U.S. FOOD AND DRUG ADMINISTRATION
CENTER FOR DEVICES AND RADIOLOGICAL HEALTH
+ + + + +
MEDICAL DEVICES ADVISORY COMMITTEE
+ + + + +
OBSTETRICS AND GYNECOLOGY DEVICES PANEL
+ + + + +
+ + + + +
MAY 17, 2005
The Panel met at 8:30 a.m. in the Walker/Whetstone Rooms of the Holiday Inn, Two Montgomery Village Avenue, Gaithersburg, Maryland, Dr. Kenneth L. Noller, Chair, presiding.
KENNETH L. NOLLER, M.D., Chair
RALPH B. D'AGOSTINO, Ph.D., Consultant
MARCELLE I. CEDARS, M.D., Consultant
AMIR H. GANDJBAKHCHE, Ph.D., Consultant
ELISABETH GEORGE, Industry Representative
MELVIN V. GERBIE, M.D., Consultant
EVELYN R. HAYES, Ph.D., R.N., CS-FNP, Member
PAULA J.A. HILLARD, M.D., Member
YULEI JIANG, Ph.D., Consultant
THOMAS M. JULIAN, M.D., Consultant
HUGH S. MILLER, M.D., Member
CHRISTINE MOORE, Consumer Representative
JOSEPH S. SANFILIPPO, M.D., Consultant
RUSSELL R. SNYDER, M.D., Consultant
JONATHAN W. WEEKS, M.D., Member
MICHAEL T. BAILEY, Ph.D., Executive Secretary
NANCY C. BROGDON
JULIA CAREY-CORRADO, M.D.
BRANDON D. GALLAS, Ph.D.
SUSAN GARDNER, Ph.D.
GENE PENNELLO, Ph.D.
JOYCE WHANG, Ph.D.
RONNIE ALVAREZ, M.D.
BRENT BLUMENSTEIN, Ph.D.
J. THOMAS COX, M.D.
ROSS FLEWELLING, Ph.D.
WARNER K. HUH, M.D.
JOAN L. WALKER, M.D.
THERESA WINGROVE, Ph.D.
THOMAS C. WRIGHT, Jr., M.D.
Call to Order................................... 4
Office of Surveillance and Biometrics (OSB)
Presentation -- Role of OSB in the Review of
Postmarket Study Designs
Susan Gardner, Ph.D....................... 9
Colin Pollard............................ 15
Open Public Hearing
Dr. Diane Solomon........................ 21
Dr. Mark Schiffman....................... 27
Theresa Wingrove, Ph.D................... 37
J. Thomas Cox, M.D....................... 43
Ross Flewelling, Ph.D.................... 49
Joan L. Walker, M.D...................... 55
Thomas C. Wright, Jr., M.D............... 60
Brent Blumenstein, Ph.D.................. 74
Warner K. Huh, M.D....................... 81
Joyce Whang, Ph.D....................... 108
Julia Carey-Corrado, M.D................ 113
Gene Pennello, Ph.D..................... 129
Panel Discussion.............................. 167
Question 1.............................. 188
Question 2.............................. 205
Question 3.............................. 209
Question 4.............................. 210
Question 5.............................. 216
Question 6.............................. 225
Question 7.............................. 233
Question 8.............................. 241
Question 9.............................. 246
Question 10............................. 250
Open Public Hearing
Dr. Mark Spitzer........................ 260
Dr. Mark Schiffman...................... 266
Sponsor Final Comments........................ 275
Panel Deliberations and Vote.................. 284
CHAIR NOLLER: Good morning. I'd like to call this meeting of the OB/GYN Devices Panel to order. My name is Ken Noller. I'm the chairperson of this panel. I'm an obstetrician-gynecologist generalist. I work at Tufts University School of Medicine.
Everyone in the room, if you haven't done so, please sign in on one of the attendance sheets that are on the tables outside the doors. I will note for the record that the voting members present constitute a quorum of the panel as required by 21 C.F.R. Part 14.
We will start in just a moment. I'm going to ask everyone at the table to introduce themselves. I'd ask you to state your name, your area of expertise, your position, and affiliation. We'll start at this end, please.
MS. BROGDON: Good morning. I'm Nancy Brogdon. I'm not a member of the panel. I am the director of FDA's Division of Reproductive, Abdominal, and Radiological Devices.
DR. SANFILIPPO: I'm Joe Sanfilippo, a reproductive endocrinologist, the University of Pittsburgh School of Medicine.
DR. GERBIE: I'm Melvin Gerbie. I'm at Northwestern University Medical School in gynecology.
DR. JIANG: Yulei Jiang. My interest is in computer-aided diagnosis and radiology. I'm here temporarily. I'm assistant professor at University of Chicago.
DR. GANDJBAKHCHE: My name is Amir Gandjbakhche. I am a section chief in the National Institutes of Health. My area of expertise is biomedical optics.
DR. MILLER: Hugh Miller. I'm a maternal-fetal medicine specialist. My affiliation is with the University of Arizona, and Obstetrics Medical Group.
DR. HILLARD: My name is Paula Hillard. I'm at the University of Cincinnati where I do pediatric and adolescent gynecology.
DR. BAILEY: I'm Mike Bailey. I work for the Food and Drug Administration. I am the executive secretary for this panel.
DR. CEDARS: Marcelle Cedars. I'm a reproductive endocrinologist at University of California, San Francisco.
DR. WEEKS: Jonathan Weeks. I'm a maternal-fetal medicine specialist at Norton Healthcare in Louisville, Kentucky.
DR. JULIAN: Tom Julian, gynecologist, University of Wisconsin.
DR. D'AGOSTINO: Ralph D'Agostino, biostatistician, Boston University.
DR. SNYDER: Russell Snyder. I'm the division director of gynecology at University of Texas Medical Branch in Galveston. I'm a general OB/GYN. I've also fellowship trained in gynecologic pathology.
DR. HAYES: Good morning, Evelyn Hayes, University of Delaware, Department of Nursing.
MS. MOORE: Good morning. I'm Christine Moore, consumer member of the panel, actually appointed to the gastroenterology panel and on loan here. I am the retired dean of student services from the Baltimore City Community College.
MS. GEORGE: I'm Elisabeth George, vice president of quality regulatory from Phillips Medical Systems, and I'm the industry rep.
CHAIR NOLLER: Thank you. Now for any press that might be present, your FDA contact is Colin Pollard. Colin, would you raise your hand? Fine.
Now, we run these meetings according to a relatively tight schedule. We have a lot to accomplish today. We want everybody to be heard. What we would also hope is that everyone would stick to their timeframe, those of you that have been assigned times, so that the panel has its time to discuss matters. We also would ask all of you to turn off your cell phones and your pagers, please. Also, you may address the panel only from the table. If you wish to speak, you need to be recognized by me. Once I recognize you, you need to come forward, sit at the table, and then speak.
I'm going to turn the meeting over now to Mike Bailey.
DR. BAILEY: All right. First I'd like to give you a little update on upcoming meetings for the panel. The next meeting of this panel will be on June 23, 2005. The remaining tentative panel dates we have in addition to that one are for August 15-16, and November 15-16 of this year.
Transcripts of today's meetings will be available from Neal Gross & Company. And information on purchasing videos of today's meeting can be found on the table outside the door. Presenters to the panel today who have not already done so should provide FDA with hard copies of their remarks, including overheads, to Michelle Byrne. Michelle, would you stand or raise your hand? And she will collect these from you at the podium if I have not received them in advance of the meeting. Dr. Noller?
CHAIR NOLLER: Thank you, Mike. Dr. Susan Gardner, of the Office of Surveillance and Biometrics, will now give a presentation to the panel regarding the role of OSB and the review of postmarket study designs.
DR. GARDNER: Thank you. Good morning. Due to a small communication glitch, you're going to have to follow your hard copy if you'd like to see a copy of my slides. Or you can just watch me, whichever you prefer. But it's short, so.
I'm going to tell you about a major programmatic change in CDRH. And operationally, that is the move of the condition of approval study program from the Office of Device Evaluation to the Office of Surveillance and Biometrics. Briefly, the Office of Surveillance and Biometrics is involved both in the premarket and postmarket activities of the Center. We're involved in the premarket activities by virtue of the fact that the statisticians are in our office, and you hear from the statisticians at all your panel meetings. And also the epidemiologists are in our office, and you'll be hearing more from the epidemiologists as time goes on. We're also responsible for signal detection of adverse events by use of various monitoring tools, including the MDR system, the Medical Device Reporting System, and our MedSun program, which is the Medical Device Safety Network. We're responsible for analysis of the signals that we see, the characterization of product risk, and the coordination of Center response to major postmarket issues, including risk communication to the health care professionals. And we're also responsible for the interpretation of the MDR regulation.
The legislative authority for condition of approvals comes from 21 C.F.R. 814.82, and it says the post-approval requirements can include continuing evaluation and periodic reporting on the safety, effectiveness, and reliability of a device for its intended use. So that's our legislative basis.
So a couple of years ago we decided to do an internal evaluation of our condition of approval study program, and take a hard look at how we were doing. We looked at all the PMAs that were approved through 1998 and the year 2000. That was 127 PMAs, and 45 of those had condition of approval studies. We then went and -- for the original purpose of the study was to look at the quality of the condition of approval studies we had ordered. And the problem was that we found that we didn't have a very good tracking system, and we were unable to locate a number of these studies and find out whether they were completed or not. We realized that often the lead reviewers went on to have other jobs. And so the lack of having a standardized tracking system, and having people move throughout the program left us, again, without a good way to track these studies. And we also realized that the lack of premarket resources made it difficult for the folks in that shop to do this job because they're really involved in the premarket piece. So again, we transferred the program to OSB.
The goal of the transfer, of course, was to improve the program, and to get better postmarket information as a product enters the market, which is our chance to see how the product -- what happens in real world use of the product. We want to better characterize the risk-benefit profile that we learned about premarket, and of course add to our ability to make good scientific decisions. So on January 1 this year we transferred the program. But I will say this was done after having a 2-year pilot that had already been carried out. So many of the procedures that are now in place, again, have been piloted for a couple of years.
The first thing we did was develop and institute an automated tracking system, which has gone live very recently. So all condition of approval studies that have been ordered starting January 1 are now in our tracking system for follow-up.
We've also added an epidemiologist to PMA teams where we think that it's likely there's going to be a condition of approval study as an outcome. Adding the epidemiologist means that there's someone on the team that's tasked with coming up with a postmarket plan during the pre-approval process. Because these are obviously tied very close, hand in hand. The epidemiologist will take the lead in developing a postmarket study plan, and also in developing a postmarket study if that's going to be the case. They're going to have a lead in the design of the study protocol, and in the evaluation of the study results as the study goes on. They'll continue to work with the PMA team throughout the process.
And now we also want to make sure that industry and CDRH has motivation for continuing to get these studies done, and to do a better job than we think had been done in the past. We think if we do a good job of formulating the postmarket question, and we have a really solid study design, that everybody is going to be more motivated to get these studies done, and to follow them and track them as we go on. We also think that having acknowledgement of the studies, and having feedback on the studies to industry, and again, having conversations within CDRH will motivate everybody to get these studies done, and to make sure that the feedback is important.
We're going to have a website in CDRH where we're posting the study results, and the status of the studies as we go along. And we also have an authority under Section 522 to mandate postmarket studies if we have outstanding public health questions that we think need to be answered. This authority also gives us the ability to add civil money penalties if the studies aren't done. So hopefully we don't have to go down that path, but if we do, we have the authority to do that.
So what impact this will have on the advisory panel. First of all, during the approval process, postmarket questions come up rather naturally during discussion. And sometimes we'll have important questions that we're going to lay out and ask you for feedback on. So again, as you go through this process, as postmarket issues come up, we ask for your feedback, and we ask for your advice on how -- and possible approaches for getting these studies done.
And finally, partly because you've asked for it, and partly because we think it's important, we're going to continue to add to the program feedback to you on condition of approval studies either by industry and by FDA. So after a product has been approved, if there is a condition of approval study, someone will come back to the panel at an appropriate period of time and give you an update on how the condition of approval study is going. Thank you.
CHAIR NOLLER: Thank you. Colin Pollard, Chief of the Obstetrics and Gynecology Devices Branch would now like to make some introductory remarks to the panel. Mr. Pollard?
MR. POLLARD: Thank you Dr. Noller. And I'll try to be brief and keep us on schedule. Good morning, ladies and gentlemen of the panel, and distinguished audience. For the rest of the day we will be hearing about and listening to you deliberate about a PMA for what for FDA is a first of a kind technology using optical diagnostic spectroscopy to help detect cervical disease.
This is an effort that dates back many years. The panel actually visited this technology in a generic fashion several years ago, and helped us write a guidance document. The sponsor, who you will hear from today, interacted with us using early collaboration formats that are now available to industry. And you'll hear from both them and from FDA reviewers about this. We look forward to your deliberations, and I would just say in the interest of time we have asked both the FDA reviewers and the sponsor to, when responding to questions, to be brief and succinct, to give you the maximum amount of time for deliberations. So with that I thank you very much and we look forward to your discussion.
CHAIR NOLLER: Thank you, Colin. For panel members, by the way, you need to be relatively close to these microphones to be heard. Let me turn the meeting over to Mike Bailey again for a moment.
DR. BAILEY: All right. We need to come back to and read the -- sorry. First we're going to read the conflict of interest statement into the record. The following announcement addresses conflict of interest issues associated with this meeting, and is made a part of the record to preclude even the appearance of an impropriety. To determine if any conflict existed, the agency reviewed the submitted agenda for this meeting, and all financial interests reported by the committee participants. The conflict of interest statutes prohibit special government employees from participating in matters that could affect their employers' financial interests. However, the agency has determined that the participation of certain members and consultants, the need for whose services outweighs the potential conflicts of interest, involved is in the best interest of the government. Therefore, a waiver under 18 U.S.C. Section 208 has been granted to Dr. Joseph Sanfilippo for his consulting interest for which he receives less than $10,001 in an unaffected subsidiary of the parent of a competing firm. Because of a part of his grant to his institution for the sponsor's study, Dr. Sanfilippo may participate but not vote in today's deliberations. Copies of the waiver may be obtained by submitting a written request to the agency's Freedom of Information Office, Room 12A-15 of the Parklawn Building.
We would like to note for the record that the agency took into consideration certain matters regarding another panelist, Dr. Ralph D'Agostino, reported past personal interests and current employee's interest with firms at issue, but in matters not related to today's agenda. The agency has determined, therefore, that he may participate fully in the panel's deliberations. In the event that the discussions involve any other products or firms not already on the agenda for which FDA participant has a financial interest, the participant should excuse him- or herself from such involvement, and the exclusion will be noted for the record.
With respect to all other participants, we ask in the interests of fairness that all persons making statements or presentations disclose any current or previous financial involvement with any firm whose products they may wish to comment upon.
And second, I will read the appointment to temporary voting member status. Pursuant to the authority granted under the Medical Devices Advisory Committee Charter, dated October 27, 1990, and amended April 20, 1995, I appoint the following as voting members for the Obstetrics and Gynecology Devices Panel for the duration of this meeting on May 17, 2005. Dr. Marcelle Cedars, Dr. Ralph B. D'Agostino, Dr. Amir Gandjbakhche, Melvin Gerbie, Dr. Yulei Jiang, Dr. Thomas Julian, and Dr. Russell Snyder. For the record, these people are special government employees, and are consultants to this panel or another panel under the Medical Devices Advisory Committee. They have undergone the customary conflict of interests review, and have reviewed the material to be considered at this meeting. This was signed by Daniel Schultz, director, Center of Devices and Radiological Health, on 4-20-2005. Dr. Noller?
CHAIR NOLLER: Thank you. We will now proceed to the open public hearing portion of the meeting. We have 30 minutes set aside this morning for open public hearing and we have some time this afternoon. So far there are three declared speakers. Two of these speakers will present during this session now, while the third presenter has requested to speak during the afternoon. Prior to hearing from the first speaker, I will read the open public hearing statement.
Both the Food and Drug Administration and the public believe in a transparent process for information-gathering and decisionmaking. To ensure such transparency at the open public hearing session of the advisory committee meeting, FDA believes it is important to understand the context of an individual's presentation. For this reason, the FDA encourages you, the open public hearing speaker, at the beginning of your written or oral statement to advise the committee of any financial relationship that you may have with the sponsor, its product, and if known its direct competitors. For example, this financial information may include the sponsor's payment of your travel, lodging, or other expenses in connection with your attendance at the meeting.
Likewise, FDA encourages you at the beginning of your statement to advise the committee if you do not have such financial relationships. If you choose not to address this issue of financial relationships at the beginning of your statement, it will not preclude you from speaking.
The first two speakers, Drs. Diane Solomon and Mark Schiffman, have asked to speak together. Dr. Solomon, you have 15 minutes allotted for this presentation.
Dr. SOLOMON: Good morning, panel and guests. I'm Diane Solomon, and this is my colleague Mark Schiffman. We're from the National Cancer Institute. We're also co-directors of the ASCUS/LSIL Triage Study, which was an NCI-sponsored randomized clinical trial. I have no personal financial affiliations with either the sponsor or direct competitor. I believe I can speak for Mark in that regard. In terms of the ALTS Clinical Trial, I would like to disclose that ALTS did receive support in the form of equipment and/or supplies at below cost from Digene, Cytyc, TriPath, and National Testing Laboratories.
In terms of ALTS, I'm sure many of you are very familiar with the trial. I would like to just very briefly go over some aspects of ALTS, and then present some data on the relative contributions of a first versus a second biopsy taken at colposcopy during the ALTS trial. ALTS included just over 5,000 women who enrolled with a community cytology diagnosis of either ASCUS or low-grade squamous intraepithelial lesion. ALTS was comparing three different management strategies for women, one being immediate colposcopy -- and this was included, actually, as a gold standard for detection of disease, and was to be the benchmark by which the other arms of the trial were to be evaluated. The second arm was HPV triage, which depended on the results of an HPV test to determine whether or not women went for colposcopy. Third arm of the trial was conservative management, where women were followed with cytology, and a high grade finding on cytology is what then referred women on to colposcopy.
And colposcopy, clinicians were instructed to biopsy any suspect squamous intraepithelial lesion. The number of biopsies taken were at the clinician's discretion. Biopsy results were read by the clinical center pathologist. And it's that result that was used in the management of women during the course of the trial. So when Mark presents data on colpo biopsy results, those are the results as interpreted by the clinical site pathologist.
Women were followed for two years intensively with cytology, as well as HPV testing, and cervicography every six months for a time period of two years. At their exit from the trial, an exit colposcopy was done on all women, regardless of their prior management. And we had a fairly low threshold for LEEPing women to assure that no one left the trial with undetected disease. So women who had evidence of just persistent low-grade disease were also LEEP'd.
Our findings, just to touch on some of the summary findings. No test, triage test, was completely sensitive for detecting what we call cumulative CIN3. And when we talk about cumulative disease for women, we're talking not about the clinical center pathologist's read of individual colpo biopsy results. We're talking about the review of all histopathology taken for a woman. And the review of all of that histopathology by a pathology QC group that reviewed all histopathology for all women at all clinical centers.
The finding, which was rather surprising, was that colposcopically directed biopsy was much less sensitive than we previously had assumed. We also found that undetected CIN3 did not regress, whereas CIN2 showed significant regression. And I'd like to show this diagrammatically, or with these histograms. What we have here -- is there a pointer? What we have here, if you see the histogram on the left, the numbers of CIN3 that were detected in the immediate colposcopy arm. The colors show when the CIN3 was diagnosed. Blue is during the enrollment period, red is during follow-up, and yellow is at exit. You can see in the immediate colposcopy arm just about half of CIN3 is detected at enrollment. This performance was much less sensitive than what we had assumed would be the case, and I think highlights that colposcopy really is a weak link in this whole chain of events from screening to detection, diagnosis, and management. Although all women went for colposcopy in the IC arm, only just over half the cases of CIN3 were detected at enrollment. In the conservative management arm, many fewer cases were detected at enrollment, and we actually anticipated this because we had a fairly high threshold for referral to colposcopy in this CM arm. A woman had to have a high-grade cytology for a referral to colposcopy.
But although the enrollment detection was lower in the conservative management arm, we see that there was some catch-up with more cases detected during follow-up. But actually over a third were not detected until that safety exit colposcopy. But what this shows is even though there was delay in detection of CIN3, by the end of the trial we had equivalent numbers of CIN3 in the immediate colposcopy, HPV triage, and CM arms, showing that CIN3 that was present, even if undetected, did not regress. We feel that CIN3 is actually a fairly good proxy of a true pre-cancer.
And I want to contrast that with the findings for cumulative CIN2. Again, this is the most severe diagnosis for a woman during the two years of the trial, being CIN2. And we see that in the immediate colposcopy arm, a majority of the cases were identified at enrollment. In the conservative management arm, many fewer cases were diagnosed at enrollment. And there was some catch-up at follow-up and at exit, but there's still, at the conclusion of the two years of follow-up was a deficit of CIN2, we presume attributable to the fact that undetected CIN2 actually regressed. So CIN2, in contrast to CIN3, is really not a pre-cancer category, but probably a mixture of some true high-grade disease, but also low-grade disease that looks more severe morphologically, but that is destined to regress.
I'd like to now turn the presentation over to Mark Schiffman, who's going to present some data on comparing the first and second biopsies taken at colposcopy.
Dr. SCHIFFMAN: Thanks for listening. Well, every new technique that's added to a first technique will increase sensitivity. Just, it's an axiom. So the question is compared to what? And so we're trying to present some baseline of what a second biopsy adds when it's done just by usual colposcopic direction, without any kind of adjunctive device.
To inform today's deliberations we evaluated the contribution of first and second biopsies when two are taken in ALTS. As an introductory observation, we saw that the first biopsy tended non-significantly to be more sensitive than the second. However, the severity of the two biopsies was positively associated. And that's -- a weighted kappa of 0.37, given the overall inherent uncertainty of histopathology, is pretty strong. We focused on the second biopsy and its effect -- sorry for the typo -- on clinical management, particularly on sensitivity, on adding to sensitivity.
So first. There were 719 paired biopsies in ALTS. Here's a cross-tabulation of the clinical center pathologist's diagnoses of the first and second of the paired biopsies. And we're going to be focusing on the top row in particular. As Dr. Solomon noted, the clinical center diagnoses directed patient management, while the pathology QC final diagnoses were used as study endpoints. Sixty-six additional cases were called CIN2 or CIN3 by the clinical center pathologist because of the second biopsy alone, 53 plus 13. Please note that not all clinical diagnoses of CIN2 are worse, or eventually considered CIN2 or CIN3 final cases by the pathology QC group. So, on any table, even though you might see a CIN3, or a CIN3-CIN3, it could still be downgraded in the eventual review if the entire panel of core pathologists saw it and decided it should be downgraded. In particular, CIN2 is not optimally reproducible in line with its regression potential, as Diane showed.
So let's look at how the 66 splayed out in terms of what they were eventually called. Here are those cases eventually downgraded and called less than CIN2. That meant they were either nothing, or low grade at worse, in the final analysis. So there were -- you see that 13 of the first biopsies that were less than CIN2 were picked up by a second biopsy as CIN2 or worse. Thirteen were called CIN2 and they were downgraded, showing that a lot of those were eventually not called as worthwhile finds. In fact, they were considered false positives. And this would be a decrease in specificity. And the decrease in specificity on a percentage basis was not that high. So we're not really going to be talking about decreasing in specificity as a main problem in our view in the trial. It's certainly on the order of a few percent in this context of triage. We're going to be talking more about sensitivity.
CIN2 is really problematic, and the reason it is because we're really not sure whether these are improved sensitivity or decreased specificity in that the actual cases, the work at picking them up is really not that certain in young women like in ALTS. If CIN2 is considered to be a valid endpoint, the second biopsy increased sensitivity by 35.3 percent among those women with CIN2 undiagnosed initially by the first biopsy. But we observed that the kappa for the clinical diagnoses of the first and second biopsies among this heterogeneous group was basically null, it was -0.02. The first and second biopsies were almost interchangeable in severity of diagnosis. And so since CIN2 is characterized by error, we relied on CIN3 mainly.
And this is the heart of the presentation. With CIN3, you certainly want to find all these. They were confirmed CIN3 final diagnoses. And so you see 27 additional pickups. Now that 27 is the important numerator of additional pickup for a colposcopically directed second biopsy in people who took two biopsies. The first biopsy tended to be more sensitive. P was less than 0.1, was almost significant, suggesting the lesions were easier to identify initially than for CIN2. However, the second biopsy still increased sensitivity by 49.1 percent among those women with CIN3 undiagnosed by the first biopsy.
Now, if you see among the 152 total CIN3's that were found in the context of paired biopsies, you see that 97, 47 plus 50, had been picked up by the first biopsy already. So the absolute and relative gain are the comparison of the 27 to the 97, either in absolute terms as an addition, or relative terms. So it's either 27 divided by 152, or 27 divided by 97. We summarize that here. The initial reduction in false negatives we called was 49.1 percent. But then if you talk about absolute increased sensitivity, it's 17.8 percent. Relative increased sensitivity, 27.8 percent. And graphically, we present this so you can see it. The first biopsy captured the bulk. The second biopsy contributed, and then some remained, about an equal percentage, undetected.
Although we don't have time to present the data, we saw no important differences in our results when we stratified by whether the colposcopist was a gynecologist or a nurse colposcopist.
So, in summary, first of all, both Dr. Solomon and I see a clear need for improving performance of colposcopy and localization of biopsy site. There's no question about that. That's a major concern. Our concern is that any additional test, any additional biopsies, will increase sensitivity to some degree. And we wonder for any device what the proper comparison is. We believe a proper comparison will be not random punches of completely normal epithelium, because with any kind of training, a colposcopist will direct the biopsies to at least non-normal looking epithelium that could be a possible lesion. And we tried to give you some sense of what that would do in the absence of a device as a form of control, if there's an adjunctive cling. And hopefully we've indicated what the yield would be, providing a point of comparison for your deliberations. Thanks very much.
CHAIR NOLLER: Thank you very much. That final question will be addressed by FDA in their presentation. Does the panel have any clarification? We don't really want to discuss the matter at this point, but is there any question to address to Dr. Schiffman or Dr. Solomon? Mel?
DR. GERBIE: Is there any effort to add improving performance of pathology at the initial reading of these?
DR. SOLOMON: As a trained pathologist, let me take that one.
CHAIR NOLLER: Thirty seconds, please.
DR. SOLOMON: We did have a pathology QC group that was a panel composed of experts in pathology. What we found was that while there were differences in the way the CC read versus the path QC, when we went back and correlated many of these diagnoses with HPV results, we found that the pathology QC group did read a little higher, meaning they tended to call things CIN3 versus CIN2, but that there was no significant increase in sensitivity for the path QC group compared to the CC group. In other words, there's variability and non-reproducibility in pathology, and that's simply part of the art of pathology.
DR. SCHIFFMAN: But the CIN3 is a good diagnosis. It's at the CIN2 level that we have concerns.
CHAIR NOLLER: Any other -- and please, as you start to speak, introduce yourself. I forgot to give your name for the transcriptionist. I have one question. Most colposcopists, if they're going to take two biopsies do the posterior one first. So if you do the anterior one, the blood runs down over the site. And so I don't -- it doesn't make sense to me that the order made would have been the more severe versus the second. Did you look at site?
DR. SCHIFFMAN: We asked -- specifically the instructions were worst site first.
CHAIR NOLLER: To do the worst site first? Thank you. Any other clarifications? Thank you.
DR. JULIAN: I have one quick question.
CHAIR NOLLER: Introduce yourself.
DR. JULIAN: Tom Julian, University of Wisconsin. What was the quality control in the colposcopists in the study?
DR. SCHIFFMAN: The quality control on the colposcopists.
DR. JULIAN: The qualifications. How did you know you had adequate colposcopists?
DR. SCHIFFMAN: Well, there was a three-person colposcopy quality control group, including Daron Ferris, Tom Cox, and Lou Burke. And there was an initial training period, the CRA exam was administered, there was retraining, or an attempt to bring up to speed people who didn't do well on CRA. That was not an exclusion criterion if they failed. At that point there were continued going around in ALTS to -- several times a year, I don't remember exactly how many, with in-services, and basically hands-on training by the three people. Each one divided the clinical centers into particular areas of responsibility, and they continued to travel like that throughout the entire study. There was also offsite review of DenVu images. If there was a QC safety net meeting, something looked bad that hadn't been called bad, it was signaled immediately back to the clinical center, and then there was an attempt to make that known to the -- not an attempt. The clinician was told that that was a problem. So we attempted quite rigorously to introduce an element of offsite static image validation. But as you know, static images alone are difficult.
CHAIR NOLLER: Thank you. One last one.
DR. SANFILIPPO: Joe Sanfilippo, Pittsburgh. Did that evaluation include all co-investigators for the testing process, and the so-called reeducation process, or just the PIs in the study?
DR. SCHIFFMAN: Everyone. There were over 40 colposcopists with widely varying ranges of experience, from nurse colposcopist fellows. We did not allow, I do not believe, unless there's something that slipped by, residents to participate. They had to be on the fellow level, experience colpo, nurse colposcopists in places where nurse colposcopists are used, attendings, junior and senior attendings, and they were all part of the same process.
CHAIR NOLLER: Thank you. Does anyone else from the public wish to speak at this time? Seeing no one, we'll proceed to the sponsor presentation.
Now, let me just remind public observers that while this meeting is open for public observation, public attendees may not participate except at the specific request of the panel. The sponsors have been given and agreed to limit their presentation to one hour and fifteen minutes. We hope that you will do so. The first speaker from MediSpectra is Dr. Theresa Wingrove, Senior Director, Regulatory and Clinical Area. Dr. Wingrove?
DR. WINGROVE: Thank you, Dr. Noller, ladies and gentlemen of the panel, FDA staff, and distinguished audience. My name is Theresa Wingrove. I do work at MediSpectra as the Senior Director of Medical Affairs there, and I do own stock in the company. And before I begin the sponsor presentation, I would like to introduce the members of our clinical research team, starting with some of the clinical investigators who will participate in the studies, Dr. Joan Walker from the University of Oklahoma, Drs. Ronny Alvarez and Warner Huh from University of Alabama, Diane Harper from Dartmouth Medical School. Also we have Dr. Mark Stoler from the University of Virginia who served as our study pathologist, Dr. Tom Cox from the University of California, Dr. Tom Wright, Columbia University. We also have Brent Blumenstein and Phil Lavin, who some of you may recognize as they've served as statisticians on many FDA panels. They provided us with a biostatistical report. And finally, Ross Flewelling, our Chief Technology Officer.
This is the outline of our presentation today. I'm going to give a brief regulatory history. Then I will be followed by Dr. Tom Cox, who has been practicing and teaching colposcopy for over 30 years. And he's going to talk about the effectiveness of colposcopy, particularly with respect to ASCUS and LSIL subjects. Dr. Ross Flewelling will give a brief overview on the scientific basis of tissue spectroscopy. Dr. Joan Walker is going to describe how the device is used, and she will give an overview of the two pivotal studies. Dr. Tom Wright, an authority on cervical disease, and is also working for us as a consulting medical director, will provide the results of Pivotal Study I. This agenda is a little out of order. He's going to be followed by Dr. Brent Blumenstein, who is going to speak to the salient points of the statistical analysis of Pivotal Study I. Finally, Dr. Warner Huh, who will present Pivotal Study 2 results. And then finally Dr. Thomas Wright will wrap things up with the risk/benefit analysis of LUMA.
This is an overview of the clinical and regulatory history of LUMA. As Dr. Pollard has indicated, our collaboration with FDA actually goes back to 1998 for our first discussions regarding the product, its possible indications, and various study designs. We submitted technical and optical safety data to FDA in 2000. After review of that data, FDA issued a non-significant risk determination letter, which meant that studies would not require an IDE, though all the studies did comply to good clinical practice, and followed the regulations for non-IDE studies.
From May 2001 to February 2002 we conducted a 753-patient pilot study. The purpose of that pilot study was to create a pathology-based spectral library to create the algorithm. During this time we remained in active discussions with FDA about the optimal study designs that would support premarket approval. We looked at a variety of different study designs, and after much deliberation with our clinical and statistical advisors, the sponsor's decision was to use the gold standard for clinical research trials, a randomized controlled trial that compared LUMA and colposcopy versus colposcopy alone.
In this Pivotal Study I trial, the colposcopist would use their colposcopic judgment and LUMA to direct the biopsy, and the comparator would be the standard colposcopic exam. We felt that this design was the least biased, most easy to interpret, most informative, and most ethical of the designs under consideration. Certainly one negative was the size and the cost of such a study. We enrolled almost 2,300 subjects, though a study of this magnitude allowed us to -- offered us the opportunity to test this device in many institutions and over 50 colposcopists.
While the Pivotal Study I was ongoing we continued our discussions with FDA. FDA's position was that the sponsor quantify the benefit of LUMA in a single arm study design in which the colposcopist went first. In this design, LUMA results would be displayed only after the colposcopist's assessment and commitment to biopsy. MediSpectra agreed to conduct this study as well, and an agreement letter was issued in February 2003 for a multi-center, single-arm study which had a sample size of 576 subjects.
Pivotal Study II began in July 2003. Later that summer we contacted FDA when we observed a lower than expected disease prevalence in the PSI study. This lower prevalence caused us to modify slightly the thresholds, and obviously a lower disease prevalence also meant that a higher sample size for PSII would be necessary. And that sample size was increased by 40 percent to 788 subjects. The study was initiated in July 2003. It was terminated in October of 2003, and you are going to have a separate slide on that question. In June 2004, the PMA was filed using the data from PSI and PSII. FDA granted expedited review status for this application because this represents a breakthrough technology with potential meaningful benefit in the diagnosis of pre-cancerous lesions.
Now I would like to make the following key points with respect to the Pivotal Study II termination. The study was terminated by an independent board of directors for financial reasons on October 16, 2003. At that point the company had expended about $50 million in clinical research, and the remaining financial resources were limited. And in 2003, there was a very unfavorable investment climate, suggesting that future financial challenges lie ahead. So the only alternative was to terminate that study. I would like to make the point that neither the board of directors nor the company were aware of any endpoints at the time of the decision. In fact, most of the data was still in progress. So the analysis of the primary endpoints was done once, and only once, and that was after the study termination decision. The board of directors, however, had seen the results of PSI, and based on the strength of those study results recommended the PMA go forward with both clinical study data.
This is the original indications for use. Its fairly short. It has been revised, and I'd like to move to what is now the proposed LUMA indication for use. And I will read that to you. LUMA is indicated for use as an adjunct to colposcopy in identifying high-grade disease in CIN2/3 patients referred to colposcopy with an ASCUS or LSIL cervical cytology result. LUMA is not intended to replace colposcopy. A thorough colposcopic evaluation with the identification or selection of biopsy sites must be performed independently and prior to viewing the LUMA results.
I'd like to note that there are two major changes in this indication since the original indication. First of all, FDA has asked and we agreed to emphasize and clarify the adjunctive nature of this product. The second change was to limit the indication to ASCUS/LSIL subjects. It was our position that ASCUS/LSIL subjects would be expected to derive the most benefit from this device, and this is where the greatest clinical need was.
Now I would like to introduce Dr. Tom Cox who will elaborate further on colposcopy performance in ASCUS/LSIL patients.
DR. COX: Good morning, I'm Tom Cox, gynecologist at University of California, Santa Barbara. I have been reimbursed for my time and for my travel. I have a very small equity share in stock in this company. And I'd like to thank Dr. Solomon and Schiffman for making my job somewhat easier in terms of bringing attention to the fact that the sensitivity for colposcopy for high-grade disease is not ideal, and that colposcopy is quite subjective. It's for that reason I think that we need an adjunct to colposcopy to help improve its sensitivity and decrease its subjectivity. And I believe the data that you'll hear on the LUMA device today will be convincing that this is an adjunctive procedure that can do these things.
I have to differ somewhat from the previous discussion from the ALTS data regarding taking more biopsies because I think that we as colposcopists in the trial and always have the option of taking more biopsies. And in the trial, the imperfect sensitivity was demonstrated despite the option that colposcopists had to take as many biopsies as they want. I think it's very difficult to change the way we practice colposcopy, but I think having a device that brings our attention to areas that we may have missed colposcopically, and reminds us that there may be other areas to biopsy is likely to be the one thing that might improve sensitivity.
It's obvious that the goal of cervical cancer screening in the United States is to detect CIN2/3 and to treat it. That is how we have lowered the rates of cervical cancer in this country by about 70 percent. And it's our duty when we see an individual with an abnormal Pap, try to make as sure as possible that when they leave our examining room that we have detected all relevant CIN2/3.
Now, in the United States, the majority of high-grade disease is detected in follow-up, at ASCUS and LSIL cytology. That is our most low-grade Pap abnormalities. It is the largest amount detected not because the individual risk for each woman is that great, or certainly not nearly as great as it is for AGUS or HSIL, but it is the preponderance of high-grade disease because of the sheer dominance of these two Pap abnormalities in terms of numbers, because they make up nearly 90 percent of the abnormal Paps that we see in the United States. So it just makes it more difficult to find the high-grade disease when you have a much larger number of women that are normal, but a significant number who have high-grade disease buried within that group.
ALTS recognized this. The study was set up as a randomized controlled trial to evaluate the management of women with ASCUS or LSIL. And the beauty of this trial in particular is it had a 2-year follow-up, which gave us the opportunity to try to determine how much CIN2/3 was still either missed or newly incident to be found over the following two years. And it was determined that initial colposcopy identified only 60 to 70 percent of the cumulative CIN2/3 found over the two years in the ALTS population. And in fact the main ALTS study conclusion was that the imperfect sensitivity after colposcopic referral in all arms is of concern. And there have been a number of Papers sine then that expresses this concern.
If we look at just the patients that were sent directly to colposcopy, that is the arm that was immediate colposcopy for ASCUS and LSIL, we see that initial colposcopy detected CIN2/3 in 14 percent of the women, and over the 2-year follow-up, another third or 7 percent were found to have CIN2/3, for a total of 21 percent, and 79 percent of the referrals, of course, were normal.
I think that it's important to remember a couple of things on this. One is that with 2 to 2.75 million women referred annually in the U.S. to colposcopy for these Pap abnormalities, 7 percent is a significant burden of high-grade disease that needs to somehow be detected in follow-up. And with high levels of loss to follow-up, I think this is a significant risk for some individuals. Secondly, as we look at the LUMA data, we have to remember that the incremental increase in detection of high-grade disease that we could possible achieve is probably limited by this upper limit of 7 percent detection of CIN2/3 that was found over the two years in ALTS.
Other randomized trials have showed similar results. This study by Dr. Alvarez who is here in the audience today was a randomized trial 11 years ago in which 375 women with an SIL Pap were referred either immediately to colposcopy with biopsy or to LEEP. LEEP, of course, ascertains the majority of the cervix and is a truer standard for how much disease is really on the cervix. And you can see that there is approximately a 40 percent deficit in the colpo with biopsy group over the LEEP only group, again indicating the imperfect sensitivity of colposcopy.
So in summary, in terms of missed cases of CIN2/3 we can say that approximately 60 percent of CIN2/3 in the U.S. is found in follow-up with ASCUS and LSIL Pap results, making this the most important group of Pap triage; that one-third of the CIN2/3 is missed at initial colposcopy by experienced colposcopists. And we have to remember, all our trials are always done in centers with high levels of -- colposcopy clinics with high attendance, and that the majority of colposcopists in the United States do not have the opportunity to have this level of experience. So this sets the bar much higher for an adjunctive test, because we presume we're probably catching more in these experienced colposcopy clinics than we do in general colposcopy. And two-thirds of the remaining missed CIN2/3 in the ALTS trial was only detected during follow-up, and many of these only with exit LEEP done very liberally for standards that we might not do in the routine community. So many women certainly must remain at risk in routine care.
And finally, when you look at the data for the LUMA trial, keep in mind that the 7 percent is likely to be the upper limit of possible CIN2 detection so that when you look at the percentage that LUMA provides as increased detection, keep it with that bar in mind. And I'd like to introduce Dr. Ross Flewelling who will talk about technology overview. Thank you.
DR. FLEWELLING: Good morning, I'm Ross Flewelling, the Chief Technical Officer of MediSpectra. I'd like to briefly review the basic science and the mechanism of action underlying our optical detection system. I would like to review the algorithms used in the system, and our non-clinical testing.
The scientific basis of tissue spectroscopy in the LUMA system, it's been well established over a number of years in that there's a great literature, which it's been extensively investigated for more than 20 years in over 200 peer-reviewed publications. There are currently three FDA-approved optical devices that assess in vivo tissue dysplasia in endoscopy and bronchoscopy. The LUMA system uses three combined optical measurements, fluorescence excitation, white light backscatter, and video imaging. The fluorescence excitation is from a 337-nanometer laser that excites native fluorescence in the tissue, which we pick up in the entire visible spectrum, from 320 to 720 nanometers. We also have white light backscatter or reflectance, diffuse reflectance, which we also collect the data in the full spectrum. And we do video imaging. Video imaging is used to provide an image of the system, to align and focus the system, to identify and control for motion, as well as to identify obstructions such as blood, mucus, vaginal wall, and so on.
The system stands 100 millimeters off of the cervix in a non-contact configuration, and the complete scan takes 12 seconds. The scanning probe head is shown on the right. A disposable cover is placed on the probe head, and a bar code allows for unique identification. It's positioned 100 millimeters from the cervix, outside of the speculum. The system is shown on the right. Inside the console are all the light sources. The light is conveyed out through fiber optics to the probe head. Mirrors are used to scan the cervix, collect the light back into the probe head, back through fiber optics, into spectrometers in the base system. On the upper left is shown the cervical scan itself. It's a 25-millimeter diameter area, 499 points at 1-millimeter resolution in a dense hexagonal path array. We do full spectroscopy at each point, and again that takes 12 seconds.
On the left is an image taken with the system in two different areas determined by pathology. In the upper left is a normal squamous area determined by pathology that gives rise to the spectrum on the right, the blue curves, the fluorescence on the top, and the reflectance on the bottom. In the area near the cervical os is CIN2/3 by pathology is shown on the red curves. And for comparison you see, and this is a consistent observation, in normal squamous tissue and fluorescence is significantly higher than the diseased tissue. In reflectance it's the reverse. The CIN2/3 has higher reflectance than normal squamous tissue, and that's consistent with acetowhitening effect.
The biological basis as contained on this slide, the areas in blue, are the ones that we're particularly interested in with our system. In reflectance, or white light backscatter, we're looking at light scattering and hemoglobin. So hemoglobin would be associated with vascularity or angiogenesis, changes associated with dysplasia. Also, cell nuclei and organelles, the size, the crowding, the chromatin content of the cells, lead to higher light scattering, acetowhitening effect, which we pick up in reflectance. In fluorescence, what we see are both functional components and structural components in the tissues. In particular, in our spectra, with 337 nanometer excitation, we believe we're looking at NADPH, a metabolic protein, and collagen, a structural component.
We then take that information and we use a multi-spectral, multivariate integration of those three optical measurements. Again, the video imaging is used for multiple purposes, to align and focus, to control for motion, but we also use it in the processing to identify non-epithelium. We look for the smoke tube, the vaginal wall, fluids, foam, and mucus. And that's identified by video imaging.
We use the fluorescence and reflectance spectroscopy in an integrated algorithm to identify tissue. So we've developed a classification algorithm, but on implementation it is a deterministic system that provides the output shown on the right, which is an image of the cervix where blue areas with the false color overlay, with the blue indicating high-grade disease, or high probability of high-grade disease. This algorithm was established through a pilot clinical study conducted in 2001. It involved colposcopy referral patients, 433 patients with one or more biopsies, an additional 133 patients with LEEPs. Those LEEPs were sectioned into 12 different slices, and the pathology was read along the full surface. So we could correlate detailed pathology to the optical measurements themselves. In total, we had pathology-qualified measurements on over 1,500 individual samples that could be correlated with our optical measurements into five diagnostic categories: CIN2/3, CIN1, metaplasa, normal squamous, and normal columnar, although in the end we are only identifying CIN2/3 as distinct from the other categories. And that was published last year in the Gray Journal.
In establishing the algorithm, as with any diagnostic device there's a tradeoff in sensitivity and specificity for the system. The LUMA algorithm was optimized to provide enhanced sensitivity with comparable specificity compared to colposcopy. The algorithm was then locked. We conducted Pivotal Study I. The algorithm itself consisted of over 70,000 lines of code, and during the two years it took to plan and execute Pivotal Study I we did identify some coding errors during that time. Those were fixed before we started Pivotal Study II. And then we conducted Pivotal Study II. We have reviewed those changes in greater detail with the FDA, and we are in agreement that they did not materially affect the outcome of Pivotal Study I.
And finally, we conducted a battery of tests in non-clinical testing, environmental testing against recognized standards that are shown on this slide. Now both electrical and optical safety against recognized standards also shown here. Now I'd like to introduce Dr. Joan Walker, who was a principal investigator for our studies.
DR. WALKER: Good morning. My name is Joan Walker. I'm a Section Chief of Gynecologic Oncology at the University of Oklahoma. I have no financial ties to MediSpectra except that they've paid for my time and travel for preparation of this meeting.
I worked with Mark and Diane on the ALTS trial from 1995 to 2000 as a clinical center PI, and then after that trial was over began working with MediSpectra on their trials. And I assert my purpose here is to show how wonderful this device is at helping me to become a better colposcopist, identifying alternative sites for biopsy in hopes that I can detect that disease that I was previously missing and hopefully won't have to watch that lady die of cancer 10 years afterwards.
I'm going to go over the LUMA system and go over the clinical trials briefly. The clinical use of this LUMA device shown here is going to be shown in a video. That video was shortened considerably so you won't have to sit through 15 or 30 minutes of colposcopy. I'll demonstrate the system setup, how the patient is prepared, how the scanning steps are integrated with the colposcopic examination, and then how those LUMA results are visualized.
The LUMA cervical imaging system is a self-contained mobile unit. A picture of it is going to be shown here. That device has a scanning probe which is on an articulating arm which allows you to move that probe into position to visualize the cervix. The probe is stored in this calibration port, and it also is calibrated for each individual patient. The patient information is placed in the system with the keyboard. And this is a completely hands-free device with foot pedals which can control the device. Each patient has a disposable probe cover which is applied to the probe, and then it is calibrated before each exam.
The patient is then prepared as a usual colposcopic exam. Speculum is placed in the vagina, the cervix is bathed in vinegar, the vinegar timing device is initiated on the instrument by pushing the foot pedal, in order to make sure that enough time has elapsed to allow the development of that acetic acid picture. Then the probe is visualizing the cervix, and then it is aligned with some special alignment dots. They have three alignment dots which have to be placed inside of the yellow focus circles in order to get the appropriate image, and then if the acetic acid timer has passed the 30-second mark but before the 120-second mark, the probe is asked to take the scan, and then the probe is pushed aside, and the usual colposcopic examination proceeds as you would ordinarily do. So the patient's cervix may be reapplied with vinegar, the examination can take as long as it needs to in order to find the appropriate places to biopsy. The colposcopist then biopsies as many times as is necessary in order to do the usual colposcopic examination.
Then they visualize the scanning results. And this is an image that's just without the scan on top. And then when you ask the machine to show you the scan, you can see blue highlighted areas that you may not have noticed at your initial colposcopic examination. And so then you can direct additional biopsies on top of the ones you already selected based on how intense this blue light is showing. So you might choose to biopsy here, and you might choose to biopsy here. You also can flip this on and off with the foot pedal in order to identify anatomic landmarks, in order to get the biopsy placed in the proper location on the cervix using the colposcope.
I hope that I have demonstrated how easy this instrument is to use. It is a non-contact device. It's easily integrated into your everyday practice. It has a very simple-to-use clinical interface, and there's a very straightforward interpretation of displaying these results.
I am going to now briefly go over the studies that we've conducted. There were two separate studies performed, 2,500 and 2,600 patients were enrolled. There were two completely different study designs which ended up with similar results. They were conducted over 15 months in 13 different practice center using 52 colposcopists. There were local full IRB approvals, and this was a non-significant risk study, and was IDE exempt. The inclusion criteria were age greater than 18 years or age of consent, the patients were referred in order to detect CIN2/3 in patients that had abnormal Pap smears, ASCUS, LSIL, HSIL, or even cancer. The exclusion criteria were pregnancy, heavy menstrual flow, or a biopsy if it had been performed since the last Pap smear had been performed. The Pap smear must have been at least seven days old but no greater than six months old. And this is the sites of all of those study centers, showing that it is very representative of the United States population. And these are the PIs at each of those sites, many of which you may be familiar with. The colposcopic characteristics that doctors -- there were 46 OB/GYN physicians, 15 of which were GYN oncologists, many of which were also on the ALTS trial, three family practice physicians, three nurse practitioners. All investigators were required to have experience and training in colposcopy.
The pathology material was blinded. It was sent to two pathologists who had to adjudicate their slide review. If they disagreed, a third pathologist was brought in. The diagnostic categories of each biopsy was no evidence of disease, CIN1, CIN2/3, or cancer. But the patients were categorized as either having CIN2/3 or not. These are the pathologists who visualized these specimens.
I will now introduce Dr. Wright. Thank you.
DR. WRIGHT: Good morning, I'm Tom Wright. I'm the Medical Director of MediSpectra. I have a financial interest with stock in the company.
What I want to present are the results of the Pivotal Study I. This study was a dual arm randomized trial. All patients entering into this study were randomized to receive either -- the cervix was washed with acetic acid. They then performed a LUMA scan. The patients were then randomized to either have that scan blinded, or to have the scan un-blinded. In the arm in which we had a blinded scan, the clinicians performed routine colposcopy as they would normally do in practice. In the arm in which was LUMA with colposcopy, the clinician saw the scanned results. At the time, they selected sites for biopsy. So they performed routine colposcopy while looking at the sites, and would take additional biopsies as needed.
Patients were referred to colposcopy because of an abnormal Pap result. This included ASCUS, LSIL, or HSIL. It's important to note that we began the study in 2002. Not all sites were separating ASC-H away from ASCUS, and therefore we have a mixture, with some patients listed as ASCUS only, and you have some ASC-H's also. This was prior to the widespread adoption of HPV-DNA testing, so we do not have HPV testing results on these patients. Randomization was done by entry Pap to ensure balance between the arms. Study endpoints were true positive. That was a patient with at least one biopsy diagnosis of CIN2/3. True positive rate is directly proportional to sensitivity and false positives, which are patients with all negative, no CIN2/3 biopsies. So a patient with CIN1 would be considered to be a false positive.
Our study hypothesis for true positive rate was that the true positive rate is greater for LUMA with colposcopy than for colposcopy alone. Therefore, the lower bound of the 95 percent confidence interval needs to be greater than zero to meet the study hypothesis. Our study hypothesis for false positive rate is that the false positive rate for LUMA with colposcopy is no more than 8 percent greater than for colposcopy alone. Therefore, the upper limit of the 95 confidence interval for the difference between the arms needs to be no more than 8 percent.
We also determined that we needed to perform joint consideration of true positive and false positive rates. This was stipulated in the initial protocol, although we did not specify the method. True positive and false positive rates will be simultaneously considered to determine that the incremental gain of true positives merits a change in the false positives. So to look at the tradeoffs.
We enrolled 2,299 patients. This was a very large study. Only 113 were excluded. The major reason for exclusion was incomplete pathology. Leaving us with 1,096 patients in the colpo arm, and 1,090 in the LUMA arm. When we look at patient characteristics across arms you can see that the two arms are well balanced. Identical median ages. We have a strong minority representation in this trial, and balancing of Paps as to ASCUS, LSIL, and HSIL across the two arms.
This slide shows true positive outcomes by arm, where are comparing the true positives in the LUMA plus colposcopy versus colposcopy alone. And if we look overall at the difference, we had a 1.9 percent increased true positive rate in the LUMA plus colposcopy compared to colposcopy alone. This was not as high as we expected. The reason we do not believe it was as high as expected is when you look in the HSIL arm, what you see is no increase in true positives among the LUMA patients. When you look, however, at the ASCUS/LSIL arm, what you see is that we have a 3 percent increase in true positives in the LUMA plus colposcopy arm compared to colposcopy alone.
We recognize that combining the ASCUS/LSIL strata is in fact a post hoc consideration. Having said that, there is a very strong clinical reason for combining patients with ASCUS and LSIL. First, we know that the disease prevalence among patients with ASCUS and patients with LSIL is quite similar. They were almost identical in this study, the TP, true positive, rates. These are markedly different than what we see in patients with HSIL. In addition, as you've already heard from Dr. Cox, the ASCUS/LSIL population is a predominant population which is seen at colposcopy. It comprises approximately 80 percent of all patients referred to colposcopy. And the clinical management guidelines are quote different between patients with ASCUS and LSIL, and patients with HSIL. Patients with HSIL tend to get a LEEP irrespective of what is found at colposcopy, whereas patients with ASCUS and LSIL do not.
I've shown you already the true positives in terms of absolute difference. What I'd like to now do is focus on true positives in terms of relative gain. When we look at the ASCUS/LSIL category, there were 99 women with biopsy-confirmed CIN2/3 in the colposcopy-only arm, and there were 126 women, 27 additional cases of CIN2/3 identified in the LUMA plus colposcopy arm, giving you an absolute gain of 3 percent, and a 26.5 percent relative gain. If you remember back to the discussions of ALTS, what was shown though, approximately one-third of the high-grade cases were missed at the initial colposcopy. Here we are seeing a 26.5 percent relative gain in detection of true positives.
This slide plots the standard error bars, looking for the Pivotal Study I true positive outcomes overall and in the ASCUS/LSIL group. And what you can see is that in the ASCUS/LSIL group, the 3 percent absolute gain does not quite meet statistical significance. It is very close, but it just simply does not quite meet statistical significance. When we look at Pivotal Study I in terms of false positive outcomes, overall we meet the study hypothesis. However, when we look at the ASCUS/LSIL group, again, we do not meet statistical significance, but it is very close again.
Another way to look at false positives is to look at the mean biopsies which are taken with and without LUMA. When we look either overall or in the ASCUS/LSIL strata and we look at the differences in the mean biopsy rates between the two arms, what we find is only a 0.27 or a 0.31 increase in the number of mean biopsies per patient. So the number of biopsies in the LUMA arm is greater in the ASCUS/LSIL strata by only 0.3 biopsies per patient.
When we look at positive predictive value, which is another way of looking at the efficiency of biopsies, what we see is that in every single case, in the LUMA and colposcopy arm, we ask the clinicians to designate whether or not an individual biopsy was taken on the basis of the LUMA result, or on the basis of the colposcopy result, or on the basis of both. When we look at the positive predictive value of biopsies which were taken because they were only colposcopy positive, we found that the positive predictive value is 10.8 percent. When you look at the biopsies which were taken only on the basis of the LUMA result, the colposcopy was not considered to be abnormal, they had a positive predictive value of 8.6 percent. This is greater than what has been seen in studies which have looked at random biopsies.
In addition, and very importantly when we talk about the benefits of using LUMA, when we look at the positive predictive value of biopsies that were both LUMA-positive as well as colposcopy-positive, they went from 10.8 percent for colposcopy only to 18.1 percent for those biopsies which were also LUMA-positive. So a significant increase in the positive predictive value of biopsies which are taken on the basis of also being LUMA-positive.
I'd now like to walk you through a couple of cases, because for clinicians I think cases are really important to see what we're talking about. Here you can see the colposcopic image in the patient referred for ASCUS. You can see there's some very faint acetowhitening on this case. Very small areas of acetowhitening. The colposcopist did not think these were significant, and did not feel that they needed to take a biopsy. Here you can see the LUMA overlay, and you can see areas of blue present in the 7 o'clock position, and a biopsy taken at 7 o'clock turned out to be CIN2/3. The clinician did not feel they needed to take a biopsy.
When we look at another case, this is a case with a reasonably broad area of acetowhitening present around much of the cervix. Again, this was a patient with an ASCUS Pap smear. The colposcopist -- and remember, these are all trained colposcopists. Many of these were colposcopists who participated in the ALTS trial. They felt there was no need for a colposcopically directed biopsy in this patient. However, when you look at the LUMA overlay, you can see that LUMA identified a number of areas of dark blue in this patient, indicating high probability for a disease, and a LUMA-directed biopsy taken at 10 o'clock turned out to be CIN2/3.
I'd now like to turn to false negatives. Remember that the LUMA algorithm was tuned in order to try and make it equivalent to colposcopy in terms of sensitivity and specificity. And we do not expect that every true positive will be positive using the LUMA device. Just like colposcopy, there are going to be cases which are not positive. There were 126 true positives in the ASCUS/LSIL arm in the LUMA -- in the ASCUS/LSIL strata, in the LUMA plus colposcopy arm, and 26 of those 126 were colposcopy-positive but were LUMA-negative. So 20 percent of them. You need to recognize, however, that there were also 15 cases which were LUMA-positive but which were colposcopy-negative. 11.9 percent. And it's because of this that we feel that LUMA's role is as an adjunct to colposcopy. The role of LUMA is not to replace colposcopy.
In terms of safety, I think it's critical that we recognize there were absolutely no device-related effects out of 2,299 patients. We had four patients with minor adverse events. These are events typical of what we see in any colposcopy which can occur.
We preformed additional analyses, looking at learning curve, cervicitis. No significant effect of those were seen on LUMA performance. We've also done random biopsy comparison. We've looked at colposcopists and site variability, joint analyses, and age.
I'd like to focus first on age. The FDA and MediSpectra performed post hoc analyses to evaluate any possible age effects. The FDA performed a Bayesian analysis that identified a possible age effect in PSI. This was very sensitive to the particular age cut, and it was the only analysis which it found any age effect for LUMA. MediSpectra has analyzed the data using multiple methodologies, and did not identify any age effect, including looking at age as a continuous variable. And important to recognize, in PSII, which you are going to hear about in a minute, there was no age effect which was seen.
The reason that I think we have to be cautious about evaluating the effect of age is that here you can see when you make the cut at less than 21, it appears to be a much greater impact of LUMA than at the other ages. However, when you look at the distribution of patients in our study, what you find is that it's very skewed, and there are a lot of patients between the ages of 18 and 24 who are enrolled into this study. If instead of making the somewhat arbitrary cut at age 21, instead we divide the population into the true tertiles in terms of the population and break it at 23 and at 33 into true tertiles, what we see is that their LUMA has an increased TP rate across all three age groups.
Another way of looking at this is to actually remove patients under the age of 23 and only look at the patients who are over the age of 23. And what we find is that the true positive, the increase in true positives is 3 percent among women over the age of 23, which is identical to what we see with the overall population, as well as a group of women under the age of 23.
Finally with respect to age, we've looked at this in terms of true positive rate. We recognize that this is not adjusted for prevalence of disease, but you can see that there is a similar true positive rate in the LUMA-only increment across all ages. So in summary with respect to age, we believe that LUMA has shown the ability to identify true positives across a wide range of ages. Multiple statistical tests performed by the sponsor have not found an age effect, and no age effect is seen in PSII.
Colposcopist/site variability. Analysis supports pooling for true positive outcome across sites and colposcopy. The sponsor has assessed confidence intervals using fixed and random effects, and it appears that random effects confidence intervals are slightly wider.
Joint considerations. We pre-planned to perform joint consideration of true positive and false positive increase in order to weigh tradeoffs. These considerations take into account the fact that the three possible outcomes of true positive, false positives, and no biopsies are interrelated, not independent. And what we find in these joint consideration analyses is that a statistically significant difference between arms for both the overall and the ASCUS/LSIL group.
So to conclude, our true positive rate increases in the ASCUS/LSIL strata are 3 percent, which corresponds to 26.5 percent relative gain. Our false positive rate in the ASCUS/LSIL strata is moderate. It is only 3 percent. And we see with the use of this device an increase in positive predictive value from 10.8 percent for colposcopy-only biopsies to 18.1 percent for biopsies which are identified by both LUMA and by colposcopy, indicating that this device is identifying sites of high-grade disease. So in overall conclusions, we found a substantial increase in the relative number of true positives. LUMA clearly identifies cases of CIN2/3 which are missed by colposcopy. You've seen two of those. LUMA improves positive predictive value compared to colposcopy. There are absolutely no safety concerns, and only a minor increase in false positives. And we believe that Pivotal Study I shows that the risk/benefit profile of LUMA is compelling.
We are now going to go into further depth on the joint inference testing. Dr. Brent Blumenstein.
DR. BLUMENSTEIN: I'm a consultant to MediSpectra, but I do not own stock. So I'm going to talk a little bit about an analysis that we did that focused on the outcome for a patient as a trichotomy; that is, the patient is either considered to be true positive, false positive, these two being a result of having had a biopsy, or no biopsy was taken for the patient.
The original protocol specified separate analyses for TP and FP. And all of the results that you've seen up until now do those separate analyses. But the protocol also talked about the idea of doing joint analyses. The trichotomist analysis that I'm going to show you now accounts for the outcome dependencies between these three outcomes. The confidence intervals and other things like that that you've seen up until now do not take into account those outcome dependencies, and are therefore broader than they need to be because they are insensitive to the value of the other parameter being discussed. The joint inference is true to the original objectives of the protocol, and that is to assess the LUMA benefit relative to an expected increase in the number of biopsies.
The data structure in the trichotomist analysis is just a 2x3 frequency table. There are two arms by three possible outcomes. The sum of these three frequencies to the total for that row is the thing that induces the dependency of these outcomes. The samples for the arms, M, the N patients in the colpo only arm and the N patients in this arm, are independent, but the samples this ways are not independent. The endpoint dependency is induced by the sum requirement.
The 2x3 table can be analyzed simply as a contingency table using chi-square or exact tests. We use a more sophisticated modeling methodology called the Grizzle-Starmer-Koch method for things like addressing interactions, and so forth. This is the 2x3 table for all strata in PSI. And this is the trichotomist analysis of this. The Cochran-Mantel-Haenszel test, which is a stratified test in this table, gives a p-value of 0.0112, and an exact test that's not stratified gives a p-value of 0.0132. Both of these are significant, and we prefer to look at these as gatekeepers for subsequent analyses, although these do establish a result in PSI that's not specific to any kind of outcome, but rather tests whether there's homogeneity in the probabilities within the arms. And this is part of the purpose of this analysis is to assess the between-arm difference, and then to focus on what those between-arm differences mean.
Now this is the detailed analyses that were done following this initial analysis. This first row just repeats the previous slide. These are the p-values that you saw for the overall test. At that point after we have done this we wanted to find out if there was an arm-by-strata interaction. That is, testing whether the arm effect that is -- the arm differences are uniform across the strata. We find that we had a p-value of 0.0175 here, indicating that there are significant arm-by-strata interactions. Now you've already heard about the clinical rationale for treating the HSIL patients differently, and this provides the statistical rationale for that.
So we chose to drop the HSIL data, and focus on the remaining data, that is the ASCUS and LSIL. The first step in that was to assess whether there was an arm-by-strata interaction within the remaining data. And the p-value for that is 0.1737, indicating that there is no evidence of an arm-by-strata interaction. Finally, the p-values for the remaining data, without HSIL, the primary analysis of between-arm differences are given here. Now these p-values are measured against a standard of 0.00714, which is 0.05 divided by 7. This correction in the significance level to be used is made because we have now re-stratified the data in order to focus on the relevant strata.
The consequences of this joint test is that we have between-arm differences, and now we have to characterize these between-arm differences with respect to measures of effect. We can choose two measures of effect. You've seen previously the TP rate difference, and the FP rate difference as measures of effect. The joint p-value is not dependent on our choice of measures of effect, however. So we can choose different measures of effect to understand the between-arm differences. One of the things I'm going to show you in a minute is an odds ratio. This is something statisticians like but other people don't like so much, but there's a particular insight that I want to communicate to you about that.
The joint tests say that one or both measures of effect between the arms differ. It's important now to assess the direction and magnitude. Now this is a contour plot. On the horizontal axis is the FP difference. On the vertical axis is the TP difference. And if you can think of coming out of this screen are the probability of every possible table that is in that 2x3 contingency table, the probability of every possible table comes out from this. And these contours show the boundary such that everything outside of this inner boundary here has a probability of 0.05. Everything outside of this outer circle has a probability of 0.00714. This is the adjusted criteria for significance. The result of this trial is right here. This is the 3 percent TP difference and the 4 percent FP difference. And you can see that it is outside of the boundary, the 0.00714 boundary, and that's why it's significant. You can also see that this outcome for this trial is in the upper right quadrant, which is what you would expect for an adjuvant therapy. It's not over here, it's not down here, it's not down there. It's here, which means that the TP difference is positive, favoring LUMA. The FP difference is as expected, higher than zero, but it doesn't approach the 0.08 criterion that we had in the protocol as being undesirable. So this is the representation of the outcome of the trial with respect to TP difference and FP difference. Two measures that are imposed on the 2x3 table that was analyzed.
We an choose two other measures. Here are some odds ratios. Now, I know that most of you are going to have a little trouble with this, but I'll -- bear with me. The TP odds is basically related to the TP rate difference, and it's just the ratio of the odds of having a true positive between the arms. And this is 1.31, which is in favor of the LUMA arm, a number greater than one favors LUMA. Now, the interesting one is this. This is the TP to FP odds, basically the ratio of the TP outcome within an arm to the FP outcome in an arm. And it represents the efficiency with which the biopsies taken yield a TP result. Now, if we take the ratio of those odds, we get 1.19. This is a relative efficiency measure of the biopsies taken. And the fact that this relative efficiency measure is greater than one is an indication that the biopsies taken in the TP arm are more efficient at identifying true positives than are the biopsies taken in the colpo only All right. So these measures are just alternative ways of viewing the outcome from the trial after we have tested for between-arm differences.
So, the joint inference provides strong evidence that there's a significant difference between the arms. The methods that we used here matches the actual data. While it wasn't pre-specified exactly in the protocol, it does match the actual data, it's more rigorous than the methods that we previously presented, and it's true to the original objectives of the protocol. So the end, finally, depending on which measure we use, the direction and magnitude are consistent with LUMA patient benefit.
Finally, we also see from the joint analysis that we have a significant result both for overall, that is including all strata, and after we have eliminated the HSIL strata. Thank you. Now Warner Huh will present the PSII data.
DR. HUH: My name is Warner Huh. I'm a GYN oncologist at the University of Alabama at Birmingham. I served as the PI for Pivotal Study II at our institution. I also was a study colposcopist for the ALTS trial at the UAB site. I've been reimbursed for my time, for preparation, and travel to this meeting, and I do not have financial interests or stock in MediSpectra.
Pivotal Study II is different from Pivotal Study I in that it's a single-arm internally controlled design trial. I think it's important for the panel to know that each patient served as their own control in this study. So after subjects were consented, the cervix was washed and acetic acid was applied, a LUMA scan was performed, but blinded to the investigator. At this point, standard-of-care colposcopy was performed with biopsy commitments, and these biopsy commitments were annotated on the screen, and locked into place, and could not be changed later on. And at this point, the scan was un-blinded to the investigator, and this committed to take biopsies at LUMA-indicated sites. These were also locked into the computer and could not be changed later. All biopsy sites were revealed. They were asked to take all biopsies, and these were sent off to the reference pathologist.
I'd like to briefly just kind of go over the annotation sites and how this worked in Pivotal Study II. Again, this is an image of the cervix after acetic acid had been applied. At this point, a blinded LUMA scan has been performed, and a colposcopy is performed at this point. In the yellow circle was an area that was deemed to be clinically or colposcopically significant to the investigator, and this was locked into place and annotated. At this point, the scan was revealed to the investigator, and asked to take a LUMA-indicated biopsy, and note this on the screen as well. And this is marked by the green circle. And again, this was locked into place and could not be changed later on.
All biopsy sites were revealed. Investigator was asked to take a biopsy at the yellow circle, and the green circle, and these biopsies were sent off to the pathologist. And we'll review this case later on in the presentation.
In terms of subjects and controls, all patients were referred to colposcopy for an abnormal Pap smear, including those patients with ASCUS, LSIL, and HSIL. And as previously discussed by Dr. Wright, these ASCUS patients include those patients with ASC-H and ASCUS under the new terminology. And again, each patient served as her own control in the study, unlike PSI.
In terms of the study endpoint definitions. The LUMA true positive rate is defined as follows. Incremental true positive patients are colposcopy patients without CIN2/3 as a percent of the total population. For the LUMA false positive rates as follows. The incremental false positive patients are among colposcopy no-biopsy patients, again as a percent of the total population. And these constituted the study endpoints.
The following is a diagram breaking down the true positive definitions. After enrollment, all patients underwent colposcopy, and these patients were split into those patients who were colposcopy-positive, those patients who identified with CIN2/3, and colposcopy-negative. And this includes those patients without biopsies, and those patients without CIN2/3. The LUMA increment is based on the colposcopy-negative patients only, as shown in this diagram, and are broken into LUMA-positive and LUMA-negative patients. Again, LUMA-positive patients are those patients who are identified with additional cases of CIN2/3.
The study hypothesis are as follows. The LUMA incremental true positive rate was set at least 2 percent for the overall population. In other words, the lower 95 percent constant was greater than or equal to 2 percent. For the LUMA incremental false positive rate, this was set at less than 15 percent for the overall population. In other words, the upper 95 percent constant was set at less than 15 percent for this study.
In terms of the patients enrolled and excluded, there were 227 total subjects that were enrolled in this study. Fifteen percent, or 34 subjects, were ultimately excluded, with the most common reason being incomplete pathology. I think it's important for the panel to note that there was a lot more stringency in terms of the pathology, and if the pathology was incomplete for any reason that subject was ultimately excluded from the final analysis. And in terms of the patients analyzed, the overall majority of these patients had ASCUS and LSIL, with the remaining 14 percent of patients having HSIL.
In terms of the patient characteristics, the mean age for this study was 28.5 years. And in terms of race breakdown, we had an excellent distribution in terms of minorities and other racial breakdown. This is similar to that seen in Pivotal Study I. And again, in the Pap smear distribution, the overwhelming majority of patients had ASCUS and LSIL. It is important to note, however, in the HSIL group we had fewer HSIL patients than that seen in Pivotal Study I.
Reviewing this slide again, I think it's important that the panel note again, the LUMA increment is based on only those patients who are colposcopy-negative. We did not include the colposcopy-positive patients. And again, it's broken down into LUMA-positive and LUMA-negative. To go over the numbers briefly, there were 193 subjects that were enrolled in this study. Again, they all underwent colposcopy. Forty-one patients had a positive colposcopy for CIN2/3. The remaining 152 subjects either had no biopsies or no evidence of CIN2/3. There were nine additional patients with CIN2/3 on the LUMA increment, and 143 patients were LUMA-negative.
When you use the protocol definition of a denominator of 193 subjects, the LUMA increment rate was 4.7 percent for this study. So in terms of overall true positive outcomes based on the original protocol definition, there was a 4.7 percent LUMA increment compared to that of 21.2 percent for colposcopy. And our overall study hypothesis was net greater than 2 percent. Again, the p-value for this was 0.0164.
Even though the original protocol definition was set as the denominator of 193 subjects, or all subjects involved, we actually felt that a more perfect definition would be those patients who would be eligible. In other words, those patients who were colposcopy-negative, not the colposcopy-positive patients. And when you use 152 for the denominator, the eligible population rate for true positives in the LUMA increment was 5.9 percent.
Again, we feel that the eligible population is probably the more appropriate definition, but irrespective of which definition you use, both the total population and the eligible population met its study hypothesis, and the following is a depiction of the standard error bars for both groups.
I'd like to review some case studies as Dr. Wright had done for Pivotal Study I. This was reviewed earlier in the presentation. Again, this is a patient who was referred to the study with an ASCUS Pap smear, and underwent colposcopy and a blinded LUMA scan. Again, there's an area that was felt to be clinically significant right here by the colposcopist, and as such had annotated a colposcopically interesting biopsy site with the yellow circle. But if you note, an additional LUMA-indicated biopsy site was taken from the blue spot, which is directly adjacent but does not overlap the yellow circle. I think in reality, this probably represents one entire lesion, and two separate biopsies of that lesion. And of note, at the yellow site, the colposcopy site there was no evidence of disease, and at the green circle there was evidence of CIN2/3.
The following is another example. This patient was referred to the colposcopy clinic with an HSIL Pap smear, and again, underwent a blinded LUMA scan and standard care colposcopy. And in this example, at approximately 11 o'clock, the colposcopist felt that this was a clinically interesting or colposcopically significant lesion that merited biopsy. But of note on the complete opposite side of the cervix, perhaps an area that the colposcopist would not have picked up, was a LUMA-indicated biopsy site. And in the yellow circle, again, this was no evidence of disease, and in the green circle was consistent with CIN2/3.
MediSpectra and the FDA had agreed to adjust the overall true positive null hypothesis based on the prevalence of disease from 2.5 percent to 2.0 percent. This was based on interim data from Pivotal Study I. In that sense, the sponsor also applied the prevalence adjustments to endpoints across study strata in proportion to the study prevalence assumptions. And to review again, the overall was set at 2.0 percent, HSIL was set at 4.0 percent, and ASCUS and LSIL was set at 1.5 percent.
When you actually review the true positive outcomes for Pivotal Study II, both for overall and the ASCUS/LSIL strata, again as previously mentioned, we met our study hypothesis for overall group either when you use the total population definition or the eligible definition. For ASCUS and LSIL, we met our study hypothesis when you use the eligible population definition with a set rate of 1.5 percent. This was not met for the total population definition.
Again, the following is a depiction of the standard error bars for the total population and the eligible population. And to reiterate, again for the eligible population we met our study hypothesis.
The following is an example or a review of the relative percent in gains for the two different groups and for the strata involved. Again, for the total population under original protocol definition for the overall group, we demonstrated a 22 percent relative percent gain in the detection of CIN2/3. The ASCUS and LSIL strata we demonstrated a 25 percent relative gain. For the eligible population of device performance, for overall ASCUS/LSIL we demonstrated 27.8 percent and a 29.2 percent relative gain respectively.
In terms of false positive outcomes, when you actually review the overall group, including the ASCUS/LSIL and the HSIL strata, we did not meet our false positive hypothesis, and for the overall group was 18.1 percent.
In terms of mean biopsy rates, overall roughly one additional biopsy was taken in this study. And this is also seen with the ASCUS, LSIL, and HSIL strata. And I think it's important for the panel to recognize that this is largely a function of the study design. What I mean by this is that the investigators were mandated and asked to take an additional biopsy at a LUMA-indicated site. This was actually specifically in the protocol as such, and as such would probably explain why there's an additional biopsy in this study.
In terms of safety, and I'd like to underscore this, and this cannot be overstated, there were no device-related effects seen in the 227 subjects that were involved in the study, similar to that seen in Pivotal Study I. There were two patients with minor adverse events, but these are commonly seen in patients who are undergoing the standard-of-care colposcopy.
So in conclusion, LUMA identified significantly more CIN2/3, the false positive rate increase was expected, largely given to the study design, but was not felt to be clinically excessive. Again, there were absolutely no safety issues related to this device, and PSII confirms the highly favorable risk/benefit ratio of LUMA as previously seen in Pivotal Study I.
I'd like to turn to Dr. Wright again for the conclusion of risk/benefits of LUMA.
DR. WRIGHT: I'm going to try and put this all into perspective, and try and show you. We've got two trials. Multiple studies, I think, have shown that a single colposcopy misses 20 to 36 percent of CIN2/3. So there's clearly a clinical problem. Multiple studies have also shown the potential for increasing the detection of CIN2/3 using spectroscopic methods. So the objective of both of our studies was to measure the adjunctive benefit of adding LUMA to a standard-of-care colposcopic exam. So it's only looking at adjunctive benefit.
When you look at the two studies combined, we have enrolled 2,526 patients, 52 colposcopists. These studies had to be large because when you remember to the ALTS trial, only about 7 percent of patients referred with ASCUS/LSIL have missed high-grade disease. So we had to do large studies. However, the overall size of these studies, their ethnic and racial mix, and their geographic distribution clearly assures that device performance has been evaluated in an intended-use population. And I really want to stress there were no device-related adverse events in either study. There are no safety issues that we have identified with the use of this device.
When you look at the two studies together, they show a remarkable consistency in outcomes. When you look at the absolute TP, true positive rate gain, patients with CIN2/3 who were identified using LUMA in Pivotal Study I it's 3 percent, and depending on which definition you use it's 3.6 percent to 4.2 percent in Pivotal Study II. Two studies with completely different study designs, and we find exactly the same result. When you look at the relative true positive rate gain, you find 26.5 percent versus 25 or 29 percent. Again, exactly the same relative gain obtained with the use of this device.
We also have shown that the positive predictive value of biopsies, which are obtained based on colposcopy with LUMA are statistically greater than what we find with those obtained by colposcopy alone. And you saw in the joint inferences talk that these biopsies are more efficient when taken with the use of LUMA.
When you look at the statistical outcomes, and you look at Pivotal Study I, the true positive outcomes, what you find is that the 3 percent absolute gain is almost statistically significant. It isn't quite there, but it's almost. When you look at Pivotal Study II in terms of the true positive gains for the population overall, using either definition of measuring device performance, total population or eligible population, we are significant. When you look at Pivotal Study II, looking at the ASCUS/LSIL population, what you find is that in the eligible population, which we feel is the appropriate population, we are statistically significant in terms of our true positive outcome.
When we turn to false positive outcomes, what we find is overall in PSI we hit the significance. But when we focus on the ASCUS/LSIL group, again, we are almost statistically significant. The bar just crosses 8. When we look at the study hypothesis results for Pivotal Study I, and I want to focus on the ASCUS/LSIL population, what you find is that in terms of the study hypothesis for true positives, we do not meet statistical significance, but again we are very close. When you look at false positives, we do not meet it, but again we are very close. And when you use joint inference testing, we do meet statistical significance for Pivotal Study I, both overall as well as in the ASCUS/LSIL population, which is our intended-use population.
When we look at the same table for PSII, and we look overall, and in ASCUS/LSIL, we meet our statistical hypotheses for true positives, but we do not for false positives. We believe this is due to the instructions to the clinicians to take a biopsy whenever possible. We ended up with one additional biopsy per patient, but only one additional biopsy per patient.
So when you take the two studies together, I think you can clearly take home that both studies found a 3 to 4 percent absolute gain, which translates to a 25 percent relative gain in the detection of CIN2/3, in the ASCUS/LSIL stratum, and both found clinically acceptable rates of false positives and false negatives, given the adjunctive use of this device.
I want to take you back to the slide that Tom Cox showed you for the ALTS results. Looking at patients with ASCUS/LSIL in the immediate colposcopy arm, where there was 21 percent high-grade disease in this arm, 14 percent of it was identified at the first colposcopy, 7 percent was identified on follow-up. We do not know the precise prevalence of CIN2/3 in our population, but when you look at what we found in Pivotal Study II, we found 14 percent CIN2/3 identified by the initial colposcopy, and LUMA identified 4 percent more, suggesting that we are detecting about half of the cases which are missed with regular colposcopy.
So multiple studies have shown that a single colposcopic exam misses 20 to 36 percent of CIN2/3. Adjunctive use of LUMA identifies 25 percent more cases of CIN2/3 than colposcopy alone. And it's important to recognize that this identifies these patients here and now, at the time of the examination, so hopefully we will have lower rates of loss to follow-up of patients with true cancer precursors. And therefore we believe that LUMA will provide an important clinical benefit to women with ASCUS/LSIL cytology results. Thank you very much.
CHAIR NOLLER: Thank you. First time in the history of the panel I suspect the sponsor finished a little early.
CHAIR NOLLER: Seventy-five seconds early. Thank you for keeping to time. We are running just a few minutes ahead of schedule, and we don't want to discuss the product at this point, but if there are points that need clarification, if panel members have a question or two to address to the sponsor. And I will open it.
It's not entirely clear to me, if you look at your Slide 97, patient disposition summary, you enrolled 227 individuals, you excluded 34 for various reasons, and analyzed 193. But then you did an eligible patient calculation that lost 41 individuals, and it wasn't entirely clear to me why you excluded those 41 for some of the analyses, the 152 eligible participants. Who were those that were dropped? Who would to address that? Please state your name as you speak.
DR. WINGROVE: Theresa Wingrove from MediSpectra. Could you clarify again, the question was?
CHAIR NOLLER: Well, if you look at your Slide 97, you have 193 patients analyzed, but then later you talk about eligible population that is 152 patients. I just want to know who those -- when you're talking about eligible population.
DR. WINGROVE: Right, that's actually the wrong slide. Can you pull up the box plot? The total for protocol population is 193 subjects after exclusions. The number that you are referring to I believe, when that slide is up, represents the number of subjects who were not found to be CIN2/3 on colposcopy. Okay, so those are -- actually the next slide -- those are your colpo-negative subjects.
CHAIR NOLLER: Thank you.
DR. WINGROVE: That is the pool of subjects for which --
CHAIR NOLLER: Thank you, that's all I needed to know. Are there other clarifying questions? Yes, again, give your last name before you speak.
DR. D'AGOSTINO: D'Agostino. I'm just curious in terms of the interpretation, and it's going to be something that we're going to have to grapple with, and I'd like to hear what the sponsor has to say. Isn't another interpretation of the data that in the Pivotal Study I the false positive failed, so you have a negative study. Excuse me, the true positive failed so you have a negative study, and that moving on to the joint analysis doesn't really necessarily clarify the issue. It's a post hoc analysis which has major interpretation problems, and one of the reasons we don't tend to see this too often in practice is because after you see a significant result it's very hard to interpret why the significant results in the joint.
So anyway, you have a negative because of the true positive. In the false positive we have a result that's in the favor of the company, but there's an 8 percent margin, so you don't produce more than 8 percent false positives. And where did the 8 percent come from?
Another question in terms of this -- again, I'm trying to grapple with how I'm going to interpret the data. This is a large study, and with a large study you tend to say that I'm going to see significance, and you didn't see significance. So I'm carried away, and I mean this sincerely, in Pivotal Study I, that I have a negative study with a large sample size that I should have somehow rather seen things work out well and they didn't. And the presentation of all the rates are all based on -- or less of non-significance. With the second pivotal study, I see some sense coming out in terms of what you're saying if I'm willing to shift my criteria for added yield using the LUMA from 2.0 to 1.5 for a particular subset. And also if I shift the definition of the analysis sets from this all to eligible. I mean, these all have a reasonableness to it, but they're all post hoc, and once you start getting into that you destroy the statistical interpretation. So I'd really like to hear the company, if possible, give some response to the negative studies, and the need to redefine things in order to really get what I don't still see as necessarily a clear picture.
CHAIR NOLLER: Dr. Wright, would you like to address that, please?
DR. WRIGHT: Tom Wright. Yes. I think it's important to put this into clinical perspective. Pivotal Study I was originally designed believing that we would identify a positive effect of LUMA in patients with HSIL as well as in patients with ASCUS/LSIL. Once we found that HSIL, we did not have a positive effect of LUMA, that removes approximately half of these two positives out of the HSIL study. So it means there are fewer true positives to reach statistical significance.
The other point in terms of false positives is that the 8 percent bar was really a clinically decided upon bar. And it's very arbitrary. False positives in colposcopy are not a real issue. We have done studies looking at different colposcopists. There's wide variations in different colposcopists. We picked an 8 percent bar based on an arbitrary criteria, that is 8 percent clinically any different than an 8.5 percent.
DR. D'AGOSTINO: That's part of my concern, that you win on the arbitrary number, you lose on the true positive number. So much is driven by maybe the 8 percent --
CHAIR NOLLER: We don't want to get into a discussion.
DR. D'AGOSTINO: I don't either, I'm sorry.
CHAIR NOLLER: Thank you for addressing the question, Dr. Wright. Are there questions? Let's take a couple of questions, but let's finish everything in five minutes, and then we'll take a break. Yes, Amir?
DR. GANDJBAKHCHE: Dr. Gandjbakhche. My question is related to the spectroscopy in the Slide Number 21, and especially in the Paper that the first author is Dr. Huh. And it's provided in the material. There are Figure 1A and B that shows the fluorescence intensity and the reflectance intensity obtained thousands of sites. The mean value seems is there. Could you provide us the standard deviation of these spectra?
And the second one is could you put some numbers on your ordinates, because I want to see the relative differences between those spectra. Which when you don't put any numbers, number of counts, or anything, that's the relative differences between the spectra and the scene. This is one of my questions.
CHAIR NOLLER: If you're prepared to answer this question now, do so. If you'd like to do it later we can arrange. Which, later? They'll answer later. Another question. Dr. Julian?
DR. JULIAN: I have just four simple yes or no questions. One of the major side effects of colposcopy in numerous studies is the anxiety level of the patients. Was there any measure of the anxiety level with the introduction of this machine and the extra time?
CHAIR NOLLER: Dr. Wright.
DR. WRIGHT: Tom Wright. No.
DR. JULIAN: Does it identify atypical vessels? That's the most common mistake made at colposcopy, the failure to identify atypical vessels. Does it do that?
DR. WRIGHT: No. Colposcopy needs to be performed. This is an adjunct.
DR. JULIAN: I understand from the materials it can't be used in the vagina. Does it always fit the cervix? Is there a limit to the size of the field of this machine?
DR. WRIGHT: This device stands back from the vagina, and it has a 2.4 sonometer diameter. So if there's an exceptionally large cervix, the whole cervix would not be entirely seen, although the central transformation zone would.
DR. JULIAN: Are these biopsies, can they be done in real time, or do you match it up with the image on the screen, you know, kind of guess or estimate where it is?
DR. WRIGHT: What we ask the clinicians to do is to look at the screen -- first they perform standard-of-care colposcopy. That's very important. They select the sites that they would biopsy. They then look at the image. They will see additional sites based on the device output that they want to biopsy, and then they do their standard approach to taking biopsies, which for most colposcopists is to look through the colposcope, also looking up at the screen in order to make sure that they are oriented and back.
DR. JULIAN: But it's a static image. It isn't like an ultrasound where like you're doing amniocentesis or?
DR. WRIGHT: No, it's a static image at that point.
DR. JULIAN: All right.
DR. WRIGHT: Thank you.
DR. JULIAN: Thanks.
CHAIR NOLLER: Was there another question. Russ? Dr. Snyder?
DR. SNYDER: Russ Snyder, OB/GYN. One simple question. There I know were studies done looking at the effect across the menstrual cycle. There was some mention about looking at the effects of an immediately performed Pap beforehand. And I know for Pivotal Study I you used an exclusion of seven days that a Pap couldn't have been done. But is an immediately performed cervical cytology completely, you know, excluded prior to performing the LUMA?
DR. FLEWELLING: Ross Flewelling. We did a series of studies to look at the effect of an immediately preceding Pap on the output of the device. What we found in that study was that the result of doing the Pap results in excessive bleeding, additional bleeding of the tissue, and that bleeding can lead to obscuring some of the cervix. So the effect was not in the spectroscopy, apart from the presence of blood, and there is some obscuring. And we did not address that, but you might have saw the gray cross-hatching in some of those images. That's what we label as being indeterminate. And so if there's excessive blood, we'll pick that up in the spectroscopy, and we'll label that as indeterminate. So it limits the area in which the device can be effective.
CHAIR NOLLER: I think at this point we'll take a 10-minute break. I'd like to ask FDA to be ready to present at 10:50, please.
(Whereupon, the foregoing matter went off the record at 10:41 a.m. and went back on the record at 10:56 a.m.).
CHAIR NOLLER: Well, the best laid plans. We got delayed five minutes by moving the room and bringing in chairs, but hopefully it's more comfortable. I would like to ask anyone that needs to leave the room to now please use the back door instead of the center door, and try to not come and go anymore than you have to.
We will next move on to the FDA presentation. And they also have one hour and fifteen minutes, so we should finish no later than 12:10. And the first presenter is Dr. Joyce Whang.
DR. WHANG: Good morning, I'm Joyce Whang. Dr. Virmani and I are the lead reviewers for FDA on this PMA. To begin FDA's presentations on this PMA, I will provide an overview of our review process. Then our medical officer and biostatistician will provide further details of their analyses of this submission.
Let's start with a brief regulatory history of the device. In August 2000, FDA issued a determination letter indicating this device would be regulated as a Class III device under the premarket approval process. FDA considered two types of indications, one for triage following an abnormal Pap smear result, and one for use as an adjunct to colposcopy. FDA indicated the PMA should be supported by results from a randomized controlled clinical trial for each proposed indication. The trial should compare device results to results from histopathological examination of biopsy tissue samples. The device's underlying algorithm must be locked before the pivotal study was begun, and all study subjects must not have been used earlier in the development or training of the algorithm.
In February 2001, FDA decided that the planned use of this device in clinical studies was what we called non-significant risk. In particular, FDA found that patients would not be put at serious risk. We considered the possibility of physical injury resulting from use of the device, and the impact of study design on patient management. Because of this designation, the clinical studies being described today did not require regulatory approval from the FDA via the IDE process that is often used with premarket clinical trials.
However, MediSpectra did meet with FDA in a number of meetings targeted toward developing agreements as to what would be needed to demonstrate the safety and efficacy of this device. There was no agreement letter for Pivotal Study I. However, there was an agreement letter issued February 2003 which defined a single-arm study, Pivotal Study II, which you are hearing about today. At that time, the proposed indication was as an adjunct to colposcopy in identifying regions of the ectocervix that most likely represent CIN2/3+ for colposcopically directed biopsy.
The PMA was received June 18, 2004, and expedited review was granted. The current proposed indication has been refined as is being presented by other speakers today.
MediSpectra has described the LUMA cervical imaging system to you. It has three primary components, the console and accessories, the illumination probe, and the disposable probe cover. The console contains the computer, control electronics, lasers, flash lamps, and spectrometer. The illumination probe is connected to the console by an articulating arm. For each patient, the illumination probe is covered with the disposable probe cover, and calibrated before being placed at the speculum.
The device uses visible and ultraviolet light to illuminate the cervix. A scan of the cervix, which requires approximately 12 seconds, provides fluorescence and reflectance spectra at 499 distinct sites. Based on the scan, the device creates a color-coded image that indicates the likelihood of CIN2/3+ disease at different locations of the cervix.
Now that you've heard about the device's regulatory history, I present the review team. As you can see, a number of people have been involved in the ongoing review of this PMA application in the areas of clinical, statistics, epidemiology, electro-optics, algorithm development, software, electromagnetic compatibility, electrical safety, disinfection, mechanical design, manufacturing, and bioresearch monitoring.
Here are the primary aspects of our review. I will summarize the findings regarding the first four topics listed here: electro-optics, algorithm development, bioresearch monitoring, and manufacturing. The last two subjects, clinical and statistical reviews, will be addressed by other speakers.
The electro-optics review included safety of the exposure levels of visible and ultraviolet light, optical performance of the device, and calibration methods. We considered the possible impact of medications on LUMA's optical measurements, and the possible impact of cervicitis on device effectiveness. We also considered the possible interactions between UV light and HPV. All of these review concerns have been adequately addressed.
Regarding the algorithm, we looked at how it was developed, how it works, and how it has been changed since the pivotal studies began. We especially looked at the changes made to the algorithms between Pivotal Studies I and II. We concluded that use of the revised algorithm would not have reduced the improvement of true positives seen with LUMA in PSI, although it may have made the increase of false positives seen in PSI slightly larger.
For bioresearch monitoring we looked at study execution, including recordkeeping and administration of informed consent. There were inspections of clinical sites and of the sponsor's facility for records related to the conduct of the clinical trial. There are no outstanding concerns about the reliability or integrity of the scientific data submitted in the PMA.
For manufacturing, we looked at compliance with design controls. The manufacturing facilities were also inspected. On the basis of the information reviewed, the firm has been found to have an acceptable GNP program.
The clinical and statistical reviews will now be addressed by Dr. Carey-Corrado and Dr. Pennello.
DR. CAREY-CORRADO: Thank you Dr. Whang. I'm Julia Carey-Corrado, and I'm going to present FDA's clinical review of this PMA. From time to time I'm going to draw the panel and the audience's attention to the discussion questions that were in the materials that were distributed this morning, just to help you focus on what we'll be talking about this afternoon.
The proposed indication for use is as follows. LUMA is indicated as an adjunct to colposcopy in identifying high-grade disease, CIN2/3+ in patients referred to colposcopy with an ASCUS or LSIL cytology test result. LUMA is not intended to replace colposcopy. A thorough colposcopic evaluation with identification and selection of biopsy sites must be performed independently of and prior to viewing the LUMA result.
I realize that I've repeated the indication for use. You've already heard it today, but I think it bears repeating, and I'd like to just highlight three aspects of the indication for use. As Colin Pollard mentioned earlier, this type of technology has previously been evaluated for different uses, including triage of abnormal Pap smears to determine who goes to colposcopy. And I just want to make that clarification here. This is not triage, of course, it is once the patient is in colposcopy. As has been mentioned multiple times, ASCUS is inclusive of atypical squamous cells of undetermined significance and also cannot exclude H cell.
And the indication for use reflects something that in FDA we refer to as the always/never rule. This is not an official rule, this is just an approach that we take to reviewing certain devices. And what we mean by that is as follows. We mean that the clinician must always identify colposcopically directed biopsy sites prior to viewing the LUMA image, and never eliminate a colposcopically directed biopsy after viewing the LUMA image. Another way of putting this simply is always commit, that is to the colpo biopsy, and never look back. And incidentally, FDA has applied this approach to indications for use of other diagnostic devices.
We have already heard that there were two clinical trials of the LUMA device. And with this slide what I'd like to do is just compare and contrast them, because it really is important to look at them individually and to understand how they differed. With respect to study design, let's see if I can do this. Pivotal Study I was a two-arm, it was a randomized and controlled clinical trial. Pivotal Study II was a single-arm paired design with the subject serving as her own control. The treatment group in Pivotal Study I was -- by "treatment group" I mean where was LUMA used. It was used in one of the arms simultaneously with the colposcopic evaluation. In Pivotal Study II it was used following a colposcopic evaluation.
Now, I already mentioned what we describe as this always/never approach. That approach was not followed in Pivotal Study I, but it was followed in Pivotal Study II.
The primary endpoints were true positive rates for both pivotal studies. The numbers are slightly different. Well, the numbers are frankly different in Pivotal Study I compared to Pivotal Study II. In Pivotal Study I, the hypothesis was that there would be a statistically significant increase in the true positive rate in that combined arm, LUMA plus colpo compared to colpo only. And it bears mentioning that it was anticipated that there would be an increase in false positive biopsies in both of these studies. And as Dr. Wright mentioned, it was agreed that approximately 8 percent would constitute a clinically acceptable false positive rate, but there was nothing magic about that number.
The false positive hypothesis, as well as the true positive hypotheses, were different in Pivotal Study II compared to Pivotal Study I. You'll note that in Pivotal Study II we were looking for a statistically significant improvement over 2 percent in the true positive rate increment, and we were looking for the upper bound on the confidence interval for false positive to be less than 15 percent instead of less than 8 percent.
The sample size was obviously very different in the two studies, almost 2,300 in Pivotal Study I. And although Pivotal Study II was ultimately planned to evaluate 788 subjects, the study was stopped early at around 29 percent. That is 227 subjects who were enrolled. The last difference I want to make here, the last contrast I want to highlight is the Pivotal Study I was not the subject of an agreement letter between the sponsor and FDA, but the Pivotal Study II was the subject of an agreement letter.
And I've already discussed the hypotheses for Pivotal Study I. True positive statistically significantly greater than zero, and the false positive would be significantly less than 8 percent.
I'd like to take a moment to talk about the definitions in Pivotal Study I. We're looking at true positive rates and false positive rates. And it bears mentioning that these rates, the denominators for both of these rates are the total number of subject, as opposed to the total number of disease subjects. Because with the two study designs, it was impossible to know with certainty the disease status of all the subjects.
This slide is to make the point that although we've heard sensitivity mentioned already, netiher study was designed to determine the sensitivity, in the classical sense, of LUMA to detect CIN2/3+ in patients who are disease-positive. To obtain that measure directly, the gold standard would have to have been LEEP, that is Loop Electrosurgical Excision Procedures, or some kind of surrogate, possibly long-term follow-up. But the studies were not designed to obtain that data.
Pivotal Study I was stratified, the randomization was stratified by colposcopist as well as by Pap stratum to get an equal distribution across the two study arms. Subject accountability is described in this slide. And one of the things that we like to do when we look at subject accountability is take a look at the numbers of subjects who are lost from each arm of the study. And we can see that within the entire patient population, 49 subjects were excluded in the LUMA plus colpo arm, and 37 were excluded in the colpo only arm. The distribution was 35 excluded in LUMA plus colpo within that ASCUS and LSIL referral population, and also 26 were excluded within ASCUS and LSIL. So again, the point I'd like to make with this slide is just to help you see how many patients were lost from each arm of the study, and then make the point that patients could be excluded both prior to randomization and post randomization. The primary reasons were device malfunction that occurred in 1.6 percent of cases, and "no majority pathology" diagnosis which occurred in 1.8 percent of cases.
A minor point here is that the "no majority pathology" exclusion was slightly different between the two pivotal studies. In Pivotal Study I that subject was excluded only if all biopsy results were lost. And in Pivotal Study II, it was as if one was lost.
This slide presents the results for the entire study population for the two primary endpoints, the true positive and false positive. The observed difference, or the observed improvement in the LUMA plus colpo arm was 1.9 percent, which was greater than zero. However, the lower bound on the confidence interval was -1.5 percent. Hypothesis was not met.
With respect to false positive, the observed increase in false positive between the two arms was 3.1 percent greater in the LUMA plus colpo arm. The upper bound on that confidence interval being 7.2, that hypothesis was met because the 7.2 percent upper bound was less than 8 percent.
At the ASCUS and LSIL evaluation, we've heard it was unplanned to combine those two Pap referral strata. And the company has presented their justification for this analysis. And so we're just going to look at these numbers again. Within the ASCUS and LSIL subgroups, the observed difference, or the observed improvement in picking up CIN2/3+ in the LUMA plus colpo arm was 3 percent. It did not meet the statistical hypothesis that the lower bound on that confidence interval would be greater than zero.
And with respect to false positive, again we see that there was an increase in false positive. It was 4 percent, which in and of itself was obviously less than 8. However, the upper bound in the confidence interval was 8.5 percent. So strictly speaking, those hypotheses were not met per statistical analysis.
The last thing I'd just like to mention is that although you've seen these numbers before, and a lot more numbers as well, this is important because the ASCUS/LSIL subgroup is the group for whom the indication for use has been written, and it is to that referral group that the device would be targeted.
We also did a false negative rate analysis. Our analysis is mathematically we did a slightly different calculation from the company, but fundamentally our results are not too different. We saw a false negative rate for LUMA of 23 percent, compared to a false negative rate for colposcopy of 15 percent. And the point that we want to make here is that the LUMA device does not see some of the lesions, the CIN2/3+ lesions that the colposcopist sees. And similarly, the colposcopist doesn't find all the lesions that LUMA detects. So both the colposcopy procedure and the LUMA device have a false negative rate associated with them.
And now I guess I'd just like to draw your attention to the fact -- before I start talking about Pivotal Study II, I'd just like to draw your attention to the fact that Panel Discussion Question 1 has three subparts, and those subparts cover a lot of what I've presented regarding the outcomes for the total patient population, the outcomes for the ASCUS and LSIL referral group. Discussion Question 2 is related to something I have not talked about. Dr. Gene Pennello will talk about it after I finish, and that is the issue of a possible age effect. And I'm raising it now because it is going to be discussed in the context of the results of Pivotal Study I.
So now I'm going to change the topic and turn to Pivotal Study II. This is the slide that I showed before. On the right-hand side I've just highlighted Pivotal Study II design. It was a single-arm study with the patient serving as her own control. The colposcopy was performed with commitment to biopsy sites that could not be reversed. Only after this was done, the LUMA image was viewed. And this is, again, follows this always/never approach that is sometimes taken with diagnostic devices.
The threshold for success, the mathematical thresholds were slightly different. The true positive rate increment had to be statistically significantly greater than 2 percent here, instead of greater than zero percent in Pivotal Study I. And the false positive rate, the upper bound on that had to be less than 15 percent for study success. And as we have already heard, the study was stopped early, but we're going to talk about the results for the 227 subjects who were enrolled.
FDA did agree with the company on the study design here, and I guess what I'd like to say is that one of the reasons was we wanted to really see what the difference was, what added, what incremental benefit LUMA contributed after a thorough colposcopy was done. And we felt that this study design allowed us to better evaluate that.
The study design does require additional biopsies. There was some discretion left in the investigator, but fundamentally the investigators were told you should perform an additional biopsy based on the LUMA overlay if the LUMA overlay indicates that there is tissue that is suspicious for CIN2/3.
Because the study design essentially was going to force an improvement, we felt that we should raise the bar a little bit above zero percent to 2 percent. And because it was requiring additional biopsies, or strongly urging additional biopsies, we felt it was acceptable to increase the incremental false positive rate to 15 percent.
This is an accountability slide to help you understand how many patients were excluded in this study. And for the entire patient population 34 were excluded. Within the ASCUS/LSIL referral group 26 were excluded. As you can see, many of these were due to one pathology biopsy having been for some reason uninterpretable, or unavailable.
And we know that the study was terminated early, so let's just go ahead and look at the results. Once again, you've seen the results before, but you've seen also a lot of sets of numbers, and I want to focus just for a moment on the numbers based on the original agreement letter with FDA. For the entire patient population, the incremental increase in diagnosing CIN2/3 was 4.7 percent, and that was the result of nine additional subjects who were found to have CIN2/3+ on the basis of the LUMA contribution. Because the hypothesis was that it would be better than 2 percent, and the lower bound on that confidence interval was 2.2 percent, that hypothesis was met. In the false positive analysis, we saw an observed increment in false positive of 18.1 percent, and that did not meet our hypothesis that it would be less than 15 percent.
Within the ASCUS/LSIL subgroup, we can see that the observed LUMA increment was 3.6 percent, and that was derived, that study population, from six additional subjects who were referred for ASCUS or LSIL Pap results. The lower bound on that confidence interval was 1.3 percent, and that did not meet the original hypothesis. Dr. Gene Pennello is going to discuss further the different thresholds for what that hypothesis should have been. The company has presented a justification for lowering it from the original target from 2 percent to 1.5 percent, but there are additional analyses that have been done as well, and Dr. Pennello the statistician is going to talk about these analyses. With respect to false positive, the increase in false positives during that LUMA increment was 20 percent increase which did not meet the original study hypothesis.
Panel Discussion Questions 3(a) and 3(b) ask the panel to discuss the results of Pivotal Study II. And as I mentioned, Dr. Pennello was going to discuss some additional analyses performed by the sponsor on both of these endpoints so that we can all understand what data are being presented, and decide, or at least consider how they should be viewed.
There's also a false negative analysis that can be done for Pivotal Study II. It's a frustrating one because it's a biopsy-level false negative analysis. And although the LUMA false negative analysis at the biopsy level was 22 percent, nevertheless there were nine additional subjects evaluated. And the sponsor's conclusion is that the subject level -- or I'm sorry, the biopsy-level false negatives are not easily or directly translatable to subject-level outcome. So I'm just showing you these results because I talked about false negative in the context of Pivotal Study I, but it's less easy to interpret in this study.
The adverse event profile for both studies was as follows. There were two reports of cramping, and some subjects made more than one complaint. So the numbers in parentheses done pertain to subjects, but rather just to complaints of different events. So there were two complaints of cramping, one of vomiting, one of weakness, one of vaginal bleeding, three of fainting, one complaint of abdominal pain, and one complaint of dysuria.
The most stark way to present the results of these two studies is as follows. Both studies had colposcopy-primary endpoints. Neither PSI nor PSII met both of those endpoints. We can see that for all subjects, Pivotal Study I succeeded in the false positive category. Pivotal Study II succeeded in the true positive category for the overall subject population. And then according to the original protocols, both studies failed within the ASCUS/LSIL. However, the confidence intervals -- I'm not presenting confidence intervals here. I'm simply presenting the most, again, the most stark analysis of the data.
I think that it's important to note that the two pivotal studies were very different in design. Pivotal Study II was performed consistent with the current indication for use. That is, with the always/never rule in place. We should also mention that the current target population, the ASCUS/LSIL was not in either original protocol as a primary endpoint.
So, given the details of the results of these two studies, what we'd like to ask the panel to consider with respect to risk/benefit is the following. And that is whether there is an incremental increase in the detection of CIN2/3+ attributable to LUMA compared to colposcopy alone, and whether this incremental increase justifies any additional biopsies, including false positive biopsies, and the risk of false negative. Now, when I say the risk of false negative, that would only apply strictly if the device were performed simultaneously, or instead of colposcopy. If it is performed according to the always/never rule, the panel can decide whether it poses any risk.
And the risk/benefit analysis is the subject of Panel Discussion Question 7. And at this point I'm going to turn the podium over to Gene Pennello, our FDA biostatistician, who's going to describe additional and in more depth analyses the has performed on these clinical trial data. Thank you.
DR. PENNELLO: Good morning. My name is Gene Pennello, and I am the FDA statistical reviewer for this device. And I wanted to highlight some statistical design and analysis issues for you for both of these two studies that were done on this device.
Here's the outline of what I want to present. What I'd like to do first is talk a little bit about diagnostics and how they're normally evaluated, and compare that to what we have here in these two studies. And talk a little bit about the study designs, and then go through some analysis issues of the two studies, and some other remaining issues that involve both studies. And then I want to summarize.
First, for diagnostics, you have two outcomes. The diagnostic test can either be positive for disease or it can be negative for disease. And what we'd like to do is evaluate the tradeoff between detecting the disease, which means you want to test positive when the disease is present, versus falsely detecting the disease, which means you test positive when the disease is absent. And just to give you the toy example of an unacceptable tradeoff, suppose you want to LEEP everyone indiscriminately, then everybody tests positive. You can detect every disease with the LEEP in every patient that has it, but you would be subjecting everybody without disease to that LEEP. So I think that would be clearly unacceptable tradeoff.
LUMA's indicated as an adjunct to colposcopy, and what we have is we can think about the test as either at the subject level or the biopsy level. So at the subject level this subject is test positive if any biopsies are taken, and it's test negative if no biopsies are taken. You can also think of it at the biopsy level. Are the biopsies are test positive and the rest of the cervix is test negative. But what we'd like to emphasize is that clinically it's more important to detect more CIN2/3 subjects than more CIN2/3 biopsies. It doesn't do any good if colposcopy has already -- it doesn't do any good to detect more CIN2/3 biopsies if colposcopy has already detected that the patient has CIN2/3 with other colposcopy-indicated biopsies.
So the subjects have -- so the analyses, the primary endpoints were at the subject level. And at the subject level you really have three outcomes, true positive, false positive, or negative. And by true positive, biopsies are taken so that the test is positive, and some are CIN2/3. And a false positive is defined as biopsies are taken, you test positive, but none are CIN2/3. A negative means you're test negative and no biopsies are taken. There's some limitations to this. In the false positive, subjects are not very well defined because you could have CIN2/3 in places other than the ones that you biopsied, so in that sense you really -- that patient would be a false negative rather than a false positive. So biopsy location matters. And we don't know anything about the accuracy of the test negatives because there's no follow-up on these patients and no biopsies are taken.
What we have in the two studies is this endpoint pair on which to assess a tradeoff, and that's the subject true positive rate and the subject false positive rate. The number of true positive subjects divided by the total number of subjects, and the number of false positive subjects divided by the total of subjects. Now, these rates I want to make clear, these are not the same as sensitivity and 1 minus specificity, even though in the literature, true positive rate is often regarded as synonymous with sensitivity, and false positive rate is regarded as synonymous with 1 minus specificity. We can't estimate the sensitivity and specificity in these studies because we don't know all the patients who are diseased and all the patients who are non-diseased. And that's how you would define sensitivity and specificity based on those numbers. So what we have to deal with is this tradeoff in these special definitions of true and false positive rates, comparing LUMA plus colpo versus colpo alone.
So as I've mentioned, neither specificity or sensitivity can be estimated in Pivotal Studies I or II. Another traditional endpoint pair I've mentioned is the positive and negative predictive values. And we can't assess the tradeoff here either because although positive predictive value can be estimated, negative predictive value cannot be.
So on to the designs of the two studies. I just want to make two remarks about the two designs. Dr. Carey-Corrado had already presented this slide comparing and contrasting the two studies. The first study, Pivotal Study I, was a randomized controlled trial, and the arms were the simultaneous evaluation of LUMA plus colpo, based on LUMA plus colpo. And then the control arm was colpo only. What I want to bring to your attention is that the colposcopist was not blinded to the arm to which the subject was randomized. So they knew which arm the subject was randomized to, and that could potentially create a bias in this study in that -- just, it may or may not have it, we don't know, but it's possible that because the colposcopist knew the subject was in the LUMA plus colpo arm, the mere presence of LUMA could have stimulated him or her to do a different colposcopy, maybe a better colposcopy in that arm, versus colpo only, or it could have, because they were relying on LUMA for example, they could have done a less vigilant colposcopy in that arm compared to colpo only, and that could depend on the colposcopist. So there could potentially anyway, there could be bias. And any difference we see between the two arms really confounds the LUMA effect with this any differential in the colposcopy performance between the arms. And I just wanted to make you aware of that.
Pivotal Study II also has a design limitation. Here in Pivotal Study II the subject was her own control. And initial colposcopy was performed followed by the LUMA increment. If the initial colposcopy was for some reason undercalled, that would bias the study in favor of finding the LUMA effect. Without presenting any numbers, we did -- what the sponsor has presented to us did not, we felt there was really no evidence of undercalling, but I do want to make you aware of this possibility.
Now, I want to highlight -- I'm going on to the analyses, and I just want to highlight some analysis issues with each study. So first, Pivotal Study I. And I want to talk about the joint analysis and the possibility of an age-related device effect. So talking about Pivotal Study I now. And just to recap, the dual hypothesis on the true positive and false positive rates was not met for both the primary analysis. And if you applied that hypothesis to the subgroup of ASCUS/LSIL to which the sponsor is seeking an indication, you would also -- the dual hypothesis is also not met.
The sponsor has presented a joint analysis, and this is an unplanned analysis in which you look at the three possible outcomes jointly together. You don't look at separately the true positive rate and the false positive rate, but you look together at the three rates in both arms, the true positive rate, the false positive rate, and the negative rate. So I'm just providing some numbers for both all subjects and for ASCUS/LSIL of these three rates. And the p-values of the test they presented show that in fact for all subjects there is a statistically significant difference between the arms when you look at all three rates jointly. There's heterogeneity in those rates. And also, if you look at the ASCUS/LSIL subgroup, the p-value there is 0.001, and that also indicates statistically significant heterogeneity. However, this kind of test is not specific on the primary hypotheses of interest, the true positive hypothesis and the false positive hypothesis. So we don't know based on these tests whether the true positive difference is significant, and whether the false positive difference is no more than 8 percent. Those were the two primary hypotheses in this study. So it's really rather vague, the inference based on the joint analysis.
I want to talk about a possible age-related device effect, or arm effect. In the original PMA, the analyses were stratified by three age groups, age less than 21 years, age 21 to 29 years, and age greater than 29 years. And we noticed there was quite a bit of difference between -- the difference between the arms in the true positive rate depended on which age group you were in, at least the observed difference. And the blue line here shows that for age less than 21 years, and I'm going over the Pap strata of ASCUS/LSIL and HSIL, that you get a large difference between the arms. The true positive rate was larger in the LUMA plus colpo arm than in the colpo only arm. And this is consistent across the strata. And then the last point here is for all subjects. And I would -- these circles here I want to say are proportional to the sample size, just to give you an indication of how many subjects we're dealing with in each of these groups.
There didn't seem to be much going on in the other two age groups in terms of an increase in the true positive rate. The sponsor did a number of analyses when we pointed this out to them, and the conclusion that they have made is that age did not have a significant effect on the arm difference. They did look at other age cuts, not the ones that were in the original PMA, but they looked at tertiles, as they presented, and for other age cuts, the age affected is less pronounced. They did present today a plotted true positive rate as a function of age for, I think, Slide 70 in their presentation, and they showed that the true positive rate in the LUMA plus colpo arm was pretty consistent across age. The problem with that is that, and they did mention this, is that the prevalence, the CIN2/3 prevalence varies by age. And so the true positive rate would expect it to be lower for lower age groups, because there's less CIN2/3 to detect. The prevalence is lower. So the fact that you have a consistent true positive rate across all the age groups that they were presenting in that Slide 70 may indicate, if you factor in prevalence, that they actually are doing better in young age groups. And there also was no comparison with the true positive rate with colpo only in that slide.
They did a Grizzle-Starmer-Koch model to investigate the age by arm analysis. And this was on the LSIL/ASCUS population only. They quickly excluded HSIL based on an army by Pap stratum interaction. And they presented some plots in Amendment 7 to the PMA that compared the fitted values of the model to the observed values. And when I look at those, and it's just a plot, and just eyeballing it, I see some lack of fit that's attributable to this possibility of age by arm interaction.
And they also did a logistic model which, instead of looking at age groups, they modeled the true and false positive rates. It's a function of continuous age. And they looked at a linear term in the model for age, and there was no age by arm interaction in the linear -- linear age by arm interaction. But they did not investigate a quadratic effect, so possible curvature in the relationship between the true positive rate difference and age was not investigated.
We did our own analysis on the original three age groups in the PMA, and it's a Bayesian trinomial or multinomial logistic model. These were the effects that were included. In this analysis we did see a significant arm by age interaction. And I don't want to alarm anybody by the word "Bayesian". There's nothing funny here. I didn't add any information from some prior distribution into this, I just wanted to get some exact analysis, or some way of investigating this age by arm interaction effect. But what the Bayesian analysis allows you to do is actually look at the primary hypotheses of the study, which is that the true positive rate difference is greater than zero, and the false positive rate difference is no more than 8 percent, and calculate the Bayesian probabilities of these within the age groups and by Pap strata. And if you look at the under 21 years of age group, the probabilities are nearly 1 or 100 percent within each of the LSIL, ASCUS, and HSIL groups. And the probabilities for the false positive rates, the difference being no more than 8 percent are quite high too. And in fact, within the under 21 years of age group, you meet both hypotheses for ASCUS and LSIL according to this analysis. I mean, if the threshold would say that the probability had to be at least 95 percent, you almost meet it for LSIL. You don't see anything like that in the other age groups. The probabilities are not that high. So this disagrees with the sponsor's analysis.
So I want to go on to Pivotal Study II and talk a little bit about -- highlight some issues there. And again, just to remind you that the dual hypothesis here was different because of the way the study was designed, it was a colpo only followed by the LUMA increment. So you were going to take additional biopsies based on LUMA. The true positive rate could not decrease, the incremental true positive rate could not decrease, so that's why they the hypothesis is 2 percent, not zero percent here. And the dual hypothesis was not met for all subjects, and in this ASCUS/LSIL subgroup it was also not met, if you applied the same hypotheses to the LSIL/ASCUS.
I want to be clear on what it means to be an incremental true positive and false positive according to the protocol. So you cannot be an incremental true positive if the initial colpo indicated biopsies in which some of them were true positives. So if based on initial colpo the subject was a true positive already, you could not get an incremental true positive. You could only get an incremental true positive if based on initial colpo, the subject was a false positive or a negative, in which you didn't take any biopsies. And so, based on the false positive and negative outcomes on initial colpo, nine of those subjects were changed to true positives based on the LUMA incremental biopsies, the 7 plus 2 here. And so that's where the incremental false positive rate comes from. It's 9 out of 193.
Now, you could not get an incremental false positive unless the subject was negative based on initial colpo. In other words, no biopsies were taken based in initial colpo, but then you took some biopsies based on the LUMA results, and then all of those were negative, and so the subject became a false positive. And there were 35 of those, so the incremental false positive rate is 35 over 193.
Now, there were some revisions to the protocol analysis. These were unplanned, and I refer you to Question 4 for that, the Discussion Question 4 you'll be discussing later. There were two changes. There were target value changes for the ASCUS/LSIL population and for the HSIL, which I don't have here. And also, this reduced denominator, or what the sponsor termed the eligible population denominator for the true positive rate.
So for ASCUS/LSIL, first the target value was changed from the original 2 percent to 1.5 percent for an incremental true positive rate. And correspondingly, the threshold for the false positive increment was changed from 15 to 16.1 percent. And the rationale is that these rates, true and false positive rates, are a function of prevalence, and if the prevalence is lowered, as it would be, as you would expect if you go from all subjects to the ASCUS/LSIL population, you would expect the incremental true positive rate to decrease, and the incremental false positive rate to increase. So there was a rationale for doing this. However, if you use these two revised target values and apply them to the ASCUS/LSIL analysis, it still fails to meet the dual hypothesis. And these are the numbers that you've already seen. The lower bound on the confidence interval for the incremental true positive rate is 1.3 percent, and so it does not meet the 1.5 percent threshold. And similarly for the incremental false positive rate.
The second change was this reduced denominator. The denominator was changed from all subjects to all subjects minus true positive subjects already detected by initial colpo. So what I already mentioned is that you can't get an incremental true positive if initial colpo had already detected CIN2/3. So those subjects were eliminated from the denominator. FDA believes this analysis is invalid, unless you revisit the true positive and false positive hypotheses. This is a different denominator. And the agreement letter that we had with the sponsor was based on the full denominator of all subjects, and not on the reduced denominator. And you would have to revisit these hypotheses in order to come up with an acceptable tradeoff when you change the denominator.
However, if you do change the denominator, these are the results for the ASCUS/LSIL population with the revised target value of 1.5 percent. With the original denominator you can't show at least a 1.5 percent incremental true positive rate. But with the reduced denominator you can because 1.5 percent is lower than the 95 percent confidence interval.
You can do the same thing correspondingly with the false positive rates. You can reduce the denominator to only those subjects who could have gotten an incremental false positive. And the denominator gets reduced from 167 to 50 here. And the incremental false positive rate is quite large if you take the same approach. And I should have emphasized here that for the true positive, the denominator is reduced in the ASCUS/LSIL population from 167 to 143.
Now some issues with both studies I want to go over, biopsy frequency in both studies, subgroup testing with respect to the ASCUS/LSIL subgroup, site and colposcopist variability, and biopsy-level positive predictive value.
This is a chart of, well, if you recall, the true positive and false positive increases were observed to be larger in Pivotal Study II than in Pivotal Study I. Now that could be due in part to the difference between the two studies, and the number of biopsies that were taken, and the number of subjects that were biopsied. So I'm trying to give you an outline here of this. For Pivotal Study I, in the colpo only arm, the mean number of biopsies was about one. In the LUMA plus colpo arm it was 1.3. In Pivotal Study II, for initial colposcopy it was about the same as the colpo only arm in PSI, it was 0.9. The LUMA increment added, on average, another biopsy. And so the total number of biopsies was 1.91 versus 1.3 in the LUMA plus colpo arm PSI. So there were more biopsies taken. Also, the percentage of subjects that were biopsied was larger in PSII than PSI, 92 percent versus 82. So this in part could explain the larger true positive and false positive increases in PSII compared with PSI.
Now I'd like to talk a little bit about subgroup testing, and in particular this ASCUS/LSIL population for which the company is seeking an indication. The secondary analyses were pre-specified in the protocol for ASCUS, LSIL, and HSIL. There was a secondary endpoint that said that the analyses would be stratified by these three groups of Pap strata, but not for combinations of these groups. And if you look at all possible groupings of the Pap strata, there's actually seven. If you include all subjects, there's an ASCUS, LSIL, and HSIL. There's also three combinations of two of this strata, one of which the company's looking at, which is ASCUS/LSIL.
Statistically when you look at subgroup testing, what you would like to do is control overall with the multiple tests that are being made, you want to control overall the chance of any falsely significant result to 5 percent. One way to do this, if you're going to do the primary analysis, make it a 5 percent level test of the primary analysis, one way to control the overall chance at 5 percent is a gatekeeping approach in which if the primary analysis succeeds, in other words you find significance, then you're permitted to test the subgroups. And you would be permitted to test the subgroups at a level that could be considerably less than 5 percent. It would depend on the number of subgroups you're testing. Now what's happened in these two studies is the primary analysis failed. And so according to the gatekeeper analysis, you wouldn't even be permitted to test the subgroups. And if you were permitted to test the subgroups, the level of the test would be less than 5 percent. Now, what you've been presented here today is testing of the ASCUS/LSIL subgroup at a level of 5 percent, not anything less than 5 percent. And statistically with this gatekeeping approach, you wouldn't even be permitted to go down to the ASCUS/LSIL subgroup and test it. Even if you tested at 5 percent the ASCUS/LSIL subgroup, the dual hypothesis is not met. All this aside, there may be clinical reasons for looking at the ASCUS/LSIL subgroup, so that's why it's being considered.
We did ask the company to look at site variability and colposcopist variability. I've got some site variability results here. The difference in the true positive rate was not -- there was no significant variability in the difference in the true positive rate between the arms in PSI. The p-value was 0.457, and there also wasn't any significant variability in the difference in the true positive rate in the PSII and the ASCUS/LSIL group. There was some statistically significant variability in the difference in the false positive rates. And the analyses being presented are simply pooling the rates. They're not considering that these differences might be different among the sites or among the colposcopists. And if you did analyses that allowed, for example, the colposcopist to have his or her own LUMA effect that could vary by each colposcopist, that would increase somewhat the confidence interval width.
And finally, there was per biopsy analyses done. So this is not on a per subject level but on a per biopsy level, and a look at the positive predictive value, which is the percent of biopsies that are confirmed to be CIN2/3. And in particular, in PSI, there was a number of results that were presented to us, but this highlights some results. In PSI, in the LUMA plus colpo arm, the colposcopist was asked to annotate whether the biopsies were based on colpo alone, or based on LUMA alone, or based on both. And if you look at the positive predictive values for these three, they're given here, you can see that for biopsies that were based on colpo alone, the positive predictive value is 17.5 percent. And that jumps to 28.8 percent when they were not only based on colpo but also LUMA found these biopsies. And the difference is statistically significant. It's an interesting finding. If LUMA was not doing anything other than randomly selecting places to biopsy, you would expect these two positive predictive values to be the same, but they're statistically difference. However, it doesn't really get at the indication for use of this device, which is an adjunctive indication. You don't really care about the biopsies that colpo has already identified. What you care about are the ones that LUMA has additionally identified, and that's the middle column here for LUMA only, and the positive predictive value is 11.3 percent. When you compare that to the literature, at least what we've been presented, it seems to be larger than random biopsies that have been done in difference studies.
But it's hard to even really assess what this means, because for LUMA only biopsies are based on LUMA regions that have been identified, but there's some discretion by the colposcopist on which regions which she's going to biopsy, and where within those regions to biopsy. So there's some colposcopy skill being built in to this positive predictive value. So we don't really know whether this -- it doesn't cleanly say what the LUMA positive predictive value is. There's some colposcopy skill involved in this result.
I also want to mention there is no negative predictive value that can be estimated from these studies. So there's no tradeoff between positive predictive value and negative predictive value. That's not available. All we're looking at is positive predictive values.
And finally, LUMA could be doing better than random on the biopsy level, but that doesn't necessarily mean that more CIN2/3 subjects are detected. That's the whole reason why we looked at subject level results as primary endpoints.
So in summary, netiher PSI nor PSII was a success in both the true positive and false positive endpoints because they both had to be met, and they either both failed, or one passed and one failed. And so for all subjects it wasn't a success, and it also wasn't a success if you applied the same hypotheses to the ASCUS/LSIL population. The ASCUS/LSIL subgroup was unplanned, and it made some statistical inference problematic. Combining the two studies is statistically difficult. Because they're designs are fundamentally different, it's hard to know how to come up with a biostatistical model that would allow you to combine them in a reasonable way.
There did seem to be an age-related device effect in PSI. The true positive rate difference seemed to be a lot larger in the younger age group than in the other two age groups that were presented. This result was not replicated in PSII. We didn't find any evidence of this in PSII, maybe because of the smaller sample size, but we just don't know.
And finally, as Dr. Carey-Corrado mentioned, you could attempt to get a subject level false negative rate for LUMA based on colpo as a reference for detecting CIN2/3 subjects. And when you use that colpo as the reference, you get a LUMA false negative rate of 23 percent. And this is based on the PSI study results. So that concludes my presentation, thanks.
CHAIR NOLLER: Thank you very much. Thank you also for staying within the time limit, well within it. We have a few minutes now before we break for lunch, and I would like to first ask the panel if they need any clarification of any of the issues that were raised by FDA, or any questions to address the FDA directly. If not, we will then go back to asking questions of the sponsor. Yes, sir.
DR. GANDJBAKHCHE: Amir Gandjbakhche.
CHAIR NOLLER: Closer please.
DR. GANDJBAKHCHE: In the algorithm you mentioned it is a black box. What do you mean by that? In the executive summary.
CHAIR NOLLER: The executive summary prepared by FDA. What page is that?
DR. GANDJBAKHCHE: It is Page 11.
CHAIR NOLLER: Page 11 of your summary. Black box.
DR. GALLAS: So when it comes down to it, the algorithm -- I'm sorry. I'm Brandon Gallas. I did do the algorithm review. I have an applied math Ph.D. And so when I say it's a black box, there are many components to the algorithm. There are weights, there are statistical classifiers, there are masks, there are lots of pieces to this problem that you can't track one thing from the beginning to the end in a real coherent way where if you know the features coming in, certainly they progress down to the final output. But to really know all the interactions of all those components in that algorithm, it's just too complicated. So at that level we can't make any decisions about the effectiveness of the device just based on looking at the algorithm. So when I say that, I'm saying we need to rely on the clinical study to make that evaluation of effectiveness.
CHAIR NOLLER: Thank you. Other questions for FDA? Yes, sir.
DR. D'AGOSTINO: D'Agostino. This is a question for Gene. Gene, as you pointed out, and as was pointed out by all the FDA presenters, the studies were negative in terms of the statistical criteria. Sometimes when they're negative, and there's an issue of maybe we should consider further, there's a shift to the sort of clinical aspects where the effects seem reasonable, and what kind of power do you have. Did you do any power calculations for this study? It's negative by the statistical analysis, but is it doomed to negativeness because the sample size was too small? There was over 2,000 subjects in the first sample, so it's -- intuitively I would say if there's something going on the statistics should verify it. Did you do anything to sort of move on that line? Do you understand my question?
DR. PENNELLO: Yes. Before these studies are conducted, we always ask the sponsor to do a power analysis so they will adequately size the study. And the size of the study is based on the assumptions of the effect that you expect to see in the studies. And evidently, the effect was not as large as what was assumed, and that's why you got a negative result in the true positive difference for Pivotal Study I.
I would say that Pivotal Study II, the sample size was determined to be smaller in part because you're using the patient as your own control, and so you're eliminating the variation between patients when you're making the comparison.
DR. D'AGOSTINO: We oftentimes hear in the discussion now if only I had another 500 subjects I would have had significance. I'm not sure I see any of that going on here.
DR. PENNELLO: Well the true positive rate difference was in the right direction, so conceivably if you sampled long enough you might see a statistically significant result.
DR. D'AGOSTINO: Yes, but we don't have it. Thank you.
CHAIR NOLLER: Dr. Miller?
DR. MILLER: My questions aren't specifically directed to FDA. They're more directed to the company. Do you want to proceed?
CHAIR NOLLER: Any more FDA questions before we go to the company? Let me first -- did you have an FDA question? Yes. Dr. Jiang.
DR. JIANG: Yulei Jiang, Chicago. I have a question about Slide Number 36, to compare positive predictive value. So this compares in the PSI trial of the colpo only when the colposcopy says it's colpo only versus when the colposcopy says it's both. A more natural comparison to me would be comparing the two arms, where one arm is colpo only, the other arm is colpo plus LUMA. Do you agree? I think that's more natural, to me, but the comparison here is different. Can you elaborate on that?
DR. PENNELLO: Yes, that would be an interesting comparison. I don't remember what the positive predictive value is for the colpo only, but I don't believe it was 28.8, but I'd have to look.
CHAIR NOLLER: No other FDA questions? We'll go back for the sponsor. First, do you want to address the question about the units for the spectroscopy, or do you want to answer new questions?
DR. WINGROVE: I think we'd like to defer that question till after lunch.
CHAIR NOLLER: Okay. Dr. Miller, what are your new questions, please?
DR. MILLER: First of all, Hugh Miller, maternal-fetal medicine. Couple of questions. The issue of the LUMA device not being as good as expected in detecting the HSIL lesions kind of got glossed over. And I wondered why the company, you know in retrospect, even if it's a post hoc sense, why it wasn't as well detected, since a lot of the basis presumptions were based on that.
Secondly, I wonder if there was any evaluation in the first trial, if there was any evaluation of the colpo only arm since LUMA measurements were done, or a scan was done, even though the investigator was blinded. I'm assuming that there was some mapping of where the biopsies were done, and there was some potential for looking at where the colpo only directed biopsies were done relative to what the LUMA evaluation would have been. And although you don't have the follow-up, there may have been some correlation possible, looking at those evaluations.
There's a lot mentioned about the fact that this trial was done in so many sites, but in reality it was really only one non-academic site. And the presumption is that this technology would be used across the country among a much broader array of colposcopic experience. So I'm curious about what kind of further sense you have of how this technology would be used in environments where there might not be as good colposcopic experience. And there doesn't seem to have been good address to the level of colposcopic experience. I noticed that when the FDA asked that question, that was again kind of glossed over. I understand that there were attendings, and nurse colposcopists, and an array, but I mean I wonder if you have some sense of this relative to years of experience, or some other measure of experience?
CHAIR NOLLER: Would the company, the sponsor, like to address those three questions, please? And then we will probably be at our lunch break. Introduce yourself as you speak.
DR. WRIGHT: Tom Wright. I would like to address first the issue of different colposcopists. We had a wide range of colposcopists who were involved in this study. Determining expertise of colposcopists though is very difficult. A lot depends on experience, the number of cases, the complexity of the cases. And we looked to see whether or not there was a variability with respect to colposcopists, and we did not see significant difference in device performance. But again, it's very hard to determine who is an expert colposcopist, who is a less expert colposcopist.
In terms of looking at the performance of this device in different clinical settings, let's say settings where you have high quality, expert colposcopy. We believe that our clinical sites uniformly were very high standard colposcopy sites. Many of them were high volume, many of them were members of the ALTS study, and we would expect to see, using an objective measure for a highly subjective art such as colposcopy, increased performance using perhaps less skill colposcopists. But we don't have data to support that. That is just a hypothesis, that the better the colposcopist is probably the less benefit you would see of this device. But we can't document that.
Dr. Warner Huh wants to address the issue about why we did not see as much effect in HSIL as we initially predicted.
DR. HUH: Again, my name is Warner Huh. When the study as originally designed, it was obviously designed to include all patients. But what we learned from the trial, and this is actually borne out from the literature as well, is that those patients who were referred with an HSIL Pap smear are more likely to have a clinically obvious and evident lesion, one that's clearly evident and consistent with CIN2/3 to the colposcopist, compared to that with someone with ASCUS and LSIL. So if you're going to biopsy that clinically evident lesion, it's hard to expect that you're going to improve detection above and beyond what the colposcopist is going to find on their exam. So that's probably the main primary reason why there wasn't any difference seen in the HSIL group.
CHAIR NOLLER: Thank you. Maybe one quick question we have time for. Dr. Snyder?
DR. SNYDER: Dr. Snyder, again, OB/GYN. In Pivotal I, one of the things I had the most trouble getting my arms around was the fact that in the colpo plus LUMA arm, the colposcopists were asked to retrospectively identify whether the basis for their biopsies was based on the colpo, based on the LUMA, or based on both. And my question is what were the specific instructions that were given to the colposcopists in making that ascertainment afterwards, and would there not be a tendency for people to say, well, both things figured into my decision?
CHAIR NOLLER: Dr. Wright?
DR. WRIGHT: The clinicians were instructed specifically, once you have looked and performed your colposcopy, and you look at the images of the LUMA on the screen, and you have decided that you are going to biopsy a given area, was that due to the fact that there's acetowhite, or vessels, or any of the standard colposcopic lesions to biopsy there. Was it due to the fact that that co-localized with the blue spot, or was it due to the fact that you only saw a dense blue region which was not associated with acetowhitening.
DR. SNYDER: What if they were separate but actually in close proximity?
DR. WRIGHT: In close proximity. In that case we really did not monitor pictures on the colposcopist. The instructions were to do as we said. If they had an acetowhite lesion with a little bit of blue in it, and they took a biopsy adjacent to directly where it's blue, would they have classified that as a both, or as an independent. They were instructed to do that as independent, but again, those are kind of subjective calls.
CHAIR NOLLER: Thank you. Yes, Dr. Weeks.
DR. WEEKS: Jonathan Weeks, maternal-fetal. I'm a little unclear on the Pivotal Study II early termination at about 227 patients. The protocol, did it not, state that an interim analysis would be done at 100 to 150 patients? There seems to be a disconnect there, and I'm a little -- I just can't seem to understand how that came about.
DR. WINGROVE: The protocol did specify an interim analysis, I think 150 subjects or so. The reason we didn't stop at that number to do the interim analysis is because the clinical sites were coming online in sequence, and we were concerned that the first two sites that were coming on would disproportionately affect our interpretation. So we wanted to have all of the sites, or a lot of sites online before we did an interim analysis. So we delayed the interim analysis for another month so that I think six or seven of the sites were online, and we had data from a representative sample before we did the interim analysis.
CHAIR NOLLER: That was Dr. Wingrove. Dr. Gerbie, you had a question?
DR. GERBIE: M. Gerbie from Chicago. Back to continue with numbers. PSII, it was decided you needed X number of numbers 500. This was decided it wasn't enough, you needed 700. And then we ended up with 200 in the proximate. How do you justify any of the three numbers and then the conclusion?
CHAIR NOLLER: Dr. Wingrove.
DR. WINGROVE: I'm going to play statistician here. Our original protocol made assumptions regarding prevalence of disease, the distribution of Pap SIL in that population, and generated a null hypothesis based on assumed presumption of disease. And under those three assumptions, we came up with the number of 528.
Shortly thereafter, after the study was ongoing we did have the Pivotal Study I data, and what was clear from that data was that our prevalence was a lot lower than we had predicted, from 33 percent to 27 percent. So as a result of that, we had to make adjustments. We clearly did not want to increase the sample size to a greater number, but based on what we had seen that would be a necessity. So that's what we did. We looked at the prevalence, adjusted the sample size accordingly, and changed the thresholds accordingly as well.
DR. GERBIE: But you didn't follow them. You increased.
DR. WINGROVE: Absolutely right. We increased the sample size.
DR. GERBIE: I mean, you had the necessity, just said, to increase it, and then you went below the initial.
DR. WINGROVE: That's what our study protocol would have predicted, but we did terminate the study, not for any reasons related to any study findings, but simply out of financial necessity.
CHAIR NOLLER: I think we'll break there. Dr. Wright, quick?
DR. WRIGHT: Just a clarification on that. Tom Wright. The board of directors decided on purely financial reasons due to the investment environment, the amount of money which has been spent already on Pivotal Study I, to stop funding clinical studies. So it was a purely financial decision. We had not seen any of the data from PSII at the time that decision had been made. We had seen data from PSI. It was purely financial.
CHAIR NOLLER: Thank you. We'll break now. We will -- wait a minute, not quite yet. We will convene at exactly 1 o'clock. The panel members, your lunch will be in the restaurant. Tell the host or hostess you're a panel member and you'll be directed to our secure site. And panel members, please remember to refrain from discussing these matters during the break. Thank you.
(Whereupon, the foregoing matter went off the record at 12:17 p.m. and went back on the record at 1:05 p.m.).
CHAIR NOLLER: Okay, I'd like to address the panel first, please. We will have questions of the sponsor and FDA now for a few minutes. I do want to focus everybody on what we have to do. Over the next couple of hours, our main job is to go over the questions that have been posed by FDA. Remember that our job is to give FDA advice, and they have asked specifically that we give them advice about nine different things. You have the questions in front of you. So that's the main job that we have this afternoon.
Now, in trying to be able to answer those questions, if we haven't heard from either the FDA or the sponsor what we need to be able to answer those questions, we should address some more questions to FDA or the sponsor. But we also don't want to get into a big discussion with either the FDA or the sponsor. We're supposed to be discussing this among ourselves. But I know that there are a few items that need clarification, and people have suggested to me they would like to ask more questions. So we want to certainly be able to ask all clarification questions, I just wanted to make that distinction. Once we finish with the questions of FDA and the sponsor, then we will start our own discussion among ourselves about the questions that FDA has posed.
And I'd like to open this by asking the sponsor if they have an answer to Dr. Amir's -- I can't say your last name -- Amir's question of this morning.
DR. FLEWELLING: Yes, thank you. We do not have a slide on it, but I think you were asking about the error bars on the spectra that appeared in the publication. And actually we do have copies of that if you want to turn to it in your Volume 1 of your materials. The page number is 4-002 and 4-003. It's the small number in the right-hand, Volume 1, Section 4, Page 2 and Page 3. And I'll wait for you to get that. The number in the lower right-hand column, the Bates number there is 4-002. That's correct.
I apologize to the audience if you don't have those materials, but those are the spectra. And there on the right you see the standard error bars are shown. They're quite small. Those are the population standard errors. Does that address your question? Was that what you were asking about? That's the fluorescence spectra, and on the following page are the same for the reflectance spectra. And those are identical to the spectra that were in the published article. Does that answer your question?
DR. GANDJBAKHCHE: Yes, this is very -- they're so small that I haven't seen them.
DR. FLEWELLING: Yes, that's right.
CHAIR NOLLER: Microphone please.
DR. GANDJBAKHCHE: And the second question was the relative changes that you can see. The numbers, that you can say that it's 20 percent different, 5 percent different, or?
DR. FLEWELLING: I'm not sure I'm following you on that. Could you repeat that?
DR. GANDJBAKHCHE: If -- depending on the numbers that you put on a Y axis, you know you can have a 1 percent relative change, or 200 percent relative change.
DR. FLEWELLING: Oh, I see. I see what you're saying.
DR. GANDJBAKHCHE: Yes, I wanted ?.
DR. FLEWELLING: The reason why, those are absolute units, but they're in arbitrary counts. They're essentially proportional to counts on the detector. So those are actually in linear absolute units. So you could add any scale you want. It's directly proportional to the counts on the --
DR. GANDJBAKHCHE: Sorry about that, I don't buy that. What do you mean? You are showing a Y axis.
DR. FLEWELLING: Yes.
DR. GANDJBAKHCHE: And this Y axis is the number of counts? What are these number of counts?
DR. FLEWELLING: Oh, I see. I don't have that number. It's directly --
DR. GANDJBAKHCHE: Roughly the relative changes that you see, especially in the second figure, Figure 5.
DR. FLEWELLING: Yes.
DR. GANDJBAKHCHE: How much is between, for example, CIN2/3 and CIN1?
CHAIR NOLLER: If the Y axis for 1 was from 1 to 100, would those differences be one one-millionth of one, or would it be 40?
DR. FLEWELLING: Well, on the Figure 5, which is the reflectance spectra, the scale here would be about zero to 40 percent diffuse reflectance. So the difference between those two is probably on the order of I would say 2 to 4 percent. The mean differences.
DR. GANDJBAKHCHE: What is the noise level of your detectors?
DR. FLEWELLING: It varies a little bit over wavelength, but the overall noise level of the system, the signal-to-noise is about 200 to 1.
CHAIR NOLLER: Does that answer your question?
DR. GANDJBAKHCHE: Roughly.
CHAIR NOLLER: Roughly. Thank you. What questions do other members of the panel have? I'll start here. Dr. Cedars. By the way, we no longer -- the panel doesn't have to give their names, just people from the audience.
DR. CEDARS: I had a question for the sponsor, going back to the lack of benefit with the high-grade lesions. And your explanation had to do with the fact that those frequently have visible lesions and so colposcopy alone was good for that. Is it also variable based on the prevalence of the disease in the population? So in other words, you are going to have a much higher prevalence of having CIN2/3 in your high-grade Paps versus your ASCUS and your low-grade. And so if it is related to the prevalence, and if our screening criteria are going to change with HPV screening, and you are going to potentially have a population getting colposcopy that has a higher prevalence of true disease, will that decrease your incremental performance with the machine?
DR. WRIGHT: Tom Wright. Two issues there. One is that clearly the prevalence of disease in patients with HSIL is different than ASCUS and LSIL. What we found in our studies was that in HSIL Pap, about 50 percent of patients had a CIN2/3. With an LSIL or an ASCUS, it was around 10 to 11 percent had a CIN2/3. So prevalence changes dramatically.
One of the reasons that we feel that colposcopy may perform differently is expectation of whether or not a high-grade lesion is present. If you are in a clinical setting where you expect that there's going to be a high-grade lesion, you certainly may look at little more carefully than you do if there's a lower expectation. And that may explain, in part, why we did not have as much of an effect as we expected in HSIL, in PSI, because clinicians expected to find the disease. So that's certainly one issue.
The other issue you asked was what would the impact of HPV. And I think the role of HPV, certainly as we were using it today, in reflex management of patients with ASCUS, is to identify a subset of patients who have the same risk for having CIN2/3, as do patients with LSIL. That's what a major finding of ALTS was, was that by identifying those HPV-positive patients in ASCUS, they had a similar risk as patient with LSIL for having CIN2/3. In our study, we found that the patients with ASCUS had a similar risk for CIN2/3 as the patients with LSIL, which is one of the reasons we felt very comfortable pooling them together into a single strata. Did that answer your question?
DR. CEDARS: Yes.
CHAIR NOLLER: Thank you. Any other questions, Dr. Cedars? I think I'll just go around the table. Dr. Weeks? Any questions? Dr. Julian? Dr. D'Agostino? Dr. Snyder?
MS. MOORE: I have one.
CHAIR NOLLER: Okay.
MS. MOORE: Given the fact that your geriatric population is increasing, I was wondering if any attention had been given to that older patient. I notice in Pivotal Study II the upper age was 64. I don't remember what the upper age limit was for Pivotal Study I.
DR. WALKER: Hi, I'm Joan Walker. There's no question that colposcopy is more difficult in women who are older. And I think in fact the LUMA will not perform as well in women who are older also. And that's unfortunate just because the transformation zone recedes up into the canal of the cervix with age. And so it is well known that you can have an abnormal Pap and not be able to find the lesion in older women. And that makes them have to go to cone, or another excisional procedure in order to find the lesion. And we do not expect LUMA to overcome that basic problem.
CHAIR NOLLER: Thank you. Ms. George?
MS. GEORGE: Yes, I actually have two questions. The first is you identified your second-most exclusion criteria as device malfunction. And I just wanted to confirm whether you meant the FDA's definition of malfunction based on the fact that you only had a very few number of adverse events. So I wanted to confirm that definition.
DR. WINGROVE: The definition for a malfunction --
CHAIR NOLLER: Name please.
DR. WINGROVE: Theresa Wingrove. The definition for malfunction was that the device failed to meet specifications. For example, in the most common pre-exclusion for device malfunction in PSI was the device failed to calibrate.
CHAIR NOLLER: And did you have a second question?
MS. GEORGE: Yes, I had one other question. And this may be my lack of clinical knowledge on this, but are clinical decisions only made through the use of the video, or retained version in the video, or is it also something that would be made on a paper version? And a second part to that is what is your retention storage factor on these views and screens?
DR. WRIGHT: We are telling the clinicians to do a complete colposcopy, identify sites where they would normally biopsy, once they have done that to then look at the LUMA image. It is a static image on a computer monitor. And based on that image, select additional sites for biopsy. We are not expecting them to go back to a paper copy, or to see any printout there, to do it off the LCD at the time. In terms of storage, Ross is going to address what the storage capacities are.
DR. FLEWELLING: We do not --
CHAIR NOLLER: Name, please.
DR. FLEWELLING: Ross Flewelling. We do not at this time provide any output of the stored data. So that's not available.
CHAIR NOLLER: Dr. Sanfilippo?
DR. SANFILIPPO: Yes, I have a technical question, and that's related to, as I understood it, the fiberoptic system. I want to draw an analogy to fiberoptics in an operating room, where with time, any of these fibers are no longer viable, et cetera. And I'm curious to know, (a) are there any long-term studies, and (b) is does the calibration at the beginning of this compensate if some of these fibers are broken, not viable, et cetera?
DR. FLEWELLING: Yes, Ross Flewelling. In terms of the fibers, we have done life testing on those fibers. I believe it's something like over 10,000 shots. And we have shown that it lasts for the lifetime of the product, which is at least five years.
In terms of actually monitoring, we have a hierarchy of calibration standards. Of course we calibrate it coming out of the factory, but we have a required maintenance at least once per year where we actually test the system against NIST traceable standards. And then at run time we put the probe into the calibration port, we go through a series of statistical tests, not only to calibrate the system, but also to verify that it's within its normal boundaries of operation. So those are the different procedures we have for verifying that it's continuing to perform.
DR. SANFILIPPO: Thank you.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: I'd like to -- a couple of questions, please. Number 1, relating also to the age at either N, since young women 18 and under were excluded, will the recommendation for its use not include people in that age group?
DR. WRIGHT: We are only asking for labeling in patients who were studied, so the age will be 18 and over.
DR. GERBIE: The likelihood is that people are -- I'm sure Dr. Hillard has more to say, but it's kind of difficult technically to separate 17 -- we probably should be doing less colposcopy on these young women under any circumstance, but since the findings even with the ASCUS and low-grade were more prominent in the young woman, now you're excluding a young age group. Am I not mistaken?
DR. WRIGHT: Our indications for use, Theresa? Could you, just in terms of the specific wording? The indications for use actually does not have the specific wording for age cutoff. And typically things are looked at in studied populations. Certainly I would think if the panel had a recommendation that you would want to see younger patients, would be -- there's no reason to expect this device -- we see no reason to expect this device to perform differently in women of reproductive ages, whether or not they are 18, 24 or 30. We've shown in PSII no effect of age. We've looked at age multiple ways. Only the Bayesian analysis of the FDA can find an effect of age. When we exclude women under 23, we see the same absolute gain in true positives in women over 23. So we just don't see an effect of age. And there's no biologic reason to expect one.
CHAIR NOLLER: Dr. Gerbie, we have a question on age too, and I think we definitely want to discuss that as we go on. Did you have another question?
DR. GERBIE: Let me ask another question. We know that approximately 50 percent of CIN2's will regress spontaneously. We know we have a 4.7 percent increase in true positives, of which significant numbers of those are CIN2, which will regress spontaneously, particularly in the younger age group. And that's why I have trouble with Slide Number 75 saying this is a compelling improvement.
DR. WRIGHT: The natural history of CIN2 is that some of the CIN2's regress, we agree. I think the ALTS data you saw in the presentation this morning clearly shows that some CIN2's regress. However, some CIN2's do not regress, and most people believe that good clinical cutoff for high-grade versus low-grade disease in the U.S. is at the boundary between 1 and 2. Having said that, CIN2 per se is a relatively uncommon diagnosis. In the ALTS data, and we have Dr. Stoler here who has done the pathology for ALTS, and he can address the percent of CIN -- no?
CHAIR NOLLER: We want to just answer Dr. Gerbie's question. We don't want a discussion of CIN2.
DR. WRIGHT: Okay. Clearly, CIN2 is an uncommon diagnosis, pure 2.
CHAIR NOLLER: Thank you.
DR. WRIGHT: Did that address your?
DR. GERBIE: Thank you.
DR. WRIGHT: Thank you.
CHAIR NOLLER: Thank you. Give the sponsor a second.
DR. COX: Dr. Gerbie, that's a good question, but when we look at --
CHAIR NOLLER: Who are you?
DR. COX: Oh, I'm sorry. Dr. Tom Cox. Dr. Gerbie, when we look at the ALTS data, that 7 percent increase obviously accounted for some regressed CIN2, which I think was estimated to be about 40 percent in the ALTS trial, and probably for some new incident CIN2/3 which we can't really estimate. So I think that when we talk about this 4 versus 7 percent, we're really dealing with a fluctuating, coming and going disease which we can't totally account for, but that that's a real gain. I don't think that it's -- because out of that 4 percent, if we have some that regress, we probably would have had some new incident disease too.
CHAIR NOLLER: Thank you. Any other questions, Dr. Gerbie? Dr. Jiang? Dr. Amir?
DR. GANDJBAKHCHE: I have two questions. The first one is your method, the strength of your method is quantitative. And you claim that there will be no subjectivity of anybody who will calculate the numbers or the colors that you are putting on the screen. If you have such a clean spectra, which they don't overlap, how you claim that you have false positives? What are the reasons that you have a false positive? This is the first question.
The second question --
CHAIR NOLLER: Actually, why don't you -- let's do that one, and then the second.
DR. FLEWELLING: Yes, Ross Flewelling. As you saw those spectra, those were the standard errors, and so of course the population distributions do overlap. And as with any classification method, there is a tradeoff between true positives and false positives, between sensitivity and specificity. So we optimized the system to have enhanced sensitivity and comparable specificity when compared to colposcopy.
DR. GANDJBAKHCHE: Okay. My second question is are you using the whole wavelengths, or you have principal components, and if so, what are the choice of your wavelengths?
DR. FLEWELLING: Yes. We certainly collect the entire visible spectrum from 360 to 720 nanometers. However you're absolutely right. When we go to looking at the principal components or the feature extraction, we narrow down the wavelength range. The wavelength ranges we actually use are typically between 400 and 600 nanometers for that part of the analysis.
DR. GANDJBAKHCHE: And why?
DR. FLEWELLING: Excuse me? And why?
DR. GANDJBAKHCHE: And why?
DR. FLEWELLING: Because in the early pilot development of the algorithm we found that that was the optimal wavelength. The signal-to-noise falls off below 400 and above about 600, and we have not found that to be useful in terms of signal-to-noise and the overlapping of the spectra in that wavelength range.
DR. GANDJBAKHCHE: And functionally speaking, you know you are doing functional image, fluorescence or reflectance. What are the species that you are interested in wavelengths, that you think that you are taking out to discriminate between different tissue types?
DR. FLEWELLING: Actually, I would not say we're doing functional imaging. I believe that what we're doing is we have an image of the cervix, and we're overlaying a diagnostic evaluation at the 1-millimeter level that's been smoothed. So we're not necessarily claiming any -- unless I misunderstand your use of the term "functional". Anything further?
DR. GANDJBAKHCHE: No, thanks.
CHAIR NOLLER: Dr. Miller, any questions?
DR. MILLER: I had asked a question earlier about whether or not there was any comparative evaluation in the Pivotal Study I of the colpo only arm for the patients who had a LUMA scan, but there was no mention of how that LUMA scan might or might not have been used. And that question wasn't answered.
DR. WINGROVE: This is Theresa Wingrove, and we did not use that LUMA scan. We had it stored, but we didn't use it. It would have been useful if you can locate the biopsy to that scan then it would have had some limited use, but we didn't do any analysis with that data.
DR. MILLER: Was there another? The other question I had was what do you anticipate the extent of training that will be required to educate a clinician that's not apprised of this trial to be up and running? How will that take place?
DR. WRIGHT: There are going to be two components to our training program as envisioned. One will be onsite training at the time of device installation, and there will also be a hands-on interactive looking at pictures, et cetera, in order to interpret the displays. What we found was that this was very effective with our investigators. They were able to learn to use this device in a short period of time. And we also looked at learning curve. Was there a difference between the first 10 patients and additional patients with investigators as far as their detections. And we found no learning curve after this. So we believe this training program will be effective. It's really reasonably simple. You have blue dots. There are a few areas you want to exclude. Other than that, it's pretty simple.
CHAIR NOLLER: Dr. Hillard?
DR. HILLARD: I have a question about age and a comment that, first of all, that one of the things that has changed since the initiation of these studies is the guidelines for initiation of cervical cytology screening. So my question is do you have any information at all about the percentage of adolescents in particular who would have met the newer screening guidelines recommendations for initiation of cervical cytology screening three years after first intercourse after coitarche. Do you have that information?
DR. WRIGHT: We know the age range in Pivotal Study I was 15 to in the 70s. I showed you the prevalence curve. The big increase in patients is around age 18. There are very few 15- and 16-year-old patients.
DR. HILLARD: So you don't specifically have that information?
DR. WRIGHT: We don't have analysis done specifically for women under the age of 18. You've sent the under 21 analysis. We also showed you the under 23 analysis. In terms of time since having first intercourse, we did not get that information and did not analyze it.
DR. HILLARD: Thank you.
CHAIR NOLLER: Thank you. Any final questions to the sponsor? If not, our job now is to discuss the questions that were raised by FDA. And you will notice that the first seven questions deal with safety and effectiveness, 8 and 9 deal with label and training, and 10 post approval study. And you all have these in front of you, so I'm not going to read them in their totality. I'll try to paraphrase them so we can move on and discuss them.
Question 1 deals with -- 1(a) really -- deals with the Pivotal Study I. And the entire population, not just the ASCUS/LSIL population. And the question here, or what we're asked to do is discuss the strength of the findings of this study relative to the proposed indication. And the proposed indication, let's remember, is as an adjunct to colposcopy for ASCUS and LSIL Pap smears only. They're not asking for a statement concerning its use in HSIL patients. Dr. Gerbie, do you have any thoughts?
DR. GERBIE: The numbers are given in absolute and then in relative improvement in the true positives. But it seems counterintuitive that when you have two techniques, and the second technique is supposed to add to the improvement, when you look at them individually the first technique is better. In other words, the colposcopy alone figures, the true positives were higher numbers than the LUMA addition. And I know you're not proposing that they be done together, but from the way I reviewed this in PSI, the figures were better in true positives. And along with that you had, again, the relative improvement of about 20-some percent of true positives as compared to the true for actual 4.7. But I don't see the same comparison with the false positives. The false positives are given as an absolute number, and not as a relative number.
CHAIR NOLLER: Any panel, more discussion on 1(a), thoughts about it? Yes.
DR. D'AGOSTINO: We're talking about 1(a) and looking at this particular?
CHAIR NOLLER: Yes.
DR. D'AGOSTINO: Where does the colpo look better in the true positive. There is a slight increase in the LUMA. Am I missing something? It's not statistically significant.
DR. GERBIE: No here, the LUMA plus colpo. But in the --
DR. D'AGOSTINO: Oh, the LUMA alone.
DR. GERBIE: Yes, when looking at the two individuals.
DR. D'AGOSTINO: Yes. I found it hard that they could really sort out that information the way they ran the studies also.
CHAIR NOLLER: The other thing is too when you use two tests, even if it's the same test, if you use it twice in a row, you know, the second time you'll pick up some more. If you do Pap smears 10 minutes apart, the second one will pick up a couple that the first didn't. So anything you add will add some, as long as it has some relevance to the disease under study.
DR. GERBIE: I think also an important point here is that the true positive is not statistically significant. I mean, it just can't.
CHAIR NOLLER: I have some trouble with the original population, even discussing at all. Since that isn't the indication that's being asked for, it seems to me to focus not the ASCUS/LSIL results really are the ones that we should look at. If they're asking for indication for everybody, then the original population would be fine, but it seems like it's the subgroup that's the one to focus on.
DR. D'AGOSTINO: But this is what the study -- it's a good question, but this is what the study was designed to look at.
CHAIR NOLLER: That's right, it was.
DR. D'AGOSTINO: It is a post hoc. So by moving, as you're suggesting, it moves us from the original protocol and we're now ready to say we're willing to have flexibility in terms of discarding somewhat the original protocol.
CHAIR NOLLER: (b) sort of follows on this, too. Really a question about whether it's clinically relevant to focus on the ASCUS and LSIL.
DR. MILLER: It seems to me, I mean, based on the comments that have just been made, that the issue at stake here is in this relatively low-risk population for which we know that there is maybe 30 percent or so that are missed, would another directed biopsy by a skilled colposcopist be more likely or less likely to yield a positive result versus a LUMA-directed biopsy for that third or so. When they looked at the number of biopsies, about -- the LUMA group got about a third more biopsies. So for the colposcopist, is the colposcopist going to do a better job, or is the colposcopist going to do a better job with the LUMA at his or her side collecting that second biopsy. And that's probably a question to direct to the colposcopists on the panel.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: In looking at some of the CDs at the various examples given, I came to the same question. If I had a concept of just taking another biopsy from an area that with a low-grade ASCUS Pap smear I may take only one or two biopsies. Now I'm saying I don't have a LUMA, but I'm going to take an additional biopsy. I'm going to take it from another area that probably wasn't as significant to me. I must say that looking at the areas in the annotated pictures with the LUMA addition, they would frequently be from areas that it would not have taken. So it's more than a random biopsy, but it may or may not hit the area that the LUMA suggested.
CHAIR NOLLER: Amir?
DR. GANDJBAKHCHE: I'm not an MD, but by looking at the number of false positives for colpo and LUMA, it seems that the doctors are aggressive to do a lot biopsies, and they get a lot of false positives. By looking at that, if there is an area that is not picked up by colpo, and can be picked up by LUMA plus colpo, although I don't understand the science behind it a little bit, but that's a plus, I think, knowing that you are doing a lot of biopsies. Why not do another one that can help the patient?
DR. D'AGOSTINO: But doesn't the data say that they're getting more false positives but they're not increasing the true positives, in this ALTS data?
DR. GANDJBAKHCHE: I don't know. These numbers, 2 percent, 3 percent, 8 percent, I am not a physician. But as long as I see that it's the patient is there, they are doing the biopsy, and they see some other places that can be disease, you know, I'd say okay let's do it. If you look at the mean number of the biopsies, you know, it was around 1.03 or 1.9, and I don't talk about the cost, I don't know about it, but doing another biopsy in the region that the colposcopy hasn't seen, that's I think a plus.
DR. D'AGOSTINO: But wouldn't you want their data to show that by doing that extra they improved the true positives?
DR. GANDJBAKHCHE: The true positive for me is an absolute number. It's three women, it's five women, it's 10 women that maybe they will be saved from getting cancer. This is my view. I am not a statistician, I don't like too much statistics in this way, but these are my views by looking at the things.
DR. D'AGOSTINO: I mean, you know, but if the data doesn't show that by taking extra -- more declarations of positive that you're producing more true positives, I think we're hardbound to argue that the extra is helping.
CHAIR NOLLER: I'm a little bit blind, so if you'd raise your hand and I'll try to take Julian and Miller and then Snyder, looks like.
DR. JULIAN: Well, this morning --
CHAIR NOLLER: In the microphone, please, Tom.
DR. JULIAN: I believe it was Dr. Huh from the University of Alabama presented these examples of areas that colposcopists didn't biopsy. And I believe they're the same slides that appear in the information here. I would have biopsied those areas because you have an abnormal Pap smear, and you have acetowhite epithelium. The supposedly expert looked at this and said this isn't worth biopsying. That's not what colposcopy was designed to do. You don't predict CIN1, 2 or 3 from the colposcopy. What you do is it's an adjunct to biopsy. Colposcopy is an adjunct to biopsy. If you do not biopsy abnormal areas you will miss disease. And the false impression that you're an expert and can tell this is just that, a false impression. And these slides, if you show them again, I will show you. You can blind me and I will show you were we will have biopsied, and on at least two out of three of them there was acetowhite epithelium that did not get biopsied. They just chose not do a biopsy on the basis of colposcopy. That's a little concerning. So I think this, if nothing, it forces you to do a biopsy instead of relying on your opinion which was wrong.
CHAIR NOLLER: Dr. Miller?
DR. MILLER: I was just going to -- I think the comment was made that there was no benefit. And you know, again, I grant you that they came to this conclusion late, but they've now decided that this technology best serves the ASCUS and LSIL group. In that group there was a benefit, there was a 3 percent benefit of true positives for those additional biopsies that were taken.
DR. D'AGOSTINO: But it's still not significant. It's still not statistically significant. I mean, numerically. You see, my concern is that once you shift these subpopulations and so forth, clinically they may make a lot of sense, but we wouldn't be talking about it unless the data did show that it looks like it's in right direction. And so what we're doing is taking a post hoc analysis, a post hoc look that came out in the direction and now is being presented to us. We have no real way of judging from a statistical point of view, and you can discard statistics if one likes, but we have no real way of judging from a statistics point of view that what we're seeing is real or just a random fluctuation. And so you have to be very much driven on somehow or other clinical intuition and clinical expertise, and then also that the non-significant results they have here from a statistics point of view is somehow or other irrelevant to our discussion.
DR. MILLER: But, I mean, you're the statistician in the group, but the boundary of 5 percent is just a boundary. And this, you know, yes it doesn't achieve the boundary, but their point was that that in combination with the second pivotal study both lead in the direction that there is some additive value. The question is is the additive value this technology, or is the additive value just that the technology -- the additive value is that the technology spurred the clinician to do a second biopsy, and that that biopsy.
DR. D'AGOSTINO: But the second study showed an additive benefit if you're willing to change the target from 2 percent to 1.5 percent, and willing to change the definition of the group you should be looking at. I mean, it's a very uncomfortable route that we have to take.
DR. MILLER: Okay.
DR. JULIAN: Just to reinforce one thing, it doesn't necessarily spur the clinician to do a second biopsy. Not all these were second biopsies. Sometimes they chose not to do a biopsy and were wrong. Because they're not all second biopsies.
CHAIR NOLLER: I have Dr. Snyder, Dr. Jiang, and then Dr. Gerbie.
DR. SNYDER: In response to I think Amir's question about there's some incremental improvement, but it's not reaching statistical significance. From a clinical standpoint this is not the last opportunity to pick up somebody's disease. In other words, these patients got here because of the screening test, and that's the cervical cytology. And then they are subjected to another screening test, and that's what colposcopy is. And we're looking at the addition of a third layer of screening test, which is the LUMA. These patients, and we know that the tried and true Pap and colposcopy are going to miss some patients. But these patients are going to continue in a more heightened surveillance scheme with at a minimum repeat cervical cytologies. And the time course for a lesion to progress from CIN1 to cancer or CIN2 to cancer is years and years. So I do think we should hold to the science, that there should be a statistically significant increase if we're going to add a third screening test, knowing that we're still going to have opportunity to pick up disease in these patients. Am I making myself clear?
CHAIR NOLLER: Let me suggest, let's have the two other people I've already noted, and let's sort of think about (b) and (c) a little bit and move on a little. Dr. Jiang?
DR. JIANG: I guess one of my comments has to do with increase of sensitivity -- sorry, increase of the true positive rate. I understand it is short of significance, but it's very close, so in my mind it's very close to make it. But then related to that, the flip side of that, I have a question to the clinicians. It seems clear to me there's an increase in false positive rate. Clinically, whether that's significant. I came from a background in mammography. In that area, false positive is an issue. Whether it's an issue here that I need to consider.
CHAIR NOLLER: Mel, you want to address that, plus your comment? You're a clinician.
DR. GERBIE: No, I can't answer that question. My mind is thinking about what I wanted to ask.
CHAIR NOLLER: What was your comment?
DR. GERBIE: Well, it goes along with Dr. Julian and Dr. Gandjbakhche, the false negatives. And maybe Dr. Wright and Dr. Cox as clinicians, and Dr. Huh. If one is going to continue to do colposcopy without changing the paradigm of deciding what should be biopsied by colposcopy and adding the LUMA, then one would biopsy the white epithelium, the abnormality that you see. I know that when I do a biopsy and I get a false positive, if I biopsied and I get nothing to me that was a false positive reading of colposcopy. But if I get -- if I think this is low-grade, the Pap smear's low-grade, I think colposcopically it's low-grade, and I biopsy it and it comes back as low-grade, to me and to most people that is not a false positive. And I think this is a statistical difference. And I'm not sure you're suggesting, I don't hear you suggesting that you change that idea, saying you don't need to biopsy what you see on colposcopy, just biopsy what the blue dots say. I don't think you're saying that. So therefore, you're still going to be biopsying the low-grades because some significant numbers of those were higher grade. And you're happy to say, `Mrs. Jones this is a low-grade lesion, this is likely to go away.' Or it didn't show anything, which I told you ahead of time it might not. But these are not true false positives anymore, when you really look at what the real world says.
CHAIR NOLLER: Let me make a comment about your question about the biopsies. It's very different from mammography because the morbidity of a needle biopsy or some sort of guided biopsy in mammography is much more morbid. The cervical biopsies fortunately really carry virtually zero morbidity once the speculum's already in place, you're doing a colposcopy already in the office. To do a second biopsy is minimally uncomfortable, and really has no sort of problem with care. So whether you do one or two biopsies is immaterial, really. It's like getting stuck, apparently, stuck for a blood draw extra. So it's not, if you will, a big deal, I think even to the people there, once they're in the office and having this done, if they have two biopsies instead of one.
DR. JIANG: What about no biopsy versus one biopsy?
CHAIR NOLLER: I think the big inconvenience it seems is having to come to the office and have the exam, the colposcopic exam. The biopsy, fortunately, if you keep your instrument sharp isn't terribly uncomfortable. Do other people agree with that? I have never had one.
DR. GERBIE: It's still easier and more pleasant for a patient to say you didn't need any biopsy.
CHAIR NOLLER: Right. That's right. And the anxiety of waiting for the biopsy to come back, of course.
DR. GANDJBAKHCHE: May I say something? I think the good thing about this system, the thing that I like about it, is it's adding another kind of imaging to the eye. The eye also is one of the best imaging systems, it's a log system, and you can look, you guys all, they can look, and try to identify things. If the system that they are claiming is quantitative, and using some other feature of light, I think this is an -- and I think this will be the trend soon. In this sense, it's two imaging systems that they come together with two imaging methods, one the eye of the physician and the second one is a quantitative method.
CHAIR NOLLER: Let me move us on a little bit. We've spent half an hour on (a) of the first of 10 questions. But I think this discussion is the thread that runs through here. (b) and (c) really deal with is the ASCUS/LSIL the right population for this technique to be used in. Dr. Cedars?
DR. CEDARS: I guess I still want to go back to the prevalence issue, particularly because of the statistical significance is close but not quite, and we keep getting to a smaller and smaller select subgroup. It makes me a bit uncomfortable. And given whatever you want to say about the age distribution, but given the presentation by the FDA in terms of the age, and that almost all if not all of your true positive increase was in the very young population, and if we're going to be changing our criteria in terms of screening. And it seems as though the sort of lower risk population you go into, you get away from the high-grade into the ASCUS, you get away from the older women and into the younger women, the more benefit you derive from your additional screening tool, then I wonder as we start to not screen some of these younger women, as Dr. Hillard was saying, will that have an impact on the incremental benefit? And since you're so statistically marginal looking at all of the patients, if you screen less younger women, will you have even less significance? That concerns me.
CHAIR NOLLER: Comments on that? Any other comments about the ASCUS/LSIL? Is this the right group?
DR. MILLER: Yes, I mean I actually -- my question would be, on the dark side of the force. Is there any likelihood or propensity for the clinician to become lazy in the implementation of this technology, and just basically look at the blue dots, and forget about the careful colposcopic exam upon which this was all predicated.
CHAIR NOLLER: Can we hold that till a later question? That's a very pertinent comment, but I think we'll get to that a bit later. Let's don't forget that, that's important.
What about the age, Question 2, the observed age difference in PSI. Again, we're still in PSI. True positives was observed primarily among the patients less than 21. Clinical implications of this finding. Does it matter what age? Yes.
DR. D'AGOSTINO: Could you remind us, the sponsor remind us, was the age cutoffs part of the stratification at the beginning of the study? I got lost in whether the 21 came completely post hoc or if it was a stratification.
CHAIR NOLLER: They were stratified in PSI but not PSII as I remember, and Dr. Wright, can you answer that in a sentence?
DR. WRIGHT: Age was not part of the stratification. The 21 came from a -- we stated we would do an analysis based on tertiles, and 21 was the expected tertile pre-study.
DR. D'AGOSTINO: So it was really post hoc.
CHAIR NOLLER: Thank you. Other comments about age? Dr. Hillard?
DR. HILLARD: My concern about age is that we do have some evidence that lesions are more likely to regress among adolescents. And so picking up more lesions isn't necessarily a good thing among adolescents if they are more likely to regress. And we do have data that suggests that that is true. So I am concerned about the age issue overall.
CHAIR NOLLER: Yes.
DR. SANFILIPPO: I would share the concern just by history. I mean, with the changes in the Bethesda classification in adolescents specifically, it's my observation that the number of LEEPs in other procedures has markedly decreased. And then by adding this parameter and finding that many more CIN2's, et cetera, theoretically we could be back to Square One again, where I know Paula and I have kind of worked at that goal to try to have longer follow-up and less invasive procedures. So I think that that has to be clarified. And maybe we need studies, very simply, that focus on patients who are 18 and younger, to address that question.
CHAIR NOLLER: One of the things we can consider is a recommendation around age, be it on either end. And I don't know if anyone wants to think about that, consider that. When we get to the end that might come up.
DR. GERBIE: I can't imagine somebody having this instrument, knows that it might be a benefit, and saying I've got a cervix here, I'm going to use it. I see a potential problem in the older person. I don't know how old that is anymore, but it's older than me.
DR. GERBIE: Where it's almost certain colposcopy will be unsatisfactory. And the fact that colposcopy is unsatisfactory, and that nothing is shown on the LUMA process still means that the normal procedure needs to be done in that patient, whether it's an endocervical evaluation, or some kind of endocervical evaluation, or even conization still has to be done. It can't be ignored. I think we need to make sure that the clinicians understand that.
CHAIR NOLLER: Ms. George?
MS. GEORGE: Isn't some of that part of your basic clinical practice decisions that you guys would be making anyway during this process, that whether the labeling clearly delineates, any indications for use clearly delineate that. I think you started to say it, that if you had a patient that presented themselves that you felt there would be value in using this, that you would end up going for that, whether they were 21 or 15 or 75.
DR. GERBIE: Well, I would -- it would be unlikely for me to say no I'm not going to use that for you, it was only tested in women of such and such an age. There's the possibility, why not. Since it is non-invasive, since I've -- there's no downside to it, then I would most likely use it. I'm concerned that people will totally come to rely on it as "colposcopy lite".
CHAIR NOLLER: Let's look at Question 3, because let's move on to Pivotal Study II. And to begin with, for the entire study population, the true positive endpoint was met but the false positive endpoint was not met. And we're asked to discuss the strength of these findings relative to the proposed indication as an adjunct to colposcopy for patients with ASCUS and LSIL. Part (b), then of course deals with the subpopulation. But let's think about Pivotal Study II. Any comments that are different from those that we just discussed in Pivotal Study I? Yes, Doctor.
DR. D'AGOSTINO: Just that, you know, here they pulled out of the air the number 2 percent that somehow or other added the increment should be greater than 2. And I'm uncomfortable in what sense, what does 2 mean. And if you look at the confidence intervals, we saw the reverse of what we saw before. Here we have on the 3(a) that we have statistical significance but it hovers. Instead of being 2 it's 2.2. And so I think should we have a discussion, or is there a need for discussion on what the 2 percent means?
CHAIR NOLLER: Let me. The 2 percent and the 15 percent were agreed upon between the sponsor and FDA as the criteria, through a long discussion and so forth. And I think that we could discuss whether or not they're appropriate. Maybe they were, maybe they weren't. But that was the standard to which the sponsor was to be held.
DR. D'AGOSTINO: That's fine if we take that posture, I have no problem with that. Then the sponsor later on wanted to switch to 1.5.
CHAIR NOLLER: That's our next question. That's 4.
DR. MILLER: But wasn't that agreement in conjunction with a different sample size agreement? And to what degree do we need to look at that agreement when, you know, for whatever reason, in this case a financial decision, the sample size was altered.
DR. D'AGOSTINO: It's interesting that they changed the sample size but didn't change the margin that they were looking at. I don't want to open up as a big discussion for us, but in terms of making sense out of this, in terms of what we call statistical significance or not, these issues do sort of roam around here.
CHAIR NOLLER: Other comments?
DR. GERBIE: Once again you have a 4.7 percent chance of finding something that the colposcopy didn't show versus the 18 percent. But looking at the number of patients, nine and 35, and taking individual patient, you have to tell her that I've got four times the chance of not finding anything extra by doing these additional biopsies.
DR. CEDARS: Well and also I don't think it's legitimate in PSII to go back to the original population if you've already sort of changed your criteria, and you're looking primarily at the ASCUS and the LSIL. Then you've got to go down and look at the secondary, you've got to go to 3(b) where it wasn't significant even by the 1.5 percent. You really can't, I don't think, look at the total population because you've changed your indication.
CHAIR NOLLER: Other comments? Looking at (b) also, as we just started to, that subpopulation again is the indication they're asking for. Any discussion on that before we go to the changes in the true positive? Dr. Snyder, then Dr. D'Agostino.
DR. SNYDER: Let me just make sure I'm at the same point. This gets back to the question, I didn't comment on it earlier, when you asked is this subset the appropriate set to be looking at. And from a clinical standpoint I think it is, because for the high-grade lesions, on a clinical standpoint, if we can't explain where a high-grade lesion is coming from, then we're going to do further evaluation of some kind. So the population that altruistically we would like to not miss a CIN3 lesion is this population. And then when we look at it, when they went specifically at this population, we didn't meet statistical significance either in the true positive or the false positive. But I think it is the right group that we would like to be targeting if we're going to add in another screening tool, another layer of screening.
DR. D'AGOSTINO: I was just going to say when we move to this, we have to then look at two no's, true positive and false positive. Both say no.
CHAIR NOLLER: Ms. George?
MS. GEORGE: If I remember correctly, looking at the slide set, the subset of indication for use was only decided more recently, that it was not the subset originally when they first started both of the two studies. It seems like that the indication for use, that that subset was not made until June of 2004, which would have been after the studies. So I think looking at both sets of data is why it's included. But again, it's the more narrow focus of the indication for use is really a more recent decisionmaking.
CHAIR NOLLER: And clinically, when you think about it, if it's high-grade, if you don't see something, you have to go on and do something more. If you do see something, you have to go on and do something more. So the one that gives us trouble are the ASCUS and LSIL. So that indication seems to me to make more sense than the whole spectrum. Anybody else have a problem and think it should be the whole spectrum? I see a lot of -- I guess we're pretty well agreed on that, that the LSIL/ASCUS is the right one.
Question 4. The sponsor proposed two changes in the analysis of the true positives for PSII compared to the original protocol. The first was reducing the increments needed for success. This is where we go from the 2.0 to the 1.5. And assigning success criteria to each Pap substrata, and reducing the denominator for calculating the true positive rate increment by excluding true positive patients who were identified by colposcopy. And when you make these changes, the study does meet the revised TP requirements for ASCUS/LSIL, but it doesn't for the FP requirement. What do we think about that? Perhaps our statistician could address that first.
DR. D'AGOSTINO: Well, if I look long enough I can make anything positive. What is the a justification for switching to the 1.5. What is the justification for the denominator. I mean, there's an on the face argument that the denominator should change, that you should remove those that you've already sort of clarified. But it wasn't done before, and my concern is that when you're doing everything post hoc, the only reason we see it before us is because it happened to turn out in the direction that the sponsor would like. We don't really have a way of judging how significant this is. It's just a set of data that sits before us.
CHAIR NOLLER: Other comments?
DR. SNYDER: I'll go back. You said we would discuss this again later. But I do think there may be some tendency, like in the Pivotal Study II that as a colposcopist we all like to think that we're perfect. But we really know we're not. But if as a colposcopist I'm challenged to pick the one single most abnormal appearing area out, I'm going to do that. And if I know that I've got now LUMA, you know, that's going to add to my ability to do that second biopsy, which I may have done, knowing that I wasn't going to put another diagnostic tool in front of me. When I teach colposcopy, I just stress the fact that it's a screening tool, and if you've got a 360 degree lesion, that adequate sampling is not going to be one biopsy. But I do think there may be a tendency, knowing that now we're going to swing another instrument in there, that I might be interested in saying, gee, you know, if found that abnormal area, what's the LUMA going to tell me. And then even if the LUMA doesn't tell me, I might go biopsy that second area again. And I might do a third biopsy based on the LUMA.
CHAIR NOLLER: Actually, that leads right into 5, because then here we talk about -- or the question deals with the false negative rates. In PSI, the estimated false negative rate for patients for LUMA is 23 percent if you use colposcopy as the gold standard, meaning that 23 percent of the CIN2/3 lesions found by colposcopy were not found by LUMA. In PSII, 22 percent of the true positive biopsies identified by colposcopy were not identified by LUMA. This goes onto the always/never rule we'll talk about in a minute. But there you have the estimated numbers. LUMA isn't perfect. Colposcopy isn't perfect. So the question is is the combination, the increase, worth the extra biopsies and so forth. More thoughts?
DR. GANDJBAKHCHE: I can say something.
CHAIR NOLLER: Amir?
DR. GANDJBAKHCHE: I think the whole story is here, you know, the thing that's exciting that there are lesions that were identified by one and not with the other, and vice versa. It showed that this is really an adjunct.
CHAIR NOLLER: Complementary.
DR. GANDJBAKHCHE: Yes. This is the whole story here. And I don't know about 1 percent, 2 percent, and so on.
CHAIR NOLLER: Dr. Snyder, then Dr. Julian.
DR. SNYDER: You know, I'm trying to figure out, not being a statistician, and I'd like to have somebody really comment on that. Because your point's very well taken. You know, again, when we're talking about screening tools, and we know that the Pap's going to miss patients, the colpo's going to miss patients, and we know LUMA's going to miss patients. And so you're right. If we find a third screening tool that's going to complement all of this, then I would very much welcome that. And if we knew that we were close now to picking up 100 percent of patients we probably wouldn't even be having this discussion. If the test was so complementary that we were in that slide where there's that little tail end of 7 percent. If we were eradicating just about all of this, we would probably already be on our way home. So I need a statistician to tell me how much are we obliged to rely on these SAT-defined statistical parameters?
CHAIR NOLLER: Let's hear from our statistician, then Dr. Julian, then Dr. Weeks.
DR. D'AGOSTINO: The point of the fact is that the statistical significance isn't there. And so the only way you get a comforting situation is by redefining targets, and redefining your denominator. So it's not -- I have to throw back to the panel, it's not really statistics anymore. Statistics said you should've stopped in the previous question.
And again, just to carry this one bit further, when you look at these intervals, we were saying before that it's almost statistically significant. If you look at the true positives, for example, the criteria for deciding that there's an additional yield with the LUMA was 1.5. The confidence interval starts at 1.6. The confidence interval is not very comforting in terms of have we even done a good job with the 1.5 percent. So the statistics would just put an end to the business. I feel the statistics would say you don't have right for that comfort.
CHAIR NOLLER: Dr. Julian?
DR. JULIAN: I've been trying to look at the numbers for this, and I think that at least from my perspective we might be losing the big picture here. The screening for this cervical disease is actually an attempt to detect, prevent, or treat early cancer, not CIN2/3 in reality. That's what Pap smears are for, and Pap smears have done that with a 70 percent efficiency in the last century. Now, Dr. Cox said I believe there are about six million abnormal Pap smears. If we say that 15 percent of those are CIN2/3, you're talking about a group of about 900,000 people. Now, you could get rid of one-third of those with HPV testing, right? Okay. And that would leave you about 200,000 people. So if you look at the cervical cancer statistics, there aren't 200,000 people a year, you know, of these 200,000 missed CIN2/3's that get cancer of the cervix. There's about 10,000 right now, about 10,000 on the latest statistics. Of the 10,000 that go on to get cervical cancer, half of those had never been screened with anything in their life. Nothing, zero. So that leaves you a group of about 5,000 people that are getting screened. Now of these, two-thirds are going to survive because of the treatments we have that exist today. Taking the number down lower, that you benefit maybe 2,000 people. Of those 2,000 people, another 5 or 10 percent of those have adenocarcinomas, or if they're very young they have such aggressive lesions they can't be treated at all. And if you assume the machine increases the pickup by 2 to 5 percent, you're talking about somewhere between 25 and 100 lives, from the whole process. Are you going to screen 900,000 people to save 25? I don't know. But cancer death is the reality. It's not really CIN2/3.
CHAIR NOLLER: Thank you for that perspective. That's interesting. Dr. Weeks?
DR. WEEKS: Again, I'm a maternal-fetal medicine person, so I don't do colpos anymore. But my concern with this is certainly the statistics don't look compelling. And against the background of Dr. Julian's excellent sort of broad view, my concern is will there be a tendency to swing in the LUMA. Really over time will colposcopists lean more and more on LUMA rather than really being diligent in their colposcopy. And as a stand-alone test, colposcopy we see is better than LUMA alone. So the combination of the statistics being less than compelling, and the whole question of how important it is to pick up in CIN2 in young women, some of the other issues we've addressed. And I think the tendency, and we're hearing that these colposcopists were very, very skilled, this technology to be introduced into an environment where there won't be as much diligence. I just am very concerned about the risk-benefit ratio.
CHAIR NOLLER: Dr. Hillard?
DR. HILLARD: I was actually going to say something very similar, which is I'm concerned about the psychology of being a colposcopist who will then have the tendency potentially to rely on LUMA and recognizing that LUMA is going to miss this 22 percent. So I do have concerns, and concerns about actual use, and what this scenario would play out.
CHAIR NOLLER: Dr. Cedars?
DR. CEDARS: I have another statistical question for our statistician, which gets back to Question 4 in whether or not the two changes are valid. The change to 1.5 I think is fairly arbitrary. But reducing the denominator with the true positive increments on one level, if what you want to know is incrementally how many more true positives do you get, then I think it's legitimate, I would think, to take the true positives from colpo out, except for the fact that if people start to use LUMA as a stand-alone, or if the colposcopic skills decline, you're missing some true positives with your LUMA alone, and so they really do belong in the denominator. So I could sort of statistically, I guess, slash clinically make an argument either way. And I don't know which would be stronger.
DR. D'AGOSTINO: Well, I agree with everything you've said. And I think that, you know, the protocol said we're going to have the denominator as everyone. And once you start changing that, there was a discomfort on my part, and a discomfort on the FDA's part. And I think the discomfort comes in the issues that you're raising, is how then solid is the denominator. If the denominator is everybody, then there's no question. Once you start removing some subjects, then you really aren't being able to follow. And the point you raised, that if the strategy for using the test changes, then the analysis we're looking at basically has very little help to us, even if it was the pre-designed analysis. So I'm very concerned that we get the results that one would like to see by these changes which are not pre-specified and do have the implications that you're describing.
CHAIR NOLLER: Let's take this a little farther down in Question 5 with the always/never rule. It's always to be used with colposcopy, you never eliminate a colposcopically directed biopsy. If you think you should biopsy something you still do, even if the machine says not to. And do we believe that the clinical impact of the false negative is adequately mitigated by the indication statement, which says you use it as an adjunct to colposcopy.
I will interject here something. As I read through the sponsor's material, it said that, but I didn't see any clear statement `Thou shalt not use this stand-alone.' And when we get to labeling, we might want to consider something stronger in their wording if we do vote for approval. But what does everybody think about to continue this false negative? Is it covered in the indication where it says it's to be used always with colposcopy? Tom?
DR. JULIAN: If it doesn't find abnormal vessels, and it probably won't tell you when you have a verrucous carcinoma with keratin from, you know, the reason for the acetowhite versus leukoplakia. It won't tell you about gross lesions. I think that yes, I mean you'd have to use it as an adjunct rather than a primary evaluation tool.
DR. SNYDER: It's also not going to assess adequacy, you know, of ability to be sure you've seen the entire squamocolumnar junction, nor is it going to tell you anything about lesions extending into the canal. Can't use it with an endocervical speculum.
CHAIR NOLLER: Certainly the colposcopists will have to decide whether it's satisfactory or not, yes. Other points about that?
DR. JULIAN: Cervicography, which has the same sensitivity, doesn't evaluate the canal at all.
CHAIR NOLLER: Number 6.
DR. D'AGOSTINO: Can I just add?
CHAIR NOLLER: Oh, yes.
DR. D'AGOSTINO: I'm not sure where it was supposed to come up and so I may be saying something out of order, but back to the comment, the question that was made about the changing of the denominator. I haven't really had a comforting discussion about the expertise of the people who were in the study. I think the sponsor said that these were really super experts. I always, you know, one of the things when I designed these two studies, is the protocols look very much alike but one group is really expert and the other is more like the group that may in fact run the procedure, and so forth. And as you start asking the question what do these numbers mean, and moving things out of the denominator and the always/never, I think they become much more compelling, much more important, as we talk about the expertise of these individuals, and how do we generalize these studies that are before us. And I don't know where that's going to come in, but I think that's going to be -- it should be an important consideration.
CHAIR NOLLER: Yes, Dr. Sanfilippo, then Dr. Gerbie.
DR. SANFILIPPO: Apropos to that comment, just drawing an analogy to the ALTS trial where there was an education curve of testing, and then basically I feel comfortable with that, seeing that there was some level of education. I didn't hear that similarly with this product, that there was either -- not necessarily a testing per se, but some signoff where each of the investigators were demonstrating a particular level of competence.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: Well, I think that Dr. Huh's discussion as to the level of colposcopists in this group showed a fairly wide range, all, though, with experience and I would think above average knowledge of the average clinician. But I would think that this process should only help the less expert person. They still have to follow their rules, and they may not have followed them yesterday, and they may not follow them tomorrow. But if they don't follow the rules, they're going to get in trouble.
CHAIR NOLLER: Other comments? Do we think this -- yes?
DR. JULIAN: I have to disagree in that I think the experts are less likely to follow the rules, make up their own -- It's true in colposcopy, they do less biopsies. And as Dr. Huh showed on these slides here, they skip lesions they should have biopsied, because they knew what it was. Whereas the basic tenets of colposcopy says you never know what it is by just looking at it. Is that true?
CHAIR NOLLER: Maybe you don't.
DR. JULIAN: They brought me here because I'm not an expert.
CHAIR NOLLER: Yes you are, and you're making some very good points Tom.
DR. GERBIE: Well, I think you can look at it both ways. I know the patients that we see as a referral center, a non-expert referral center, come from patients who are by and large over treated prior to our seeing them. I think that's the major problem that referral centers have is over-treating, and all of ours are from non-academic or large community institutions.
CHAIR NOLLER: Does that fit in with our discussion of safety here? I think it's clear that the machine doesn't electrocute you, or anything like that. We don't have any fear of infection. But as far as safety, is this going to be used in a way, or potentially easily used in a way that's going to hurt women? If so, then that's important for us to decide. Russ?
DR. SNYDER: And tell me if this is not pertinent, but I totally agree with Dr. Julian's earlier comments, his summary picture. The only thing that's been shown to reduce the mortality odds ratio of dying of cervical cancer is the Pap smear. We don't even have that data for liquid-based cytology. But we don't have that data for colposcopy either. And I think it's pretty much apparent that this machine doesn't have any safety concerns. I do have some concern that could the one thing that is a good screen, you know, I shouldn't use that term. But I would worry that are people going to be less apt to do appropriate cytology screening with this device. And I don't know if that's appropriate for this committee to discuss at all. Certainly -- and it's a mixed bag. Some people do an immediate repeat Pap prior to the time of colpo, and some people don't. There's pros and cons for that. With this machine we are taking out one other added safety check, if for no other reason screening for some sort of endocervical atypia to do a repeat Pap. You can't do that with this machine. Are others going to be inclined to want to follow patients with LUMA? I don't know if that's appropriate.
CHAIR NOLLER: Ms. George?
MS. GEORGE: I think people should revisit the definition of safety, and it does link back specifically to when accompanied by adequate directions and warnings against unsafe use. So that I think that we do need to remember that when we go to the later questions where we specifically talk about the labeling. Because if we feel that the device in general is safe, we want to make sure that there's that linkage back to the labeling.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: I know that in the study a Pap smear would preclude doing the test, but is that going to be a rule, that a Pap smear cannot be taken at the same time?
CHAIR NOLLER: Well, it couldn't be taken before. You could do it after.
DR. GERBIE: You certainly can't do it after because you've already done acetic acid. But can it -- can we ask the sponsor? It cannot or it was not?
DR. CEDARS: I thought they said that it could not because it would stir up bleeding and it would affect the reading.
DR. SANFILIPPO: And that was the criteria for the 7-day interval before the procedure was done, for that very reason.
CHAIR NOLLER: Just the scraping could change the reflectance.
DR. GERBIE: That's an elephant feeling here. We can have an answer yes or no.
CHAIR NOLLER: Let's ask the sponsor yes or no? Can a Pap smear be done before it, or is it a thou shalt not?
DR. WRIGHT: Tom Wright. Taking a Pap smear immediately before you do the scan can result in bleeding, and we do not want a Pap smear done immediately before the scan.
CHAIR NOLLER: Thank you.
DR. GERBIE: Kind of a follow-up is that the objection about bleeding, is heavy bleeding at the time of LUMA. Light bleeding, as with colposcopy, is still a subjective choice. One might or might not do it. Is that the same here? Are we worried about actually causing bleeding from the cervix that would obstruct the --
CHAIR NOLLER: Wait just a moment. Would you just address the question whether it can be done if there's any bleeding? Just that question.
DR. WALKER: Well, what I wanted to comment on --
CHAIR NOLLER: This is Dr. Walker.
DR. WALKER: -- is that you can assess the endocervix after you've completed your colposcopy. And many people use a cytobrush in the endocervix in order to assess the endocervix. And that can be done after the LUMA's been utilized.
CHAIR NOLLER: Can LUMA be used if there is any bleeding?
DR. WALKER: Only if you can clear it off. So you have to use vinegar to clear it off. In some patients you can, in some patients it keeps re-bleeding.
CHAIR NOLLER: Thank you. Okay. Should we move on a bit? We have a few more things to cover. Number 7, did the planned and unplanned analyses of PSI and II demonstrate the safety and effectiveness of this device for the ASCUS/LSIL population? That is, do the probable benefits to health from use of the device outweigh any probable risk? We'll talk about that first. Now we're down to the nitty gritty. Tom?
DR. JULIAN: I think one of the things that is questionable in terms of benefit and risk is that it appears that one of the groups that was most highly identified were the young patients with CIN2/3 lesions. And what will now happen to these patients that would have been followed?
CHAIR NOLLER: Microphone.
DR. JULIAN: I think they'll get an excisional procedure once CIN2/3 is found, which is something that -- I mean, we have two experts in pediatric adolescent gynecology sitting here, and neither one is me, by the way. And maybe they could speak to that. But I think it'll increase the number of LEEPs they get, which indirectly is a safety issue.
DR. SANFILIPPO: That was exactly my concern before, and I felt that at minimum to conduct the studies, and perhaps to identify that in the conditions of labeling. Because I share your exact concern, and there's history to support your statement. And with the change in the Bethesda classification, we have markedly decreased the aggressive, LEEP, and other procedures, and that was my point. I don't want to go back to that.
CHAIR NOLLER: Dr. Hillard?
DR. HILLARD: Well, clearly the rationale behind changing the guidelines about initiation of cytology screening was related in part to the concern about overly aggressive evaluation, management, and treatment. And patients getting LEEP'd and re-LEEP'd and overly problematic procedures. So that is a part. The other aspect of that is of course the fact that -- two things. One is that these lesions are more likely to regress in adolescents, and so that certainly is a concern. I am very concerned about this population, given what I am seeing right now. Those guidelines are not even now being followed, so I am seeing many patients who are having Pap smears done at a time prior to -- and of course the clinician should be allowed to use his or her judgment as to whether it's appropriate to do an initial screening. But I think that you start down that path, and you have an abnormal result, and you then follow it up with colposcopy, and then you get led down the path. And I believe that you're correct. I'm very concerned that this would lead to more procedures that do have the potential for harm in the future in terms of possible risk of pre-term delivery, et cetera. So I'm concerned where we may be led more down the path.
CHAIR NOLLER: Dr. Snyder?
DR. SNYDER: I not only share Joe's and Paula's concerns, but I'm petrified of them. And that is where we don't -- we don't want to find disease we don't need to treat in that age group. But I ask -- I didn't comment earlier on the age effect. Where I'm seeing this is we've got sparing statisticians. We've got, you know, this one model from the FDA and we've got the sponsor's model. And I don't, you know, we have to put that in perspective because from an actual clinician's standpoint, I don't see why if both techniques use acetic acid, why there should be any difference in our ability. from a colposcopic standpoint it doesn't matter. So why with this reflectance and fluorescence would there by really any difference in somebody less than 21 versus somebody that's 25 or 35.
But so then I get down, if we really don't know which side is correct, maybe the problem is we don't have enough numbers, and we don't have enough information like among these really young women with the high-grade lesions, what percent of them are CIN2, what percent are CIN3. I mean, nothing here stratified patients by that classification. And I see why they didn't. The reason we have low-grade/high-grade is treat and not treat. But now as we're understanding more of this CIN2 is going to regress, more of the higher grade lesions in younger women are going to regress, more HPV is going to regresses in the younger women, maybe it is pertinent to have some more data.
CHAIR NOLLER: Dr. Gerbie wants to make a comment. Let me just mention that we can later on decide as we're thinking about conditions, we can think about age. And perhaps something, you know, we can think about whether or not it should be used under age 18, or not under age 18 pending more data, or postmarket analysis and so forth. We can address issues of age in our deliberations because I think it's clear that we have very little information in women under age 18 from these studies. Dr. Gerbie?
DR. GERBIE: Well, we really should be discussing the cytologic management, not the histologic or colposcopic management. The thrust should be in not doing cytology on these young people.
Secondly, once colposcopy is chosen, then I think you're obligated to make a decision how far do you want to go to make the diagnosis. Some people kill the patient to save their life. I think we can go back a few steps. From hysterectomy, back to cone, to LEEP, et cetera. But if you decide to take a biopsy, and even if you get a biopsy of CIN2 or CIN3 in a young person, you still have the choice of how you're going to manage that patient. And how you're going to follow the patient if you choose a conservative management. You're going to put a better follow-up on that patient. It doesn't automatically say that yes, she has a CIN2/3, that now you need treatment. So there's an education thing. I don't think we can stick everything into one sock here.
CHAIR NOLLER: Amir?
DR. GANDJBAKHCHE: This question is for the chairman. Could you suggest to change their data analysis to get more effective?
CHAIR NOLLER: Can we suggest to change the data analysis? Certainly we can suggest that, but it won't be done today of course. But it would be one of the things that we could talk about. Actually, I don't really want to address that now because what I would say in response to that might prejudice our deliberations later, but we might talk about that later. But I'll try to remember that. If I don't, remind me. Dr. Cedars?
DR. CEDARS: Well, I have two comments. One is that I think it'll be very difficult, and I realize this isn't really the sponsor's responsibility, but it gets back to Dr. Miller's dark side. I think it's going to be very hard to say you have this machine in your office, but if somebody's under a certain age you can't use it. I just don't see that that's reality. That's not what's going to happen.
The second issue I have is while I agree with Dr. Gerbie that your choice of what to do once you have even a histologic finding is dependent upon you, except that now you have a young woman who thinks she has something really bad. And so you're also going to be driven by what she thinks she needs, and she wants this bad stuff gone. And so even though your advice might be to be conservative, and to watch this, and to follow it, the acceptance of someone who thinks they have a cancer, or pre-cancer, or some bad thing is to cut it out, get rid of it. And so I still think you're going to expose a lot of young women to procedures that they might not need.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: Well, I just feel that sticking your head in the sand is never the right thing to do, and never to presuppose what a patient wants. And I've gotten in lots of trouble over the years doing that. And we see patients on both sides, patients who insist on being treated for low-grade lesions, and patients who insist on not being treated for high-grade lesions when I really want to treat them. So the younger person, she gets a vote. And I've been surprised sometimes, but if it is a higher grade lesion, she, and if there's a parent around, has to be made to understand that the recommended treatment for a high-grade lesion is to be treated. We can watch this if this is what you want, but we have to watch it closer, and you have potential for going in either direction.
CHAIR NOLLER: Let's -- we've talked about (a), 7(a) quite a lot. 7(b), though, is a little different. It says, "Will use of the device provide clinically significant results." I.e., does it really matter if we pick up a few more. And we sort of talked about that. Does anybody have a comment specifically about that? And clinically significant results could be, as Tom Julian talked about, how many deaths from cancer will be prevented, if any, or it could be on the other side, you know, how may CIN1 lesions are gone. I don't know, what do people think? Yes.
DR. D'AGOSTINO: Quite often in these discussions you like to say we have statistically significant results. Now we want to interpret the confidence intervals, or the point estimates. We have to remember in this discussion we don't have statistically significant result. So you're pushing aside, which I'm not saying we shouldn't do, but we're pushing aside the statistical significance and focusing on these numbers, which may not have really sort of scientific validity in terms of a statistical interpretation.
CHAIR NOLLER: Other comments? If not, let's move on to Number 8, labeling and training. Does the panel have any comments on the labeling and instructions for use provided by the sponsor. This is where we talk about things like thou shalt never use without colposcopy, and age, and so forth. You've all seen the proposed labeling that was in the sponsor's manual. What things should be added or subtracted from that? Assuming that it's approved, of course. Then it would -- yes.
DR. SANFILIPPO: Well, I would clearly state not for use in patients under 18 years of age, as one of the criteria. And once again, that's prerequisite to designing the studies and getting that information. So I would personally like to see that.
CHAIR NOLLER: How do other people feel about the lower age limit?
DR. GERBIE: I think it does complement the recommendations of the cytologic evaluation. And I think together it gives more strength to not doing the Pap smear, and not doing the procedure.
CHAIR NOLLER: Yes, Dr. Weeks.
DR. WEEKS: I could certainly support an age limit of less than 18, but I would still have quite a few concerns about the young women who are evaluated at less than 21 years of age because of the number that may be exposed to LEEPs, and the impact on future fertility, childbirth, et cetera. So I think the study didn't include patients less than 18, clearly labeling should prohibit or discourage the use of the device in that age group. But I still think there's a great concern up till 21 years of age at least.
CHAIR NOLLER: Dr. Hillard, then Dr. Sanfilippo?
DR. HILLARD: I would agree with the concerns about under 21. I certainly would agree with 18, but under 21 given the questions about the statistical analyses related to under 21, and in addition the concerns about the practical utility.
CHAIR NOLLER: Dr. Sanfilippo?
DR. SANFILIPPO: The other category that we have not discussed today is pregnancy, and that was one of the exclusion criteria. And I think that has to be, again, clearly stated that there is no data in terms of its efficacy with pregnancy.
CHAIR NOLLER: Good point. Dr. Hayes? Microphone please.
DR. HAYES: We've discussed the younger client. I'd like to turn our attention also to the older client, that the sponsor also spoke with the older people were not in the study. And there was indeed an indication about how it really would not be appropriate most likely for the older client. And I think that needs to be stated.
CHAIR NOLLER: Ms. Moore?
MS. MOORE: I was just going to ask Dr. is it Gerber?
CHAIR NOLLER: Gerbie.
DR. GERBIE: Gerbie.
MS. MOORE: Gerbie, yes. When you said -- no you weren't the one, I don't think. Whoever it was who said that this should -- the label should say not to be used for patients under the age of 18. Do you use the word "recommended" so that that would give then the physician who is working with that young person some kind of wiggle room?
DR. SANFILIPPO: Again, I would tend to think that that becomes the FDA's decision. I would see where you're coming not recommended. My bottom line point is I'd like to not see it used, period, until we get adequate information. So I suppose if I had one word, it'd be stronger than "not recommended". Not to be used. That's my personal opinion.
MS. MOORE: Okay.
CHAIR NOLLER: Dr. Gerbie?
DR. GERBIE: Unless you have evidence that there's a difference in the biological behavior of a CIN2/3 lesion diagnosed with conventional colposcopy versus LUMA, then I don't see how you can make a recommendation for the difference in age.
CHAIR NOLLER: Yes, Ms. George?
MS. GEORGE: I have two questions on the Pap smear, and also on colpo. Are there limitations imposed on those already, and if there are not, what would prevent a doctor who already has a patient under the one who's 15-years-old to just not, if they've done it for the last six months on all their other patients, to just go ahead and do it?
And then the second thing is that there is the limitation -- it is not listed as a warning, or caution, or prevention -- in their labeling with regards to pregnancy. And they have quite a long list, and that may be what we want to consider is that it's just limitations. But again, practice of medicine, whether it's a warning, a caution, unless there is a full stop prevention ability to prevent you doctors from doing something, you're going to do whatever the hell you want.
DR. MILLER: We pretty much do whatever the hell we want even when there is a warning.
DR. MILLER: As terbutaline will attest to, which is still rampantly used even though the FDA has told us we shouldn't be using it.
CHAIR NOLLER: Other comments about this before we move on? Let's look at Number 9. Does the panel have any concerns about the training needed to use this device safely and effectively?
DR. GERBIE: Can I go back to the instructions again? I'm sorry.
CHAIR NOLLER: Sure, go ahead.
DR. GERBIE: I think exclusion/inclusion criteria for a study are different than inclusion and exclusion for clinical use. And I think these are probably more stringent in the study than they would necessarily have to be, some of them, in actual practice. I mean, many of them are probably the same, but some of the others don't fit.
CHAIR NOLLER: How about Number 9, training. Do you have any comments about that?
DR. GERBIE: Yes. It looks like I could learn how to use this.
DR. GERBIE: With lots of training.
CHAIR NOLLER: It looks pretty easy if you're already a colposcopist. And they demonstrated no learning curve. Anybody? Joe, was it you -- who had the comments about learning before?
DR. SANFILIPPO: Yes, I did, of the investigators, i.e., drawing a parallel to the ALTS trial. But I think I'm becoming convinced that the learning curve is a very short one. So I don't have a problem with it.
CHAIR NOLLER: Yes.
DR. MILLER: I guess I would just go back to kind of the reverse thinking, which is the training seems straightforward, and in this day of technological elevation to its heights, again I'm concerned about the bias that this new technology somehow takes the clinician into the realm of thinking of this as the gold standard of colposcopy and undermining their clinical skills and their clinical suspicion that basically no stone should be overturned; that now if the LUMA says -- I just worry that the false negative results of the LUMA will get missed in the glamour of this nice big machine that's got a nice LCD monitor, and you know, basically points you right to the bad parts.
CHAIR NOLLER: Let me -- yes. First.
MS. GEORGE: Just a quick comment on that. I'm assuming that Pap smears came out before the colpo.
CHAIR NOLLER: Yes.
MS. GEORGE: And so the same type of thing, that do people not do Pap smears at all and just go right to colpo?
CHAIR NOLLER: No. Screening is still Pap smear.
DR. GERBIE: But it certainly happened with the LEEP procedure, let's excise every patient.
CHAIR NOLLER: One thing, and I've been sitting here thinking about whether to -- I think this is an appropriate comment. In the labeling, there labeling can have an effect on use in the sense that if it says it's not to be used under age 18, it's unlikely that there will be any reimbursement for its use under age 18. So in fact the labeling can have some effect on clinical practice, whereas if it doesn't say that, it may very well be used, or not in pregnancy, or over age 70, or whatever. But reimbursement does dictate, particularly something where there would be some cost of using the instrument. We aren't supposed to talk about cost, but I think it's appropriate in this case probably.
Dr. Snyder looks skeptical.
DR. SNYDER: The only thing I thought about, I think a lot of us in academic centers are all hooked up with digital cameras hooked up to our colposcopes, and monitors for the patients to look at and everything. I'm not sure how prevalent that is on the outside. There may be a temptation to want, since one part of this is just a plain image of the cervix, and an easy way to print out something for a chart, that even if it said not under 18, somebody may want to get a picture for the medical record.
CHAIR NOLLER: Mel?
DR. GERBIE: Sort of along the same line, can this instrument, might it be used as a colposcope itself? Because -- can I ask someone to tell me why not? I'm looking at a magnifying system, that acetic acid has been put on the cervix, and it's bright lights, and I'm getting an image. Why can't I use that?
CHAIR NOLLER: In two sentences.
DR. WRIGHT: In two sentences. Dr. Julian is totally right. The resolution of this device is not sufficient to allow us to look at atypical vessels, and that is key with colposcopy.
CHAIR NOLLER: Let's look at Question 10. Does the panel have concerns about any issues that should be addressed in the post-approval study. Some of you are new to the panel. This supposes that we recommend that it be approved, and FDA approves it. And this would be making suggestions for things that should be followed along afterwards. Are there things that we would like the sponsor to keep checking on? For example, say it gets approved for use over age 18. We might suggest that they start a study in younger or -- 70 patients, patients over 70. Are there any sort of those sorts of issues that should come up? This is a little different from conditions we'll talk about later. Yes.
DR. GANDJBAKHCHE: I have a question, and the question is as follows. That if you approve a PMA, is everything locked, for example the design. They should use the same design, the same algorithm, and so on and so forth? They cannot change after the premarket approval the design, the head or anything? I don't know about that.
CHAIR NOLLER: Colin? Can you answer that, please? Or Nancy?
MR. POLLARD: The simple answer is to make changes like that the company, if the device were to be approved, the company would need to submit a PMA supplement describing those changes, and providing the appropriate verification and validation data to support that change.
CHAIR NOLLER: It appears that we're sort of out of discussion for awhile. We have a number of things to do after this, but I think maybe -- let's take our break a little bit early. It's 2:52. Let's try to be here at 3:00.
(Whereupon, the foregoing matter went off the record at 2:52 p.m. and went back on the record at 3:08 p.m.).
CHAIR NOLLER: Take your seats, please. Close the doors, please. I'm going to look to one of our panel members to raise two issues, please. Nancy?
MS. BROGDON: Thank you. There are two questions that have come in discussion with FDA staff, and we wanted to ensure that the panel covers these questions. One is a little bit of follow-up on the age questions, and Dr. Pennello has something he'd like to discuss there.
DR. PENNELLO: Hi, Gene Pennello. We are hearing from your discussions a suggestion to exclude some of the younger women from the indication. And I just want to remind that, in least in our analysis of PSI, it seemed like much of the effectiveness was in the young age group, and so removing that group, you may be removing a lot of the effectiveness of the device.
CHAIR NOLLER: Panel open for discussion. Paula?
DR. HILLARD: I guess I would just say that, yes, you may be looking at the effectiveness of the device. From your analysis, you're saying that it was more of a difference among the youngest group. Am I correct in understanding that? And that may well be true in terms of pickup of CIN2/3. CIN2 may be more likely to regress among that age group. So are you really having the same degree of benefit by doing that pickup to begin with in that youngest age group, as opposed to an older age group?
CHAIR NOLLER: I guess sort of another way to say it is if you take out under-21, there's so little pickup over 32 that why would you have the instrument to begin with. Is that what I'm hearing? Was that your point?
DR. PENNELLO: That was basically my point. You might be in danger of losing any of the effectiveness that LUMA might have by removing that young age group.
CHAIR NOLLER: More discussion. Dr. Miller.
DR. MILLER: Well, I guess in the way I'm thinking about it it takes us back to another way of looking at the risk/benefit ratio. We've already, in looking at the different questions that have been put before us, we've already raised real question about whether there's any benefit. Certainly there doesn't appear to be a statistical benefit, and we're pushing that line, saying that there may be some benefit in the subpopulation. Although there doesn't seem to be overt risk for using the device, there are adjunctive risks. In other words, over-utilization of the technology, over-dependence of the technology, populations that may be adversely affected. So I'm thinking about it from a risk/benefit perspective, wondering there's not so much benefit, and there's ever growing risk. And the benefit, you know, potentially is going down if we take out the younger group.
And although we haven't emphasized it, as the other MFM specialist on the panel I'll just say that over-utilization of LEEPs in young people who have reproductive life ahead of them, and have not really exercised their reproductive life is a real problem. There are some good technologies available to us to assess that problem, but the application of those technologies is not clear yet. And certainly we potentially set young people up for more invasive procedures, and more cumbersome pregnancies.
CHAIR NOLLER: Yes.
DR. D'AGOSTINO: Could someone remind me, the endpoint is CIN1, CIN2/3+, and so forth. Did the data get collected by that grouping, or was the data actually collected by the individual categories? Where I'm heading is if we say what else should be done with this data, can the sponsor actually go back and sort out the CIN2's versus 3's?
CHAIR NOLLER: That certainly should be possible.
DR. D'AGOSTINO: We have it combined.
CHAIR NOLLER: Do you want to address that?
DR. WRIGHT: After much discussion with a variety of people, we collected data as CIN2/3 in the aggregate. So that is the reference pathology.
CHAIR NOLLER: Thank you. So it's not possible. Nancy, you had another point you wanted to bring up?
MS. BROGDON: Yes, our other question was whether the panel feels there's been enough discussion about the instructions given to the investigators during the two studies, based on the colors shown in the display, and where and whether biopsies were taken.
CHAIR NOLLER: I had some concerns about that too. I heard a couple different things. One was that the -- and maybe I heard it wrong, but that the investigators were asked to take a biopsy based on the LUMA image in PSII. So if they decided to do one for colposcopy, then they had to take another one. I'm not sure that's true, but I'd like to hear.
The other thing is that the pictures we saw had nice blue areas, but this is a continuum yellow to blue, somehow. And what were the instructions given to the investigators about where they should biopsy. If there's no blue, were they still supposed to biopsy a yellow site, or not at all? Dr. Wright, can you address that? Both issues, please.
DR. WRIGHT: Right. Instructions to the investigators were that they were to biopsy a blue site, and they were to biopsy yellow sites if clinically indicated. So if a blue site was there it was to be biopsied. If it was yellow and it correlated with something they saw colposcopically, any degree of acetowhitening, we would expect them to do a biopsy.
One of the points with respect to the difference between PSII and PSI is that in PSI they looked at both images, the colposcopy and the LUMA images, and decided where to biopsy. So if there was an acetowhite area and they decided to biopsy in the center of it, and it was also blue but had a periphery of it, that would have been considered meeting the requirements for having taken a blue target. In PSII, a large lesion, if they wanted to take a colposcopically directed biopsy of it, if they could fit in a second biopsy of the blue spot they were instructed to fit in the second biopsy, which is one of the reasons there were so many biopsies in PSII.
CHAIR NOLLER: Thank you.
DR. WRIGHT: Thank you.
CHAIR NOLLER: Panel understand that? Anybody have concerns about that? Before we go on to the next section, do people have clear in their mind whether or not you feel the data support the use of the instrument? Dr. Gerbie?
DR. GERBIE: We didn't spend a lot of time, maybe don't want to, on 7(b), clinically significant results. And this has come up in a lot of lay press about medications. A new medication has come out, and yes it's safe, but is it worthwhile. Are we going to be held to the same criteria when we say clinically significant results. Statistically we look like we're not making it, statistically, but statistically and significant. And to an individual patient, as most of us know, statistics and significance get thrown out. I'm not sure we've discussed this enough.
CHAIR NOLLER: More discussion.
DR. D'AGOSTINO: Just to remind, the point I was making is that the usual discussion is you want to see the statistical significance before you start talking about the clinical significance. And so we're jumping over the statistics completely, and that leaves us with being in a very precarious position to even talk about the data in any meaningful way.
CHAIR NOLLER: Russ, could you? Tom?
DR. JULIAN: Talking about the statistics?
CHAIR NOLLER: No, we're trying to decide whether use of the device will provide clinically significant results. Will it help women?
DR. JULIAN: Well, I've been thinking about this somewhat differently, because to me the study itself and the analysis of it's kind of like quicksand. If you can walk on water, going across quicksand is just a stroll. If you can't, which I cannot, I can't really make heads or tails of it. I think what this does do is it is an adjunct to colposcopy. And I think that colposcopy is going downhill rapidly because there are more and more practitioners being trained in it with less and less cases for practitioners, many of whom have never seen cervical cancer. And this may have some real value for the under-armed colposcopist, so to speak.
CHAIR NOLLER: Other comments? Let's move on. We're now at the point where we have our second open public hearing. We have received a request from Dr. Mark Spitzer to speak during the hearing. Dr. Spitzer, we'd like to ask you to disclose any commitments you have, or we encourage you to mention any sponsorships that you have. We have, let's see, we've given you 5 to 6 minutes for your statement.
DR. SPITZER: Thank you, Dr. Noller. My name is Mark Spitzer. I am a Professor of Clinical Obstetrics and Gynecology at the Weill Medical College of Cornell University, and I have been practicing colposcopy and teaching it on a national and international level for over 20 years. I have taught on and behalf of and developed educational material for the American College of Obstetricians and Gynecologists, and the American Society for Colposcopy and Cervical Pathology. I am currently the president-elect of the ASCCP, although I am not here representing them, or any other organization. I am just representing myself, and these statements are my own.
As a matter of financial disclosure, I participated in a mock panel review for MediSpectra in preparation for this meeting, and I was paid for that participation, but I have no financial or consulting relationship with the company, and I am not being paid nor are my expenses being reimbursed for this meeting. I have also done research for Polytechnics Corporation, which is developing a competing product.
Briefly, my message is that we have trained more and more colposcopists over the years. Each practitioner is doing fewer and fewer colposcopies, a point that Dr. Julian made just a minute ago. The result is that we have many inexperienced practitioners whose level of skill is marginal at best. Anything that would assist these novice colposcopists in identifying significant disease can only be viewed as a benefit.
Please allow me to make some specific observations. Skill in colposcopy and the management of abnormal Pap smears requires three elements, knowledge of management algorithms, knowledge of the underlying science, and pattern recognition skills. The most important way to gain these skills is through clinical experience. Over the last 20 years, we have improved our didactic education of colposcopy, and we have also trained many colposcopists, including gynecologists, family practitioners, nurse practitioners, et cetera. This was invariably done in referral centers, where a large number of colposcopies were done. These trainees went on to practice colposcopy in their communities, and improved their patients' access to care, and reduced waiting times for examinations and treatment. However, in the process the referral centers disappeared, and today's trainees often do not have access to such centers, and so they have only limited clinical experience and continue to have limited experience in clinical practice. You would expect that these practitioners would not be very good, and in my experience they are not.
This observation is supported by research. In 2001 we published a survey done on behalf of the American Society for Colposcopy and Cervical Pathology of residency training directors in OB/GYN and family practice. Four hundred and eighty-five program directors responded. On the basis of their estimates, and the colposcopy experience of graduating residents is quite limited, especially in family practice. By the time they completed their training, only 15 percent of family practice residents performed more than 10 evaluations of high-grade lesions, and only half had evaluated more than 10 low-grade lesions.
The estimates for OB/GYN residents were much higher. While there is no data available on how many colposcopies it takes to develop basic competence in colposcopy, various published estimates range from 50 to 150 colposcopies, with at least 10 high-grade lesions. In 2004, we presented our experience with an objective online examination to assess resident skills in colposcopy. In the process, we surveyed both the residents and their program directors. While 92 percent of the program directors thought their residents were adequately trained, only 70 percent of the residents felt adequately trained. The exam scores correlated better with the residents' assessment. In other words, the residents who felt they were poorly trained did poorly on the exam. I feel that this supports my conclusion that we are turning out many inadequately trained colposcopists.
Colposcopy practice by inexperienced practitioners leads to missed or delayed diagnosis and treatment. Moreover, inexperienced colposcopists are also likely to take additional unnecessary biopsies, order unnecessary tests, or do an unnecessary cone biopsy or LEEP because of a lack of correlation between their colposcopic impression and the histology resulting from a failure to perform a biopsy at the appropriate area. The Canadians and British solved this problem by limiting colposcopy to select numbers of experts. It's not possible in this country.
The literature shows that about 65 to 70 percent of all CIN2 and 3 originate in women with ASCUS and LSIL smears. Current guidelines direct that most of these women would not have LEEPs. As the number of ASCUS and LSIL smears rise, and the guidelines change to have even fewer of these women have LEEPs, more CIN2 and 3 will be missed unless we improve the quality of our diagnostic testing, specifically colposcopy. By the way, it would not be reasonable to solve the problem by creating algorithms that manage ASCUS and LSIL more aggressively, because that would just result in more over-treatment.
The ALTS data shows that with good colposcopy, and we assume that they have good colposcopy, about one-third of the CIN2 and 3 lesions were missed at the initial colposcopy. One can only wonder what that number would have been with colposcopy as it is practiced in the general community. The only reason that we're not seeing a rise in cancers is because we're doing way too many LEEPs for CIN1, and because the disease takes so long to develop.
With that background, any technology that would assist the novice colposcopist in identifying CIN2/3 can only be seen as a benefit. We should not stifle the development of new technology and new approaches to the diagnosis of this condition. While I will not comment on the merits of device, I am sure that we can all remember the initial attempts at the development of HPV-DNA testing. They were not very successful, nor was the test very useful. However, with time, improved tests were developed on the foundation of the older ones, and today, HPV testing is a fundamental part of our management strategy for the diagnosis of cervical pre-cancer. It's my feeling that we should give new technologies that are safe and demonstrate reasonable effectiveness a chance. Thank you for your attention.
CHAIR NOLLER: Thank you, Dr. Spitzer. We have one more request to address the panel, and that's from Dr. Schiffman. He's been granted two minutes. Just like Congress, two minutes.
DR. SCHIFFMAN: I testified against HPV testing in the mid-`80s. And then it testified for it when it was ready, years later. I don't understand why we are allowing some pretty pictures of a couple of anecdotal cases and a trial that fails all of the known requirements of an FDA panel to go to the point where people are saying that need overwhelms our data rules. I've spent 20 years evaluating techniques and technologies that are always pressing forward, and yes, it's hard to get venture capital money, but there are rules as to when FDA gives approval. And now I believe they're being bent past breaking, and it's quite of interest to me, given that we spent $30 million ALTS, and a lot on others to try to validate when a technique is really useful for national use, to see that something that really fails all the different criteria is being evaluated based upon the need for improving colposcopy. There's a great need to improve colposcopy. This technique has not shown in any way that it is better than just forcing additional biopsies. And by forcing additional biopsies, more disease is detected, sensitivity increases, specificity decreases. It's a no-brainer to me, and I don't understand what's going on.
CHAIR NOLLER: Thank you. Nancy?
MS. BROGDON: I suggest you may want to ask Dr. Schiffman on whose behalf he's speaking. He spoke on behalf of NIH this morning, and you may want to ask that now.
DR. SCHIFFMAN: Those were personal views based on 20 years of doing this.
CHAIR NOLLER: Thank you. It's not the official NIH position. Is it correct to comment on those? Yes, it's okay. Yes, comment.
DR. D'AGOSTINO: I think the open hearing comments were quite useful. My concern is we don't have the study to evaluate what was being suggested. The study doesn't have people with poor training, it has people with extremely good training and so forth. We don't know what LUMA is going to do in the presence of people who are degrading or have very poor performance. And so to use these studies as a basis for approval because of this need is even worse than overriding the statistical lack of significance that we have in the data.
CHAIR NOLLER: This is part of our panel discussion. Dr. Cedars?
DR. CEDARS: Well, I just wanted to agree with that, and just say that my concern is that if the problem is poor colposcopic skills, that that's what should be worked on. And even though you may have as best, even if it were statistically significant a 2 to 5 percent increase in true positives, you have a 20 percent incidence of false negatives. And so I have a concern saying that this is going to help people who can't do colposcopy.
CHAIR NOLLER: Amir?
DR. GANDJBAKHCHE: Yes, but I think the point of Dr. Spitzer that I liked was you need another eye, a different eye. And I think it's different from the subjective assessment of a physician. And this is the thing that they claim they are able to do. Do some kind of quantitative measurements.
DR. D'AGOSTINO: But the study didn't have people with poor skills.
DR. GANDJBAKHCHE: I -- sorry? Say it again?
DR. D'AGOSTINO: The studies we have before us was not designed to test and to show that the device is useful in the poor-skilled individuals. I mean, it's a different --
DR. GANDJBAKHCHE: I haven't voted yet. I don't know what I will do. But the point I am making is this is two views for the future of the medicine. That you have a quantitative way to assess things that you see in the yellow, or blue, or whatever, and they are based on some spectroscopy. It's not the color. I hope it's not the color, and I will tell when I will vote how I see the future of this kind of device. But you should have in mind that these are the major new way to look at things. These are related to the metabolic activity of the claim, or some species that are behaving differently in different tissue types.
DR. MILLER: Yes, I just wanted to, in listening to some of these latter comments, it occurred to me that if we were to move forward, and we were to consider labeling, we might want to consider how this technology would be used in the lesser skilled colposcopist in terms of faced with the pretty pictures, I in my mind have kind of assumed that the biopsy sites would be really the blue sites. But really, one could conceive of someone saying, well you know, you've got yellow here, and some green there, and some blue there. We'd better nail all of those sites, and now we're talking about really a very different application than what was studied, but I think has to be a consideration when this is released to a market of much more variably skilled colposcopists, who might be prone to biopsy all of the areas that are illuminated.
CHAIR NOLLER: Dr. Cedars and then Dr. Weeks.
DR. CEDARS: Again, I think we're arguing a different study. I mean, if the argument is that we need accessory tools like this for the people who are less skilled, then the study should be LUMA alone to colposcopy alone. And that's not the study. If it's done as an adjunct it's done as an adjunct to people who have good colposcopic skills. So if you remove the good colposcopic skills, then you're looking at a totally different tool that wasn't tested.
CHAIR NOLLER: Dr. Weeks?
DR. WEEKS: Well, at this point I would just reiterate probably what everyone else has said. But again, I'm bothered by the fact that we haven't proven statistically that there is benefit, and I think that's extremely important. And then we're talking about putting a tool in the hands of a group of investigators that we haven't studied at all. Third, we really haven't, I don't think there's been any consensus that we will really improve the health of a significant number of women, but there is a very real potential that we're going to increase the number of procedures that we're performing in these women, and those procedures could have some adverse health consequences.
CHAIR NOLLER: Dr. Jiang?
DR. JIANG: I'm just going to make a comment, speaking my own mind here. I think the data presented in front of us are not perfect. I think one reason for that, maybe it's difficult to do these studies, to get the data that we want. And there's a lot more additional study are required to answer those questions. But I think a decision has to be made at some point about that. So I guess I'm saying that we have to make a judgment based on imperfect data.
CHAIR NOLLER: Thank you. Mel?
DR. GERBIE: A few months ago I read about an FDA-approved device for treating fibroids. The headlines, of course, say FDA approves it, and then halfway into the article it mentions significant failure rates in a certain period of time. And I'm saying how can you approve this and know that it's got a significant failure rate over a period of time. So I'm not sure what our job is totally here. I don't know of any studies or devices in the last X number of years that weren't studied by authorities, by fetal monitors read by people who have more experience. So everything is done by people who have more experience, and it goes out into the countryside, and works its way through, and some continue, whether there's evidence, like a fetal monitor, that it does anything. I don't think you're going to teach people how to do mid-forcep rotations today, even though it's maybe significantly easier than doing the high C-section rate we have. It just isn't going to be done.
So what we're trying to do is see what's going to be best for patients. And maybe the ideal situation is to train specific number of expert colposcopists, devise referral centers and systems for that. I don't think that's going to happen. We train our residents. They get state licenses that says most of the time they can practice medicine in all its specialties. And the female patient wants the best care she can get. And I'm not saying this is a pro or con. I think this is just the difficulties that I'm having with the decisions.
CHAIR NOLLER: Thank you. Let me ask if there are any other members of the public that wish to speak at this time? Seeing none, the open public hearing part of this meeting will be closed.
We will now go to the part of the meeting where both the FDA and the sponsor have approximately five minutes to make their final comments. During these comments, we will not question them nor have interaction with either one. The FDA first.
MS. BROGDON: FDA staff have no comments.
CHAIR NOLLER: Thank you. Sponsor? And as you come up, please introduce yourselves.
DR. ALVAREZ: My name is Ronald Alvarez. I'm the GYN oncologist division director at the University of Alabama at Birmingham. I have no financial interest in MediSpectra but I was a PI for several studies, and have been provided support to travel to this FDA panel meeting. I appreciate the opportunity to make a few closing remarks on behalf of the sponsor.
We've all spent a considerable amount of time reviewing the intricacies of an enormous amount of data on a novel optical imaging device to be used as an adjunct to colposcopy. And I repeat, an adjunct to colposcopy, not a replacement to colposcopy.
I hope that we would take a few minutes to just step back and gain some perspective on what we are trying to accomplish here today. Although we've seen a significant reduction in the incidence of cervical cancer in the United States since the introduction of the Pap smear, there are still many areas in this country that harbor high incidences of cervical cancer, many states like my own. There are a number of different avenues of research that are involved in improving the prevention of cervical cancer. Clearly, a need for behavior modification, improving access to care for those, particularly in underserved populations. There's clear need to identify new biomarkers that could be used in cervical cytology to identify those patients at highest risk. And we've heard today how important it is to improve our management guidelines, and to educate physicians about the implementation of these guidelines, particularly in adolescent populations.
Likewise, there is clearly a need to improve upon the current methods of evaluating a patient with an abnormal Pap smear. This has really been the primary focus of our discussion today. I'd like to recall that there was a time when patients with an abnormal Pap smear all underwent conization, or multiple blinded and random biopsies. Colposcopy was developed to facilitate the evaluation of these patients, to direct the clinician to biopsy the areas most likely to harbor disease, and to guide the physician with respect to further management of the patient. It has remained the gold standard for evaluating patients with an abnormal Pap smear for over 50 years.
For so many years, we all believed that colposcopy was a very accurate method to detect significant cervical dysplasia. However, we have come to understand from a number of studies, including the ALTS studies, how poorly accurate this device is, how insensitive it is for detecting significant disease. As you heard often today, over 60 percent of patients with CIN2 present with an ASCUS and LSIL Pap smear, and in the ALTS study, one-third of these patients were missed for colposcopy. I hate to inform you that myself and many of the people that are here with us today were those so-called expert colposcopists that participated in that trial.
Today we've discussed in detail emerging optical technology that provides us quantitative information and has been demonstrated to improve upon the performance of colposcopy when used as an adjunct, once again, and not as a replacement for colposcopy. We have reviewed in detail two studies that show a similar and repeatable magnitude of effect with respect to detecting CIN2/3 in a patient population that this is most relevant in. These studies included a large group of institutions and colposcopists, and a large patient population that was geographically and demographically diverse. Specifically, we've seen an approximately 25 percent increase of detection of CIN2/3 when LUMA was used as an adjunct to colposcopy compared to when using colposcopy alone, many that were performed by these so-called experts.
Were these studies perfect? The answer to this questing is clearly no. Who wouldn't redesign a study, once completed, particularly when the results of that study come to fruition. I caution you that we must be able to recognize in any study the potential clinical significance of an intervention when the results of a study do not meet the strict preset statistical design placed on the initial protocol. Much like we must recognize whether an intervention is of no clinical significance in spite of reaching statistical significance in appropriate design trials.
Hundreds of studies have demonstrated the potential benefits of spectroscopy in the context of cervical neoplasia. Most of these have actually involved very small populations of patients. The studies today presented represent the largest to date to evaluate this emerging technology in a very clinically relevant manner. Designing, implementing, and analyzing the perfect study to evaluate this technology in a strict ASCUS/LSIL population would require enormous patient and financial resources that quite frankly not even the federal government could currently afford. And we'll perhaps miss a potentially unique opportunity to begin to incorporate this technology to improve the performance of colposcopy.
Could we achieve the same results with additional biopsies? Dr. Solomon and Schiffman have alluded to this fact. However, are multiple biopsies the current standard of care for patients with ASCUS and LSIL, or should they be mandated? In fact, in the ALTS trial, a little over 10 percent of over 5,000 patients entered into this trial had more than one biopsy obtained. Should more than two biopsies be mandated in this population to even further increase the sensitivity for detection of CIN2/3? It seems to me that this policy is retroactive, is moving the field back to a time where we did a conization on all patients with an abnormal Pap smear.
What about the discomfort and other side effects associated with additional biopsies? We all have done colposcopy and biopsies. And I'm sure that many of us have patients that we recognize how uncomfortable this procedure is. Perhaps you can believe that you will ultimately detect this CIN2/3 in a patient with an ASCUS or LSIL Pap smear the next time she comes in for an evaluation.
Well that may indeed be the case if we could be assured that patients were 100 percent compliant with follow-up. This is far from the case, particularly in patients at highest risk for developing cervical cancer. Indeed, this was the case with the ALTS trial, in which compliance was about 85 percent in an environment that was rich with infrastructure and incentives to ensure compliance. My colposcopy clinic patients at UAB are in general from lower socioeconomic groups, have poor understanding of the importance of cervical cancer screening, and have limited resources to access cervical cancer clinic facilities.
At my institution, we schedule 120 patients with an abnormal Pap smear from the local health departments each week at our colposcopy clinic for either colposcopic evaluation or a loop procedure. Over half to these patients fail to keep their appointments each week, and there are many who are never evaluated or appropriately treated. Thus, relying on patient follow-up to address deficiencies in colposcopy I think is fraught with false security, and doesn't directly address the problem of colposcopy.
In summary, we have demonstrated that the LUMA device when used as an adjunct to colposcopy is safe, easy to utilize, and improves the detection of CIN2/3 by nearly 25 percent in the most relevant patient population, namely those with an ASCUS and LSIL Pap smear. The information presented represents years of diligent work that only begins the important research that needs to be done in this area. Is there likely some other method presently available to address this deficiency in such a forward-thinking manner? There is none to my knowledge. Is there going to be a need for training physicians in the use of this technology? Certainly that is the case, and the company is certainly committed to doing that in the appropriate fashion. Will there be a need to ensure that we address the issues regarding management of patients, particularly the younger patients, with respect to once they have had a diagnosis? Certainly that is the case. Will there be improvements upon this technology? We will certainly see additional enhancements.
My colleagues and I believe that this will be the first of many innovations that, consistent with the initiatives within the NIH roadmap, will allow us to improve upon our ability to image the cervix and diagnose more patients at highest risk for developing cervical cancer, namely those with ASCUS/LSIL Pap smears that harbor CIN2/3. With over 2 million women referred with these abnormal Pap smears annually, can we afford not to adopt this emerging quantitative optical imaging technology that will provide true health benefits for a significant number of patients at risk for cancer? Certainly we know that that is the question that you will have to address today. Remember that approximately 20 percent of patients who develop cervical cancer do so as a result of failure to follow up, and as a result of inappropriate management. I thank you for your time and your attention.
CHAIR NOLLER: Dr. Wingrove, I am told you can do this in a minute?
DR. WINGROVE: That's correct. I just want to echo some of Dr. Alvarez's thoughts. We do believe we have a study which has -- two studies which have consistently found the same finding, a 25 percent increase in CIN2/3 detection. We believe that the medical community should have the opportunity to use this advice, but we have heard your recommendations. We have heard your concerns regarding training and regarding labeling, and we would welcome any further recommendations you would have on how you think that would best improve or address your concerns. We have every intention of working with medical societies like the ASCCP to ensure that the training is optimized and consistent with their recommendations. Thank you.
CHAIR NOLLER: Thank you. The panel will now proceed to the deliberations and vote. And as a prelude to this we'll ask our secretary Dr. Bailey to read the panel options and the votes. I ask the panel to pay particular attention to this part.
DR. BAILEY: The Medical Device Amendments to the Federal Food, Drug, and Cosmetic Act, as amended by the Safe Medical Devices Act of 1990, allows the Food and Drug Administration to obtain a recommendation from an expert advisory panel on designated medical device premarket approval applications (PMA) that are filed with the agency. The PMA must stand on its own merits, and your recommendation must be supported by safety and effectiveness data in the application, or by applicable publicly available information.
The definitions of "safety", "effectiveness" and "valid scientific evidence" are as follows. "Safety" is defined as a reasonable assurance that the device is safe when it can be determined based upon valid scientific evidence that the probable benefits to health from use of the device for its intended uses and conditions of use when accompanied by adequate directions and warnings against unsafe use outweigh any probable risks. The definition of "effectiveness" is a reasonable assurance that a device is effective when it can be determined, based upon valid scientific evidence, that in a significant portion of the target population the use of the device for its intended uses and conditions of use when accompanied by adequate direction for use and warnings against unsafe use will provide clinically significant results. "Valid scientific evidence" is defined as valid scientific evidence from well controlled investigations, partially controlled studies, studies and objective trials without matched controls, well documented case histories conducted by qualified experts, and reports of significant human experience with the marketed device from which it can fairly and responsibly be concluded by qualified experts that there is a reasonable assurance of the safety and effectiveness of a device under its conditions of use. Isolated case reports, random experience, reports lacking sufficient details to permit scientific evaluation, and unsubstantiated opinions are not regarded as valid scientific evidence to show safety or effectiveness.
Your recommendation options for the vote are as follows. One, approval if there are no conditions attached. Two, approvable with conditions. The panel may recommend that the PMA be found approvable subject to specified conditions, such as physician or patient education, labeling changes, or a further analysis of existing data. Prior to voting, all of the conditions should be discussed by the panel. The third option is not approvable. The panel may recommend that the PMA is not approvable if the data do not provide a reasonable assurance that the device is safe, or the data do not provide a reasonable assurance that the device is effective under the conditions of use prescribed, recommended, or suggested in the proposed labeling.
Following the voting, the chair will ask each panel member to present a brief statement outlaying the reasons for his or her vote. Dr. Noller?
CHAIR NOLLER: Thank you. And now as we go through this, the format is fairly rigid. And I would ask that when we get to the discussion if you would raise your hand high so I can see you, and I will try to take you in order as I see you.
The first think we need to do is to obtain a main motion. And I'm going to ask the group if there is a motion to recommend approval, approval with conditions, or not approvable.
DR. GERBIE: Question.
CHAIR NOLLER: Dr. Gerbie has a question.
DR. GERBIE: Can you state what are we -- what is the proposal. Excuse me. What exactly is the proposal?
CHAIR NOLLER: The proposal is to recommend to the FDA that this device is safe and effective as described, and I guess that's really it. Is that right? Because ?. Is that right?
DR. GERBIE: My question is, is this specifically for ASCUS, ASC-H, low-grade, or not?
CHAIR NOLLER: That's the -- yes. That's true. That's what the sponsor has asked for. That is the indication, specifically for ASCUS and LSIL. Nancy, is there anything more there I should say that I messed up on?
MS. BROGDON: No, I think you got it.
CHAIR NOLLER: Okay. So I'm looking for a motion to approve, approve with conditions, or not approve. I can't make the motion. Dr. Cedars?
DR. CEDARS: I would put a motion to not approve.
CHAIR NOLLER: We have a motion not to approve. Did I hear a second?
DR. SNYDER: Second.
CHAIR NOLLER: Please say your name when you second.
DR. SNYDER: Dr. Snyder.
CHAIR NOLLER: Dr. Snyder seconds. Now we'll have discussion of the motion. Who would like to speak first? Let me ask Dr. Cedars to speak first since she made the motion, then Dr. Snyder, and we'll go from there.
DR. CEDARS: I make the motion number one because I don't believe that there is scientific evidence to support efficacy, certainly not statistically significant, and because of concerns specifically regarding the youngest population in which there appeared to be the greatest benefit, based on the data that's available.
CHAIR NOLLER: Dr. Snyder?
DR. SNYDER: I'd like to start off by saying I'm really excited about this technology. I'm optimistic about this technology. And I just, you know, pray that it's going to get to the point where it's going to be able to maybe even replace colposcopy. And I will always worry about stifling ongoing research because of decreased funding. And that could be construed to not be in the public's best interest, but I also feel strongly that neither is it in the public's best interest to put a device out there that doesn't meet the rigors of good clinical evidence, and science-based evidence. And that concludes my comments.
CHAIR NOLLER: Other discussion? Yes.
DR. D'AGOSTINO: Having said so much earlier, I have to say something to repeat myself. But I don't see the statistical significance there, and it's very bothersome. And the more we talked, the issues of the young group being the group that may in fact have the CIN2's where there's going to be potential regression, the expertise of the people who are using this when it goes out, if it does go out to be used. There's just so many issues about the data, and the implications of this being approved that I support the non-approvable.
CHAIR NOLLER: Other discussion? Before we take a vote, I will ask the three members who have participated in the panel discussion today but who will not vote to make a statement. Our public member, please? Into the microphone.
MS. MOORE: I guess I start with a question. I wanted to know if -- that I don't expect you to answer, it's just one of these questions that I have in my mind. Does statistical significance override clinical significance. That was one of the questions that I thought about.
And then the other comment is if this device is to be used as an adjunct, as we have been told, it is to me another tool in the diagnostic process which I think could be of benefit. So as the consumer representative, it would seem to me that with the suggestions that have been offered today, that this should be approved with conditions. And if I were a voting member, I think that's what I would vote for.
CHAIR NOLLER: Our industry member, please.
MS. GEORGE: Very similar to what Christine has just said is that I, having actually sat in their shoes and created many of these 510(k) PMA's myself, I think they've done an excellent job of clinical study. They've really taken a significant amount of time to pull all of the information together. I too would have chosen the approval with conditions, specifically using some of the discussions of going back and doing more analysis of data, because I think there is a significant amount of information and data they have here that could be sliced and diced in many different ways. I know that the statisticians say that you can put anything together to make it look good, but I think that, you know, inversely as well maybe they'd go back and find out that it wouldn't be as good, and they wouldn't want to put it out there.
And then secondarily, I think that also as Christine stated, I think that a new technology, an added technology that doesn't harm the patients, because I think that there was no evidence here that anyone was getting hurt by anything of this, that there was really only evidence that it either supported or potentially assisted in the clinical decision process. So I too would have gone with approval with conditions.
CHAIR NOLLER: Third member. Dr. Sanfilippo, do you have any comments?
DR. SANFILIPPO: I want to just say that certainly clinical application, and clinical importance is absolutely paramount. And I am very impressed with the technological advances bringing us here today, because it represents some tremendous thought, ingenuity, et cetera. While I don't have a vote, I do feel this technology one day will certainly help clinicians. And personally, I would just like to see some additional data, and I'm hopefully that perhaps somehow it occurs, because I do think that there is some clinical application at all age ranges. Unfortunately, I didn't see that today.
CHAIR NOLLER: Thank you. It has been moved and seconded that MediSpectra's premarket approval application P040028 for the LUMA Cervical Imaging System is not approvable. In a moment I will ask everyone on the panel for their vote. You have three options, to vote for the motion of not approvable, against the motion, or you may abstain. I will start with Dr. Gerbie. Dr. Gerbie, how do you vote?
DR. GERBIE: I vote against the motion.
CHAIR NOLLER: Against the motion. Dr. Jiang?
DR. JIANG: I vote for the motion.
CHAIR NOLLER: Votes for the motion. Amir?
DR. GANDJBAKHCHE: No, against the motion.
CHAIR NOLLER: You vote against the motion. Dr. Miller?
DR. MILLER: I vote for the motion.
CHAIR NOLLER: Miller votes for the motion. Dr. Hillard?
DR. HILLARD: I vote for the motion.
CHAIR NOLLER: Dr. Hillard votes for the motion. Dr. Cedars?
DR. CEDARS: For the motion.
CHAIR NOLLER: Dr. Cedars votes for the motion. Dr. Weeks?
DR. WEEKS: For the motion.
CHAIR NOLLER: Dr. Weeks votes for the motion. Dr. Julian?
DR. JULIAN: For the motion.
CHAIR NOLLER: Dr. Julian votes for the motion. Dr. D'Agostino?
DR. D'AGOSTINO: For the motion.
CHAIR NOLLER: Vote for the motion. Dr. Snyder?
DR. SNYDER: For the motion.
CHAIR NOLLER: Votes for the motion. Dr. Hayes?
DR. HAYES: For the motion.
CHAIR NOLLER: Votes for the motion. It's the recommendation of the panel to FDA that the MediSpectra PMA P040028 for the LUMA Cervical Imaging System be not approved. The motion carried 9-2, 9-2 with no abstentions.
I'm now going to ask each panel member the reason for his or her vote, starting with Dr. Gerbie, and we will go around the table. Following this, we will go around the table a last time, asking each person to state what the sponsor would need to do to make the PMA approvable. Dr. Gerbie, why did you vote as you did?
DR. GERBIE: I voted against the motion because I feel that the technology has shown enough promise to get this into the public arena for further development. I realize there are shortcomings in the statistical analysis. There are many small points I disagree with in the presentation, including even the last statistic of the 25 percent increase, but I feel that for women's health issues, that this is an appropriate technology, that we should look at the positive aspects of this rather than the statistical analysis.
CHAIR NOLLER: Thank you. Dr. Jiang?
DR. JIANG: I voted for the motion. To me it seemed to be this device can pick up additional CIN2/3 lesions, but then there's downside. There's additional biopsies of patients who otherwise would not have biopsy. And there's potential of over-treatment. But really, the issue was the motion raised that the biggest benefit is in younger patients, and there's some argument on that, but the data seems clear to me. And that caused me to think the clinical significance is really the problem.
CHAIR NOLLER: Thank you. Dr. Amir?
DR. GANDJBAKHCHE: My motion would be approval with condition. I will explain why. Because this is maybe the first device that we see that's quantitative. And it's very important to assess these needs for physicians. I think the data they provided showed some promises. But my condition would be more data analysis. Open their black box and let FDA and other experts look at this, and see if the way that they are looking at the data can be improved. And I'm sure it can be approved. They were under tight conditions because they were doing the PSI and PSII. They could not unlock their algorithms. They can do that, and I am pretty confident approval with condition, that opening the black box will for sure bring new sites for this kind of technology.
CHAIR NOLLER: Thank you. Dr. Miller?
DR. MILLER: I voted for the motion because ultimately I wasn't persuaded that this technology, although exciting and also of interest to me personally, would provide a significant benefit over and above a skilled colposcopist doing an adequate number of biopsies. So I echo some of the other sentiments stated before, that I think this is a technology in motion. I recognize the difficulty and the expense in getting to this point. But I think there are real risks, and the risks are over-treatment, over-diagnosis, and also potentially a dumbing down of a skill set that we need to be drumming up.
CHAIR NOLLER: Thank you. Dr. Hillard?
DR. HILLARD: I voted for the motion, and for a number of reasons. I do think it's an exciting technology. I think it has potential, and the concept of putting together a more objective evaluation of colposcopy I think is very exciting. So I think it is exciting.
I am concerned, however, because I start from the assessment that the results did not meet the preset criteria for statistical significance. And there certainly are others on the panel and in the room who have more expertise in statistics, but I think that I have to rely on this as a starting point. So I have those concerns.
I am concerned about that I'm not sure whether an increase in detection of the CIN2/3 that is present with the device is that much better than an additional biopsy. So I have those concerns. I'm concerned about the over-detection of disease that may not be clinically significant. And so while I'm not concerned about the safety of the device itself in how it is used, I am concerned about, and I'm concerned that an increase in false positives can lead to over-treatment that may potentially harm women. And for all of these reasons I felt compelled to vote as I did.
CHAIR NOLLER: Thank you. Dr. Cedars?
DR. CEDARS: I just want to echo what everyone else has said. I think this is very exciting technology. I have concerns about approving it because it's exciting technology until it is able to pass the burden of proof. And I feel like that's what's not yet been done.
CHAIR NOLLER: Thank you. Dr. Weeks?
DR. WEEKS: Likewise, I wasn't convinced by the statistical analysis. I believe that the statistical analysis was set up in a way to reveal important clinical changes. The potential for harm, I think, is significant. While I think some of the speakers' comments about losing patients to follow-up in some ways can be compelling, but even if such a technology were used, and we diagnose a young woman with a CIN2/3, they still have to have follow-up. Unless we're going to embark upon some of those early treatments that, again, I'm very concerned about.
CHAIR NOLLER: Thank you. Dr. Julian?
DR. JULIAN: I have great respect for the investigator sponsors.
CHAIR NOLLER: Can't hear you, Tom.
DR. JULIAN: I have great respect for the investigator sponsors. I know many of them. I feel that LUMA may have great potential. It's an intriguing device. I don't think there's any doubt that it's safe. And I think in a society where we have an everything must be done mentality about public health, it has much potential. But the potential is in finding a small number of potentially problematic lesions, but probably not in preventing cervical cancer, or preventing cervical cancer in such small numbers as to not be truly beneficial.
I think the potential that should be studied is the benefit to non-expert colposcopists, rather than expert colposcopists. If it could be shown that this was effective at improving the marginal, or the early colposcopist, it would have real value. But the overwhelming data in my opinion presented by the FDA and two of the three statisticians here convinced me that it had not met rigid enough criteria to be a benefit.
CHAIR NOLLER: Thank you very much. Dr. D'Agostino?
DR. D'AGOSTINO: I also am excited about the technology. I just wanted to make sure it's clear that people, or maybe they don't understand, that I spend most of my research life on the Framingham Study. It's a cardiovascular study, but we spend lots of time looking at non-invasive technologies, subclinical markers, and this is exactly, this idea of new technologies, and discriminate functions, and classification functions is exactly what I do all the time. That doesn't mean that we get excited about a technology and therefore say it's ready for primetime, it's ready for use. And I think when they went to the studies, and interpreted the studies, and ran the analysis of the studies, they left some of the statistical significance issues aside, which I think we correctly picked up in terms of emphasizing them.
And also, I think that I was speaking mainly about the statistical significance during the day, but I think the clinical significance, I'm not overwhelmed by the notion of these false positives, and also the idea of the regression of the CIN2 without us really having a clear indication of what's going to happen to these individuals. And I don't think the study addressed those issues. So it's not just the statistical significance that led to my decision.
CHAIR NOLLER: Thank you. Dr. Snyder?
DR. SNYDER: I'm going to reiterate also much of what some others have said. And first off I too want to say that I have the absolute utmost respect for the individuals that are involved in this trial, and involved with advising this company. I think women's health owes very much to these individuals that have published, that have done research, and have additionally done teaching to our specialty, and ancillary health care providers. And we owe a great deal to them. And I also feel comfortable that they're going to stay involved in the cutting edge of research, and to how we can further decrease cervical cancer and mortality. And I want to see them do further development and refinement of this.
If I could have recommended approval of this device to stay only in the hands of the clinical researchers that were doing this, I would have done that. But as I see it, that my job on this panel is to say whether we are going to recommend that something comes out with FDA approval or not. And as I see it, patients regard that in very high esteem. And if we say this device is ready to carry the seal of the FDA, then patients are going to go out there, and they're going to hear about it, and they're going to feel like they are possibly getting inferior care if their health care provider doesn't have access to this piece of equipment. And then the next thing is that clinicians are going to be compelled, irrespective of the science, to be able to provide this technology, because that's what their patients are demanding.
And I just did not feel like we had met the rigors of science, which we are supposed to be guarding at this point in time. I don't think that just another tool is a fix for inadequate training, and I don't think it's -- well, let me, I think one of the other things you said I could do is suggest other areas of research. I would love to see this go head to head with colpo. I'd like to see it --
CHAIR NOLLER: Actually, we may -- let's do that in the next round.
DR. SNYDER: Okay.
CHAIR NOLLER: Thank you. Dr. Hayes?
DR. HAYES: Yes, I voted for the motion because I do not believe based on the data that we had that it met the definitions of safety and effectiveness. There was not compelling data support. And that realistically there was a high probability that there would be potentially over-treatment and inappropriate treatment.
CHAIR NOLLER: Thank you. Since the panel voted to recommend that the PMA is not approvable at this time, we must now identify what the panel believes is needed by the sponsor to make the PMA approvable. We're trying to help both the sponsor and the FDA. And before I ask the voting members, let me ask our non-voting members what they think. Ms. Moore?
MS. MOORE: Well, of course since I mentioned the labeling regarding -- well, not the labeling. I mentioned the question of age earlier on in the discussion. And I talked about the senior citizen, or the geriatric patient, but since then we've brought up that about the adolescent. So then I would add that to my comment.
And then to satisfy the statisticians, I'm not one -- I don't know exactly what to tell them, but I think that they would have to do something with maybe a population of people, and then to set their statistical goals, and in some way meet those goals.
CHAIR NOLLER: Thank you. Ms. George?
MS. GEORGE: I think as everybody has said, focusing on the statistical aspects of it, getting more data, looking at the data they already have to be able to help support the concerns that everybody has identified. I think that additional focus on the training and the labeling. I think the labeling, one of the things I noted was I had mentioned it earlier, that there's limitations, but they're not really clear whether they're things that they should or should not do. So maybe addressing that a little more clearly.
And then lastly, the training. I know I had some -- I did wonder a little bit about the reader variability and I think a few people have brought it up with the skill level of the people that were involved, that they may have been very skilled, so maybe there variability is even more so a risk. I think Dr. Julian mentioned that, that sometimes the more skilled they are, the more variability there could be there. So I think that that needs to be addressed.
CHAIR NOLLER: Thank you. Dr. Sanfilippo, what should the sponsor do to make it approvable?
DR. SANFILIPPO: I would just say personally witnessing a secretary who works with me, and has worked with me for many years, and is as we speak dying of cervical cancer, this becomes something very important to me, and in a sense near and dear to my heart. And my communication is that I just hope you all will continue with what is very much cutting edge. And if I were to have some communication, it would be the experimental design going forward. And again, I would strongly encourage you to give due consideration. I'd like to see studies in individuals less than 18 years of age. I'd like to see this in the hands of residents, and I'd like to see this in the hands of attending physicians who, if you will, are remote from their educational curve, i.e., residency. And I'd like to see head to head comparison with colposcopy. And again, to right out of the gate initially look at the labeling recommendations, restrictions if you will. And in a sense, if those four criteria were present with appropriate significance, then I think we could move forward.
CHAIR NOLLER: Dr. Gerbie, you voted against the motion, but what advice would you give the sponsor?
DR. GERBIE: I'm afraid I don't have a lot of advice. And I realized, as did many of the people who voted the other direction, that there's potential here. I've followed some of this information from publications in the journals, and from other researchers, and think that we need somehow to keep the stimulus going. I think it would be very difficult to devise a different trial that's better than the trial that was done. The numbers are -- 4.5 percent is still a relatively small number at this condition. And to get to a point that is still treatable in a conservative fashion, with the main argument being that to avoid loss of patients. Perhaps the low-grade patient who remains low-grade over a period of time with no obvious findings is a small subset for this, but it looks like that would be a lesser market.
CHAIR NOLLER: Thank you. Dr. Jiang?
DR. JIANG: I guess two comments. One has to do with technology. It seemed to me the device had some potential to increase true positive, but that potential is quite small. So from the technology standpoint, if that can be improved, that would be great. On the other side, reduce the false positive.
And the other point I think that's more important is clinically understand the age effect, whether there is an effect, and whether the device can benefit people that -- not the younger group.
CHAIR NOLLER: Thank you. Dr. Amir, you voted against the motion, but what do you have to suggest?
DR. GANDJBAKHCHE: I don't think that the idea to do another clinical trials and so on is right. Because they have done a very good job on that. It's impossible to get better than this. However, being in the biomedical optics maybe for 20 years, the devil is in the algorithm. Because there is a lot of data processing that should be done to obtain the kind of images that you see there. And I recommend to the company to actually look at this, go back to their data, and try to re-analyze it in some other way. Open up the box to the FDA, they will ask some experts to come and look at it, and I'm pretty sure that these numbers that you see here in the false positive/true positive will change, I assure you.
CHAIR NOLLER: Thank you. Dr. Miller?
DR. MILLER: Yes, maybe to echo some of those thoughts. I think fundamentally the goal should be to enhance that which distinguishes the LUMA assessment from the colposcopic assessment, so that ideally you increase the true positive and lower the false positive rate. And some specific ideas would be to, number one, maybe do some more careful analysis of what happened in those cases where the LUMA identified a lesion that the colposcopist didn't think existed. You know, we had some commentary about, you know, there was some controversy, difference of opinions, but it was a sense that there were some examples of patients where the LUMA took the colposcopist to an area of the cervix that would not have been felt to be suspicious by clinical assessment. So what was it about those cases in your cohort that maybe could change the algorithm, or enhance the product.
The second point was I think there was a feeling on the panel, certainly I had a feeling that because the CIN2 and the CIN3 were lumped together that it diluted the value of the diagnostic tool; that it would have been of greater value if the LUMA device had better distinguished between 2 and 3, because we would have been more persuaded by value in the 3 category. And we were told by Dr. Wright that it wasn't possible, but I assume that you have these block specimens. I assume that it might require having a pathologist look at it again, but my question would be is it possible to separate out the 2's from the 3's? You might have some power problems, but that might be worthwhile.
And the third was is it possible to go back to the PSI trial, and again try to look more carefully at the colpo only group, where the colposcopist was directed by their own intuition in terms of where to biopsy, but you have the LUMA assessment, and try to match the scan and where the LUMA scan would direct someone to biopsy compared to where the colposcopist biopsied.
CHAIR NOLLER: Dr. Hillard?
DR. HILLARD: Certainly others have made some good suggestions, and some good thoughts. I think that if one were to do additional studies, or even to try to look at this issue around this device or others in the future, being sure that you have good data for the youngest age group in collecting the data about age of coitarche and first intercourse. Certainly lesions take awhile to develop, and knowing how long it has been that that lesion has potentially been developing is potentially important. So collecting that data.
And then the only other thought I had has already been mentioned, related to CIN2 versus 3, and looking also at that data by age.
CHAIR NOLLER: Thank you. Dr. Cedars?
DR. CEDARS: I think that there actually do have to be other studies that are done, because as we've been discussing, you can sort of slice and dice the statistics in different ways, but if it's not statistically significant, in answer to your question in terms of clinical significance versus statistical significance. If it's not statistically significant, you can't even say that the conclusion is valid. And so it doesn't really matter whether it's clinically significant or not. You can ask the reverse, but I think it's very hard to try to say something's clinically significant if it's not statistically significant first because that's whether or not it's valid. And so I think you're going to have trouble doing that with the data that you have.
And I particularly think, and Dr. Hillard brought this up earlier, that the criteria for screening has changed. And obviously, you can't sort of be penalized for having started the study before the new criteria. But I think that the new criteria, having changed, is going to change how people practice, or at least we hope it's going to change how people practice. And I would think that any new study is going to have to be based on the new screening criteria. And as your first study showed, you're going to be looking primarily at your low-grade Paps, your ASCUS and your LSIL, which is going to take out 50 percent of your patients, which is part of the reason you probably didn't have power when you started looking at the lower grade lesions.
So I think you almost have to do another study, using the new current screening guidelines, including age, and HPV status, and looking at specifically just your low-grade Paps. And then you're going to have to set hard statistical guidelines that you're going to meet. And I think in terms of opening up the black box, if you open up the black box and you change the algorithm, then you've got to do the study all over again anyway. I mean, you can't really -- because you don't have biopsies.
CHAIR NOLLER: No discussion, please.
DR. CEDARS: You don't have biopsies in those spots. So I think you really are -- I mean, I think it's very exciting technology, but I really do think that you're going to have to do another study.
CHAIR NOLLER: Thank you. Dr. Weeks?
DR. WEEKS: Well, I don't know how practical a suggestion this is, but it sounds like a number of panel members think more study is needed. I'd simply like to see Pivotal Study II completed, because if it were completed with the N that was estimated in the beginning, then I think the data, if it holds true, if this trend holds true, would be much more compelling.
I agree with Dr. Miller that it'd be interesting both for PSI and II to look back at the slides and try to separate out CIN2 and 3. For the package material, or teaching material, I think it's an important question, what are clinicians going to do when CIN2 specimen, or CIN2 is discovered while using LUMA. Perhaps some of the magnates, or the practitioners and researchers who are in the forefront can put together some sort of teaching material that could be available to the clinicians online to hopefully encourage them not to start off on a path of doing multiple LEEPs and surgical procedures, particularly on young women. And that's the end of my suggestions.
CHAIR NOLLER: Thank you. Dr. Julian?
DR. JULIAN: Well, I have to reiterate that looking at the group of investigator sponsors here, they're among the best in the specialty, and I'm not really qualified, in my opinion, to tell them how to do research. If I had to make a single recommendation, it would be to demonstrate that the LUMA improves outcomes in non-expert colposcopists, not in expert colposcopists. That is the target group for this device. Rehashing the information presented here will bring up the same problems that we had today, I think.
CHAIR NOLLER: Thank you. Dr. D'Agostino?
DR. D'AGOSTINO: I think re-analyzing the data that you have is going to be very important. I think the CIN2's versus CIN3's, if it's possible to make that sort, and also look at the situation where the LUMA was taken but then not read, if you can get some data, and whatever else you can do to get data from the existing studies. I do think you're going to need another study, and possibly can design the next study on a sort of smaller population, a target population, some population where you think the LUMA, from analyzing this past data, where you think the LUMA might actually make a big contribution so that there's sort of enough evidence for the FDA and for an advisory committee to say there's something going on here.
I do think the issues that were just raised, though, will haunt you, and I'm not sure how you're going to handle that without a really large study. I do think it's expert-driven in this study, and the open hearing just made me all the more concerned that if the study looks similar to what you have now, how are you going to generalize it. So I do think you have those issues, but in terms of getting approval, getting the FDA to be convinced that there's something going on here, you may be able to, from looking at the past data, get a well directed new study with clear hypotheses, clear statistical procedures, and not a huge study to show that there is a benefit, and then you face the other issues as time goes on.
CHAIR NOLLER: Thank you. Dr. Snyder?
DR. SNYDER: Like Dr. Julian, I don't really feel qualified to be telling them what to do. I do think that if there's any way to refine the ability of this to detect abnormality, through different changes in the algorithm, through more just direct comparison from one biopsy to another biopsy, it would be wonderful. I'm not sure how to do that.
CHAIR NOLLER: Thank you. Dr. Hayes?
DR. HAYES: Well, in appreciation of the richness of skill on the team and our discussion today, and some of the recommendations regarding education and labeling, and almost like lessons learned. And that's a retrospective look. As we go forward, I do think there needs to be another study, but operationalizing some of the suggestions I think will make it ever better, and address the questions would be a good thing to do, and I would encourage them to do that.
CHAIR NOLLER: Thank you. One last bit of housekeeping here. Please leave all your materials on the table. FDA will pick them up. Nancy Brogdon, any final comments for us?
MS. BROGDON: Only that FDA staff would like to thank the panel for your time and efforts and your expertise. Thank you.
CHAIR NOLLER: I would like to thank all the panel members. I think everyone was very thoughtful today. They did their homework, and I greatly appreciate it. It made my job easier. This 69th Meeting of the Obstetrics and Gynecology Devices Panel is now adjourned.
(Whereupon, the foregoing matter was concluded at 4:31 p.m.).