3. Regulatory Considerations

In view of the complexity of adaptive designs, the sponsor should communicate with the regulatory agency as early as possible during the trial design in order to allow sufficient time to improve the plan accordingly.

5.1 Communication With The Regulatory Agency

For adaptive designs aimed at exploratory objectives, it is not necessary to communicate with the regulatory agency. However, communication with the regulatory agency is needed if the trial may affect the safety of many subjects, eg, a master protocol trial with a large number of subjects; or early development is aimed at exploratory objectives, which may later evolve to confirmatory studies. Usually, it is necessary to communicate with regulatory agency in advance for adaptive designs in confirmatory studies such that there is sufficient time to consider the suggestions, concerns, and/or opinions of regulatory agency in the early stage of design, especially for designs that are complex and/or utilize new methods. Any documented agreements with regulatory agency should be reflected in the amended protocol.

5.2 Documentation Requirements

The documents to be submitted by the sponsor should contain all the theories, literatures and data used to support the use of the adaptive designs for review by the regulatory agency. Preparation of the documentation should focus on the pre-specified adjustment plan and comprehensive discussions of the clinical meaningfulness, validity, integrity and feasibility.

Medical meaningfulness is an important factor in judging whether the use of the adaptive design is appropriate. The documentation should contain sufficient evidence to support clinical meaningfulness of the trial results after appropriate adjustments. For example, after one or more adjustments, interpretation of the trial results may become quite difficult, or the trial results may eventually reach a statistically positive result without clinical meaningfulness.

Validity applies mainly to the statistical methods, of which the most important criterion is whether the statistical methods used can control the overall type I error rate at a 2-sided 0.05 (or one-sided 0.025) level. The documentation should include the pre-specified adjustment plans, all adjustment procedures and details, and all references cited. If the adaptive design is extremely complex and there is no specific theoretical formulation, it may need to be illustrated by simulation methods. The sponsor needs to consider during the planning stage whether the simulation results can be independently verified by a third party.

Integrity pertains to the operation and conduct of the trial, of which the criterion is that the design used will not introduce bias due to the trial operation or conduct. The documentation should include all operational procedures, especially how to set up a firewall to ensure that the analysis results will not be disclosed. Other relevant guidelines may be referred to for adaptive modifications that the data monitoring committee are in charge of.

Feasibility is aimed at assessing whether the potential adaptive modifications planned can be implemented in practice, which requires the sponsor to make a comprehensive evaluation.

The above are only the basic contents that should be included in the documentation. If the sponsor thinks that there are other materials that would facilitate communication with the regulatory agency, those may also be submitted.

5.3 Other Considerations

In principle, the plans for adaptive modifications in an adaptive design must be prespecified within the trial protocol and statistical analysis plan before the clinical trial starts. In general, non-prespecified modifications to a trial are not recommended. However, in the practice of clinical trials, it is sometimes necessary to make ad-hoc modifications based on the data.  After careful consideration, valid ad-hoc modifications to a trial may be acceptable given those modifications do not compromise the validity, integrity, and feasibility of the trial, and appropriate communication with the regulatory agency to obtain confirmation in advance is required.

In addition, making certain modifications to an ongoing clinical trial based on external data is not considered an adaptive modification, but should be reflected in protocol amendments, which are communicated to the regulatory agency in a timely manner. There are many situations where the protocol is amended based on external data: for example, a trial in which the drug is too toxic for patients who are marker negative or newly completed trial(s) of drug(s) in the same class demonstrated  that the effect is only in marker positive patients, thus, the target population needs to be modified to marker positive patients only; newly completed trial(s) of drug(s) in the same class demonstrated that the choice of the primary endpoint is not appropriate, or newly published corresponding guidelines have recommended another primary endpoint definition, which requires modification of the primary endpoint; change in the standard of care treatment requires modification of the control group treatment; or a trial needs to be terminated early because it cannot continue enrolling patients. The sponsor should pay particular attention that these modifications are based on external data only, but not on results from the ongoing trial itself.

Reference

1. CDE.Guidelines for the Establishment and Implementation of the Clinical Trial Data Monitoring Committee.2020

2. Bauer P, Köhne K. Evaluation of experiments with adaptive interim analyses.Biometrics 1994;50: 1029 – 1041, rectioninBiometrics1996;52: 380.

3. Bauer P, Kieser M. Combining differentiated phases in the development of medical treatments with in a single trial.Statistics in Medicine1999;18: 1833-1848.

4. Bauer P, Posch M. Letter to the Editor.Modification of the sample size and the schedule of interim analyses in survival trials based on data inspections by H. Schaefer and H.- H. Mueller, Statistics in Medicine 2001;20: 3741 – 3751. Statistics in Medicine2004;23: 1333 – 1335.

5. Chang M. Adaptive design method based on sum of p-values.Statistics in Medicine 2007;26: 2772 – 2784.

6. Chen C, Li X, Li W, Beckman RA.Adaptive Expansion of Biomarker Populations in Phase 3 Clinical Trials.Contemporary Clinical Trials2018;71: 18-85.

7. ChenJYH, DeMets DL, Lan GKK.Increasing the sample size when the unblinded interim results is proming. Statistics in Medicine 2004;23: 1023-1038.

8. ChowSC, Chang M. Adaptive Design Methods in Clinical Trials.CRC Press;2nd edition 2011.

9. Chow SC, Shao J, Wang H, Locknygina Y. Sample size calculations in clinical research. Chapman & Hall/CRC, 3^rdEdition2018.

10. Cui L, Hung HMJ, Wang SJ.Modification of sample size in group sequential clinical trials.Biometrics1999;55: 853-857.

11. EMA Scientific Guidelines: Reflection paper on methodological issues in conventional clinical trials planned with an adaptive design.https www.ema.europa.eu/en/human- regulatory/research-development/scientific-guidelines;2007.

12. FDA Drug Guidance for Industry.Adaptive Designs for clinical trials of drugs and biologics. https www.fda.gov/drugs/guidances-drugs/all-guidances-drugs2018.

13. Friede T, Kieser M. Sample size receiving in internal pilot studies: a review.Biological Journal2006;48: 537 – 555.

14. Friede T, Parsons N, Stallard N. A conditional error function approach for subgroup selection in adaptive clinical trials.Statistics in Medicine2012;31: 4309 – 4320.

15. Friede T, Stallard Nigel.A component of methods for adaptive treatment selection.Biological Journal 2008;50: 767 – 781.

16. Gould AL.Interim analyses for monitoring clinical trials at do not material effect the Type I error rate.Statistics in Medicine1992;14: 1039-1051.

17. HochbergY.A sharper Bonferroni procedure for multiple tests of signature.Biometrika1988;75: 800 – 802.

18. Jenkins M, Stone A, Jennison C. An adaptive seamless phase II/III design for oncology trials with subpopulation selection using related survival endpoints.Pharmaceutical Statistics2011;10: 347 – 356.

19. Lan KG, DeMetsDL.Discrete Sequential Boundaries for Clinical Trials. Biometrika1983;70: 659 – 663.

20. Lehmacher W, Wassmer G. Adaptive sample size calculations in group sequential trials.Biometrics1999;55: 1286 – 1290.

21. Maca J, Bhattacharya S, Dragalin S.et al.Adaptive Seamless Phase II/III Designs Background, Operational Aspects, and Samples.Drug Information Journal 2006;40: 463-474.

22. Marcus R, Peritz E, Gabriel KR.On closed testing procedures with special reference to ordered analysis of variance.Biometrika1976;63: 655 – 660.

23. Müller HH, Schäfer H. Adaptive group sequential designs for clinical trials: Combining the advances of adaptive and of classical group sequential procedures.Biometrics2001;57: 886 – 891.

24. Müller HH, Schäfer H. A general statistical principle for changing a design any time during the course of a trial.Statistics in Medicine2004;23: 2497 – 2508.

25. O'Brien PC, Fleming TR.A Multiple Testing Procedure for Clinical Trials.Biometrics1979;549 – 556.

26. Pocock SJ, Group Sequential Methods in the Design and Analysis of Clinical Trials.Biometrika1977;64: 191 – 199.

27. Proschan MA, Hunsberger SA.Designed extension of studies based on conditional power.Biometrics 1995;51: 1315 – 1324.

28. Rosenblum M, Van Der Laan MJ.Optimizing randomized trial designs to distinguish whichsubpopulation from treatment.Biometrika2011;98: 845-860.

29. Shih WJ.Sample size re-investigation – a journney for a decade.Statistics in Medicine2001;20: 515-518.

30. Shih WJ, Li G, Wang Y. Methods for flexible sample-size design in clinical trials: Likelihood, weighted, dual test, and promising zone applications.Contemporary Clinical Trials2016;47: 40-48.

31. Simes RJ.An advanced Bonferroni procedure for multiple tests of signature.Biometrika1986;73: 751 – 754.

32. Stallard N, Hamborg T, Parsons N, Friede T. Adaptive designs for congenital clinical trials with subgroup selection.Journal of Biopharmaceutical Statistics2014;24: 168 – 187.

33. Wang SJ, Hung HMJ, O'Neill RT.Adaptive patient design design in therapeutic trials.Biological Journal2009;51: 358 – 374.

34. Wassmer G, Brannath W. Group sequential and continuous adaptive designs in clinical trials.Springer; 2016.

35. Wu PS, Lin M, Chow SC.On sample size investigation and re-investigation adjustment for variation in congenital trials.Journal of Biopharmaceutical Statistics2016;26: 44-54.

 

36. Zhang JJ, Blumenthal G, He K, Tang S, Cortazar P, Sridhara R. Overestimation of the effect size in group sequential trials.Clinical Cancer Research2012;18: 4872-4876.

 

1） values in the two stages as p1 and p2 respectively. Compare p1 + p2 with 0.2236 for efficacy evaluation. This method has a restriction that the efficacy estimates δ1 obtained from interim analysis should not be in the opposite direction with the original one δ0 , and this method can’t reduce sample size.

 

Example 4: Two-stage Seamless Phase II/III Adaptive Design

Imagine there is a multi-center, randomized, double-blind, active-controlled, parallel-group, superiority study to assess the improvement of a new drug in some symptom. The primary endpoint is change from baseline at Week 8 in a score, which is assumed to follow a normal distribution approximately. It is planned to use a two-stage seamless phase II/III adaptive design. The phase II trial has two drug groups (high dose and low dose) and one control group with a randomization ratio of 1:1:1. Efficacy, defined as the difference of mean improvement in the score between subjects in two groups, is expected to be δ. Let one-sided α=0.025 and power 1-β (e.g., 90%). Sample size N is calculated for the comparison between one dose group and the control group. When 3n1 subjects finish their 8-week follow up, Stage 1 ends. Total sample Size is N+n1 (number of subjects in the two comparison groups N + number of subjects in the de-selected dose group in Stage 1 n1).

 

For Stage 1, let p11 and p12 be the p-value of Z-test between low dose group vs. control group (null hypothesis H011 : there is no difference between the low dose group and the control group) and high dose group vs. control group (null hypothesis H012: there is no difference between the high dose group and the control group), respectively. Closed testing procedure and Hochberg procedure is used to adjust multiplicity. The p-value of the test of no difference between either the low or high dose group and the control group (H011∩H012) is pint1=min[2*min (p11, p12), max (p11, p12)]. The p-value from efficacy comparisons between the selected dose group in Stage 1 and the control group will be p1=max(pint1, min(p11, p12)) after multiplicity adjustment. 

Example 5: Adaptive Enrichment Design

Imagine there is a multi-center, randomized, double-blind, active-controlled, parallel-group, two-stage superiority clinical trial. The primary endpoint is overall survival (OS), with the secondary endpoint progression-free survival (PFS). Assume the hazard ratio is 0.75 in the overall population HR (F), and 0.55 in the positive subgroup HR (S). Use 1-sided test level 0.025, test power 90%, and other necessary parameters to calculate the total number of deaths required in the overall population, denoted as N0.

 

The trial is planned to select the target population in the interim analysis when 40% subjects are enrolled. The duration of overall survival is quite long in this study, hence the selection of target population in the interim analysis is based on PFS. The decision strategy is: ① If estimated HR (F)<0.85 and HR (S)<0.65, the trial will be continued in both the positive subgroup and the overall population in Stage 2; ② If HR (F)≥0.85 and HR (S)<0.65, only subjects in the positive subgroup will be enrolled in Stage 2; ③ If HR (F)<0.85 and HR (S)≥0.65, the trial will be continued in the overall population in Stage 2, and the positive subgroup will not be analyzed; ④ The trial will be terminated for futility if HR (F)≥0.85 and HR (S)≥0.65. 

As to this trial, if the test in Stage 1 is based on PFS and the test in Stage 2 is based on OS, it will be difficult to interpret the meaning of p-values of two stages in the final analysis. Therefore, in this trial design, decision making of Stage 1 is based on descriptive statistics of PFS, while in final analysis, the p-value of two stages is based on OS. For trials using survival as an endpoint, no matter in which stage the endpoint event occurs for subjects enrolled in Stage 1, it should be included in the results of Stage 1 in the calculation. Otherwise, the assumption of independence between two stages will no longer be held, and type I error rate will be inflated.

 

Example 6: Adaptive Master Protocol Study

Imagine there is a superiority clinical trial which is to test a new drug in treating rare cancer patients with BRAF V600E mutation positive. The primary endpoint is the objective response rate confirmed by an independent endpoint committee, and the time to sustainable response is documented. A multi-center, single-arm basket design is adopted. Subjects enrolled in the trial must be in the late stage of the disease and have a central laboratory confirmed BRAF V600E mutations. There are five cohorts, including anaplastic thyroid cancer, biliary tract cancer, gastrointestinal stromal tumor, hairy cell leukemia, and small intestine adenocarcinoma.

 

Although all subjects are enrolled under the same protocol in this trial, each of the 5 cohorts will be considered as an independent trial with separate results to support the submission of corresponding cohorts. Because the purpose of the trial is to support the application of a new drug, the sample size must be determined in advance, and the sample size required for each cohort should be calculated separately according to the decision rule of superiority. As to the consideration  to combine data from two or more cohorts, due to the lack of sufficient data in this trial to support the investigational product have the same mechanism of action and similar efficacy in patients with BRAF V600E mutation positive, therefore, it is not acceptable to combine data from any two or more cohorts to support the application of a new drug in the corresponding pooled cohorts.

 

 点击此处，查看原文附件

Guideline on Adaptive Designs for Clinical Trials (Draft)