1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics...

33
1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER

Transcript of 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics...

Page 1: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

1

Efficacy ResultsNDA 22092 (MTP-PE)

Laura LuStatistical Reviewer

Office of BiostatisticsFDA/CDER

Page 2: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

2

Main Issues for DFS1. The Applicant’s pooled analysis not

appropriate due to treatment by regimen interaction and a comparison to an experimental arm that performed worse than standard of care.

2. Statistical significance is not reached when applying applicant’s method of pooled analysis to FDA dataset.

3. Conduct of interim analyses complicate the interpretation of DFS results.

Page 3: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

3

Main Issues for OS

1. The primary endpoint (DFS) of the study was not met.

2. Patient follow-up for OS was inadequate to perform a meaningful analysis.

Page 4: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

4

INT-0133 Efficacy Datasets

• IDM 2003 Dataset – used by COG for analyses published in JCO

• IDM 2006 Dataset

• FDA Dataset

Page 5: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

5

FDA Dataset• FDA considered data captured on case report

forms to be “primary source data”.• Where discrepancies identified between CRFs

and IDM 2003 dataset, CRF information used to determine dates of death and relapse.

• FDA did not modify IDM 2006 dataset because CRFs for additional follow-up not submitted. Therefore, FDA could not verify the accuracy of this information.

Page 6: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

6

FDA DatasetExcludes• Seven (7) ineligible patients contained in IDM

Dataset 2003Includes • Nine (9) additional events identified in review of

CRFs• Additional follow-up documented on CRFs from one

institutionModified based on FDA review of CRFs • Change in length of disease-free survival for 66 pts• Change in length of overall survival for 68 pts

Page 7: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

7

Main Issues for DFS

1. The Applicant’s pooled analysis not appropriate due to treatment by regimen interaction and a comparison to an experimental arm that performed worse than standard of care.

2. Statistical significance is not reached when applying applicant’s method of pooled analysis to FDA dataset.

3. Conduct of interim analyses complicate the interpretation of DFS results.

Page 8: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

8

Study Design

• Randomized to four study arms• One arm (Regimen AΘMTP-PE ) represents the

control arm/standard of care• Three experimental arms, each containing at

least one experimental agent• Evaluation of the efficacy of any experimental

regimen needs to be considered relative to the control regimen.

Page 9: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

9

Study Analysis Issues• Study designers considered use of pooled

analysis (A/B MTP-PE vs. AΘ/BΘ MTP-PE) for the effect of MTP-PE.

• CCG/POG discussed the risk of this study design and analysis: “We hope that interactions between MTP-PE and the alternative chemotherapy arms will be similar. In this case it will be possible to analyze the proposed study by a factorial design. If the interactions are different, it will be necessary to consider the study as if it were a four arm analysis.”

Page 10: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

10

Pooled AnalysisApplicant's Method

• Compare

Regimen AMTP-PE vs. Regimen AΘMTP-PE

Regimen BMTP-PE vs. Regimen BΘMTP-PE

separately• Pool the results in the two comparisons

(stratified log-rank test).

Page 11: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

11

Treatment by Regimen Interaction

• Hazard ratio for AMTP-PE vs. AΘMTP-PE = 0.99

• Hazard ratio for BMTP-PE vs. BΘMTP-PE = 0.62

• The p-value for the treatment of MTP-PE by regimen interaction is 0.067.

Page 12: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

12

Kaplan-Meier Curves for DFS by Regimen Based on Primary Data

Page 13: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

13

Kaplan-Meier Curves for DFS by Regimen Based on Primary Data

Page 14: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

14

Regimen B vs. Regimen A

Hazard ratio of Regimen BΘMTP-PE vs. Regimen AΘMTP-PE is 1.18. Therefore, Regimen BΘMTP-PE performed worse than Regimen AΘMTP-PE.

Page 15: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

15

Sensitivity Analysis • Conducted to evaluate impact of inferior

performance of Regimen BΘMTP-PE on pooled outcome.

• In this analysis, results for Regimen AΘMTP-PE (control arm) are substituted for the results in Regimen BΘMTP-PE (inferior experimental chemo arm) in the pooled analysis.

Page 16: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

16

Hazard Ratio

P-value

Results by substituting AΘMTP-PE for BΘMTP-PE on FDA Dataset

0.86

0.28

Results of Sensitivity Analysis

Page 17: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

17

Analysis Driven by Comparison to Inferior Regimen

This sensitivity analysis shows that the small p-value of the pooled analysis is driven by a comparison to an experimental regimen that performed worse than the control regimen.

Page 18: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

18

Invalid Analytic Method

Based on• Evidence of an treatment by regimen interaction

AND• Evidence of pooled analysis driven by a

comparison where experimental regimen (BΘMTP-PE) is inferior to control regimen (AΘMTP-PE)

A pooled analysis is not appropriate.

Page 19: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

19

DFS Results by Regimen Relative to Regimen A Based on FDA Dataset

Variable Number of Patients

Number of Events

Hazard Ratio relative to A

P-value

Regimen AΘMTP-PE

171 60 ---

Regimen AMTP-PE

165 57 0.99 0.96

Regimen BΘMTP-PE

166 67 1.18 0.35

Regimen BMTP-PE

169 46 0.73 0.11

Page 20: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

20

Main Issues for DFS1. The Applicant’s pooled analysis not

appropriate due to treatment by regimen interaction and a comparison to an experimental arm that performed worse than standard of care.

2. Statistical significance is not reached when applying applicant’s method of pooled analysis to FDA dataset.

3. Conduct of interim analyses complicate the interpretation of DFS results.

Page 21: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

21

DFS Results by Applicant’s Pooled Method Based on FDA Dataset

• Hazard ratio for A/B MTP-PE vs. AΘ/BΘ MTP-PE = 0.78,

p-value = 0.065

Page 22: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

22

Main Issues for DFS

1. The Applicant’s pooled analysis not appropriate due to treatment by regimen interaction and a comparison to an experimental arm that performed worse than standard of care.

2. Statistical significance is not reached when applying applicant’s method of pooled analysis to FDA dataset.

3. Conduct of interim analyses complicate the interpretation of DFS results.

Page 23: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

23

Type I Error Rate

It is impossible to determine the actual Type I error rate due to the conduct/timing of the interim and final analyses.

Page 24: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

24

Interim Analysis Conduct

• According to the final protocol amendment on 6/16/1997, one interim analysis performed with no detailed information for conduct or alpha spending.

• Two additional interim analyses on EFS conducted; timing not based on specific number of events.

Page 25: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

25

Final Analysis Conduct

• A final analysis was not conducted according to the protocol (approx 167 DFS events)

• IDM provided results for an analysis performed after 228 events

• If the timing of the final analysis is influenced by the results of the interim analyses, the type I error rate will be impacted.

It thus unclear what alpha should be used for the IDM analysis that included available data as of 4/9/2003.

Page 26: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

26

Final Analysis Conduct

• If a pooled analysis is performed based on 167th DFS event using IDM 2003 dataset, the results are not statistically significant (nominal p-value = 0.11).

Page 27: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

27

Main issues for OS

1. The primary endpoint of the study (DFS) was not met.

2. Follow-up on OS was inadequate to perform a meaningful analysis.

Page 28: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

28

Survival Analysis Issues

• When the primary endpoint was not met, all alpha was spent.

• Any further analysis after the study failed to win on the primary endpoint increases the type I error rate. So literally, the difference in other endpoints should not be considered statistically significant.

Page 29: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

29

Survival Analysis Issues

• Analysis plan for OS was not pre-specified. ‘Post-hoc’ analyses make it difficult to interpret the results for OS, since, by continuing to conduct tests for a treatment effect on different endpoints and/or the same endpoint a so-called “statistically significant result” (p < 0.05) can eventually be obtained even when there is no treatment effect.

• When an endpoint is selected based on the study results, the results for that endpoint are biased.

Overall Survival analyses are exploratory

Page 30: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

30

Main issues for OS

1. The primary endpoint of the study (DFS) was not met.

2. Follow-up on OS was inadequate to perform a meaningful analysis.

Page 31: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

31

Adequacy of Follow-up for OS Analyses

As of the 2003 data cut-off for OS– 22% of patients (148/678) had died per IDM’s

2003 dataset– Among 530 remaining patients who were alive

as of last contact, there were:• 8% with last contact on/before 12/31/1994• 11% with last contact on/before 12/31/1997• 51% with last contact on/before 12/31/2000

Page 32: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

32

Adequacy of Follow-up for OS Analyses

• More than 50% of the 530 patients alive at the last contact were lost to follow-up two years prior to data cutoff in 2003

• In a well conducted trial for registration with OS as a primary endpoint, FDA expects substantially less than 5% of patients lost to follow-up at data cutoff.

Page 33: 1 Efficacy Results NDA 22092 (MTP-PE) Laura Lu Statistical Reviewer Office of Biostatistics FDA/CDER.

33

Adequacy of Follow-up for Survival after Disease Occurrence

Among those patients lost to follow-up for survival,

– 26 had active disease (recurrent osteosarcoma or AML) at the last follow-up.

– These patients probably died.