Diagnostic Assays to Identify a Subgroup Likely to Benefit from a Therapy: Bayesian Models to Bridge from the Clinical Trial Assay to a Me-Too Assay via External Concordance Data

• Activity Number: 320
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Gene Pennello*+ and Jingjing Ye
• Companies: FDA and FDA
• Keywords: non-differential subgroup misclassification ; Non-differential subgroup misclassification ; personalized medicine ; tailored therapy ; measurement error ; calibration

Abstract

For some therapies, a companion diagnostic assay (or test) is used to identify patients likely to benefit from the therapy. For example, colorectal cancer patients who test negative for KRAS mutations are eligible for treatment with cetuximab. After the therapy and the test are co-validated in a randomized clinical trial, a second, "me-too" test may be developed by another manufacturer for the same purpose. Usually, however, another clinical trial is not possible to conduct. Instead, concordance between the two tests may be evaluated in a study external to the original clinical trial. We use a Bayesian approach to combine the concordance data with the clinical trial data to obtain estimates of clinical efficacy of the therapy in subgroups identified by the "me-too" companion test. In particular, we assume conservatively that misclassification of the original test result by the "me-too" test is non-differential to clinical outcome, but then use a set of priors designed to investigate sensitivity to this assumption. We evaluate our methodology in a clinical trial in which results from both tests are available.