it, 1 out of 4 patients with losartan had an adverse event characterized by hyperkalemia and 1 out of 10 had hyperkalemia in excess of 6. So, as a non-nephrologist, I wouldn't characterize that as a small frequency of adverse events. I think it's a genuine issue and I was simply asking whether there was a time course relevant to the appearance of hyperkalemia, which is a clinically relevant event.
DR. BAIN: I'll let Dr. Keane respond to that clinical question, but let me first go back to your desire to want to look at dose effect. This particular study design was not a forced titration. Patients were titrated based on their response to blood pressure. So, it's very difficult to do any type of a "randomized comparison" between dosages when they're being up-dosed based on how they're responding to therapy.
DR. BREM: So, you essentially, in asking for approval, would recommend what dose? The 100 milligrams or the 50 milligrams?
DR. KEANE: As the trial was designed, it was an up-titration to blood pressure. So, we're starting at 50 and then moving up to 100 to satisfactorily control blood pressure at 140 over 75, which is what was achieved in this clinical trial.
Let me go back to address some of the other issues regarding the clinical aspects of this. All of these patients are sick patients, as you know. This is a group of individuals with advanced renal dysfunction that frequently have problems with hyperkalemia, and so the fact that we only saw 1 in 10, for me as somebody who has been in the practice of nephrology, is almost routine. We know these kinds of abnormalities are happening. We actually aggressively treat for them and manage these patients such that hyperkalemia or related potassium issues are not seen very frequently in this patient population. So, I think that is part the management of patients with more advanced renal dysfunction and the problems of hyperkalemia are not uncommon.
DR. BORER: Dr. Hirsch?
DR. HIRSCH: Dr. Keane, I want to go back, nevertheless, and beat on the dose-response question one more time because I'm still intrigued by moving from preclinical pathophysiology to human disease. Although there wasn't a real effort to be able to evaluate the actual drug dose and response, nevertheless there is a blood pressure effect.
So, you must have had the opportunity to look at the change in blood pressure in cohorts or in tertiles of blood pressure lowering to outcome because they aren't identical between the losartan and the placebo groups. So, one might get some information. In other words -- you know where I'm going with this.
DR. KEANE: Yes. I think the issue with us is that you have to recognize that, again, blood pressure was difficult to treat in these patients. We required about three-and-a-half additional medicines on average to actually control blood pressure. So, to actually interpret any dose titration that might be occurring -- are any of these drugs going down, are they being upped -- is virtually impossible for us I think to get at this point in time any reasonable explanation or reasonable data regarding dose titration because there are so many things consistently going on within an individual patient because our goal was to lower the blood pressure and we were using three-and-a-half plus drugs to achieve that.
DR. HIRSCH: I know it's difficult but perhaps that can explain some of the regional heterogeneity. Obviously, blood pressure lowering is a key determinant of the renal outcome. So, I'm still curious.
DR. KEANE: We can probably look into that, Dr. Hirsch, and see if there's any data that we can pull out to satisfy your inquiry.
But I will emphasize again that we were trying to control blood pressure, not looking at dose-response curves, and if we wished to do that, we would have done a different trial. That's something maybe in the future that we can do. I think that's something that's interesting.
DR. TEMPLE: Are you asking about the relationship of outcome to success in controlling blood pressure or to dose?
DR. HIRSCH: The former, the success in controlling blood pressure as being a means to determine and achieving that clinical outcome regardless of how difficult it was to get there.
And that gets back in a sense to the earlier discussion of how we make adjustments, adjusting over a group mean -- I want to do a regression and say, well, on balance, the whole population had a benefit that's independent of the blood pressure. But it may really well be that those patients treated with whatever combination drugs, losartan achieved the greater blood pressure lowering in those that had a clinical benefit.
And it has impact on how I translate this in my patients if this drug were to achieve approval. Overall, if I can achieve the blood pressure lowering, one way or the other, with compliance in any region, I may achieve a comparable benefit. Blood pressure is important here.
DR. TEMPLE: So, you might do this by -- that is, compare people who got below systolic of 140 with people who didn't, things like that?
DR. HIRSCH: You can look at those who achieved target. You could do it by tertiles. You could do an actual regression within each of the treatment groups. A change in the relationship between the two treatment groups in a regression would satisfy me. A different outcome in the two treatment groups for comparable blood pressure, getting rid of that delta, would satisfy me. That would tell me it's the drug not the pressure.
DR. KEANE: We do have some data in terms of actually trying to address the issue, Alan. It is confounded by the fact that all these things are going on.
DR. HIRSCH: I know how difficult it is. I really realize that there are many things going on here. But nevertheless, if I had to look at one factor, other than region, compliance, et cetera, it would be blood pressure in a hypertensive diabetic population.
DR. KEANE: Yes, and I think clearly this gives you the ranges of blood pressure that actually we achieved in the overall trial. So, it's hard actually. It doesn't answer precisely your question in terms of how we got there, but I think, as you can see, in amongst the different regions, blood pressure was reasonably -- and this is the mean arterial pressure -- controlled.
Do we have a stratification by achieved blood pressure?
DR. BORER: While you're finding that, just to get back to Paul's issue, I wonder if you have some information about the distribution of adverse events in the treated group over time. Forget about the relation to when they were titrated up or titrated down. If you saw that the rate of adverse event occurrence was approximately similar across the duration of the trial, then that would give you some confidence that you're not seeing something happen because more drug is being given over time. So, while you're looking for the data to answer Alan's question, maybe you can see if you have that as well.
DR. KEANE: Fine.
DR. BAIN: Dr. Borer, related to that issue of adverse events, we prespecified in the protocol -- I think there were -- six prespecified adverse events, which we looked at specifically, and what we did in that analysis was a time to the first event of that prespecified adverse event. Do we have that slide?
Now, just to confuse you, we have hazard rates here rather than risk reductions, but essentially to get the risk reduction, you just take 1 minus that hazard rate and multiply by 100. So, for each one of these, we looked at was there a difference between the two treatment groups in the time to the first occurrence of the adverse event that we prespecified. And as you go down the list, which was already introduced in the main core presentation, we did find a difference in hyperkalemia and also hypokalemia.
DR. BORER: Thank you, Ray.
And with regard to Alan's question?
DR. KEANE: We'll have to look at the database in greater detail to see if we can ferret or tease out some of that information.
DR. BORER: Okay.
DR. KEANE: Would you like me to actually try to clean up some of the issues that we addressed while we get this answer for you?
DR. BORER: That would be fine. Before we leave this regional issue, though, which we've had a lot of discussion about, the FDA reviewed this data. We have a medical reviewer. We have two statistical reviewers, Dr. Hung and Dr. Chi, and they saw all these things and didn't flag these as being show stoppers. I'd like to just hear from the FDA about the FDA reviewers' conclusions about the issue of differences among regions and differences among races. Dr. Hung, are you back there?
DR. HUNG: Basically I did some analyses and tried to explain potential differences between the Asian region and the other regions. I seem to feel that for some reason the Asian people have higher baseline proteinuria levels. So, I kind of got the impression that that probably at least partially explains the differences among this potential heterogeneity. Other than that, I really cannot conclude anything else. That's my best explanation.
DR. BORER: But you didn't find that the outcome, skewed as it might have been in terms of subanalysis by region, precluded the ultimate conclusion that the drug worked and that it was reasonable to infer that the drug worked. Is that correct?
DR. HUNG: Yes.
DR. BORER: You just didn't think the regional issue was a show stopper.
DR. HUNG: Right. I don't think the regional issue is sort of a killer for the evidence. But I feel that all the evidence is not strong, although I realize that because the triple endpoint has -- remember, this is a loss of information about creatinine in some patients. So, the current analysis, which is an intent-to-treat analysis, may dilute some of the potential signal. That's my feeling. I cannot say one way or the other.
DR. BORER: Bob?
DR. TEMPLE: Well, I don't have to tell you or anybody on this committee that this is one of the big conundrums in analyzing trials, and we've had some of those before this committee. There's a strong bias supported by strongly worded papers that instruct you that if you do subset analyses, you're some kind of idiot. Yet, the plausibility and the interestingness of them is overwhelming in the other direction. They seem very hard to ignore. It's perfectly obvious that if you take the triple endpoint and remove Asia, you've got very little left. Well, as I think Paul Meyer wrote in a paper, if you take the strongest anything out of a trial, it always looks weaker. Of course, from the other end, people like to drop the worst clinic because it was obviously a bunch of shoddy practitioners and you shouldn't leave them in. And then the study always gets stronger.
So, we're tormented by this constantly. I don't know how much attention people have paid to it, but the MERIT study of metoprolol in heart failure had a strong finding overall on the combined endpoint of hospitalization and death and a very strong finding everywhere in the world but in the U.S. on death alone, a nearly 50 percent reduction in mortality in the rest of the world and a 0 percent effect essentially in the United States.
So, we said in the label very carefully, this may be true, may be not true, and have been abused all up and down the world since then for relying on a subset analysis, which everybody knows is stupid because Richard Peto shows that you can use adiacal signs and make convincing arguments and stuff like that.
In some sense, there's no really perfect answer to this, which is why you're looking at all these other subsets. You're looking at the double endpoint. It's not as strong there. That makes some sense. But there is no perfect answer to these questions. You make the observation. If there's an obvious explanation, then everybody is happy, but there never is.
I think our overall impression is that you should make major decisions based on subgroups like that very cautiously. I won't say never. I think sometimes they're so overwhelming you can't sensibly ignore them, but you do it with the greatest possible care and you try to think of everything that might explain it and look in other places and do your best. But nobody can give you a yes or no answer. Therefore, we never do.
DR. NISSEN: If I can just comment very briefly, I think that was obviously wise advice. I try to restrain myself from looking at those too. When a study is very powerful, a strong p value, robust findings, I tend to look a little harder if something is marginal. I suspect that this committee in the past -- I have only been on here a year or two -- I think probably has done that. So, where we begin to take more credence in those groups is when we have an effect that's kind of marginal and now we're trying to understand why is it marginal. That is why I probed this area. I would never have probed this area if the treatment effect had been substantially larger and the p value a lot stronger.
DR. TEMPLE: Before you leave that, it's also true that if they're really, really strong, you can almost never make them go away by looking at a set. Well, so there's a certain tautology in that. It's true. When the p value is .02, all kinds of things can make it go away. It's not that hard. There we sit stuck on this pin.
DR. BORER: Before we go on to this next issue, Tom, you wanted to make a point?
DR. FLEMING: I'd just add a little more philosophy to this. As both Steve and Bob have acknowledged, this is an issue that is extremely difficult. It's an art.
We all recognize that in all likelihood the efficacy and safety of treatments probably do differ by various patient characteristics, and yet our trials are barely powered to be able to reliably determine treatment effect in the aggregate. So, when you start breaking down into subgroups, you're inevitably going to be underpowered to be able to really reliably detect the signal when there really is a difference, so you have false negatives. But you also have a great risk of false positives, because you're testing inherently so many different hypotheses by looking in all these subgroups, that you may see something that looks like it's an effect modifier and it's spurious.
My own sense is that doesn't mean we shouldn't look at subgroups and have some general sense of whether or not this is giving us greater confidence or lesser confidence about the reliability of the results. But most would argue in most settings, if we see something that is in fact evidence of effect modification, it's generally an hypothesis generation that needs some kind of external validation.
In fact, I've always argued when I look at subgroups there are three fundamental things I look for. One is what is the overall strength of evidence here that we have. How strong is the statistical evidence. And one of the many ways of looking at this are tests for interaction. The tests for interaction on a region is a .04.
I suspect when you take into context all of the subgroups that were looked at here, the probability that you're going to see something of this order of magnitude by chance alone is not negligible.
At the same time, the reason it concerns me a little bit is I care the most about the U.S. at this point because we're an advisory committee for a regulatory authority looking at U.S. applications, and the U.S. population does certainly show less effect than the average. But basically from a strength of evidence here, this isn't a show stopper as I see it.
The second criterion is biological plausibility. How plausible is it that there really is effect modification? The example I always use is herceptin. Herceptin is an intervention that was developed based on the concept of Her-2/neu overexpression. So, if I see effect modification by level of Her-2/neu overexpression in an advanced breast cancer patient, I'm not surprised. That's highly biologically plausible.
So, a lot of the probing you've been doing, Steve, I think is very appropriate here. Is there some rationale here that explains this just beyond statistical association? Right now, from what I've seen, I don't see a smoking a gun. That doesn't mean that it's not true, but I don't see anything that's truly substantially establishing plausibility.
The third criterion is independent confirmation. There needs to be some independent confirmation. My own sense about this is looking at region, it's in most trials likely that what Peto would say or Salim Yusef or many others who've written on this, their advice is wise. Most of these signals are, in fact, spurious. We need confirmation.
My only twinge of reservation here is it seems like we've seen quite a few studies coming before us where the North American or the U.S. population shows less effect, and that just may be a spurious observation on my part. But it would be interesting to go back and look more globally. I don't generally trust the results in a single trial that there's a region effect modification, but might there be more here that if we looked more widely, as the FDA can do with its benefits of seeing so much of what's happening in research -- and of course, this is an answer that I don't want to see be proven because we really do want to be able to rely on international results, and yet we do need to also know what the truth is.
So, my own sense, in terms of confirmation, is we have this single study, and one of the limitations is it doesn't give us an ability to confirm whether there's a region effect modification for losartan in this indication. But it would be of interest to see whether, in a broader sense, other studies that are at least in related classes might show any evidence of region effect modification.
DR. NISSEN: Tom, I'm sorry. One quick follow-on. I really wasn't probing the concept necessarily that this group had some modification of effect, but I was terribly struck by the fact that so many more patients in the placebo arm in Asia discontinued than in the treatment arm. And I wondered if the behavior of the practice patterns, the clinicians, how patients got in and out of the trial, was somehow different in this region and that that's driving some of this.
DR. FLEMING: Well, I think what we're seeing, as you astutely pointed out, for the differences I think is that so many on losartan in Asia didn't discontinue. What was really distinguishing was the Asian population on losartan had a very low rate of discontinuation. Now, that might, in fact, influence enhanced efficacy if higher levels of adherence implies higher levels of efficacy.
Also though, as you pointed out, a valid concern is if, therefore, there is particular adherence in the Asian population to losartan and more so than in the placebo Asian patients, does that in any way reflect on or influence the level of follow-up for outcome. I haven't heard anything on end-stage renal disease/death outcomes, but it might for the triple endpoint. But for me, I'm not so worried about that because I'm going to look at the double endpoint.
DR. NISSEN: Right, and the reason I was probing that, Tom, is why should they be so much more compliant with losartan in Asia than they were with placebo? It's a very, very striking difference.
DR. BORER: Everybody was more compliant with losartan than placebo, according to the data. The magnitude varied but everybody was, I think.
There was one, to me, striking point that came out of your regional blood pressure data that will lead into Alan Hirsch's question that I'll let him present himself. In the Asian group, the blood pressure effect of treatment was, to me, importantly less than in the other regions; that is, in fact, although we get into the mean arterial versus systolic versus whatever issue that we raised earlier, the mean arterial pressure for the last few years in the Asian group was actually higher on losartan than on placebo. So, the fact that the effect was seen in that group as strongly as it was, despite the fact that the blood pressure was higher on the putatively active agent, is an interesting piece of confirmatory evidence to me about the blood pressure.
But Alan raised another question about the relation of change in blood pressure to outcome, and perhaps you want to ask that directly.
DR. HIRSCH: It's not easy to sit between these two sides of the table and get a point out.
Just to make one more philosophical point before leading into the data, we're obviously looking for biologic reality not for differences only in practice standards or in how we defined end-stage renal disease. I hope. Because biologic reality is if the molecule works, hopefully, more or less genomically similar, that it will work everywhere in the world. At least I am going to make that assumption for the moment.
So, to link the two arguments here for a minute, my worry is that something is happening. There's something we're seeing in this signal in whichever countries comprise Asia whereby this is a group that has -- and here's my hypothesis -- a higher baseline protein excretion, perhaps again a higher placebo discontinuation rate, hitherto unexplained, and perhaps -- because we really were not sure -- a lesser blood pressure lowering effect, and therefore the losartan group looks better there. I don't know if that's true or not, but I was trying to make a chain of biologic causality.
So, what I was trying to come up with -- I think angiotensin II is potentially a toxic agent, but this is a drug approved for blood pressure lowering in a trial that was designed to lower blood pressure. I came up with two analyses that I thought might work again.
One again is since you're titrating the target blood pressure, could we see data that simply segregates all the patients in the trial into those that achieved their target and those that did not achieve their target and then look at the effect on ESRD or any event rate by achievement of target blood pressure. If all those patients that achieve target blood pressure have the same outcome, regardless of treatment allocation, then it's not the molecule. It's the blood pressure. Everybody with me?
Or another way of looking at it is to create, again, a regression or a slope. The y axis, for example, might be a risk or a hazard ratio, and the x axis, for example, might be again a change in blood pressure because you do have sequential clinic visits every 3 or 4 months, and again a change in the slope there would also indicate an effect of the drug versus placebo.
DR. BAIN: We have not done the analysis which looks at only those people who achieved their blood pressure and then what the outcomes are because essentially we're getting into this area of two outcomes and adjusting one outcome for the other and not really doing a randomized comparison. So, it's very difficult and complex. We can do the analysis. It's the interpretation that becomes extremely difficult.
DR. HIRSCH: I realize I may be asking for something more than can occur, but nevertheless, I'm making an effort and I'll look to my other colleagues for support or not.
DR. TEMPLE: In his discussion of this, Salim distinguishes between sort of good things to adjust for and look at and bad things to adjust for. The good things are all baseline characteristics because those are at least randomly assigned or at least you're pretty sure they are. Adjusting for outcome variables is very treacherous business because one factor can have different effects on both. So, as was just explained, they tend not to do those kinds of things. It doesn't mean you can't do them, but you've got to be careful and think hard about it.
DR. BORER: With that having been said, we have several more questions still unanswered, and I think, Dr. Keane, you were going to move on to clean up some of those and we'll see if you go down the whole list here.
DR. KEANE: Let me just bring back up Dr. Bain. I think this is a very important issue related to the outcomes in relationship with proteinuria and looking at end-stage renal disease as the endpoint. I think Ray has provided us with that information now so that you can take a look at it. It really is, I think, a very important relationship.
DR. BAIN: A little housecleaning first. Table 1 on page 8, where we were discussing earlier looking at the median time, and it looked like losartan was 13.3, 1303, and placebo was 1373. Essentially what was done in this table was they went to the median on the y axis, drew a line across until they hit the two curves, and then dropped the line down to estimate the median time. It turns out that these numbers are flipped. We confirmed that losartan is actually 1373 and placebo was 1303, and you can confirm that yourself by going to that figure of our triple endpoint and drawing a line for yourself and dropping it down. You'll see that you hit the placebo cumulative incidence curve first and you drop down to 1303.
DR. THROCKMORTON: What page?
DR. BAIN: Page 8 in the FDA, table 1, statistical review.
DR. BORER: Yes. Where it shows a 70-day difference, the 70-day difference is in favor of losartan.
DR. BAIN: So, that's number one.
Going back to -- it seems a very long time ago -- a conversation, where Tom nicely laid out the different types of adjustments that are typically done. We talked about a baseline predictor, and then we talked about an effect modifier. I want to take the first one first, and we're back to the overhead.
This is a request where, remember, in the main presentation we showed the baseline prediction of the triple endpoint, and then there were a lot of requests for the prediction of ESRD, the prediction of ESRD/death, and Dr. Temple wanted a prediction of doubling ESRD.
That's what we showed in the main presentation. Remember when we got out to about 16, we drew a line over, it was like 12, 15. Here, this one is the doubling ESRD, so there's a steeper curve so that baseline proteinuria is actually very predictive of ESRD doubling, and it's most predictive of end-stage renal disease. Again, remember, that's baseline proteinuria. So, that's the answer to that question.
DR. KOPP: Just one other point. That's proteinuria defined as milligrams of albumin per gram of creatinine.
DR. BAIN: Yes. Urinary albumin to creatinine.
So, that was our baseline predictor analysis. The next one that you requested was the effect modifier analysis to see whether or not the risk reduction varied across the categories of baseline proteinuria. Slide 605.
So, what we're doing here is we have five categories of baseline proteinuria. We have the total sample size of individuals within each one of those categories. And the last column is the risk reduction. Now, remember, this is the risk reduction for the primary composite. Remember, overall the primary outcome was 16 percent. So, you can see that when you're less than 1,000, it's a risk reduction of 4.6. It goes up to 15 when you're in 1,000, and then 2,000, it's up to 17, all the way up, and when you're greater than 4,000, with 163 patients distributed between the two groups, you're at about 20 percent. So, that's the risk reduction as a function of baseline proteinuria category. So, that's housecleaning number two.
Now, one last thing that I'd like to do is something that we talked about much earlier today, which was the prespecified analysis for what we called the baseline risk score. Could I see slide 1339?
This is just a little paragraph from our data analysis plan that indicates what this analysis is. So, we were looking at the treatment effect comparing losartan to placebo, adjusted by various baseline covariates, and this analysis was done in two steps. The first thing we did was we developed a risk score. What we did was we took pooled data. We pooled the placebo and losartan data, and formulated a baseline risk score for our primary composite endpoint. Then after we did that, we took that risk score and estimated the treatment effect when we controlled for this baseline risk score.
1340, please. We prespecified in the DAP a total of 15 baseline risk factors that we would be interested in evaluating in this manner, and they are listed there.
1341. Essentially what we did for our primary composite outcome, the baseline risk score was defined as a linear combination of those covariates that were significantly selected among the covariates. The way we did that was we just did a stepwise selection procedure using the Cox regression model without treatment effect in the model, meaning we were pooling the placebo and treatment groups. Then the treatment effect was determined by performing this model with terms, including the treatment effect and the baseline score. So, that was the second part. And then we reported the results with a p value and 95 percent confidence interval.
Now, we're on to 1135.
DR. FLEMING: If you go back a slide for a second.
DR. BAIN: Okay. So, urine albumin to urine creatinine ratio. That was our prespecified stratum. So, those are the 15 characteristics, some of them continuous, some of them categorical.
DR. FLEMING: So, certainly key ones would be systolic blood pressure based on what we've seen. You're going to be identifying those that are predictive of outcome and independently predictive, and that's a very rational thing to do. At the same time, what guides some of our interest as well is whether factors are imbalanced, and one that's not as strongly predictive, if it's imbalanced, really is key. So, hematuria and systolic blood pressure would be two that we would think that there's interest in.
DR. BAIN: Okay. So, we'll take a look at those.
So, now we're on to slide 1135. So, in summary, we used the multivariate Cox regression model, pooled the treatment groups, and looked at the primary composite endpoint as a dependent variable, and selected significant predictors by a stepwise procedure. Then we calculated the risk score as a sum of the products of the significant predictors based on their regression coefficients from those that remained in the Cox regression model after the stepwise procedure.
Next slide. Now, this is two slides, but what I'm going to show you here is these are the two groups. Although we did the analysis pooled, I'm going right back to our baseline slide where, remember, we had a slight imbalance in the mean proteinuria between the two groups, and that's still there. And you can see serum albumin, serum creatinine, hemoglobin, sitting systolic blood pressure. So, on this slide you have five of the ones that remain in the risk factor model.
What I put on the right-hand side is kind of an indicator of the strength of an individual variable and its relationship with the risk score. It turns out that the strongest of our baseline risk score factors was urine albumin to creatinine ratio.
The next slide will give you the rest of the covariates that were in this model. It turns out that Latin American, yes/no, insulin use, and at the bottom here you see what I was alluding to earlier, the linear score of significant factors. The more negative that value is, the higher your risk. So, you can see that, when you do a linear combination of these risk factors, you tend to see that the placebo group is -- I'm sorry. The other way around. The more negative, the less at risk you are. Therefore, what's shown up here is the losartan group has a higher risk when you look across not just one factor, but all seven of those factors.
Next slide. So, then what we did was we took it to step two which was take that risk factor score, which is a single score for each individual and enter that into our original model, which was our primary results model, which was where we showed that 16 percent reduction. And if we now adjust for their baseline risk score, the risk reduction goes from 16.1 to 23.9.
Graphically it's shown on the next slide. There's our primary outcome above, and when we adjust it, a stronger treatment effect, less than .001.
Although we developed the risk factor score based on the triple endpoint, we actually used that risk factor score to see what effect it had on our other clinical endpoints, and that's on the next slide. You can see that you see that same effect. For ESRD, the treatment effect got stronger, and for end-stage renal disease or death, the adjustment made it slightly stronger.
DR. BORER: Ray, can I just ask about the factors you used? It seems that all of them are reasonable except perhaps for region. That sort of begs the issue. Did you sort of look at this? Latin America, yes/no, probably was down on the list.
DR. BAIN: Well, no, the original list had all regions. It just turns out that that particular region of the four --
DR. BORER: No. I understand. But you ultimately used Latin America, yes/no, in the model that you used for adjustment. Wrong?
DR. BAIN: No. Latin America went into the risk factor score. It was one of the things that you multiplied by in order to come out with a single number for a patient.
DR. BORER: Right. That's what I mean. So, it was used to adjust for risk.
DR. BAIN: Correct.
DR. BORER: And all I wanted here is that if you took Latin America out of there, you've still got a directionally similar movement --
DR. BAIN: We would have to run that analysis. We just did? The same.
DR. FLEMING: It should be.
It's certainly relevant to know that we didn't just look at proteinuria, but what's apparent from this is that when you look at those covariates that are going to fall out of a variable selection model, the one that's dominating here is proteinuria. It's very predictive of outcome, and it's the one that is influencing outcome effect because there's also this imbalance in the tail. So, this is certainly a reassuring analysis to say it's not just that we're looking at proteinuria because we could see that it had this imbalance.
It would be relevant, though -- at least for me. I'd like to see the analysis that simply adjusts simultaneously for baseline systolic blood pressure because the differences that we see over time were largely already apparent at baseline. And Bob Temple is right. One has to be really careful when you're using post-baseline values of covariates and how you interpret the results. But some of the blood pressure difference was already apparent at baseline. So, even if it doesn't show up in your variable selection model, it could be an influential factor as a confounder because of that imbalance.
My suspicion is that the most significant confounder here is proteinuria. It's increasing the estimate of effect from 16 to 23 percent or 22 percent, and systolic blood pressure at baseline I'm guessing will correct that back 1 or 2 points, and in the end you'll still have a net increase but not quite as much when you just look at proteinuria, which is more or less what's showing up out of that analysis.
DR. BAIN: So, we'll do that.
DR. BORER: While you're doing that, JoAnn, you had several questions related to looking at the double endpoint, and perhaps you want to restate them if they haven't been answered already.
DR. LINDENFELD: No. I think I've gotten an answer to those in this data.
DR. BORER: Blase, I think you had asked about the time to event when you looked at the double endpoint as an outcome, the average time to event rather than the triple endpoint. I'd like to hear that too, if you happen to have that, understanding, as we all do, that it's a poor man's way of looking at anything; that is, the average time to event among those people who had events when the event was the real hard endpoint.
DR. BAIN: You wanted it for ESRD only, Jeff?
DR. BORER: ESRD or death.
DR. BAIN: Okay. So, why don't you put up the ESRD slide from the core presentation, or whatever one.
Now, of course, in this particular outcome, we don't get to 50 percent of events on the y axis, so we can't talk about the median time. But we can talk about, for example, the time to 20 percent of the patients reaching their endpoint. Therefore, you just draw a line over from 20 and drop it down. So, it's about 2-and-a-half years for placebo and 3 years for -- so, about a 6-month difference.
DR. BORER: How about if you do the same thing since that's obviously a reasonable way to get a gestalt of the number we asked for. If you look at the post hoc analysis curves where you accounted for MI, stroke, cardiovascular endpoints, so we're seeing a net effect.
DR. TEMPLE: What are you asking, Jeffrey? They saw no difference in those.
DR. BORER: No, no. What they did was to combine ESRD, death, myocardial infarction, and stroke in one analysis.
DR. BAIN: So, Jeffrey, let me make sure I understand. Which endpoint?
DR. BORER: There it is.
DR. BAIN: Now, here you could probably maybe go up to 40 percent, although it's getting pretty thin up there. But again, you would do the same thing. You could draw a line from either 40 or 30 around, and you're going to see that there's a difference in the time to the events.
DR. BORER: Yes. We're still talking about 6 months.
DR. BAIN: 6 months.
DR. BORER: Beverly?
DR. LORELL: I think the other way, Tom Fleming's comments notwithstanding, that both Blase and I were interested in getting a feel for is among those patients who had a major heart event, what was the average time to that event.
DR. BORER: These are the people who had endpoints. So, that's about the best we're going to do, but it looks like it's about 6 months.
DR. FLEMING: Before we leave this, could I comment on these two slides, 379 and 73, maybe going to 397 first?
From a renal perspective, this is what one might expect to be the component of a composite endpoint when you look at the clinical endpoints, the component that would be presumably most specifically being targeted, prevention of end-stage renal disease. It's interesting, when one is doing studies such as these that there's a lot of wisdom for not doing a study that would have 2 years of median follow-up. And this study, in essence, has information predominantly through 3 years, a limited amount of information out to 4 years.
Just to follow up on some previous discussion, this estimate is a 28.6 percent relative reduction in the rate of failure. That's based on a weighted average of what the true reduction is at all points over time. The true reduction in the first 18 months is 0. The curves are overlapping through the first 18 months, which means that the true reduction between 18 months and about 42, when you have most of your data, is probably about 40 to 45 percent, meaning that if you do a study like this and you use the log rank and the Cox regression methods and get relative risk reductions, those estimates are a weighted average over time of what truly isn't a truly constant reduction over time. The reduction in the first 18 months is 0; in the last 18 months is 40.
So, if you had done this study with one less year of follow-up, you would be weighting much more proportionally on the 0 rather than the 40, and your estimate would have been 20 percent. If they had done another year of this trial, and these curves represent truth, the estimate would have been even greater than 28.6.
So, it's interesting to look at these curves and understand what these relative risk estimates mean, in particular when there's evidence such as this that the reduction isn't constant over time.
The second point on the next slide, if you look at slide 73, you see the same basic phenomenon. One of my questions here is this is in fact an endpoint that some of us in January had really wanted to focus on if we wanted to focus on clinical measures that were combining both the renal as well as cardiovascular elements. In the renal, we were saying end-stage renal disease/death; in the cardiovascular, were cardiovascular death, MI, and stroke. So, I'm delighted the sponsor presented this.
Essentially what we're looking at is this is made up of 47 fewer end-stage renal disease events, 18 fewer MIs, 3 fewer strokes, and 3 excess deaths, which are made up of 11 excess cardiovascular deaths, but 8 fewer of the non-cardiovascular deaths. That's a net difference of 65 events, although I can't tell how many people that is. I'm assuming it's probably about 40 to 55 fewer people have at least one of these events because this is a time to the first of those types of events analysis. Can the sponsor clarify what that is?
DR. KEANE: We're just looking, Dr. Fleming.
DR. FLEMING: You can tell me later after you have a chance to look.
DR. KEANE: It is.
DR. FLEMING: What's the answer?
DR. KEANE: Time to event.
DR. FLEMING: I can add up and see there are 65 fewer events, but I can't tell how many fewer people had at least one event. That's what this analysis is looking at. It's time to the first of those types of events. I'm guessing it's 40 to 55.
DR. KEANE: We're in the process, as I said, to get the number of patients that were involved with this.
DR. BAIN: One more housecleaning chore. Tom asked when you go and do the Cox proportional hazards regression on the triple endpoint, remember it was a 16 percent reduction. When you adjust for baseline systolic blood pressure, it goes to 21.6 in addition to baseline proteinuria stratum.
DR. BORER: My list of questions is exhausted but yours may not be. Did you have some other clarifications you wanted to give us?
DR. KEANE: Yes. If you will just give us a couple of more seconds, we'll give the response to Dr. Fleming's question in terms of the number of patients that the event curves actually encompassed.
DR. BORER: While you're doing that, what we'll move on to next, as soon as you have that answer, is the questions, and we'll structure the rest of the discussion around those.
DR. KOPP: One additional question.
DR. BORER: Two. I think JoAnn has one there also.
DR. KOPP: To return to the issue of how you define end-stage renal disease, it was suggested during the break that one additional analysis would be to ask what was the final creatinine before somebody went on dialysis and did it differ between the two groups.
DR. BAIN: Now, I believe the question was that and by region for serum creatinine. At least, that's what I have here.
Let me tell you by region these are median serum creatinines. I'm going to give you losartan then placebo. In Asia, it's 7.0 and 6.2. In Europe, it's 4.9 and 5.3. In Latin America, it's 5.2 and 6.6. In North America, it's 4.4 and 4.9.
Now, the interesting thing about those numbers is you'll probably notice that they tend to be lower in Europe and North America. Well, it turns out that the time from the serum creatinine to end-stage renal disease is actually higher in those two groups, probably driven by study drug discontinuation and then some of those patients going into telephone follow-up. So, therefore, you're not getting serum creatinines very close to their ESRD events.
DR. BORER: It seems clear, if I heard the presentation correctly that you just made, that the creatinines at endpoint were systematically higher in people on placebo than on losartan. Is that correct? Did I hear that right?
DR. BAIN: No. They go back and forth. It's true in Asia. It's not true in EU. It's not true in Latin America, and it's not true in North America. So, in Asia it tends to be a little higher.
But now, remember, we're talking about in Asia a total of 60 events; in EU, a total of 40; 60 in Latin America; and North America -- you know. So, the numbers are pretty small here. Remember, these are only people who actually had the event of end-stage renal disease.
DR. BORER: Right. Is there any reason that there should be a discrepancy as large as 1 milligram percent of creatinine between those groups? Does that have to do with timing of checking --
DR. BAIN: I'll put out a possibility there. These are people who are going to end-stage renal disease. So, it really is probably a function of exactly where you're picking them up relative to their end-stage renal disease. That's a guess.
DR. BORER: It's probably an unanswerable question.
DR. LINDENFELD: I have two questions about secondary endpoints. The first is we've seen that there's a nice reduction in proteinuria. Could you tell me if the reduction in proteinuria correlates with the doubling of serum creatinine or end-stage renal disease? Just in terms of how we use these drugs, I'm interested in knowing if the change in proteinuria correlates with the other endpoints.
DR. KEANE: It does.
DR. LINDENFELD: Strongly?
DR. KEANE: Strongly.
DR. LINDENFELD: Good.
Then in terms of cardiovascular endpoints, I know this is a difficult issue, but I'd like to hear your discussion. The study was stopped prematurely because of data with ACE inhibitors and cardiovascular mortality. Here we see very little signal for cardiovascular mortality. I understand there's a little bit shorter follow-up here and fewer numbers, but we see very little signal there. I wondered if you could just tell me your thoughts about why there doesn't appear to be much effect on cardiovascular mortality.
DR. KEANE: There are a number of things that I think are clinically relevant. First of all is the RENAAL trial is a renal protection study. It wasn't a study of cardioprotection. So, we had a smaller group of patients, 1,500. Usually when you look at a cardioprotective study, it's substantively larger. So, that was one set of issues.
And number two is actually the duration of the trial was relatively shorter as compared to most cardioprotection studies.
When we look at the type of patient we had in our study, we made a very concerted effort to not have patients that had a lot of cardiovascular disease antecedent to our randomization process for the RENAAL trial, so that we had lower cardiovascular disease manifestations or disease history. We had a smaller group of patients. We were looking at specifically enriching our population for renal events. So, I think when you look at all of that, I think that explains, at least in part, the reason why we didn't see a major difference in cardiovascular endpoints.
Again, our composite, I'll just underscore, was almost a 10 percent risk reduction for the composite. So, it was in the right direction, and as I showed in my concluding slide, there was some noise around the 0 line in terms of the point estimates for each of the components of the overall. So, I think that's most likely, as best I'm able to look at, the explanation for why there wasn't more robust changes in cardiovascular disease endpoints.
DR. BORER: We'll go on to the -- I'm sorry. Blase?
DR. CARABELLO: But from a functional standpoint, the fact of the matter is, the Data and Safety Monitoring Board said these patients should not be precluded from being on an ACE inhibitor, or maybe you could even take that as a recommendation that they should be on an ACE inhibitor. I've got a sea of folks in my hospital who are diabetics and already on an ACE inhibitor. We don't have any data about that interaction. What are we supposed to do?
DR. KEANE: I'm not sure I completely understand what question you're asking me.
DR. LORELL: May I help?
DR. KEANE: Yes.
DR. LORELL: I think I know what Dr. Carabello is getting at. I think it's a very difficult issue here.
The issue of use of an ACE inhibitor was raised in two ways in this trial. It was raised in study design by prohibition of use of an ACE inhibitor during the trial and a washout period. Secondly, it was raised, as Dr. Carabello pointed out, by the Steering Committee prematurely stopping this trial because of the issue of cardioprotection in multiple arenas of the use of an ACE inhibitor.
We've seen data presented here, very elegantly, of really an extraordinary tight link in this patient population of renal events and cardiovascular events. If my numbers are correct, 1 out of 5 of the patients in this study died, and in the losartan group, 57 percent of those were cardiovascular events. So, this is a major clinical issue for Dr. Carabello's patients and most of us around the table.
So, I think one of the real dilemmas here that was part of design and part of trial stoppage is the issue of how you think about this drug for renal protection in the absence of demonstration of a cardioprotective effect, which is what kills many of these patients. And it's very problematic. We don't have any data here in this population about combined use of ACE inhibitor and ARB.
So, I think what Dr. Carabello and I might like to hear is your thoughts in industry, were this drug to be approved, as to how you would recommend the use of this drug relative to the absence of a demonstrated cardioprotective effect as your secondary endpoint and the ethical issue that was raised by the Steering Committee in stopping this trial of potential ethical need to use an ACE inhibitor.
DR. BORER: While you're answering that, just clarify for me. I believe you had 100 patients who actually were on ACE inhibitor because the ACE inhibitor prohibition was by amendment after the trial had started, if I'm not mistaken. So, you may actually have some data that are relevant.
DR. KEANE: Actually that is not correct. All the patients prior to randomization had their ACE inhibitor or ARB stopped. Let me just make some points and then we can have some additional discussion.
We did look at interactions, as you saw, between prior ACE use, prior ARB use, and either the renal or cardiovascular endpoints, and there was no interaction, as best we could see, in that data set. So, it didn't look like prior utilization of ACEs or ARB impacted any of the results that we had.
Number three is that in terms of the Steering Committee's decision to stop the trial, obviously we didn't know what the recommendations were at that point in time. Dr. Brenner is here, if you'd like to have this discussion in greater detail. What our concern was, in a more global perspective, is that we had a placebo group of patients that were not on any AII blockade. With the Mann data that came out in the Annals last year, that raised at least a concern for us in terms of continuing the trial. And that was really what I think our issue was. It wasn't that we weren't seeing any other benefits for ARB, but the issue was that we had half of the trial that was not on any AII blockade.
Then finally, let me just make the point that you have to take our data in the context of what the study was. This was a study that was done in patients with advanced renal functional declines, type II diabetics with proteinuria, and all had fairly advanced disease. All of the other ACE data that is out there is not in this population. So, it does present us with a bit of a difficulty in terms of what should be done on an individual patient basis, but our focus really was in the patients with advanced renal disease, with proteinuria, and looking at renal outcome. So, we didn't have that data at that point in time when we stopped our trials.
DR. BORER: Steve and then Bob.
DR. NISSEN: Yes. I think we're all saying about the same thing. Let me see if I can be very precise here.
Our diabetic hypertensive patients with renal insufficiency are, by and large, all on ACE inhibitors, and they're on them for two reasons I think. One is that there's some pretty good evidence of cardiovascular protection that I think most people would generally accept, and many individuals have extrapolated from the type I data and said, well, if it works in type I nephropathy, it probably works in type II even though the agency never approved it. So, now we're confronted with if the agency approves losartan for this indication, clinicians are faced with a really big dilemma. Do you take the patient off the ACE inhibitor and switch them over to losartan? Do you add losartan to an ACE inhibitor?
I think what people are getting at and I would like to be very specific about is our fear that for patients that are stable and doing well on ACE inhibitors, because of a label change, someone is going to say to the physicians, here's the only labeled drug for renal protection in type II diabetes. Let's have you take your patient off of your ACE inhibitor and put them on losartan. Then we don't know whether the cardioprotective benefits of ARBs are comparable to ACEs or not. We don't know that.
So, this creates a huge dilemma and that dilemma was manifest by the Steering Committee of this trial feeling like they couldn't go on with the trial because they couldn't withhold ACE inhibitors from these patients. So, we're really on the horns of a terrible dilemma as a consequence of that.
DR. KEANE: Let me just call on Peter for a second here.
DR. BORER: Just before you do, Bob, did you have a comment to make first? Then, Peter, maybe you can comment.
DR. TEMPLE: I just thought it would be very helpful to pin down exactly which treatments are because we're analogizing and doing our best and which treatments are really well documented. The only data comes from HOPE. Right? Not the same population. No renal disease, but they did have macroalbuminuria. Right. So, it's those people you'd be worried about, people who are being put on some ACE inhibitor or other on the basis of HOPE or ramipril itself, which is the only one that was actually studied. So, that's the group that you're worried about. What do you do with those people now? If they turn out to have a little elevation of creatinine, do you now ignore those results and switch or things like that? Good questions.
DR. BORER: Peter?
DR. KOWEY: Let me just address this. Peter Kowey, paid consultant for Merck.
Blase, let me put this in some perspective because it's a question that's come up repeatedly in looking at this information. A lot of people have already asked this question. And the answer is that from my perspective as a cardiologist, we frequently have patients that have competing risks and have competing diseases for whom there are some therapies that actually overlap. There are situations where we have to make a value judgment as to how important the cardiovascular end of things are going to be in this scenario versus how important is the renal end of this disease in this scenario. Because you're right. We do not have data on combined use of these two drugs. So, we can't advocate that.
As a cardiologist, my answer is that the more I'm concerned about these patients being disposed to a renal endpoint from the point of view of how they look, vis-a-vis the patients that were enrolled in the trial, the more likely I am to use an ARB, and the more likely they are to look to me like a cardiovascular patient, the more likely I am to use an ACE inhibitor. But it's a judgment that needs to be made on a patient-to-patient basis.
I think it's a little tiny bit unreasonable to expect a trial like this to answer every question that can be asked about cardiovascular disease. It can't. There are data on both sides of the question for ARBs and ACE inhibitors even within this realm.
So, I understand your question but I don't think that approving this drug for this indication necessarily places patients at risk. It places doctors in a position where they have to make a clinical judgment, but you do that all the time anyway. So, that's where I think this sits.
DR. BORER: I'd like to weigh in here just for a minute because this is a drug approvability panel, advisory panel. I think that we have to consider what the FDA does and what it doesn't do, and as Bob pointed out, what the data that we're weighing in against are.
My understanding of the HOPE trial is it involved individuals who were greater than 55 years of age with known coronary disease and at least one other risk factor and that risk factor could have been diabetes. It could have been hypercholesterolemia. It could have been hypertension. And one drug was used. Actually two, if you include vitamin E, but that didn't work. So, that's the database. That's it.
Having said that, the FDA doesn't establish medical practice. Sorry, Bob. I know that bothers you.
DR. BORER: No, the FDA doesn't mandate medical practice. It may indirectly, but that's not what it does. So, in a sense this issue is not a primary FDA concern. The medical practice is determined by consensus, by advisory panels, which I'm not too happy with, and by the courts based on what evidence you can bring to bear if something goes wrong and somebody says it's your fault that it did.
Ultimately there aren't the data to draw the firm conclusions that people might draw about should you be on an ARB, should you be on an ACE inhibitor, just as Peter said. The data aren't there. You have to make your best judgment based on the data that exist.
I would like to suggest, therefore, just so we can take this issue off the table and not make it an approvability issue, when I don't think it really primarily ought to be, but it's worth a great deal of thought because it's a very important issue because people do act on the basis of just what Steve said and Beverly said -- that is, if the indication is there, then it may be malpractice not to do this -- that perhaps the FDA should consider, if everybody on the panel agrees to this, something in labeling that says what we know and what we don't know and that anything that's said about this drug shouldn't be construed to suggest that the approval, if it happens to be approved for this indication, should be construed as a mandate that it must be used, that indeed competing risks, individual patient, et cetera. This is a drug that can do certain things or maybe can't. If we find that it can, then it can.
I think that this is something that we've never considered before in a formal way, and I think considering the possibility of putting something into the label about it might help. But I think that fundamentally we shouldn't make approvability decisions based on this because the data aren't there to allow us to do it. So, I throw that out for discussion so we can get rid of it before we go on.
DR. TEMPLE: It's an interesting thought. We occasionally do say what is now known about something although we don't we do it very often. It gets very, very difficult. For example, I'm sure most people think that ACE inhibitors should be used in the population that was studied in HOPE. The fact is only one drug of that class has been studied. There isn't a second study. The result was dramatic and, incidentally, not so impressive in the United States.
DR. TEMPLE: For what you will choose to make of that.
Well, if you really wanted to set the whole stage, you'd have to write a little essay describing all those things and say what you should when one member of a class does something. How much should you believe about all the others? And then by the time you're done talking about people with other risks -- for example, if you have a little heart failure, well, there's a lot more data on ACE inhibitors than there is about AII blockers. So, that might influence you.
This is not an easy thing to write. I think you should write a chapter. But we would certainly consider trying to provide some perspective as best we can. But our trouble is we're not supposed to infer things too much. We're supposed to be even more data-dependent than you guys are, and that really means you can't speculate at all in labeling. So, it's a problem.
DR. BORER: Steve?
DR. NISSEN: The reason we have this dilemma, of course, is that a therapy, ACE inhibitors, in this population became virtually the gold standard without ever coming before the FDA for approval. That is to say, ACE inhibitors for renal protection. If you talk to diabetologists and to nephrologists -- you guys correct me if I'm wrong -- but for many a year everybody has put these patients on ACE inhibitors for their renal protective effects. That's reality and we have not changed that. So, you see the reason it's a little bit different than some other circumstances is not so much a question of competing risks, it's also a question of are ARBs as good, even though ACE inhibitors were never approved, so that what will the public policy implications be of approving under those circumstances.
DR. TEMPLE: It's not not approved. They weren't studied.
DR. NISSEN: I understand.
DR. TEMPLE: It's not the same thing.
DR. NISSEN: Okay, I understand.
But I'm saying a therapy that was neither studied nor approved has become the gold standard. So, now you've got this gold standard that's actually made of bronze, and now what we're looking at is what evidence do we require in order to set the conditions in motion for people changing that practice.
DR. TEMPLE: I will say something about that. This comes up when we try to think about how something compares to "available therapy." Well, what does available therapy mean? What people do or what we've actually written up? We actually have just put out a final guidance on this. This is important because it determines whether you're a priority drug or not and other things that you probably don't care that much about.
What we have is a strong bias toward available therapy, meaning something that has been through our review. You might think we would think that and you wouldn't be surprised. But it means someone has bothered to pull the data together and put it forth, and you get extra credit for having done that. Our inclination is to think that approvals and things like that should take into account what has been well documented, but probably shouldn't worry too much -- you may choose to worry about it; that's your privilege -- about what might be true but nobody has ever bothered to study or been able to study. It would be hard to study those things now. That's probably the reason nobody does it.
In labeling and other places, we do tend to focus most on what's actually known and studied and been before us and been reviewed by expert committees and less on things that might be true but haven't been studied.
DR. BORER: Dr. Haffner, did you have a comment before Paul and Beverly?
DR. HAFFNER: Yes. First of all, I understand I have been active in the ADA in professional practice, and this is a very difficult issue. In fact, the ADA came out with new recommendations on their professional practice recommendations in January and they disagree in different position papers. The hypertension paper said that ARBs and ACEs were both first line partially because we hedged our bets for the cardiovascular issue. The nephropathy guidelines said that ARBs were first line and ACEs were a second choice. And this is actually the same professional organization.
I should mention that the entire thing is driven by HOPE data, and the HOPE study, as you know, didn't actually examine this issue. There is no ACE data in people with advanced renal failure. And if you go back to the two years before HOPE came out, in fact many people thought that ACE inhibitors might not be so great on the basis of the UK PDS data. So, HOPE isn't the only study in diabetics. The UK PDS data compared this relative to atenolol and actually atenolol did a little bit better than an ACE inhibitor in spite of the fact the ACE inhibitor is better tolerated.
So, I agree. I think ACEs are a good therapy. I don't think they have been well established in progression of renal disease, but I don't frankly know in these sorts of patients whether ACEs are better than ARBs. We don't actually have that sort of comparison, and my guess is we'll never actually have that comparison. So, I think there is some general uncertainty and you have to look beyond the most recent study. There is a long history there with some real questions involved in it.
DR. BORER: Thank you. I should point out too that the issue of approval of ramipril for the renoprotective effect that was putatively seen in HOPE was voted against by this panel when it came up a couple of years ago.
Paul and then Beverly.
DR. ARMSTRONG: In responding to your challenge, Mr. Chairman, which I think is were this drug to be approved, what would the label look like, my own view would be that it does work in a very specific population that we've heard about today and does not cause harm relative to some other issues that we would be concerned about that coexist in these patients and indeed in the broader population that we all see. So, I think crafting a label, to the extent that we know what it does or we think we know what it does in a fairly select population where a number of issues that patients receive ACE inhibitors for are not present, would be prudent. We obviously wish to avoid the issue that Dr. Carabello has raised which would be that a broad number of patients at risk to cardiovascular disease who are receiving ACE therapy now for good indications and evidence-based medicine would not be switched off them because they have co-existent renal disease.
DR. BORER: Beverly?
DR. LORELL: Thank you. Well, that really underscores a concern that I had along this issue. I think that if cardiovascular events were defined not as a retrospective, ad hoc look but as a clear secondary endpoint in this trial -- and I think one of the concerns in my mind, just as was pointed out in the presentation earlier today, is the data on cardiovascular events were inconsistent in this trial. They didn't all go the right way. The cardiovascular death event -- and cardiovascular death was 57 percent of all deaths in the losartan arm -- in fact, went the wrong way. So, I think it's an added source of some unease in this discussion that the cardiovascular death, in fact, looked to go the wrong way. The data were not internally consistent in sitting on the left of the line, and this committee really wrestled with that issue as a component of our decision regarding a different ARB for the same indication not too long ago. So, it is problematic.
DR. BORER: Yes. It's a problem. Clearly, again, the cardioprotective studies didn't involve people with near end-stage renal disease and whatever. So, it's hard to get to. But that's why I think, at the end of the day, we may have to suggest to the FDA that it do some label crafting.
DR. TEMPLE: Remind me. But except for HOPE, most of the cardioprotective effects of ACE inhibitors have been shown in people with ventricular dysfunction. Somebody needs to correct me if this is wrong, and that's not true of HOPE. So, these people didn't have that. The only figure that related to heart failure actually was going the appropriate way. So, I'm not so sure, except for HOPE, which I do find inspiring -- it's not quite clear what the effects of ACE inhibitors are in all those other settings in the absence of ventricular dysfunction. I think that's true.
DR. BORER: Yes, that's true, and I think you would have said HOPE is hopeful, had you the facility with words that Dr. Konstam does.
DR. TEMPLE: Of course, as someone who lives in the United States, I'm not sure I know about that statement.
DR. BORER: Why don't we move ahead to the questions.
But before we do that, I'm going to have to ask a point of information. It says on here that there is to be a break at 3 o'clock. Now, it's a quarter of 3:00. I'm wondering if we're mandated to take that break? We're not? We're not. Okay.
Oh, excuse me. One thing I've forgotten to do that I should have done is I didn't ask if there was any public comment. The meeting should have been open at 1 o'clock for public hearing. We didn't have any formal requests, but was there anyone who came here to make a public statement?
DR. BORER: No. The record should show that there was no request for public comment.
Now let's go on to the questions here. The Cardio-Renal Advisory Committee is asked to opine on the benefits and risks of losartan, an angiotensin II receptor antagonist, for the treatment of nephropathy in type II diabetes. Reviews of chemistry, pharmacology, toxicology, biopharmaceutics, biometrics, and clinical safety present no apparent barriers to its approval.
The committee is asked if it believes the strength of evidence for a treatment benefit supports approval.
The direct evidence is derived from one study. RENAAL enrolled 1,513 subjects with type II diabetes, hypertension, proteinuria, albumin to creatinine ratio greater than or equal to 300 milligrams per gram, and serum creatinine between 1.5 and 3 milligrams per deciliter. Subjects were randomized to placebo or losartan, titrated as tolerated from 50 milligrams to 100 milligrams, and followed for a mean of 2.4 years. The primary endpoint was a time-to-first-event comparison of losartan and placebo for death, end-stage renal disease, or doubling of serum creatinine. The result was an estimated risk reduction of 16 percent, p equals .022, with treatment groups diverging after about 6 months.
So, our first question. There were 686 total endpoint events in the placebo and losartan groups, 32 fewer in the losartan group than on placebo. One of the characteristics of a none-too-small p value is that the result is sensitive to the handling of subjects with incomplete data. And Tom actually got into that in some detail earlier. In RENAAL, there were no subjects randomized but not treated, no subjects with questioned event adjudication, and no subjects lost to follow-up for end-stage renal disease or mortality.
So, 463 subjects discontinued the drug. How were they handled? How should they have been handled? And what effect did the sponsor's rules for handling dropouts have on the credibility of the principal findings?
We can deal, I think, with all of those together. Tom, do you want to start off?
DR. FLEMING: How were the patients who discontinued drug handled? My understanding is they were all followed from that point forward. All of them were followed for the end-stage renal disease/death endpoints. As I understand, 40 percent of the person-years of follow-up subsequent to that time, though, patients were not followed for the change in doubling in creatinine time. As a result, the procedure used by the sponsor to handling those patients who discontinued was fully appropriate, giving us complete follow-up information about end-stage renal disease and death.
There is, however, a problematic issue arising with the 267 patients who did, in fact, have discontinuation of their follow-up of doubling in serum creatinine time prior to having any of the elements of the triple endpoint. While those people did, in fact, have subsequent follow-up for end-stage renal disease/death, it makes the triple endpoint analysis a bit more difficult to interpret, but the double endpoint, end-stage renal disease/death, analysis is fully free of any of that concern.
DR. BORER: JoAnn, you were the committee reviewer. Do you have any other issues?
DR. LINDENFELD: No. I have nothing to add to that. I agree.
DR. BORER: Does anybody around the table disagree with those conclusions?
DR. BORER: No? Great.
We will move on to number 2. Of the 686 primary endpoint events on placebo or losartan, 52 percent were creatinine elevation and 48 percent were death or need for dialysis. All of the treatment difference was the effect on creatinine. Was this a statistical anomaly? Was this because there were just so few clinical outcome events? Was this because the effect on clinical outcome would not be expected over 54 months? Was this because an effect on serum creatinine is a poor predictor of clinical outcome?
Why don't we stop there, and JoAnn, why don't you take those four together? Then we'll go on to number 5.
DR. LINDENFELD: Let me start off with was this a statistical anomaly. I don't believe so, and I don't believe that this was just that there were so few outcome events. I believe that the serum creatinine is a predictor of clinical outcome. I think when we look at the combination of end-stage renal disease and death, that's positive. So, I would say that, to just phrase this a little bit differently, we did see the double endpoint of end-stage renal disease and death was positive, and I think the creatinine is predictive of that. So, let me answer that that way rather than these specific questions.
DR. BORER: Okay. With that having been said, what we saw I think was that when you looked at time to first event, in fact, all the action was here, but when you looked beyond that to death or end-stage renal disease, bad things happen maybe because the disease progressed, which I think is what you're saying.
DR. LINDENFELD: Exactly.
DR. BORER: Bob?
DR. TEMPLE: It still seems strange and not fully explained although, for the reasons you just gave, not worrisome. When the first thing that happens is end-stage renal disease, somehow that's not influenced. When end-stage renal disease follows creatinine doubling, which in some sense it always must, then it turns out to be okay. There must be something in the analysis that led to that conclusion because it all doesn't make sense, but maybe it doesn't matter for the reasons you just gave. There's something odd about it.
DR. BORER: Statistical anomaly.
Before we go on to 2.5, does anybody on the committee have a different view of 1 through 4? Steve?
DR. NISSEN: I just would say if it's a statistical anomaly, if I'm not mistaken, the same thing was seen in the previous trial, the IDNT trial. So, there's obviously something biologically going on here, and I worry about these people who didn't have creatinine measured, whether somehow that's influencing it. How do you get to have end-stage renal disease before doubling your creatinine? Well, one of the ways to get there is nobody has your creatinine in hand, and so they don't know. Maybe that's what's going on here, Bob. That would be one of my guesses.
DR. TEMPLE: But you'd still expect it to be influenced. If there's an overall delay, there ought to be an overall delay on end-stage renal disease where you didn't get a creatinine just as much as where you did.
DR. NISSEN: I hear you.
DR. TEMPLE: There's something wrong with the analysis here that we haven't been smart enough to figure out. That's what I think.
DR. BORER: Tom, are you smart enough to figure it out?
DR. FLEMING: Well, I just want to take the burden off the statisticians and put it back on the clinicians here.
DR. FLEMING: I can figure it out statistically, but I don't know what the true clinical answer is.
Look at figure 379. You can't but just remember when you did look at figure 379, which is the figure that showed what was the time to end-stage renal disease distribution. Remember we were saying those curves overlapped in the first 18 months and then separated thereafter. So, statistically part of the reason that we are seeing the numbers of people who have end-stage renal disease as their first event not being different between arms isn't surprising when I look at this curve.
Now, I'm going to ask my clinical colleagues the harder question. Why is it, if there's an effect, that it doesn't show up at all for the first 18 months and then emerges thereafter?
DR. BORER: Yes. I will try to answer that even though I'm not a nephrologist.
DR. FLEMING: Before you answer it, because I really do want you to, one other aspect of this that does a bit to complicate things when we look at the triple endpoint is when you look at doubling first and then you're not necessarily following time to doubling in all people, quite frankly I have a lot of trouble understanding the triple endpoint when we have 267 people who aren't followed for one element for a significant period of time. So, I prefer to look at the double endpoint or the single endpoint here, ESRD, and in this endpoint the question I would ask you clinically is, is there an explanation that you have for why you don't prevent events in the first 18 months and then you do thereafter?
DR. BORER: The entry criteria for this study, as I understand them, was a creatinine of 1.5 to 3. Renal failure nominally is a creatinine of 6. So, nobody entered even close to renal failure and they had to at least double their creatinine to make it, and it takes some time to do that. So, I'm not at all bothered by the fact that there was a period of no effect when you look at the ESRD as the endpoint.
DR. TEMPLE: I guess one explanation is let's hypothesize that there's a delay before you accomplish much with this therapy. That means the end-stage renal disease endpoints that turned up as initial endpoints were people who turned up with that problem very early. The die had been cast, if you like, and wasn't as influenceable as later end-stage renal disease. That makes some sense. Maybe we can model that or something.
DR. NISSEN: But the data was to the contrary, though, because the higher your creatinine at the beginning, the more likely you were to see a benefit. If you look at the subgroups of creatinine less than 2 and greater than 2, almost all the benefit was in the people who started out greater than 2.
DR. BORER: Ray, would you like to weigh in here?
DR. BAIN: Yes. We did look at those 64 and 65 end-stage renal disease events that occurred without a doubling, and we looked at a number of factors to try to understand why this occurs. One of the things that was curious was you remember that the average baseline serum creatinine across all groups was 1.9. If you look at those 64 and 65 patients who went to dialysis without doubling, they had a higher serum creatinine. So, they were worse off coming into the trial and maybe had a different trajectory.
DR. BORER: Let's go on to 2.5. Is everybody else satisfied with this discussion so far?
Let's go on to 2.5. Subjects who experienced doubling of serum creatinine could later have end-stage renal disease or die. When these events are counted, the relative risk of death on losartan was 1.02 and the risk of needing dialysis .71. Are these data supportive of an effect on clinical outcome?
I think we just answered that one.
3. In RENAAL, the mean blood pressure was significantly lower in the losartan group than in the placebo group. How does one know that blood pressure alone was not responsible for losartan's treatment effects? And 3.2, is the mechanism of the treatment effect relevant to the description of trial outcomes?
DR. LINDENFELD: I'm not sure we absolutely know that this wasn't a blood pressure effect, but we saw the data corrected for differences in systolic blood pressure, pulse pressure, which made more of a difference than the mean blood pressure. Although the effect was somewhat less, the effect was still present. I feel relatively assured, based on this and other data comparing amlodipine that we've reviewed previously, that this wasn't all a blood pressure effect.
DR. BORER: JoAnn has added a new wrinkle to the discussion that we'll come back to later. Are there any other comments about that? Steve?
DR. NISSEN: Yes. I basically agree.
The difficulty I have is estimating the magnitude. Again, the reason it becomes relevant here is this issue of a relatively marginal statistical significance for the overall trial. When you've got a marginal result, then this analysis becomes very important. So, I think it's very difficult to estimate the effect of blood pressure here. They made their best guess at it. It was very reasonable and so on. But if it were a larger effect due to blood pressure, then that marginal p value becomes even more marginal. So, it becomes relevant in the context of a very subtle drug effect.
DR. BORER: Dr. Brem?
DR. BREM: Well, I'd like to come back to the dose issue again. The design of the study was to titrate the medication up to dose effect and then there seems to be a paradox that, once you've done that, there's no effect of blood pressure on renal preservation. I'm still having trouble grappling with the separation.
If the blood pressure isn't important, why didn't everybody get 100 milligrams and then you titrate the blood pressure so that you know 100 milligrams works or it doesn't work. Now, we have a third of the patients who had 50 milligrams. Again, in reference to a prior study that we reviewed in January, there was a dose dependence. So, I still come back to separating the direct effects on the kidney from those that are directly related to blood pressure.
DR. BORER: Yes, that's a toughy. For myself, I'd come back to the Asian data that we saw. The effect on outcome events was biggest and the effect on blood pressure was negligible, if any, which would be consistent.
Blase, was that your light on or Alan?
DR. CARABELLO: The blood pressure in the Asians was a little higher.
DR. BORER: But the treatment effect was less. The effect of losartan relative to placebo was the least on blood pressure.
Alan, were you going to say something?
DR. HIRSCH: I was just going to repeat my same point. I feel like I've gained some support here. Just that I think it's hard to know the blood pressure effect simply using adjustments within the study regressions. This committee has often looked at this issue and determined that in different populations one can predict different effects of a net blood pressure difference. I would simply say for the record I don't think we know.
DR. KEANE: And if I may just underscore the fact that we actually titrated or up-dosed the losartan to achieve a blood pressure effect. We weren't looking at the reverse of that. There were at least three-and-a-half additional agents that these patients were on at any given time interval. So, it makes it very difficult to tease out what actually is going on with the effect of the drug as you change the dose.
DR. BORER: Tom?
DR. FLEMING: Just maybe to draw another distinction here and follow up on what Alan is saying. There are at least two ways to look at adjustments for the blood pressure. As I see it, the particular differences are in the systolic blood pressure. Those are apparent at baseline as well as over time. The imbalances at baseline we can handle through a traditional covariate adjustment, and both that factor and proteinuria are imbalanced at baseline. As the analyses that have been presented to us show, the particular influential covariate there is proteinuria, much more so than the imbalance in baseline of systolic blood pressure.
But they started off, I think, at 153 versus 150, and over time they ended up at 142 or 143 against 140. So, the separation was maintained.
So, the other question to ask, beyond whether there was confounding because of baseline imbalance, is to what extent is the effect of treatment mediated through the changes or differences that are seen over time in blood pressure. Bob was talking earlier about the caution one has to have when you're using time varying covariates to essentially try to address the question, is the effect of a treatment on outcome entirely mediated through a marker, in this case blood pressure over time.
One classical approach statistically to do that is to use what's called a time varying covariate, not just to adjust for differences in that covariate at baseline, but differences in that covariate over time, so that at the time of any event, you take into account what the person's blood pressure was at that particular time.
The caution one has to have is that the way you interpret that, if there's no difference after adjusting for that covariate, isn't that treatment is not influencing the outcome, but the entire effect of treatment seems to be accounted for by the effect of treatment on that covariate, on that marker.
When this is done here, some of the effect is accounted for by the differences in blood pressure, although it looks like, from what I've seen, the majority of this effect is still there even when you adjust for the time varying covariate of blood pressure over time.
I agree with Alan's point about, however, as sophisticated as these analyses are, you've got to interpret them with great caution. For example, that analysis is basically assuming that if the time varying blood pressure is, in fact, influencing risk of outcome, it's specifically mediated through what that exact last measured value was. Maybe it isn't. Maybe it has more to do with what the value was over a period of 6 months before that time or a combination of that blood pressure and any other things, such as diastolic blood pressure or other factors.
So, it's statistically the best approach I can think of doing. It's what the sponsor did. It gives you a clue about whether or not the effect of treatment is at least partially mediated through these differences in blood pressure, but it's only a clue and one has to interpret it very cautiously.
DR. BORER: We've said a lot of things here and I'm going to try to summarize them just to make sure everybody agrees so that the FDA has a clear statement.
With regard to 3.1, how does one know that blood pressure alone was not responsible for losartan's treatment effects? I think the general consensus is we don't know, but we think it probably didn't because of all the things that people have said here.
And is the mechanism of the treatment effect relevant to the description of trial outcomes? Well, we don't really know that either, but we think that the outcome is not solely based on the effect on blood pressure, but we can't say that rigorously.
Does everybody accept that? Okay. FDA you've just heard it.
Comment on other secondary endpoint in RENAAL. There was a prespecified analysis of time to first cardiovascular death, nonfatal MI, hospitalization for CHF or unstable angina, stroke, or coronary or peripheral revascularization. There were 515 such events with no significant difference in the distribution between groups. Is this evidence of a lack of clinical benefit? Is it comforting that there was a lack of apparent harm? Were there simply too few events to show a meaningful effect?
JoAnn, why don't you go ahead on 4.1?
DR. LINDENFELD: I think that there are too few events here, too short a period of time, and a population that does not have as high a cardiovascular risk as we might expect. So, I would say that this is not necessarily lack of clinical benefit, it also makes it difficult to say that there's lack of harm. We see very little difference. I'm comforted that there's no trend here toward harm. I think we'd have to follow these patients longer to see evidence of clinical benefit. I'm not disturbed, from what I've heard, by the lack of clinical benefit. I think that we just would have to follow more patients for a more prolonged period of time.
DR. BORER: Is everybody in agreement with that? Are there any other comments? Tom?
DR. FLEMING: Just to maybe refine, I agree with JoAnn, just to add a bit to it in the spirit of what we also said three months ago in January.
It's very relevant to assess these cardiovascular events. Clearly, a major portion of what's happening to these patients clinically that's very important goes beyond the specific renal events into the cardiovascular arena.
If one looks at the aggregation of these six components, one sees an estimate of about a 9 percent decrease. If one had expected to have had sensitivity, I guess I would have to say to you clinically how big does the study have to be, and the answer to that depends on whether you would require a 30 percent relative reduction or a 10 percent relative reduction or a 5. If you actually required a 10 percent relative reduction to be clinically meaningful, which is consistent with what the point estimate is here, we would have needed to do a study of about 10,000 to 15,000 people. With 10,000 to 15,000 people, instead of 1,500 people, we would have been adequately powered to discern the difference between a 10 percent reduction and no reduction.
So, I would disagree with the sponsor who, at one point, said these data show no effect on these endpoints. I think these data suggest a 10 percent reduction, and it's not possible to discern whether or not that's reliable evidence of a reduction or consistent with random variability with no true reduction at all.
I also, though, break this down a little bit further to say, as we said three months ago, not all of these are equally important, and we wanted in particular three months ago to focus on cardiovascular events, MI, and stroke. When I look at those sub-elements, it's about a 5 percent relative reduction if we wanted high sensitivity. If you view a 5 percent relative reduction to be clinically relevant, you want to distinguish that from no reduction, it would take 50,000 people.
So, my overall sense here is these data are suggestive a reduction of 10 percent. That's consistent with none. At least we can rule out that there's not a 30 percent reduction. So, if you say that's what we need, then this study is conclusively ruling out the kind of benefit you would expect or you would want to see. I suspect that we wouldn't say there has to be a 30 percent reduction. On the other hand, we could rule out that there's harm by 10 percent.
Now, one of the comments that came up is look at cardiovascular death though. That's going in the wrong direction. That's 90 versus 79, 10 percent in the wrong direction. That is something to think about. That is, in fact -- and I will maybe discuss this more -- a signal that there might be an increase, but in that small subgroup, that's entirely consistent with worse than a 10 percent, maybe a 30 percent, reduction or entirely consistent with no reduction at all to a 10 percent benefit.
So, in the same sense that I would caution against looking at 50 against 68 nonfatal MIs, saying, ah, we've clearly prevented nonfatal MIs, I would caution against looking at cardiovascular deaths of 90 versus 79, saying we've done something bad there. When we start subdividing these into these sub-elements of the six, we really push the data beyond the limits of what they can reliably be telling us.
So, my overall view of this is an extension of what JoAnn has said. This study clearly should have, as it did, look at cardiovascular events. It's suggesting a 10 percent reduction, but it would have had to be 5- to 10-fold larger to be able to reliably discern whether that's truth or whether that's in fact random variability off of no difference at all.
DR. BORER: Is everybody in agreement with Tom? Are there any concerns? Paul?
DR. ARMSTRONG: Are we going to have a chance to talk about the potential for harm at some other point in the questioning sequence? I don't see it, and if you want to reserve it to the end, that's fine.
DR. BORER: We will in the very next question.
DR. ARMSTRONG: I'll keep my powder dry.
DR. BORER: Actually the one after 4.2. Once we get to 5, we get into the meat of the issue here. I'm sorry about that. We're talking about proteinuria, and I said the "meat" of the issue. Sorry.
DR. BORER: 4.2. Proteinuria, assessed as milligram per gram of creatinine was lower on losartan at all times after baseline. Additionally, the rate of loss of renal function, assessed by the slope of reciprocal of the serum creatinine over time, was significantly lower by about 13 percent in the losartan group. What did these results contribute to the confidence one has in the clinical benefits of losartan in RENAAL?
DR. LINDENFELD: These add confidence to the benefits we've seen. I think these are all consistent based on what we think the pathophysiology of this disease is, to see a reduction in proteinuria and a decrease in the slope of creatinine, and the fact that also that the decrease in proteinuria correlates with the outcome. So, I think these are all consistent and add somewhat to the confidence of the results.
DR. BORER: Are we all okay with that? Okay.
Number 5. I'm sorry, Paul, I misspoke. We'll get to the issue that you raised somewhere in here, but not quite with this one because the focus of number 5 is a little different.
Are the results of RENAAL alone an adequate basis for approval of losartan for the treatment of type II diabetic nephropathy?
A drug with a related mechanism of action, captopril -- oh, sorry. I'm moving on here.
Are the results alone an adequate basis for approval?
DR. FLEMING: Just for clarification, you were right when you told Paul that it's question 5. I assume in question 5 we now bring everything together, including safety.
DR. BORER: Yes, okay. Paul, did you want to say something about that?
DR. ARMSTRONG: Well, I wanted to reflect on the fact that in the safety presentation -- first of all, I'm convinced this drug is effective. I'm also convinced that it has the potential for harm and that in understanding that, a better appreciation of some of the issues around hyperkalemia in particular seems to me to be germane.
So, where I'm going with this, Mr. Chairman, is that the sponsor in the safety presentation suggested that there were no deaths attributable to adverse events. Yet, we saw data on two or three patients who died suddenly who had potassiums of 7 and 6 as the last measured biochemical indicators, which to me, while not establishing cause and effect, nonetheless we're not able to dismiss that the sudden deaths in those patients were not attributable to those potassiums.
Although we've heard from elegant and sophisticated consultants about the fact that hyperkalemia is common with these patients and therefore we can expect that 1 out of 4 or 1 out of 10 might have either hyperkalemia or potassiums greater than 6, many physicians treating these patients will not be as sophisticated as those in this room. So, the issue around how to balance that and to weigh it into a label and a caution is the issue I wanted to bring forward because I was uncomfortable that we have had adequate discussion around that. And I say that still with the notion that this is an effective drug.
DR. BORER: We ought to have a little bit of discussion about that. I think that we have to remember, in the context of that discussion, that in talking about risk, we have to consider risk in the context of its acceptability for the intended use, and the sponsor did give us an all-cause mortality, all-cause horror show, risk versus benefit relationship analysis. So, ultimately we're really talking about the relation of benefit to risk.
Having said that, this is an issue that we didn't really get into in any major way. Perhaps we ought to talk about that a little bit.
Bob, did you want to say something?
DR. TEMPLE: Yes. It may not be obvious from the question, but recall in January in a similar situation the issue was whether a single study with a p value in the neighborhood of .02 was sufficient evidence. So, this is intended to introduce that --
DR. BORER: Yes. We'll get to that.
DR. TEMPLE: I just wanted to be sure.
DR. BORER: It hit me over the head like a sledgehammer that that's where we were going here, but let's get the safety issue off the table first. Beverly?
DR. LORELL: Well, the discussion today has raised an issue that I hadn't thought about previously, and that deals with the time course of apparent benefit that Tom Fleming brought up where one must wait for about 18 months to see the curves begin to diverge.
We've also seen some evidence which is suggestive. Although the trial was not designed to look at this, there are some very elegant data suggesting that the likelihood of benefit is greatest in those patients who have a more severe degree of renal dysfunction, whether as estimated as a marker using serum creatinine or magnitude of proteinuria.
So, I think the risk of hyperkalemia for the general public is a weighty one in that the issue comes up, do we know whether or not we are exposing a population that is going to have much less relative benefit in terms of renal protection to a risk of hyperkalemia. I guess that here what I think we didn't quite ever see that I think Dr. Hirsch brought up is it would have been nice to see sort of the time course of time to hyperkalemia to understand whether or not the time course of exposure to risk of an event that is at minimum a huge hassle for the clinician and the patient and at worst life-threatening, follows a similar time course of benefit or whether there's a discrepancy there. I'm not sure if said that clearly.
DR. BORER: No, you did. It sounds as if we don't, at first glance, have the data to answer the question rigorously.
But before Dr. Kopp weighs in here, Paul specifically raised this issue in the context of a labeling discussion, and I would ask you whether you think that your concern here is one that rises to the level of approvability or is this something that really requires a big bold caution in the label because we just don't know and we're not sure exactly what subgroup really benefits? And obviously hyperkalemia kills. So, is this a labeling issue? Is this an approvability issue? What do you think?
DR. LORELL: I'm not sure about it. I think in looking at the numbers that I've written down during the day, there was a sizeable number of patients who had hyperkalemia of a magnitude that would, at minimum, warrant getting the patient to the emergency room to give an anti-hyperkalemic therapy. It looks like the development of even a single hyperkalemic 6.0 or higher event was somewhat predictive of death. So, I guess I would welcome discussion among others on the panel on this.
DR. BORER: Dr. Kopp?
DR. KOPP: My comment was actually not so much about hyperkalemia. I don't know if it's going to be off target, but it relates to this issue of clinical benefit and who benefits the most. It's true if the outcome is not seeing dialysis, not going on dialysis during the course of the study, the greatest benefit was in those who had the highest levels of creatinine. But if we understand this drug as slowing the rate of decline of renal function, then you can argue that the person who starts it earlier with a lower creatinine will enjoy more days of not having dialysis because their rate has slowed. So, they might enjoy 1 or 2 years of being dialysis-free, whereas the patient who begins it late with a creatinine of 3 or 4 has fewer months or years of benefit.
DR. BORER: Bob?
DR. TEMPLE: Is it a fair presumption that the people most likely -- maybe it's known from the data -- to become hyperkalemic are the people who were more impaired? I ask that because these drugs are already approved for the treatment of hypertension, irrespective of whether they prevent renal dysfunction, and they, ACE inhibitors, and beta blockers all increase the rate of hyperkalemia. We know that. It's in the label and there are warnings about it.
So, are we looking at something that's worse than usual, and if it is worse than usual, is it in the people who you think get the most benefit or is it in everybody, including the people who don't seem to benefit very much; that is, the people who aren't as impaired?
The company may have data on this.
DR. BORER: Yes. Do we have an answer to that question, Dr. Keane?
DR. KEANE: Yes. I was just going to show, first of all, one of the slides to show the overall magnitude of the potassium levels throughout the entire trial to reemphasize the fact that potassium is in the range, throughout most of the trial, in the upper 4 milliequivalents so that we're really not seeing a huge degree of hyperkalemia in this patient population overall.
DR. ARMSTRONG: Mr. Chairman, 1 out of 4 patients were hyperkalemic with losartan. So, when you say you were not seeing a lot of patients, for me 1 out of 4 is a substantial number. I'm sorry.
DR. NISSEN: This slide is not relevant here.
DR. BORER: Yes. The average values may not speak to the point.
DR. NISSEN: They don't speak at all to it.
DR. KEANE: This is the percent of patients with an event described as hyperkalemia in the losartan group over time compared to the placebo group.
DR. HIRSCH: So that is a fourth. I think our concern is not the mean but obviously those patients who lie near the limits would be at risk.
DR. TEMPLE: Is that a percent or a fraction? What is on the y axis?
DR. KEANE: That's percent of patients with an event.
DR. BORER: So, it's .2 percent.
DR. TEMPLE: Is it 3 percent, not 30 percent?
DR. KEANE: 25 percent.
DR. TEMPLE: So, it isn't the percent. It's the fraction.
DR. BORER: Beverly?
DR. LORELL: I'm not sure I know what to do with this. But I think it's of interest in thinking about this because the time course of accumulating severe hyperkalemic events starts early and progresses. So, we have, in terms of thinking about risk benefit, a period that goes for many months where there's no difference between placebo and intervention up against a period of enhanced risk. I think all would agree, relative to your question, Bob, that this is a different level of risk of hyperkalemia than is seen in the hypertension trials. In fact, the hypertension trials would have excluded many of the patients, not all but many of the patients, in this study.
DR. BORER: In all fairness, though, and without wanting to advocate one position or another, when you look at the mortality curves, they're spot on too during that period. So, it's a risk, but people weren't dropping like flies because they incurred that risk. Obviously, whoever was managing them saw the problem and dealt with it adequately. Again, I think we have to consider -- maybe we just have to keep it in mind for subsequent points on this list of questions -- that this is an open issue. Should this be an approvability issue or should it be a labeling issue?
Let's get to the intent of question number 5. This is a single study.
DR. KEANE: Can I just add so that we're clear about what the potassium values were? I think as you look at the distribution of K levels between 5.5 and up to 7, the distribution that is occurring within the vast majority of these is weighted at the lower levels, i.e., 5.5 to 6 and 6 to 6.4, towards losartan. But as you get into the higher potassium levels that we were able to determine, they're fairly comparable both in the losartan group and in the placebo group. So, I think that should be taken into consideration when one talks about what the potassium values are in these patients that are going to have hyperkalemia.
DR. BORER: Okay, thank you.
Well, the intent of number 5 really is whether a single trial is adequate as a basis for approval. We have this study alone. We saw a study in a related compound or a pair of studies in a related compound with at least some similar pharmacologic effects three months ago. We've heard about data, and this committee when it was differently constructed, considered other data that might be related having to do with a compound that has some effects on the same system, the renin-angiotensin system.
So, the question is do we have enough information from this one trial alone, and if we don't, is there sufficient information available from other sources that we have enough information in aggregate to draw a conclusion about approvability.
With that in mind, we were all sent --
DR. THROCKMORTON: Jeff, I think I'd like to break those up just a bit. Let's stay focused on are the results from RENAAL alone an adequate basis without taking into account any other data that you might want to. You'll have chances in later questions to do that.
DR. BORER: Okay. Well, then let's just do that since that's what the FDA wishes us to do.
Are the results of RENAAL alone an adequate basis for approval of losartan for the treatment of type II diabetic nephropathy? JoAnn?
DR. LINDENFELD: RENAAL, taken in isolation, I don't think would make this an approvable drug. I think the p value was not what we would usually consider for one multicenter trial. We saw that as the primary time to first endpoint, this was primarily a doubling of creatinine. My answer to this would be no.
DR. BORER: Does anybody else have any comment about that? Does anyone disagree with it? Steve?
DR. NISSEN: I don't disagree, but I want to make sure I say why I don't think it's adequate.
First of all, it is a single trial. That alone is not a bar if the single trial is very robust, but it's a single trial with a fairly marginal p value.
Tom, in all deference to your comments, I have to use the triple endpoint here. That was the prespecified endpoint. So, I focus a lot of my thinking on you get one chance up front to choose your endpoint and this is what they chose. Right or wrong, it's pretty marginal.
Then I've got all these confounders to deal with. I've got the problem that if you take out the 250 patients from Asia from either the triple or the double endpoint, it no longer is statistically significant. Now, that does not meet my standard for a robust effect when it seems so clear that it's being driven by a very small population. Similarly, there are other confounders here, like the blood pressure difference.
So, when you take a marginal p value for the principal endpoint of the trial and then you erode it with these other factors, then it doesn't rise to that level of evidence that we've typically required for approval. So, alone RENAAL doesn't meet the standard from my perspective.
DR. BORER: Does anybody else want to state an opinion about this? Beverly?
DR. LORELL: I would like to. I spent some time actually reviewing the criteria that the FDA itself has put out for a single trial because we wrestled with this issue very recently.
I think that there was very compelling data presented both from this study and in aggregate for many studies supporting a role of the renin-angiotensin system in accelerating progression of diabetic renal disease. And I doubt any around the table would disagree with that biology.
But if one looks at the guidelines from the FDA itself to both industry and those of us as reviewers for evidence of effectiveness from a single study, there are several concerns, and I agree with JoAnn.
First, with regard to the endpoints that were presented to us in trial design, there is not evidence of a highly statistically persuasive outcome. The statistical outcome for the predefined primary endpoint was marginal.
Secondly, we're advised to look at consistency across subgroups. Here I think the issue of the gnarly problem of the data from Asia is very problematic and really counts as a point of concern about consistency.
Third, very explicitly, we are advised to look in a single trial as to whether or not there is evidence of efficacy in multiple endpoints involving different events. The example that's given in the publication from the FDA is quite helpful here. In this trial, the different events that we are given to look at are end-stage renal disease and all-cause mortality. I think one could argue very strongly that doubling of creatinine and end-stage renal disease are highly coupled, and we know that because the median time from going on to end-stage renal disease after doubling of creatinine is only 30 days. So, if we look at the criteria of different endpoints regarding clearly different events, this trial does not quite make that criteria. There was not an effect on all-cause mortality, nor was there an effect on a very important predefined secondary endpoint of major cardiovascular events.
Finally, I am influenced by the concerns that the Steering Committee of this trial itself had regarding the use of this drug in the context of albeit insufficient data regarding ACE inhibitors in cardioprotection.
So, I think with the criteria that the FDA itself provides to its advisors for thinking about efficacy in a single study, it doesn't quite reach the burden.
DR. BORER: I'm going to ask in a minute whether everybody agrees with this, which I think we do. But I want to just ask for a little clarification about the last statement. What I heard -- and we have the person here who can tell us about it -- was not that the Steering Committee was concerned about people not taking ACE inhibitors, but they were concerned about half the patients not taking any drug that impacted on the renin-angiotensin system. I think we don't want to over-interpret what they did based on something they didn't mean. So, can we just have a clarification about that?
DR. BRENNER: My name is Barry Brenner. I had the pleasure as serving as the PI and the chairman of the Steering Committee.
You're correct. The late January '01 meeting of the DSMB issued a directive based on their review not only of HOPE but of another publication that was soon to be printed, the Mann paper, which was a substudy of HOPE dealing with patients who had renal disease. In that study there was cardiovascular protection with ramipril. That was the first time that a population akin to the RENAAL trial showed a benefit in a clinical trial with interruption of the renin-angiotensin system. That was late January.
Our committee convened on 10 February, very soon after learning of the concern of the DSMB. And we were blinded, although they were not of course. And we voted unanimously to terminate the trial a year before it was scheduled to be terminated because our concern was that in the placebo group there was no renin-angiotensin system blockade.
The decision was based on increasing evidence from HOPE to some extent but much more compellingly from the Mann paper, not yet published, but which we had in manuscript form, that in patients with renal disease ACE inhibitors may be effective in reducing cardiovascular events, as I say, in patients who had cardiovascular risk factors. So, it was because the placebo group had not received and could not in the trial receive blockade of the renin-angiotensin system.
Recall that when the trial began, there was virtually no evidence about protective effect on the heart, cardiovascular system, with ACE inhibitors. And the first paper that was compelling in this regard in diabetics was the substudy of HOPE, and that was in the year 2000, five years after we began. So, once we had now additional evidence, obviously only coming from HOPE -- but that's all there was -- dealing with renal disease and showing a cardiovascular protective effect, that drove us as an independent body to terminate the trial.
DR. BORER: Thank you. I've just learned that the FDA would like a formal vote on this question.
DR. FLEMING: Could we have further comment first?
DR. BORER: We sure can.
DR. FLEMING: I'd like to just add a little bit to this. This certainly to me is the most critical question on the agenda.
As my colleagues have pointed out, great emphasis should be given to the primary endpoint of the trial. There certainly is evidence of benefit here. The p value was .02 on that primary endpoint. When we saw a similar phenomenon or a similar strength of evidence in January, some of us judged that as consistent with one positive trial but not two positive trials.
Going beyond that, as has been pointed out, there are some safety issues here, most discernably or notably the hyperkalemia. There are some concerns about subgroups. As the FDA has asked us to look at consistency across subgroups, we see within the Asian population an apparent dominance in terms of where much of the positive signal is. And as has been pointed out, there are some issues to address relating to the ACE inhibitors and their efficacy and how does that complicate this. These are all areas of concern.
Having acknowledged those areas of concern, though, I think there are a number of really critical issues to consider. I'd like to turn in particular to the specific criteria the FDA has asked us to consider about whether a single study is adequate.
We, in fact, have a large multicenter trial. That's the easy part.
Going beyond that, though, of real critical importance are the criteria, are there multiple endpoints that essentially reinforce our sense of strength of evidence, and are these results statistically very persuasive findings. My own sense about that, even though it's subjective, is something I've always attributed to Ray Lipicky although he denies it's his, but I always think of .025 squared times 2 as a two-sided significance level that might be viewed as what you need to see in order to be consistent with the strength of evidence from two adequate and well-controlled trials. That's a p value of .00125.
When I look at these data and probe further, I acknowledge that -- and I'm very reluctant to move away from a prespecified primary endpoint except in those very settings where I have a philosophical strong concern about the appropriateness of the choice of that endpoint. To my way of thinking, what we're looking at here is a combination of loss of renal function and death, and we're trying to best characterize that. I'm at a loss for knowing why we have to use the component doubling in creatinine time when it doesn't take that long to follow people, if it really is going to translate into end-stage renal disease, to end-stage renal disease. Hence, as some of us at least argued last January, we were questioning the persuasiveness of the triple endpoint.
If this study had had a triple endpoint of four 0's and a 1, and the double endpoint of end-stage renal disease/death had been a .03 to .06, I would have called that at best a study that translates into the strength of evidence of a single positive study. I.e., I would have discounted the surrogate.
As a result, I think it's logically inconsistent for me, even though I'm a strong believer in adhering to the primary endpoint, that if I would not have given credence to that primary endpoint, I should focus on what it is that really matters, particularly in a setting such as this where I can't interpret the primary endpoint because they didn't follow 267 people to the triple endpoint. They did, to their credit, fortunately from my perspective, followed everybody to what really to me matters from a renal perspective, end-stage renal disease/death. When you look at that endpoint, one gets significance levels on the order of .002. That's for end-stage renal disease; .009 for end-stage renal disease/death.
That's not, by the way, what we saw in January. When we looked away from the triple endpoint into these elements last January, we were seeing less strength of evidence, which for some of us was the concern. Here we're seeing much more strength of evidence.
Furthermore, we asked, looking at a different domain, last January please don't just show us the clinically important renal endpoints. Also show us the most important, clinically important cardiovascular endpoints, which before I ever looked at these data, we had specified as end-stage renal disease, MI, stroke, and death.
When you look at those specific endpoints for stroke and death -- granted, cardiovascular deaths are notable as having 11 in excess in the wrong direction -- strokes and non-cardiovascular deaths are 11 in the right direction. So, overall, there's no difference. I'm not claiming as a result that I know that there's any benefit here on anything related specifically to mortality, but over and above those measures are the 47 excess deaths prevented on end-stage renal disease and the 18 events prevented on MI, which is a total of 65 events, which statistically is at the .003 level.
So, now I'm looking at what in January we called the clinically most important events which, by the way, are the ones we can interpret because that's where they had complete follow-up. What we see are significance levels on the order of .002, .009, .003 before there's any adjustment for an issue of baseline imbalances, which is another area that causes many of us to be greatly cautious. Nevertheless, I've been persuaded that there's something here of real relevance, and when adjusts for imbalances in proteinuria, there's another log reduction in these significance levels.
Now, is there persuasive data here when you start subdividing into two groups? Well, let's subdivide the Asian into the non-Asian populations. Now, that's in a certain sense not particularly optimal for this intervention because we did so specifically having seen the greatest signal coming from the Asian population.
But when we do so, if you go with me for the moment to what's the most important endpoint, which I don't believe is the triple endpoint -- it's the double endpoint -- when you look at the double endpoint and you adjust for proteinuria, it must be significant. Clearly the Asian subgroup is significant even without adjusting for proteinuria. The sponsor showed us the double endpoint in the non-Asian patients had a lower confidence interval. It was very close to 0 even before adjusting for proteinuria.
So, if you focus on what really matters, end-stage renal disease and death, and account for what is a pretty strong case for an imbalance at baseline adjusting for that, even if you simultaneously also adjust for the systolic blood pressure differences, you'll find subgroups in the Asians and the non-Asians that I strongly suspect and can ask the FDA to validate will, in fact, be significant in both of those groups.
So, as I look through these criteria, I am seeing statistically persuasive findings that are on the level of .00125, and we are seeing, across the endpoints that really matter, the renal endpoints alone or the renal and cardiovascular endpoints that last January we designated as the most important, there's a significant difference there as well.
What about the consistency across subgroups? That is certainly a relevant concern. There are some uncertainties, but the analysis that I was just giving of the Asian/non-Asian on the dual endpoint I think does show that there's benefit across both groups.
What about hyperkalemia? It's harder for me to assess that. We're seeing, over 3 to 4 years, a difference in 10 to 12 percent against 25 percent. But unless you would believe that that would translate into influencing negatively these hard clinical endpoints long term and, if anything, if we're going to project out long term, I'm thinking we have every reason to think the effects here are going to be even greater. These curves are diverging. I hate to put emphasis on what we haven't seen, but if we're going to extrapolate what the adverse effects of hyperkalemia may be that we haven't yet seen, and if you're going to take the liberty to do that extrapolation, you should take the liberty to extrapolate what these effects are. They're growing over time. The separation between these curves in ESRD-free survival are actually growing over time.
The last issue is the issue of how to address the ACE inhibitor. At least my understanding is if we assumed that we knew the ACE inhibitor was effective and we have a placebo-controlled trial and if in fact you look at these data and say that this is establishing that losartan is effective, it seems to me the question isn't whether you approve losartan. The question is can we motivate the conduct of a trial following the approval of losartan that would allow us to establish whether we should be using an ACE inhibitor or an ARB or the combination thereof. That seems to me the relevant follow-up question if there is in fact some uncertainty.
So, when I look at these FDA criteria, I find very strong evidence to suggest that what we are looking at as those factors that need to be met for this study to be viewed as adequate for approval are in fact met.
DR. BORER: Since we've been asked for a vote, let me translate that into simple terms. Is RENAAL alone an adequate basis for approval of losartan? Tom?
DR. FLEMING: Do you want my vote?
DR. BORER: Yes.
DR. FLEMING: Yes.
DR. BORER: Let's start at the other end of the table because we haven't given them too much of a chance, and then we'll come back around. Yes or no and give a reason.
DR. BREM: I guess yes, after that compelling discussion.
DR. BREM: I was going to say it's equivocal, but I've been certainly swayed by your arguments. So, I vote yes.
DR. BORER: Dr. Kopp?
DR. KOPP: Yes, I have a problem here too that I was of the belief that the great sin was to do other than what was laid out initially, and I think coming from a statistician, what you have said has been swaying me. I actually would have preferred to vote last and see what other people were saying.
DR. KOPP: But since I'm going at this point, I would have to say yes. And I guess the reason is because I am swayed that ESRD is important, that there was a significant difference.
DR. BORER: Bob?
DR. TEMPLE: Well, just to be sure. I'm certainly not trying to influence the answer. But alternative analyses are always persuasive to the people who like them. It's just worth remembering. I think Tom is very sensible, but he's presenting you an analysis that he finds persuasive for reasons that he gave. But subgroup analyses and alternative analyses and covariate analyses that we see all the time are always persuasive. Just don't forget that.
DR. BORER: Paul, are you persuaded?
DR. ARMSTRONG: I'm not allowed to vote with the knowledge of the field and taking other things into account. It's only on this data. And if I knew nothing else about the field, I would be optimistic but reserved and would probably vote no.
DR. BORER: Do you want to state for the record a couple of reasons?
DR. ARMSTRONG: Just the strength of the analysis and the fact that it's a positive trial, and I think it's effective. But it's one trial and the evidence isn't of the usual standard. But clearly we haven't heard the end of this discussion today.
DR. BORER: Beverly, I think you already gave all your reasons, but if you have any more to add, go ahead and give your vote.
DR. LORELL: I would still vote no. I think Tom Fleming's discussion is very persuasive, but nonetheless, it's not our job to redesign the trial and the primary and secondary endpoints that were predefined. So, using the predefined endpoints of this trial, I would vote no.
DR. BORER: Susanna?
DR. CUNNINGHAM: Well, I've been challenged all day as I've listened to all the discussion and been going back and forth and thinking about what does this mean to the person who has renal disease and what does this mean to the person who is going to be developing renal failure. I think after listening to Tom's very persuasive presentation and thinking about what it would mean perhaps to have 6 months less of renal failure, I would say yes in this case.
DR. BORER: Steve?
DR. NISSEN: I think I've already stated my reasons, and actually, Tom, although you are very persuasive, I was not persuaded.
DR. NISSEN: A lot of what I know you taught me, and I still think to go beyond the primary prespecified endpoint of the trial, you've got to have very compelling reasons to look at what you're looking at.
I do also agree with what Bob Temple said, which is I can take data and I can contort it almost any way I want and come to a conclusion.
But I look at the primary efficacy endpoint, and I really believe that it was marginal and that it was really troubling to see that it all came basically from one subgroup. I just don't think that's the compelling evidence that a single trial needs to have. And the word is "compelling," and it's just clearly to me not there from this single trial. And I reserve the right to vote otherwise based upon other information today, but on this trial alone, I don't think it comes even close.
DR. BORER: I would vote no also. I think that these are very suggestive data, that it's a positive trial, but I'm concerned about the lack of robustness as judged by the different results across different groups. Things tend to go the same way, but the magnitude is highly variable. There are a lot of confounders that I can't interpret fully. There are inadequate data to draw firm conclusions about cardiovascular events, although that is not a show stopper, and there are some concerns about safety that also aren't show stoppers but I would like to see some confirmatory evidence before I would vote yes. So, I will vote no for this trial as a single trial to be sufficient for approvability.
DR. HIRSCH: The last three speakers have mostly summarized most of my reasons for voting no.
But having the microphone on, this really was a visionary trial design when it was created. It's landmark in its ability to show a definitive, clear change in ESRD in this particular population studied.
But I really do believe it's important for us to stick to the true trial, high significance design with no homogeneities that are evident with robust, overlapping positive outcomes, especially -- and I'm bring to the table another issue -- in an era when I think all of us are seeing an effort for us to reach to molecules and to study them in a global design, bringing in multiple countries in a way that ultimately that, besides our USA mandate, will be used for global marketing.
When we have a single trial of what I think is sort of borderline significance, I think the standard has to be very high, and I'd like to caution all of us not to lean on single trials, ignoring everything else that we might know about ARBs and ACE inhibitors, before we set a new de facto standard. These standards should have the very highest level of knowledge. We're not there with a single trial.
DR. BORER: Blase?
DR. CARABELLO: Yes, I agree. I think that we've made mistakes based on single small trials. If this were 30,000 patients, that might be one thing, but I do agree that single trials are dangerous. But my major reason for voting no here is I can't get away from the fact that almost the entire benefit was concentrated in a tiny subsection of the study without even a trend in the other subpopulations.
DR. BORER: JoAnn, as our committee reviewer, you have the last word.
DR. LINDENFELD: I would say again no for the reasons that have been stated. I think this study by itself is not quite convincing enough. Also, by itself I don't think it gives us as much assurance that this isn't just a blood pressure effect.
DR. BORER: So that means we can go on to the subsequent questions.
A drug with a related mechanism of action, captopril, has an indication for diabetic nephropathy in patients with type I diabetes. The primary basis of that approval was the demonstration in a 409-subject, 2-year study, of 51 percent reduction (p equals .004) in risk of doubling of serum creatinine alone, and a 50 percent reduction (p equals .006) in risk of mortality or end-stage renal disease. Both effects were manifest in the first few months of treatment. Captopril also reduces the progression of microalbuminuria to overt proteinuria.
Are the results with captopril germane to a discussion of losartan? In particular, 6.1, is nephropathy in type I diabetes enough like type II diabetes? 6.2, are the pharmacological effects of captopril and losartan adequately similar?
DR. LINDENFELD: We discussed this in January with the same question, and I think at that time we felt that the nephropathies in these two were very similar. The patients are more different perhaps than the nephropathies in some of their other covariates.
And the pharmacologic effects, while not the same, are similar.
DR. BORER: So, then you think the results of captopril are germane to the discussion.
DR. LINDENFELD: Yes. Sorry. I do.
DR. BORER: Does anybody disagree with that? Steve?
DR. NISSEN: Yes, I guess I don't agree. While the drugs affect the same enzyme system, they attack it at different points. Gosh, I could give you innumerable examples of drugs. Even within a class, it's tough to know whether an effect is held across the class. There are some very good examples of drugs where one drug in the class actually is effective and the other drug isn't.
Now we're asked to look across two classes of drugs. The minute you broaden it to that point, I think it's a terribly slippery slope. I've had a lot of time to think about this since January, and I feel more strongly than ever that you can't extrapolate ACE inhibitor data to ARBs and you can extrapolate ARB data to ACE inhibitors. And you really shouldn't try. Even in the ACE inhibitors, there are issues of tissue selectivity, et cetera. We don't know if they're important or not, Bob, but the point is that's even within an individual class, let alone across two classes. So, I really do think this is dangerous.
DR. BORER: Paul?
DR. ARMSTRONG: In the same way that I'm unable to transfer the effects of ACE inhibitors to ARBs in cardiovascular disease, which I know more about than this subject, I would not be prepared to do it with the subject matter at hand.
DR. BORER: Dr. Brem?
DR. BREM: I would say the question before us really, is there some similarity and can you draw from that similarity -- I don't pretend to say that they're equivalent and I don't thinks that's the question before us. We're just asked is there some relevance from one to the other, and I think the answer to that is probably yes, there is. They do have some potential common modes of action.
And diabetic pathology, the actual structural changes and so on are quite similar from one to another. Many of the pathologic mechanisms involving glycosylated end-stage products and proteins and so forth are quite similar.
So, I think there is a similarity, but it's not equivalence. I think we should recognize that.
DR. BORER: That's with regard to type I and type II diabetes.
DR. BREM: And the captopril and losartan, that there are similarities that you might be able to draw upon, but they're not equivalent.
DR. BORER: I'd like to comment on that too. I agree with those who would not accept the captopril data as particularly germane. That is, I agree with Dr. Brem that there is something there. We are affecting the same system, and therefore at the very least, the effects of captopril would, for me, support the generation of a hypothesis that ARBs work in patients with diabetes to protect the kidneys. But beyond that, I'd have a hard time going because there are, I think, potentially important differences in pharmacologic effects between the two classes.
I agree absolutely with Steve. Among the molecules within a class, I think we've seen many examples of different pharmacologic effects besides the primary ones that we think about and therefore I would want to see more information than merely the captopril information to support losartan effectiveness.
On the other hand, I say again that I agree with Dr. Brem, that there's something here that supports at least the hypothesis that ARBs work, and then we have to study them.
DR. HIRSCH: One more extrapolation of that. Obviously, there are relationships between these drug classes that let us form hypotheses to lead to new clinical trials and there are similar effects. Let's ask the question. How would we all know on this panel or how would we know designing if they were similar or not? What it would take would be comparative trial data in humans. So far, what we're doing on this panel I think is taking monotherapy versus placebo in trials from different populations and different investigators and then trying to make comparisons. That's not how we'll decide if they're similar or not.
DR. BORER: Beverly?
DR. LORELL: I agree, Jeff, with your summary on this issue.
DR. BORER: Tom?
DR. FLEMING: I largely agree as well, Jeff. The FDA in a number of instances in the past, as I understand, has followed a strategy that I think makes a lot of sense. If they're working with the sponsor in setting up a development program and recognizing that we need confirmatory studies that a strategy has been at times to have two studies of the agent done in pathophysiologically related settings and if the studies are both positive, they reinforce each other and we obtain the confirmatory evidence in that manner.
If one level of germaneness would be is this in fact a study of that nature that could reinforce the RENAAL study -- and my sense is not at all -- in fact, if we believed that the captopril study provided evidence that was that relevant, I would wonder why this committee or the FDA in general hasn't approved captopril in type II diabetes, step one.
Step two is it's a different class, as has already been mentioned by my colleagues, and my clinical colleagues can address this much better than I.
So, my own sense is at a certain level it is germane, but I would consider it a pretty modest level because it is in fact a related, but still different, setting with a different class.
DR. TEMPLE: Jeffrey?
DR. BORER: Bob.
DR. TEMPLE: Just to be clear, we anticipated that people might find the single study standing on its own not a million miles away from sufficient but fairly close. This was an invitation to think about whether something in a closely related pharmacologic class tipped you over.
DR. BORER: And you've heard that it wasn't so far.
DR. TEMPLE: Yes. This was to offer you that opportunity.
DR. BORER: On the remaining four questions, we need to have a formal vote with reasons.
First, if the results with captopril are relevant to losartan, are the results on -- well, we've actually said what we needed to say. Do you want a formal vote on 7 now?
DR. THROCKMORTON: 7 has been answered to our satisfaction. If you would just go to 8, that would be terrific.
DR. BORER: Let's go to 8. Are the results of RENAAL and prior expectations derived from the captopril database an adequate basis for approval of losartan for the treatment of -- I think we've answered that one as well, but we can give a formal vote, if you like.
DR. THROCKMORTON: Yes, if you wouldn't mind.
DR. BORER: Okay. Why don't again we start at the far end of the table here. Dr. Brem?
DR. BREM: Sticking my neck out first, yes, I would vote for approval. My thinking is that while both classes are clearly different, I absolutely agree with that, and there are perhaps subtle mechanisms of action which may be different, and when we argue that type II diabetes and type I diabetes aren't exactly the same either, there's enough supporting evidence, in terms of direction and efficacy, that I would use those two together to make a story that it's worth approval.
DR. BORER: Dr. Kopp?
DR. KOPP: Well, I earlier said that RENAAL alone was sufficient, so obviously I have to say that A plus B are sufficient. I don't actually find that the ACE inhibitor data adds much.
DR. BORER: Paul?
DR. ARMSTRONG: No.
DR. BORER: Beverly?
DR. LORELL: No.
DR. BORER: Susanna?
DR. CUNNINGHAM: Yes and no.
DR. BORER: We need a comment there.
DR. CUNNINGHAM: I already said yes for the first, so that's my statement. But I don't believe that captopril is the same thing. I believe it has a different action, so I can't say yes to the whole thing.
DR. BORER: Steve?
DR. NISSEN: No.
DR. BORER: And I vote no.
DR. HIRSCH: I'd like to keep some suspense going here for a while.
The previous question was more theoretical, and I gave a very vigorous theoretical answer. When we come to practical gestalt sense of things, I am swung a bit, although I will tell you I haven't quite swung over, so I'm still a close no.
DR. BORER: Blase?
DR. CARABELLO: No.
DR. BORER: JoAnn?
DR. LINDENFELD: No.
DR. BORER: Tom?
DR. FLEMING: I've already said yes based on RENAAL. I would agree.
DR. BORER: The record should show that Tom agreed to yes/no.
Okay, we're up to number 9. Let's go through number 9 and then I want to ask a question. Number 9. In considering the approval of irbesartan for diabetic nephropathy, the advisory committee expressed interest in the program for losartan, which we did. The respective sponsors now have reciprocal agreements allowing reference to IDNT and RENAAL in support of one another's programs.
Do the findings of IDNT support the effectiveness of losartan for diabetic nephropathy? Why don't we start with that and then we'll go to 9.2 separately. JoAnn, do you want to start with that?
DR. LINDENFELD: I believe the IDNT results do support RENAAL, and I think a doubling of creatinine was the primary finding there. But doubling of creatinine in my own mind -- and we discussed this last time -- is more than just a surrogate. Doubling of creatinine is a real result that I think we can see in the clinic that applies to patients. Yes.
DR. BORER: Tom?
DR. THROCKMORTON: Jeff?
DR. BORER: I'm sorry. Yes.
DR. THROCKMORTON: I wondered if we could maybe get just a little more discussion around this question. Maybe it wasn't worded quite as well as we could have. Part of this was what's your level of comfort about thinking of two drugs from pharmacologically related classes as supporting efficacy. You can imagine some people might be uncomfortable doing that in the same sense that we heard some uncomfort about the ACE inhibitor data informing your decisions about an ARB. Maybe that's a foregone. Maybe that's easy. There's no question that it's informative and could be supportive, but if there is anything short of that, it would be interesting to hear some thoughts along those lines.
DR. BORER: Okay. I thought you wanted that in the context of 9.2, but that's fine. Bob?
DR. TEMPLE: Whenever. This is a slightly radical concept for us. That's why we sent you the evidence document which discussed it. Give me a minute to just say something about that document.
One of the things it did was describe all the circumstances -- it was our attempt to describe all the circumstances in which we would accept a single study of a particular drug and a particular use as sufficient evidence. You already referred to one. If it's really strong and the p value is out to the end of your arm and there's internal consistency, fine, we've all done that, and that's fine.
We also defined a whole bunch of situations in which one would draw further support from other controlled trials. Now, you can say that's a Bayesian prior or you can say you really have another study and it doesn't really matter. We give a whole list, studies of different doses, other regimens. There is one thing toward the end of that list that talks about pharmacologic and pathophysiologic endpoints.
Although the words aren't as clear about the present situation as one might have liked, because we didn't really think of it, there is a suggestion that when knowledge of the pharmacology, coupled with clinical evidence that that pharmacology is relevant, is present, that you might be able to rely on a single study, not no studies. We're not talking about class labeling or anything like that.
The main difference between the past cases that we considered and this one is what we had in mind when we wrote it is, oh, you've got 12 studies of ACE inhibitors. Maybe now one more might do.
In this case, there isn't any approval of any of them. So, there's a certain simultaneity that we hadn't really come to grips with. You're thinking about one study is supporting a drug, coupled with another study supporting another drug, and vice versa, sort of a crossover, where there isn't any established track record, approved drug. But we thought that was close enough to the situation we had contemplated to invite you to think about it, and we're thinking about it internally too.
It's something of a novelty, although as we've pointed out when we looked at it, we've been approving new heart failure claims for ACE inhibitors based on single studies with p values between .05 and .01 right along. We have not identified that as being based on the prior experience, but there's no other good explanation. We have not insisted on p values at .00125 or anything like that. So, we've been engaging in this without making it particularly explicit. So, this time we're making it explicit and giving you hard work.
DR. BORER: Well, we're explicitizing here. I guess that given that preamble, JoAnn, you may want to discuss 9.1 and 9.2 together.
DR. LINDENFELD: Yes. Again, I think the findings of IDNT do support the effectiveness of losartan. The captopril helps just a tiny little bit I think for several reasons. We've seen a pathophysiology here and a biologic plausibility that's consistent. We've seen a doubling of creatinine in both studies that's relatively similar, although we didn't see the end-stage renal disease in IDNT I think we've seen in this, and for all the reasons Tom said, I think we see a consistency of overall benefit here and no specific differences in the drugs that I can see that make me concerned that we're dealing with some sort of different mechanism among the single class of drugs. So, for all these reasons, I do think that IDNT supports RENAAL.
DR. BORER: So, going to 9.2.1, are the findings of IDNT as persuasive for losartan as would be a replication of RENAAL?
DR. LINDENFELD: No. I think a replication of the same study with the same drug would be more persuasive, but a little bit more. I can't put a number on that.
Let me just get the next one. A second study demonstrating losartan and slows progression I don't think would be as persuasive as IDNT. We've said that proteinuria alone we're not willing to consider as an endpoint in itself. So, 9.2.3, a study demonstrating progression to microalbuminuria would not be as strong.
And beating an active control arm in RENAAL, I think I'd have to understand exactly what you meant by that.
DR. THROCKMORTON: That was a reference to the IDNT trial where, if you remember, there was an active control arm.
DR. LINDENFELD: Yes. I think for all the reasons we said there, because it's hard to tell if an active control arm might cause some problems, that that would not be as strong either.
DR. BORER: So, it sounds as if -- I'm trying to pin you down here so we can move on to the next one with real confidence -- that IDNT is supportive of RENAAL but not quite as supportive as a second trial would have been.
DR. LINDENFELD: Right.
DR. BORER: And we'll get to the next question about whether it's supportive enough.
Tom, what do you think about that?
DR. FLEMING: I think I largely agree point by point with JoAnn's answers. My sense is clearly here what we're dealing with now is an agent from the same class and a study in the exact same setting, which adds to the relevance, but to my way of thinking clearly is not as persuasive as if we had a second trial of losartan in type II diabetes. So, it is certainly addressing some of the concerns that at least I had in the previous question, but it's less persuasive than another trial would be of losartan.
I think it's relevant here because at least some of us viewed that the IDNT trial provided a strength of evidence consistent with, just barely, one positive study. At least some of us argued that was one positive study, just barely, but not the greater strength of evidence that we would think you would have to have had to base an approval on that.
And the surrogate, microalbuminuria, study that was presented in that setting, if it were presented in this setting, as JoAnn said, I would view to be of relatively low persuasiveness. That's the type of study that adds to the sense of biological plausibility, but I want to see it confirmed with a clinical endpoint trial.
So, it's not necessarily symmetric here. Maybe I need to clarify that, but in essence what I'm viewing is that with RENAAL in hand and with IDNT being viewed as a single positive trial, it doesn't really provide the strength of evidence of a full separate trial, but it certainly is relevant and provides some additional strength of evidence to RENAAL.
DR. BORER: Blase?
DR. CARABELLO: You're asking for comment or a vote?
DR. BORER: We're asking for a vote and a reason.
DR. CARABELLO: I would vote yes.
First of all, I want to point out that being from Philadelphia, I thought IDNT was a word, like "IDNT a nice day," using it in a sentence.
DR. CARABELLO: I think these two agents are extremely close together. I think they constitute two confirmatory trials, that they are effective in type II diabetes, and I would vote yes for approval.
DR. BORER: It wasn't the approval issue.
DR. CARABELLO: I vote that they're complementary.
DR. BORER: Alan?
DR. HIRSCH: Can I make this simple? I think they are clearly complementary, but as Tom has said, ultimately it would be nice to have a single molecule studied definitively before we create a new precedent. The reason for that is, again, that we tend in our pharmacologic era to try to differentiate products even within the same class when that's to our benefit, and now we're trying to lump them when it's to our benefit. At some point, we need secure knowledge again. But they clearly are supportive.
DR. BORER: I would vote that IDNT is certainly supportive of the effectiveness of losartan. As JoAnn and Tom and everybody else said, I find the persuasiveness somewhere short of replication of RENAAL and forget about the other two. I'll tell you why.
There are some outstanding safety issues that we want resolved here, and while I can accept IDNT as sufficiently supportive of efficacy of a drug that does the things that both of these drugs do in common, so that that's nicely supportive, the safety issues are harder for me to understand. Small differences in pharmacologic effects based on molecular structure might well alter the safety profile a little bit. The magnitude of hyperkalemia that occurs, what have you.
Since those issues weren't really adequately resolved or completely resolved -- I shouldn't say adequately -- for losartan in this one trial, I am concerned that they're still not resolved when I have the IDNT data. The cardiovascular protection issues remain again. Here too, I'm not sure that two drugs of the same class do exactly the same thing.
So, if I were to, ultimately at the end of the day, say that the concordance of evidence is sufficient so that we should approve losartan for the indication that's being sought, I would write the label very carefully to indicate what we don't know and what's left out here.
I think that I would reemphasize what I said earlier, particularly in view of what I've just said, that if we vote to approve the drug, that the FDA should cause something to be written into the label that makes it clear that we are not in any way suggesting the data are sufficient to mandate that this drug should be used in preference, say, to ACE inhibitors or whatever else we may be doing. That's hard to do. I understand, but I think that that's important because of the paucity of information we have.
And finally, I think it's very important that if we believe this as a committee that we go on record saying that we do not believe that if we choose to vote for approval, that we're suggesting in any way that that mandates class labeling for ARBs. It may be that the FDA was never thinking of that. I'm sure the FDA never was thinking of that, but I personally would like to say that I don't think that that would be right.
DR. TEMPLE: Yes. I really wanted to make that clear. We're talking about when a single study of ordinary persuasiveness as opposed to superior persuasiveness would be enough. We're not talking about a no-study standard. We're more risk-averse than that.
DR. BORER: Well, so having said that, again in summary, I believe that IDNT supports losartan. I have concerns. I don't think it's the same as a replication of RENAAL or a second study with losartan. Where we end up on that depends on some of the ancillary issues about labeling that I raised.
DR. THROCKMORTON: Jeff, but what I'm hearing is most of your concerns have to do with safety exposure. And I heard some of the same things from Alan. Is that what you're suggesting?
DR. BORER: Well, I am concerned about safety. Let me say also that with regard to efficacy, I would prefer to have a second study of losartan. However, I would be willing to accept the concordance of evidence from IDNT and RENAAL as sufficiently demonstrating efficacy for diabetic nephropathy. The issue of approvability, however, relates to efficacy as related to safety, and there I have some concerns and I just want to express those.
DR. THROCKMORTON: Yes, but the only reason I was saying that is here we have a large safety database of the two compounds because they are approved. You might imagine a place where two unapproved compounds --
DR. BORER: True. Well, the fact is we're talking here about people with near end-stage renal disease, which changes the ball game a little bit, and it may be that within the dossiers that have been submitted, there's a lot of information about that. That may resolve the issue. I don't know those data because they weren't presented.
DR. FLEMING: Just one more clarification before we go on.
DR. BORER: Tom.
DR. FLEMING: A comment that Bob just made I think, at least from my perspective, it would be appropriate to just have a quick follow-up. I think what Bob was pointing out was this is the type of consideration that's especially relevant when you have a single study and it provides -- I think your words were -- an ordinary level of or what I would call just an adequate level of strength of evidence to meet what we would think of as the standard for this to be called a single positive trial. And does this evidence reinforce it at a level that would lead you to say, now we can approve, as opposed to where you might say you don't need this if you viewed that that single trial was of the strength of evidence of two positive studies.
I would just point out that I don't disagree with that. I would just point out that those of us -- and some of us have said so, that we view this IDNT trial as relevant but less than the strength of evidence that you would have had by another study, technically speaking, would argue that you could readily have a single positive trial of the agent in hand that just barely is adequate to be judged as a single positive study and, hence, a companion trial that together doesn't make it because the second study is less than the same strength of evidence.
By the way, one last point is keeping this in mind, to be consistent, if that other trial goes in the wrong direction, we would equally be weighing that; i.e., if it's relevant, it's relevant whether it's positively reinforcing or negatively reinforcing.
DR. TEMPLE: I think that's true.
We haven't asked you to compare the two studies here, but I just want to remind you the other guys beat two drugs, including one where the effect on blood pressure was almost the same. So, it has its own strengths too, but each one has somewhat different strengths.
DR. BORER: It's interesting. I think, though, I want to add something to what I said earlier in light of Tom's comment. I'm persuaded by IDNT in part because of my interpretation of RENAAL. Even though I wouldn't have accepted it as a single study that is dispositive for judgment or for approval, I do think it's better than the p equals .022 for the primary endpoint because I too am persuaded, as Tom is, that the hard endpoints are more important.
So, I give a little bit higher bounce to RENAAL than it might have had nominally and a little bit lower bounce to IDNT than anybody else might. But you put the two together, and that makes me feel reasonably comfortable about the effectiveness. And then I have the other issues that I mentioned.
DR. NISSEN: I don't think we should go here. I want to dissent a little bit on what's been said, and let me see if I can articulate why.
First of all, I think there's a danger when looking across two different trials involving two different drugs. I would just point out to you there's a lot of precedent here. We have drugs, HMG CoA reductase inhibitors. Everything looked like it was going to go in the right direction, and it pops up that cerivastatin has a huge problem. Nobody anticipated that.
DR. TEMPLE: Not an effectiveness problem.
DR. NISSEN: No, not an effectiveness problem, but a safety problem. I'm just pointing out that there are examples. There are many examples where there is heterogeneity here.
DR. TEMPLE: No one would ever argue that you learn about the safety of a member of a class from the previous members.
DR. NISSEN: I understand.
DR. TEMPLE: Beta blockers have been carcinogenic occasionally and done other things. We don't think that for a minute.
DR. NISSEN: No. I understand.
But just to say that I think we set up a precedent here if we're not careful that we may regret. I guess I want to be careful about that precedent.
Let me point out something else. If you're going to combine the data in your mind from IDNT and RENAAL, you have to combine all the data. And I would point out to you that in RENAAL, there was about a 12 percent excess of cardiovascular death, not significant, compared to placebo. Isn't that right? Wasn't it 1.12?
DR. TEMPLE: In RENAAL.
DR. NISSEN: Yes, in RENAAL. And in IDNT, there was a 36 percent excess in comparison to active control, in this case amlodipine. So, the cardiovascular deaths went in the wrong direction in both trials. So, if you're going to combine the two trials, you got to combine them on the plus and on the minus side.
Again, what it does tell me is that the when you combine those two trials, looking for cardiovascular benefit, you're now looking at 3,000, 4,000 patients. You're not looking at such a small study. So, now we're sitting here in a post-HOPE era and with all the data we have and now we've got two trials, both of which failed to show the cardiovascular events going in the right direction. In some cases like IDNT, they went in the wrong direction for one of the arms. And now I'm troubled by it all.
I don't think you can combine it for the efficacy side without looking at both the pluses and the minuses of combining the two trials, and I really do think there's a lot of risk here in taking two different drugs in the same class and saying, well, we've got trials with each of these, we're going to use them to support each other because it's going to come up again and again, unless you really want to make the standard a lot lower. So, that's my dissent.
DR. BORER: Susanna?
DR. CUNNINGHAM: I think a yes and a no. I think they could support each other, but I don't think it's as good as replication. But I think all the other points that have been made are good ones. So, it's not exactly a real clear-cut picture.
DR. BORER: Beverly?
DR. LORELL: I think that having looked at these two trials very close together is almost more troublesome than helpful. I think I agree more with Steve. Let me see if I can articulate why that is true.
I think the evidence in aggregate from both of those trials and other smaller studies compellingly argues that there is a biologic real effect of interfering with the renin-angiotensin system on progression of renal disease. However, I think there are a couple of problems when you actually look at these two trials together, and I think we can't help but do that on this panel.
One is that they are remarkably congruent in the modest nature of the effect on the primary endpoint, and in fact they're quite congruent that for the primary composite endpoint, the p value for both was .02. So, if anything, that would suggest that the effect is a fairly modest one and not a robust one.
If one tries to dissect and sort of redesign the trials, going back to Tom's analysis of the hard endpoint of end-stage renal disease, here we actually have a bit of a problem because in this trial, if one pulls that out not as the primary endpoint, it looks highly significant. In fact, if I recall my data notes correctly from a month ago, it was not significant in the other trial.
DR. TEMPLE: They didn't have the same kind of follow-up here.
DR. LORELL: I agree.
DR. TEMPLE: Up to the time of the endpoint, the results were the same.
DR. LORELL: In fact, in pulling it together, we have somewhat discordant data.
I share Steve's concern that we have two data sets with the very worrisome hint of actually a negative effect on the hard endpoint of cardiovascular death which was the major cause of all deaths in this trial.
So, if anything, I think yes, it supports an effect on the biology, but I think that putting them together actually is somewhat dissuasive for approval of this for the endpoint.
DR. TEMPLE: Can I just ask? The previous study showed no difference versus placebo on mortality, and there was a trend or maybe more than a trend, favoring the calcium channel blocker. So, are you saying that suggests a negative effect on mortality, or are you suggesting that everybody ought to be on a calcium channel blocker?
DR. LORELL: No, I didn't say that at all. Far be it for me to suggest that all these patients should be on calcium channel blockers.
DR. TEMPLE: I didn't think so.
DR. LORELL: That was not what I said. I agree with Steve's point that there is a worrisome signal regarding an event that is of immense clinical importance.
DR. THROCKMORTON: But it's just worth thinking so we understand. In the IDNT trial -- and I can be corrected if I'm wrong -- the irbesartan mortality was, in fact, lower on irbesartan than it was on placebo, 16 percent versus 14.9 percent. You're right. Amlodipine had a point estimate that was lower than irbesartan, but versus placebo, which was the comparison we had today, in fact it trended in the correct and better direction. So, it would be a wash if all you were doing was adding the two of them up.
DR. FLEMING: Just as a quick statistical clarification, Steve and Beverly, I think you raise very relevant issues about looking at each of these components, and cardiovascular death certainly is a key component. Just as a reminder, the 90 versus 79 deaths translates into an absolute 1.5 percent increase or a relative increase of about 14.5 percent, and for an event that occurs at 10 to 12 percent in the population to be able to detect that relative risk, i.e., to reliably sort out whether this is a true 15 percent increase or it's purely noise, would take 20,000 to 25,000 people. It is more of a concern seeing it in two trials of size 1,500 apiece, but even those two together are still one-eighth of the size that we would need to say anything reliable about whether that's a true increase or it's consistent with random variability.
DR. TEMPLE: Tom, let's be clear we've got the facts right. We don't think you do see it in the irbesartan trial. What you see is a comparison with a different drug, but against placebo it was actually slightly but irrelevantly better.
DR. LINDENFELD: Yes. I think just the placebo group was 16 percent in IDNT and I think 14.9 in irbesartan and 14.1 in amlodipine.
DR. TEMPLE: So, that's not seeing the same thing twice.
DR. LINDENFELD: Overall deaths, total mortality.
DR. FLEMING: I thought what Beverly --
DR. LORELL: My concern was cardiovascular deaths explicitly.
DR. BORER: Let me ask before we go on to Paul, we want for the record, Beverly, a vote on 9.1. Do the findings of IDNT support the effectiveness of losartan for diabetic nephropathy. Forget out the magnitude of the support which you've already answered in detail. But can you look at these data in looking for support for RENAAL?
DR. LORELL: I think they provide soft support in aggregate.
DR. BORER: Paul?
DR. ARMSTRONG: I side with JoAnn and I do so in the context of reassurance about ARB versus placebo in a broader pattern of evidence than what's been provided. So, when I take that into context, I'm comforted.
DR. BORER: Do you want to go to the specifics of 9.2 in addition to having voted yes on 9.1?
DR. ARMSTRONG: Just to reiterate what JoAnn has said all the way down the line.
DR. KOPP: Yes. I think the findings from IDNT are supportive.
I guess I'd like to make a point about cardiovascular mortality to remind ourselves that once somebody starts dialysis, the first-year mortality is something like 20 to 25 percent and the two-year mortality is something like 40 percent. So, if this drug is able to postpone dialysis for even 6 months, it may gain more lives because people don't start dialysis and die from cardiovascular death potentially caused during that 6 months or 2 years of treatment.
DR. NISSEN: That's why I was so disturbed by the fact it didn't go in the right direction. I would have expected that a drug that would delay renal failure would have an effect on death, and the fact that it didn't was very bothersome to me in both trials.
DR. KOPP: Yes, that's a fair point.
I guess one issue would be is the dialysis follow-up sufficient to see the full benefit, but I take your point.
DR. BORER: Dr. Brem?
DR. BREM: To 9.1, I think they are supportive in several ways. One, there is a dose dependence to the effect in the IDNT trial which we have not been able to demonstrate in this trial. Also, the blood pressure was more comparable in the IDNT trial. There was, in other words, less of a difference. So, it would support the view that if you gave enough of a sartan drug, that it would have some beneficial effect on the progression of diabetic nephropathy.
The other point that I would make is I think that the irbesartan trials now are more supportive in light of ancillary or perhaps less relevant endpoints and that is the progression of proteinuria. In this particular study, there is a clear demonstration that proteinuria is associated with the progression of renal disease. That was not able to be demonstrated on the IDNT trial satisfactorily. So, if one uses that information in context, that's further information to me that would support its beneficial effects.
So, yes, I think the two studies together are complementary. They aren't additive, but they are complementary.
DR. BORER: Having said that, why don't we start with you again. I think we know the vote because you've already given it. Should losartan be approved for the treatment of nephropathy in patients with type II diabetes?
DR. KOPP: Yes.
DR. BORER: Do you need more reasoning given everything we've heard, or can he just vote yes?
DR. THROCKMORTON: No.
DR. BORER: Okay.
DR. KOPP: Yes.
DR. BORER: Paul?
DR. ARMSTRONG: Yes.
DR. BORER: Beverly?
DR. LORELL: No.
DR. BORER: Susanna?
DR. CUNNINGHAM: Yes.
DR. THROCKMORTON: Jeff, I'm sorry. If for any reason that you don't -- well, no, that's fine. I take it back.
DR. BORER: Steve?
DR. NISSEN: No.
DR. BORER: I vote yes, but I want to expand a little bit. I think that the drug should be approved for diabetic nephropathy in patients with type II diabetes. I've already talked about my views of RENAAL, that I think it's a pretty good trial, better than the nominal p value on the primary endpoint, but not as good as two. I thought IDNT supported it to some extent, sufficient so that I'm willing to accept the effectiveness. But I'll say again I believe that there are multiple safety issues that have to be dealt with in labeling, that the mandated use of this drug has to be dealt with in labeling, the fact that it shouldn't be mandated.
And I am concerned about cardiovascular death, but I'm less concerned than some of my colleagues because we haven't looked at this population before, and I just don't know what happens to cardiovascular events in people who have end-stage or near end-stage renal disease. So, I'm really not willing to delay the drug because of my concerns, which are very real, about the cardiovascular event issue. I won't say cardiovascular death because I agree with Dr. Kopp about the tradeoff there.
Having said all that and having voted yes, I have to say that if another ARB comes to this committee tomorrow for the same indication, I may say no because I may not find the totality of data from all these trials that we've seen now -- that's two, RENAAL and IDNT -- together with whatever is presented for the new molecule sufficient to weigh me in favor of the new molecule even if the study turns out to be nominally positive for the new molecule. So, this is not a precedent. This is a specific issue.
DR. HIRSCH: Well, I would vote no, but that's in the guise that I don't believe a disapproval, if the committee goes that way, will deprive a single patient with diabetic nephropathy of access to a molecule that's available. However, if approval is recommended, then like my colleague to my left, I believe that the safety issue -- yes, sir?
DR. TEMPLE: Well, I guess we would hope, as a general matter, that you not consider such practicalities.
DR. HIRSCH: Okay, I won't.
DR. TEMPLE: We want your answer and your answer is no.
DR. HIRSCH: It remains the same.
DR. TEMPLE: But if your answer would be different if somehow they weren't available --
DR. HIRSCH: No. I hear you. My answer, Bob, is the same in the sense that I've been all day, including at the last meeting, trying to be consistent between the two meetings, like Bev, trying to look at the data from single trials, even the two trials together, and decide, fine, if this weren't available, do I have adequate data at this point to bring it to approval. I would like not to have to change my vote the next time an ARB comes. I'd like to make sure the evidentiary standard persists.
DR. TEMPLE: But I guess we hope the votes will be as if it wasn't available. Otherwise, it's sort of --
DR. HIRSCH: Well, I answered that. You may have to ask the others who aren't here to revote.
DR. TEMPLE: No. You were the only one who indicated that was influencing you. It's hard not to face reality and know that it doesn't really matter whether we approve it or not. People can use it anyway because they can read the New England Journal. But we like to try to do it completely abstractly and on an evidentiary basis and not do that because, you know, fair is fair.
DR. HIRSCH: Fair is fair.
DR. THROCKMORTON: Was your reason for raising the issue of it being available your concern with the safety?
DR. HIRSCH: Yes. My hesitation is deep and long and I'm not sure that the whole room has to hear a rehash of what you've heard, but it does get attributed to what I would see if I were merging the two trials. I'd still believe there is a safety blip that is of some concern and efficacy standards that are not quite achieved. So, it's that risk/benefit analysis we always follow here on prima facie evidence.
DR. BORER: Blase?
DR. CARABELLO: I vote yes.
DR. BORER: I think you already gave your reasoning.
DR. LINDENFELD: I vote yes. Just two quick comments.
Again, as has been said before, I think we need to try to say something that this doesn't imply superiority over ACE inhibitors, just something in there.
And the other thing in terms of safety is that I think we need to make it clear that the risk of hyperkalemia is cumulative with this drug over the course of the drug and probably always needs to be monitored ongoing over time. That's a little bit different I think than in other situations.
DR. TEMPLE: I do have to say commenting that it's not better than ACE inhibitors or that that hasn't been studied would be a very unusual thing for us to do when there's no data showing that ACE inhibitors work in this setting.
DR. LINDENFELD: That's right.
DR. TEMPLE: That's tough to support.
DR. LINDENFELD: That is tough, but maybe we can think about that. But it would be nice to not have this supplant the use of ACE inhibitors I think or not have this be a reason to do that. Maybe that's not possible, but I share that concern.
DR. TEMPLE: Also, if either of you have specific suggestions or little phrases that you think will do this, please.
DR. BORER: Off-line we can do that. I think what JoAnn is saying is the same thing that I am, that there has to be some kind of language that says we are not saying that this is the be all and the end all. Now, how to say that, I think we can figure out, make some suggestions, and you can improve upon them, and something can be done.
DR. TEMPLE: Again, though, for reasons that are not known to us, no one has bothered to study ACE inhibitors in this setting. I don't know why that is. Maybe because they're all going to go off patent soon or whatever the reason is, but nobody has studied it.
DR. BORER: Right. No, I'm not suggesting that we can say anything about the relation to ACE inhibitors. What I'm saying is we must say that we don't know what the story is with cardiovascular morbidity and mortality here, and therefore we can't say that this is necessarily the way that one must go in the individual patient.
DR. TEMPLE: We would certainly give the results that show no benefit on those endpoints. That would be part of the trial.
DR. BORER: Okay, well, I don't want to take up everybody's time, but we can write down some suggestions.
DR. LORELL: Mr. Chair, are we really giving two votes here? One is a yes/no vote if we get to have a conditional labeling, and one is a yes/no vote if we don't have any influence on labeling?
DR. BORER: No. I think we're saying we think that the drug should be approved or it shouldn't be approved and then we're giving caveats about how we think that ought to happen. We're only an advisory committee and the law is sitting across the table from us.
DR. TEMPLE: No, but suggestions on what you're thinking of are welcome. I'm just trying to signal that saying something like we don't know whether you're better than ACE inhibitors, when they don't even have that claim, would be very difficult to support against -- well, it would be difficult to convince me to do that, much less the company.
DR. BORER: I was going to make the suggestion off-line, but sometimes we have meetings or sections of meetings about issues rather than about drugs. I think this issue in so many areas may have risen to the point of needing some kind of open discussion at some other time.
Tom, I think you already voted on approvability, but do you want to restate it?
DR. FLEMING: Sure. Yes, I vote approval. And maybe I could just add one additional clarification and that is I wanted to reinforce the wisdom that I have heard today that one has to be extremely cautious when one is analyzing data in putting particular focus on primary endpoints. I want to strongly endorse that as extremely important. From a statistical perspective, when we're using a trial as a confirmatory trial, if one wants to assess strength of evidence in a confirmatory fashion, it is extremely important that we prespecified our primary hypothesis with primary analysis methods so that we can really interpret strength of evidence in that context.
Having said that, as strongly as I think statistical procedures are extremely useful in making these assessments, as to whether or not studies adequately establish favorable benefit to risk, any statistical approach must be viewed with the broadest assessment, bringing in clinical judgment.
In my own sense in this particular trial, if one views, as I do in this particular case, that the primary endpoint that was set up here has as an important component, at least a component that I view to be a surrogate, which in particular in this setting isn't intrinsically necessary to rely on because the clinical endpoints are occurring so rapidly, it's my own assessment that the evidence that I would have looked for in this trial -- even though the primary endpoint had the triple component, I would have relied on the double component anyway.
Essentially, at least in my own perspective, what one needs to avoid is a situation where you don't see what you had hoped to see in the primary endpoint and you start looking around for other ways of getting a more favorable conclusion. Clearly that's hazardous.
What I'm trying to sort through here, though, is in my own assessment, at least speaking for myself, that's not the issue. As we discussed in January, some of us had argued very strongly that in a setting such as this, the primary endpoint ought to be based on those elements that are truly the clinical elements, especially when they can be addressed in a timely way and especially in a setting such as this where it's somewhat problematic how to interpret the triple endpoint because there wasn't a complete follow-up. It creates a situation, at least for me in my own judgment, that the primary endpoint would not have been persuasive even if it had been meeting the standards for strength of evidence of two positive trials.
So, I guess what I'm trying to articulate here is just to reinforce that there is great wisdom in stating that one needs to be extremely cautious about the strength of evidence and the focus that you give to the primary endpoint, but there clearly needs to be judgment and you clearly need to look at the totality of results. So, I would like to make it clear that, at least in my own case where my view of what is clinically relevant leads me, independent of what these data would have shown, to focus on the dual endpoint shouldn't be viewed as an endorsement of readily deviating from the primary endpoint.
DR. BORER: What we've said in summary is that the majority of the committee favors approval of this drug for the requested indication. I think it's important to remember that there have been many important concerns raised and that, though the vote was clearly and decisively in favor of approval, the magnitude of the difference between those who voted yes and those who voted no may be relatively small and, therefore, that there's no binding precedent based on this vote for how this committee would view other drugs if they come with similar credentials for similar indications.
Finally, I think I can say for all of us that we voted the way we did without the expectation that another trial is going to be performed because the data didn't satisfy everybody. We voted based on what we have.
DR. TEMPLE: I just wanted to thank you. This was another difficult session, just like the one in January was. You grappled with it.
I have to tell you it's hard not to allow people to draw inferences from the fact that we're making use of data from another drug of the same class, and they will. You are, of course, free to be as situation-appropriate as you want.
But this was very helpful to us and it was a good discussion. So, we thank you.
DR. BORER: Thank you.
If there are no other questions from the FDA, I'll call the meeting adjourned.
(Whereupon, at 4:47 p.m., the committee was adjourned.)