1 / 31

Exploring early interim analyses in basket designs in Oncology

This presentation delves into analyzing basket designs in oncology, discussing Bayesian hierarchical models, recruitment strategies, and futility analysis methods. It explores the adjustment of models for target rates and decision-making processes for different cohorts based on interim and final analyses. Presented on May 24, 2019, in Lyon, France.

berthar
Download Presentation

Exploring early interim analyses in basket designs in Oncology

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Exploring early interim analyses in basket designs in Oncology Oliver Sailer, Anna Pöhlmann, Frank Fleischer 24 May 2019, BAYES2019, Lyon, France

  2. Basket designs and Bayesian hierarchical model Presentation title, date, author

  3. Basket designs • Basket trial* • One experimental treatment • (Patients with similar genomic features) • Different disease types (Renfro and Sargent, 2016) • Questions • Does the treatment work sufficiently? • Can we identify cohorts with a promising effect? * Renfro and Sargent (2016) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  4. Basket designs - analysis approaches Pooling Stratification • Assumes the same underlying response rate in all cohorts • Potentially large bias • Assumes independent cohorts • Low precision • Bayesian hierarchical model (BHM) • Assumes exchangeability between cohorts • Parameter of between-cohort variation determines the extent of borrowing Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  5. BHM adjusting for target rate (Berry et al., 2013) • Random cohort effect in terms of log-odds of response rate • Likelihood • Prior • Implementation: R interface to JAGS, R2JAGS package Exchangeabilityof log-odds after adjustingfortargetrates Informative prior on between-cohortvariability; scaleparameterchoiceseeNeuenschwander et al. 2015 Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  6. Recruitment and analysis strategies Presentation title, date, author

  7. Recruitment and analysis strategies • Interim analyses performed to limit exposure of additional patients to an ineffective drug • Futility could be based on within-cohort analysis only or borrow information across cohorts • Possible consequences of futility • Stop recruitment into cohort • Do not develop drug further for cohort (no-go) • Exclude cohort from later analyses (“test-then-pool”, Viele et al. 2013 ) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  8. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 1 cons cons cons I go go go F cons nogo I F cons nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  9. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 1 cons cons cons I go go go F cons nogo I F cons Interim futilityanalysis Decisionforcohort i useonlydataofcohort i nogoif <r responders nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  10. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 1 cons cons cons Final analysis Decisionforcohort i use all datafrom all cohorts based on posterior of BHM I go go go F cons nogo I F cons nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  11. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 2 cons cons cons I go go go F cons nogo I Interim futilityanalysis Decisionforcohort i use all datafrom all cohorts based on posterior of BHM F cons nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  12. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 2 cons cons cons Final analysis Decisionforcohort i use all datafrom all cohorts based on posterior of BHM I go go go F cons nogo I F cons nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  13. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 3 cons cons cons I go go go F cons nogo I F cons Interim futilityanalysis Decisionforcohort i useonlydataofcohort i nogoif <r responders nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  14. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else nogo nogo nogo nogo • Strategy 3 cons cons cons Final analysis Decisionforcohort i use all datafrom non-futile cohorts based on posterior of BHM I go go go F cons nogo I F cons nogo I F cons Recruitment I: Interim (herefutility) F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  15. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else • Strategy 4 Interim / final analysis Decisionforcohort i use all datafrom all cohorts based on posterior of BHM I I III … I F Recruitment I: Interim F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  16. Recruitment and analysis strategies Overall go at firstcohortgo Overall nogoif all cohortsnogo Overall consider, else • Strategy 4 X Interim / final analysis Decisionforcohort i use all datafrom all cohorts based on posterior of BHM In caseofinterimfutility, stoprecruitmentforcohort X I I III … I F Recruitment I: Interim F: Final Time Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  17. Model-baseddecisionrule Cohort i NoGoConsider Go q50,i median posteriorresponse rate cohort i c1,i, c2,icohortspecificdecisionboundary Variation qg,ig-quantileofposteriordistribution q50,i > c2,i q50,i < c1,i else q1-g,i > c2,i qg,i < c1,i else Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  18. Recruitment and analysis strategies • Strategy 1-3: Futility after 10 patients, final after 20 (per cohort) • Strategy 4: Interim analysis every 5 patients, Final at 4*20 max. Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  19. Response scenarios Presentation title, date, author

  20. Response scenarios • Positive and negative scenarios • Positive nugget scenario • Mixed response scenario Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  21. Model parameters Presentation title, date, author

  22. Model parameters • Model adjustment parameter • Controls borrowing • Target rate by cohort set to assumed response rate if drug works (as in Berry et al 2013) • Inter-cohort variability prior • = 0.5 allows for range of up to substantial heterogeneity Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  23. Operating characteristics Presentation title, date, author

  24. Operating characteristics • Decision probabilities in scenarios • Average sample size and duration Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  25. Simulation results Presentation title, date, author

  26. Scenario 1 NoGo, Consider, Go: Probability in %, N average number evaluable, t average duration (months) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  27. Scenario 2 NoGo, Consider, Go: Probability in %, N average number evaluable, t average duration (months) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  28. Scenario 3 NoGo, Consider, Go: Probability in %, N average number evaluable, t average duration (months) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  29. Scenario 4 NoGo, Consider, Go: Probability in %, N average number evaluable, t average duration (months) Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  30. References • Berry SM, Broglio KR, Groshen S, and Berry DA (2013): Bayesianhierarchicalmodelingofpatientsubpopulations: efficientdesignsofphase II oncologyclinicaltrials. Clinical Trials, 10(5), 720-734. • NeuenschwanderB, Wandel S, Roychoudhury S, Bailey, S (2015): Robust exchangeability designs for early phase clinical trials with multiple strata. Pharmaceut. Statist. 15, 123–134. • Renfro LA and SJ Mandrekar (2017): Definitionsand Statistical Properties of Master ProtocolsforPersonalized Medicine in Oncology. Journal ofBiopharmaceuticalStatistics 28(2), 217-228. • Viele K, Berry S, Neuenschwander B et al. (2014): Useofhistoricalcontroldataforassessingtreatmenteffects in clinicaltrials. PharmaceuticalStatistics 13, 41-54. Exploring early interim analyses in basket designs in Oncology, 24 May 2019

  31. Questions? Optional subtitle

More Related