1 Statistical Review Dr. Shan Sun-Mitchell

2 ENT Primary endpoint: Time to treatment failure by day 50 Placebo BDP Patients randomized Number of treatment failures 30(45%) 18 (29%) Adjusted hazard ratio (95% CI) 0.63 (0.35, 1.13) Stratified log rank test P =

3 Multiplicity Multiple analyses on multiple pre-specified and non-specified endpoints for study ENT 00-02, study 875 and both studies combined were conducted. Once the study failed on the primary endpoint, any further analyses are exploratory and statistical significance cannot be determined.

4 Post-Hoc Endpoints Survival at Day 200 Post-Transplant –Based on the timing of the transplant, not on randomization date or study drug administration –Not a valid endpoint for comparison Survival at One Year Post-Randomization –Not pre-specified for either trial –Sponsor’s p=0.04 in ENT can not be compared with the 0.05 level ---primary failed, no α left ---multiple analyses on multiple endpoints Overall Survival Post-Randomization –Not pre-specified for either trial –No planned uniform follow-up

5 Problems with Post-Hoc Analyses Study ENT failed to demonstrate efficacy based on the primary endpoint There was no uniform follow-up for patients post study treatment and any post study treatment or other conditions that may influence survival were not captured---potential bias in the analyses. All subsequent analyses are considered exploratory since there is no type I error rate left for further testing. Any subsequent analyses can only inflate type I error rate. All the p-values in the subsequent analyses are not interpretable.

6 Problems with Pooling / Meta Analyses

7 Problems with Pooling The two trials are not concurrent—10 years apart Two different study designs Primary objective and study endpoints were different The stratification factors were different Different study therapy and follow-up durations Different enrolled populations Follow-up on all patients for survival not planned

8 Guidelines on Meta Analyses ICH Guidance E9: “Under exceptional circumstances a meta analytic approach may also be most appropriate way, or the only way, of providing sufficient overall evidence of efficacy via an overall hypothesis test. When used for this purpose the meta-analysis should have its own prospectively written protocol”

9 Guidelines for Meta Analyses EMEA 2001 Points to consider on applications with 1. Meta analyses; 2. One pivotal study, Section Regulatory prerequisites of retrospective meta-analysis: “prerequisites for a retrospective meta-analysis to provide sufficient evidence for a claim include:--Some studies clearly positive”… “A retrospective meta- analysis of only two studies originally intended to stand on their own is not expected to add any useful information”

10 Post-Hoc Exploratory Analyses: Overall survival post-randomization for studies 875 and ENT Placebo BDP Placebo BDP Subjects randomized Survival status (dead ) 17(59%) 10(32%) 32(48%) 27(44%) Hazard ratio (95% CI) 0.47 (0.22, 1.04) 0.71(0.42, 1.20) Stratified logrank test* p=0.06 p=0.20 * p-value can not be compared to 0.05

11 No pooling / Meta analyses can be conducted (ICHE9 and EMEA 2001) as neither of the studies demonstrated efficacy with respect to survival

12 Summary Failed primary endpoint Further analyses of pre-specified /unspecified endpoints increases the false positive error rate Multiple analyses of multiple endpoints have been conducted Survival at day 200 post-transplant—not valid efficacy endpoint Pooling /Meta analysis not acceptable

13 Safety Clinically significant safety issue is HPA axis suppression ACTH stimulation tests done at baseline and, also day 51, if normal baseline and no treatment failure by day 50, ENT Abnormal HPA axis function day 51 in 86% of BDP patients vs. 58% of control group (77% vs. 57% in a second analysis)

14 Registration Study Failed primary efficacy endpoint Abnormal HPA axis function at day 51 was identified as a safety issue

15 Conclusion Additional studies are necessary to demonstrate efficacy and safety