Statistical Review and Evaluation

 

 

NDA #:                                              20-757

Applicant:                                           Sanofi-Synthelabo

Name of Drug:                                   Irbesartan

Indication:                                          Treatment of renal disease

Document reviewed:                          Electronic submission

Date of submission:                           August 3, 2001

Statistical Reviewer:                         John Lawrence, Ph.D. (HFD–710)

Medical Reviewer:                            Juan-Carlos Pelayo, M.D. (HFD–110)

 

 

1 Introduction

 

            This is a review of Studies CV131-048 (The Collaborative Study Group Trial: the Effects of Irbesartan on Morbidity and Mortality in Hypertensive Patients with Type II Diabetes and Diabetic Nephropathy) and EFC2481 (IRbesartan MicroAlbuminuria in type 2 diabetes).  These were the main studies conducted in support of this application. 

 

2.1 Study Design

 

1715 hypertensive patients (SeSBP > 135 mmHg and/or SeDBP > 85 mmHg in an untreated patient, or receiving antihypertensive medication) with type 2 diabetes mellitus and overt nephropathy (urine protein excretion > 900 mg/24 hours) were randomized into one of three treatment groups- once daily administration of placebo, irbesartan, or the active control amlodipine.  Patients in the irbesartan group were initially given 75 mg, and were force-titrated up to 150 mg at week 2, and up to the final dose of 300 mg at week 4. Patients in the amlodipine group were initially given 2.5 mg, and were force-titrated up to 5 mg at week 2, and up to the final dose of 10 mg at week 4.  Additional antihypertensive agents were encouraged to attain equal degrees of blood pressure reduction within all treatment groups. Downward titration was permitted per protocol at the discretion of the Investigator.  The demographic characteristics of the patients in each group appear in Table 2.1.1.  There do not appear to be any major differences in the baseline characteristics of the three groups.  There was expected to be an enrollment period of approximately 2 years and the study was expected to end 2 years after the last patient was recruited.

 

 

 

 

 

 

 


Table 2.1.1 Characteristics of the patients in the two groups at baseline.  For continuous variables, this table shows the group mean ± standard deviation. [Reviewer’s analysis]

 

Characteristic

Irbesartan

Amlodipine

Placebo

N

577

559

563

Age (years)

59 ± 7

58 ± 8

59 ± 8

Gender (Male/Female)

376/ 201

356/ 203

401/ 162

Race (Caucasian/Black/Other)

436/ 63/ 78

384/ 84/ 91

411/ 77/ 75

Baseline SeDBP

87 ± 11

87 ± 11

87 ± 11

Baseline SeSBP

160 ± 20

159 ± 19

158 ± 21

Log of baseline urinary protein excretion rate

8.0 ± 0.8

8.0 ± 0.8

8.0 ± 0.8

 

 

2.2 Planned Statistical Analysis

 

            The primary endpoint was the time to the first occurrence of doubling of serum creatinine, end-stage renal disease (ESRD), or all cause mortality.  ESRD was defined by renal transplantation, need for dialysis, or serum creatinine ³ 6 mg/dL.  The primary analysis used the logrank test to compare the distribution of the time to the composite event in the irbesartan group versus the placebo group.  The comparison of the irbesartan group to the active control amlodipine was a secondary analysis and therefore, no adjustment for multiplicity was used for these two comparisons.

 

There were four interim analyses.  The actual number and timing of these interim analyses were not specified in advance.  The stopping boundaries were defined using the Lan-DeMets alpha-spending function of the O’Brien-Fleming type.  The information times of the interim analyses and the nominal critical values appear in Table 2.2.1.  This reviewer used the Lan-DeMets software from the University of Wisconsin and found close agreement between the limits that were used by the sponsor and the limits that were found by this reviewer.

 

Table 2.2.1 Nominal critical values and information times of interim analyses. [Source: Appendix 6 of Study Report and confirmed by reviewer]

 

Interim Analysis

Information Time

Nominal critical value

Nominal alpha

1

6.76%

7.5326

< 0.0001

2

15.37%

5.0071

< 0.0001

3

33.81%

3.4047

0.0007

4

50.97%

2.775

0.0055

Final

100%

1.98

0.0477

 

 

            There was one pre-specified secondary cardiovascular composite endpoint (cardiovascular death, nonfatal myocardial infarction, hospitalization for heart failure, permanent neurological deficit attributed to stroke, or above-the-ankle amputation).  The three pre-specified secondary analyses were: comparison of irbesartan to placebo on this secondary endpoint and the comparison of irbesartan to amlodipine on both the primary and secondary endpoints.

 

 

2.3  Results

 

            A total of eight patients were lost to follow-up out of the 1715 patients enrolled.  Sixteen patients in the study were not randomized but were included as part of the treatment group corresponding to the treatment they actually were given (2 in the irbesartan group, 8 in the amlodipine group, and 6 in the placebo group).  19 patients in the irbesartan group and 11 patients in the placebo group discontinued treatment and/or regularly scheduled office visits.  However, many of these patients (6 in irbesartan, 2 in placebo) were counted in the analysis as having a primary event.  Since these are relatively small numbers, these issues are not likely to cause as much concern in this study as they would in other studies.  Table 2.3.1 shows the number of events in each of the three arms for the composite endpoint and each of its components. The Kaplan-Meier curves appear in Figure 2.3.1.

 

Table 2.3.1 Number of events in each arm, point estimates of relative risk, 95% confidence interval for relative risk of event in irbesartan group relative to placebo. [Source: Tables 10.1.1.A and B of Study Report and confirmed by reviewer]. The components are tabulated in two ways: first by the definition of the primary endpoint (i.e. counting only the first event for each patient) and second by the number of patients that had the event at any time during the study.

 

Amlodipine

Placebo

Irbesartan

RR / 95%CI/ p-value

Irb vs. Pbo

N

567

569

579

 

Primary Endpoint

(death, ESRD, 2´ SC)

233

222

189

0.80/(0.66, 0.97)/

0.0234

Decomposition of Events in Definition of Primary Endpoint

(counting only first occurrence of primary event)

Death

54

64

64

0.94/(0.67, 1.33)/0.744

ESRD

50

47

43

0.88/(0.58, 1.33)/0.542

2´ SC

144

135

98

0.67/(0.52, 0.87)/0.003

Summary of All Events Occurring at Any Time in Study

Death

83

93

87

0.92/(0.69, 1.23)/0.568

ESRD

104

101

82

0.77/(0.57, 1.03)/0.073

2´ SC

144

135

98

0.67/(0.52, 0.87)/0.003

24 patients in placebo, 16 patients in irbesartan, and 15 patients in the amlodipine group reached ESRD and doubling of serum creatinine on the same date and appear in both rows.


Figure 2.3.1  Kaplan-Meier curves for estimated event-free survival in three arms.

 

 


            From the curves in Figure 2.3.1 and the counts of number of events in Table 2.3.1, it is apparent that the patients in the irbesartan group had fewer events than the patients in either the placebo arm or the amlodipine arm. The p-value for the primary analysis (irbesartan versus placebo on time to the composite endpoint using the logrank test) is 0.0234 [Source: Tables 10.1.1.A of Study Report and confirmed by reviewer].  Most of the apparent reduction in relative risk is from the creatinine component.  Note that serum creatinine ³ 6 mg/dL is included in the definition of ESRD.  If one excludes those patients who met the definition of ESRD because of high serum creatinine, then there are numerically fewer patients in the placebo group who reached the primary endpoint due to ESRD (23 patients in the placebo group and 27 patients in the irbesartan group).  Moreover, when the components are analyzed individually, a significant difference only appears in the serum creatinine component.  Although this composite endpoint was pre-specified as the primary endpoint, one can argue that the benefit was only shown on a surrogate endpoint (SC).  One patient in the placebo group progressed to ESRD and was alive at the end of the study, but the date of the event was unknown.  In the sponsor’s analysis, this patient was included in the counts of events, but not included in the logrank analysis.  Since the patient was in the placebo group, this is a conservative way of handling the missing value and this reviewer would prefer to use a less conservative approach that adheres to the intention-to-treat principle, viz. performing an analysis that counts this patient as having an event at a time that is interval censored. As a way of checking the robustness of the primary analysis, this reviewer tried imputing 0, 700, or 1400 days as the time of the event for this patient.  In all three cases, the p-value was even more significant than the analysis that ignores this patient.  Likewise, there were two patients in the amlodipine group that had a missing date for the time to ESRD and were not included in the secondary analysis comparing irbesartan to amlodipine.  For those patients had missing serum creatinine measurements, it is possible to use interval censoring techniques or a worst case analysis (i.e. assume all patients in the treatment group had an event and those in the placebo group did not).  However, since there were so few patients (19 in the irbesartan group and 11 in the placebo group) and many of them had events, it is unlikely that these alternative analyses will be useful.  Moreover, a mixed effects model showed that the mean rate of change of serum creatinine was significantly smaller in patients in the irbesartan group relative to either placebo or amlodipine (p=0.004 and p=0.013 respectively from Section 10.4.1 of the Study Report).  Hence, it is possible that the missing serum creatinine levels, if missing at random, would bias the results against the irbesartan treatment.

 

            For the secondary analyses, there was an estimated 23% reduction in risk on the primary endpoint relative to amlodipine (p=0.0064).  Again, this apparent effect is explained mainly by changes in SC.  There was no significant difference observed on the secondary cardiovascular endpoint relative to either placebo or amlodipine (p=0.45 and p=0.69 respectively).

 

2.4 Safety

 

            The ten most common adverse events are listed in Table 2.4.1.  There appears to be a difference in the reports of edema in the amlodipine group relative to the other two groups.  There does not appear to be a difference between irbesartan and placebo in the number of any specific adverse event.

 

 

Table 2.4.1 Most common clinical adverse events; number and % of subjects.

 

Adverse Event

Placebo

N = 563

Irbesartan

N = 577

Amlodipine

N = 559

Edema

211 (37.5)

222 (38.5)

337 (60.3)

Musculoskeletal Pain

215 (38.2)

218 (37.8)

193 (34.5)

Upper Respiratory Infection

143 (25.4)

144 (25.0)

136 (24.3)

Dizziness

111 (19.7)

143 (24.8)

97 (17.4)

Fatigue

147 (26.1)

134 (23.2)

129 (23.1)

Nausea/Vomiting

111 (19.7)

112 (19.4)

108 (19.3)

Diarrhea

83 (14.7)

102 (17.7)

73 (13.1)

Headache

110 (19.5)

94 (16.3)

72 (12.9)

Cough

84 (14.9)

84 (14.6)

96 (17.2)

Abnormality Retina

68 (12.1)

75 (13.0)

52 (9.3)

 

 

 

 

 

2.5 Results by subgroup

 

            The confidence intervals for the relative risk for the primary endpoint appear in Figure 2.5.1.  Although the study was not powered to show a treatment effect in each subgroup, these confidence intervals appear to show that the possible benefit of irbesartan over placebo is greater in males than in females and is also greater among the White race than it is among Non-whites.  Moreover, the point estimate for the relative risk among North American patients in the study is 0.95.  This appears to be larger than the relative risk in other regions.  There were 204 North American patients in the irbesartan group and 197 North American patients randomized to the placebo group.

 

 

Figure 2.5.1 Relative risk by subgroups [Source: Figure 10.1.2A of Study Report]

 

 

3.1 Study Design

 

Study EFC2481 was a multinational double-blind study comparing two doses of irbesartan with placebo.  611 patients with Type 2 diabetes mellitus and SeSBP>135 mmHg or SeDBP>85 mmHg or being treated for hypertension and evidence of microalbuminuria were randomized. After a 3-week placebo run-in period, patients randomized to one of the irbesartan groups were given 75 mg for the first two weeks of the double-blind treatment period and were titrated up to 150 mg for the next two weeks.  Those patients randomized to the 300 mg dose were then titrated up to the final dose at week 4.  Subjects remained on this dosing regimen until month 24.  Table 3.1.1 shows the patient demographics.

 

Table 3.1.1 Characteristics of the patients in the two groups at baseline.  For continuous variables, this table shows the group mean ± standard deviation. [Source: Table 8.3 of Study Report]

 

Characteristic

Irbesartan 150 mg

Irbesartan 300 mg

Placebo

N

203

201

207

Age (years)

58 ± 8

57 ± 8

58 ± 9

Gender (Male/Female)

134/ 69

140/ 61

142/ 65

Race (Caucasian/Black/Other)

198/ 2/ 3

194/ 0/ 7

203/ 0/ 4

Baseline SeDBP

90 ± 9

91 ± 10

90 ± 9

Baseline SeSBP

153 ± 14

153 ± 14

153 ± 15

 

 

3.2 Planned Statistical Analysis

 

The primary endpoint was the time to occurrence of clinical proteinuria (defined as albuminiria excretion rate greater than 200 mg/min and an increase of 30% from baseline at two consecutive evaluations).  Each active treatment group was compared to placebo using the two-sided logrank test with a Bonferroni adjustment so that each comparison was done with a nominal alpha of 0.025.  The primary analysis was on the per-protocol population.  Only patients who met all inclusion criteria and who stayed on double-blind therapy for at least 3 months were included in the per-protocol population.  Also, all patients from site 1004 were removed from the per-protocol population due to significant lack of compliance.  No efficacy interim analyses were planned.

 

            The pre-specified secondary objectives were to evaluate the changes from baseline in overnight urinary albumin excretion rate, estimated creatinine clearance, von Willebrand factor, fibrogen, factor VII, plasminogen activator inhibitor-I, and lipid profile.

 

 

3.3  Results

 

            The number of events in each subgroup, as well as the p-value from the logrank test appears in Table 3.3.1.  There is a statistically significant difference between the 300 mg group and placebo (p=0.0013) and the 150 mg group showed a numerical trend toward longer proteinuria-free survival, but did not reach statistical significance (p=0.096).  The analysis on the intent-to-treat population gives similar results (p=0.0004 for the high dose vs. placebo and p=0.085 for the low-dose vs. placebo).  The Kaplan-Meier curves for the estimated probability of remaining proteinuria-free appear in Figure 3.3.1.

 

 

 

Table 3.3.1 Number of events in each arm. [Source: Tables 10.1.1.1 of Study Report and confirmed by reviewer]

 

 

Placebo

Irbesartan 150 mg

Irbesartan 300 mg

N (per-protocol population)

172

160

162

Primary Endpoint (Proteinuria)

27

16

10

Nominal p-value vs. placebo

 

0.096

0.0013*

Relative risk vs. placebo

95% CI for RR

 

.595

(0.321, 1.105)

0.311

(0.146, 0.662)

*Significant at level 0.025 using Bonferroni adjustment.

 

 

 

 

Figure 3.3.1  Kaplan-Meier curves for estimated event-free survival in three arms [Source: Figure 10.1.1.1 of Study Report].

           

 

            The results for the secondary analyses appear in Tables 3.3.2, 3.3.3, and 3.3.4.  The first two variables were only analyzed at 3, 6, 12 18 and 24 months from randomization while the remaining variables were analyzed at 1-year and 2-years from randomization.  Both doses were compared to placebo and there were no adjustments for multiple testing. The only striking difference that appears among all of the secondary parameters was in urinary excretion rate.  There were nominally significant differences in change in urinary excretion rate at every time point between both doses of irbesartan and placebo.  There were no significant changes in serum creatinine at any time point for any dose of irbesartan.  There did not appear to be any significant difference in the coagulation parameters in Table 3.3.4 although there was a nominally significant difference in  plasminogen activator inhibitor-I for the high dose relative to placebo.  Finally, there did not appear to be any significant difference in the mean change in lipid profile in Table 3.3.5 although there was a nominally significant difference in HDL cholesterol levels for the low dose relative to placebo.

 

 

Table 3.3.2 Results on secondary analyses on change from baseline in overnight urinary excretion rate. GM = geometric mean, PC = % change, SEM = standard error of mean. [Source: Tables 10.2.1.1A of Study Report]

 

 

 


Table 3.3.3 Results on secondary analyses on change from baseline in serum creatinine. GM = geometric mean, PC = % change, SEM = standard error of mean. [Source: Tables 10.2.2.1 of Study Report]


 


Table 3.3.4 Results on secondary analyses on change from baseline in von Willebrand factor, fibrogen, factor VII, and plasminogen activator inhibitor-I. [Source: Tables 10.2.3.1 of Study Report]

 

 

aUnadjusted raw mean change for vWF and Fibrinogen; geometric mean % change for Factor VII and PAI.
Table 3.3.5
Results on secondary analyses on mean change from baseline in lipid profile. [Source: Tables 10.2.4.1A of Study Report]

 

 

aGeometric mean for triglycerides; raw means for everything else.
3.4 Safety

 

            The ten most common adverse events are listed in Table 2.4.1.  There appears to be a difference in the reports of dizziness and diarrhea in the 300 mg group relative to placebo and numerically more reports in the 150 mg group relative to placebo.  There does not appear to be a difference between irbesartan and placebo in the number of any other specific adverse event.

 

 

Table 3.4.1 Most common clinical adverse events; number and % of subjects.

 

 

 

3.5 Results by subgroup

 

            The estimates of the relative risk in subgroups appear in Table 3.5.1.  The point estimates are always less than 1 and appear to be particularly small in females, patients under 59, and patients with normal creatinine clearance (>107 ml/min/1.73 m2).

 

 

 

 

 

 

 

 

 

 

 

 

Table 3.5.1  Estimates of Relative Risk in subgroups from Cox model in per-protocol population [Source: Figure 10.1.2C of Study report]

 

 

 

            The results by country are summarized in Figure 3.5.1.  There is one point in the graph for each country where there were a sufficient number of patients recruited so that an estimate of the hazard ratio could be obtained.  The y-axis represents the estimated log-hazard ratio for irbesartan relative to placebo on the primary endpoint.  The x-axis represents the standard error of this estimate.  Countries with more patients have more information about the hazard ratio (smaller standard error), and therefore appear on the left side of the graph.  The dashed lines represent the overall pooled log-hazard ratio as well as the 95% confidence bands (pooled estimate ± 1.96 standard errors).  Since all of the points lie within the confidence bands, there are no countries that have significantly different treatment effects than what one would expect given the pooled results.  The only country in North America where patients were recruited in this study is Canada and there were too few events among Canadian patients to estimate the relative risk in this subgroup.

 


 


Figure 3.5.1  Plot of log-hazard ratios for primary endpoint by country.

 

 

4. Conclusions

 

 

            Two placebo controlled studies appeared to show positive findings on the primary endpoints for irbesartan 300 mg in the populations studied. Study CV131-048 showed a statistically significant increase in the time to a composite endpoint (doubling of serum creatinine, ESRD, all-cause mortality) with a p-value of 0.023 and also appeared to show superior efficacy to the active control amlodipine 10 mg.  Most of the apparent benefit seemed to be in delaying increases in serum creatinine.  The effect of irbesartan on the primary endpoint appeared to be smaller in North America than in other regions.  Study EFC2481 showed a statistically significant increase in the time to occurrence of clinical proteinuria with a p-value of 0.0013, but failed to show a statistically significant difference between a lower dose of 150 mg relative to placebo (p=0.096).  There were positive findings on some of the secondary endpoints: a 23% reduction in relative risk in the irbesartan group relative to the active control amlodipine on the primary endpoint in Study CV131-048 and nominally significant differences in change in urinary excretion rate at every time point and in change in plasma activator inhibitor-I in Study EFC2481. Diarrhea and dizziness were reported more often in the irbesartan groups than in the placebo groups in both trials, but there did not appear to be significantly more reports of any other adverse event in the irbesartan groups in these studies.

 

 

 

 

                                                                                    _________________________

                                                                                    John Lawrence, Ph.D.

                                                                                    Mathematical Statistician

 

 

This review consists of 13 pages of text, tables, and figures.

 

 

 

Concur:            James Hung, Ph.D.

                        Acting Team Leader, Biometrics I

 

 

 

                        George Chi, Ph.D.

                        Division Director, Biometrics I

 

cc:                    NDA # 20-757

                        HFD-110/Dr. Lipicky

                        HFD-110/Dr. Pelayo

                        HFD-110/Mr. Fromme

                        HFD-700/Dr. Anello

                        HFD-710/Dr. Chi

                        HFD-710/Dr. Hung

                        HFD–710/chron

 

 

LAWRENCEJ/594–5375/report.doc/5/3/2003