All Times EDT
Keywords: two-stage randomization, assay assessment
We describe the design and analysis of a study with two objectives: 1) evaluate a new assay that identifies patients appropriate for an approved drug, and 2) compare the drug to a comparator drug. The study needed to yield an estimate of the concordance table for the new and old assay, estimate the difference in positive predictive value between the new and old assays, estimate the treatment difference for patients who qualify for treatment based on the new assay, and estimate the treatment difference for patients qualifying based on the old assay. Due to the high cost of the assays and an expected 50% screen failure rate, patients were randomized to receive one of the two assays during screening. If a patient was positive on the assay, they were then randomized into the drug portion of the trial. For patients receiving the sponsor's drug, the alternative assay was also performed (in a blinded procedure). We show this design can produce estimates almost as precise as a much more expensive design that collects both assays during screening. It is much more precise than designs that regard the two randomly assigned assay groups as separate and independent.