Adjusting for selection bias in assessing treatment effect estimates from multiple subgroups

Abstract

This paper discusses a number of methods for adjusting treatment effect estimates in clinical trials where differential effects in several subpopulations are suspected. In such situations, the estimates from the most extreme subpopulation are often overinterpreted. The paper focusses on the construction of simultaneous confidence intervals intended to provide a more realistic assessment regarding the uncertainty around these extreme results. The methods from simultaneous inference are compared with shrinkage estimates arising from Bayesian hierarchical models by discussing salient features of both approaches in a typical application. K E Y W O R D Sselection bias, shrinkage estimation, simultaneous confidence intervals, subpopulations

A comparison of estimation methods adjusting for selection bias in adaptive enrichment designs with time‐to‐event endpoints

“…Other selection rules based on conditional power or predictive power could have been chosen, however similar conclusions are expected 18 . Besides, some works in literature have also set up the ground for further exploration, for example of the calculation of appropriate confidence intervals via simultaneous inference, 19 bootstrap resampling, 10 confidence regions based on orderings 12 or simultaneous inference based on the duality between hypothesis testing and confidence intervals 9 . All of these methods have advantages and limitations, and we refer to literature for further information.…”

A unified Bayesian framework for bias adjustment in multiple comparisons from clinical trials

“…Figure B1 shows estimates without adjusting for multiple comparisons, estimates with adjustment for multiple comparisons assuming the true treatment effects from a normal distribution, and estimates with adjusting for multiple comparisons using the proposed BMA method. mutate(model = rep(c("raw","normal","GM2","GM4","GM6","GM8","GM10","BMA_best","BMA_avg"),each=68), BMA-mod Bayesian model averaging method over the five Gaussian mixture models defined in Equation (5).…”

“…Glimm provides a comprehensive review of existing work in handling selection bias in clinical trials. 5 Estimators based on frequentist and empirical Bayesian methods have been extensively studied. [6][7][8][9] The drawback of these approaches is they either assume the true values are the same (group sequential design) or the true values are from a normal distribution.…”

“…The study team is often tempted to focus on those promising results and choose to move the corresponding treatment, subgroup, and endpoint forward for further investigations. Glimm provides a comprehensive review of existing work in handling selection bias in clinical trials 5 . Estimators based on frequentist and empirical Bayesian methods have been extensively studied 6‐9 .…”