Determination of Sample Size for Validation Study in Pharmacogenomics

Pharmacogenomics aims at co-development of a drug that targets a subgroup of patients for safety and efficacy and a device that identifies the responder group through their genetic variations. Development of such a prognostic device includes a training stage and a validation stage. The transition from the training stage to the validation stage typically involves change of platforms as a subset of potential genetic markers predictive of drug response are identified in the first stage and only those are used in the second stage. With the change in consideration, this paper concerns how to determine sample sizes for the validation stage to meet pre-specified sensitivity and specificity requirements in order to avoid futility of pharmacogenomic development. In particular, taking microarrays as a medical device, which measure gene expression levels, we show how to decide the numbers of subjects per group, replicated samples per subject, replicated probes per gene for the validation experiment. The change of platforms is taken into account in the sample sizes calculation by statistical modeling. Our formulation of sensitivity and specificity requirements calls for estimation of both measures. Lower bounds with carefully calibrated confidence levels can give appropriate sample size to meet the requirements. The procedure is illustrated in a proof-of-concept mice experiment.