Subgroup analysis of treatment effects for misclassified biomarkers with time-to-event data

TL;DR

This paper introduces a novel EM-based method for analyzing treatment effects in subgroups defined by misclassified biomarkers in time-to-event data, addressing the challenges posed by non-proportional hazards and measurement error.

Contribution

It develops a mixture Cox model approach with EM estimation and profile likelihood confidence intervals to accurately assess treatment effects despite biomarker misclassification.

Findings

Method performs well in simulations with close to nominal coverage.

Applied successfully to renal-cell cancer trial data.

Addresses non-proportional hazards in subgroup analysis.

Abstract

Analysing subgroups defined by biomarkers is of increasing importance in clinical research. In some situations the biomarker is subject to misclassification error, meaning the true subgroups are identified with imperfect sensitivity and specificity. For time-to-event data, it is improper to assume the Cox proportional hazards model for the effects with respect to the true subgroups, since the survival distributions with respect to the diagnosed subgroups will not adhere to the proportional hazards assumption. This precludes the possibility of using simple adjustment procedures. Instead, we present a method based on formally modelling the data as a mixture of Cox models using an EM algorithm for estimation. An estimate of the overall population treatment effect is obtained through the interpretation of the hazard ratio as a concordance odds. Profile likelihood is used to construct individual and simultaneous confidence intervals of treatment effects. The resulting confidence intervals are shown to have close to nominal coverage for moderately large sample sizes in simulations and the method is illustrated on data from a renal-cell cancer trial.