Premium
Two‐stage marker‐stratified clinical trial design in the presence of biomarker misclassification
Author(s) -
Zang Yong,
Jack Lee J.,
Yuan Ying
Publication year - 2016
Publication title -
journal of the royal statistical society: series c (applied statistics)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.205
H-Index - 72
eISSN - 1467-9876
pISSN - 0035-9254
DOI - 10.1111/rssc.12140
Subject(s) - biomarker , covariate , stage (stratigraphy) , oncology , statistics , computer science , predictive marker , medicine , artificial intelligence , machine learning , mathematics , biology , cancer , paleontology , biochemistry
Summary The marker‐stratified design (MSD) is an important design to assess treatment and marker effects in personalized medicine. The MSD stratifies patients into marker positive and marker negative subgroups on the basis of their biomarker profiles and then randomizes them to the standard treatment or a new treatment within each subgroup. The performance of the MSD can be seriously undermined when the biomarker is measured with error (or misclassified). A recently proposed analytic method corrects the biomarker misclassification in the MSD under the assumptions that the biomarker classification rates are known and no other covariates need to be adjusted. We propose a two‐stage MSD to relax these assumptions. We analytically investigate the bias in the estimation of prognostic and predictive marker effects and treatment effects caused by biomarker misclassification in the presence of covariates, and we propose an expectation–maximization algorithm to correct such biases. The design does not require prespecification of the misclassification rates and can incorporate any covariates that potentially confound the prognostic and predictive marker effects and treatment effect. Numerical trial applications show that the method has desirable operating characteristics.