A Bayesian basket trial evaluates the same target across five tumour types. Four small cohorts have similar response rates, but one larger cohort has

Question

A Bayesian basket trial evaluates the same target across five tumour types. Four small cohorts have similar response rates, but one larger cohort has no responses. What is the central benefit and principal calibration risk of a hierarchical model?

Accepted Answer

Adaptive borrowing can improve precision in concordant small cohorts, but excessive shrinkage can create false-positive conclusions in a genuinely inactive heterogeneous cohort

Answer

No information is shared, making it identical to five independent Simon designs

Answer

The model can estimate efficacy without prespecifying priors

Answer

All cohorts are pooled completely, so heterogeneity cannot affect inference

Answer

Bayesian borrowing guarantees control of frequentist type I error without calibration

Bayesian Basket — SCE Medical Oncology MCQ

A Bayesian basket trial evaluates the same target across five tumour types. Four small cohorts have similar response rates, but one larger cohort has no responses. What is the central benefit and principal calibration risk of a hierarchical model?