Type I Error Control and Interim Monitoring for Co-Primary Hypotheses Involving a Subgroup

Presented by Jiayi Hu

Masters Candidate in Biostatistics

Plan B Adviser: Lianne Siegel

The recent growth of immunoglobulin-based therapies has motivated clinical trials testing primary endpoints both in the overall cohort and in subgroups of patients, such as in patients without specific antibodies at baseline. Existing multiple testing methods in clinical trials often ignore the natural correlation between test statistics in such contexts, resulting in overly conservative type I error control. Motivated by OTAC, an ongoing Phase III trial evaluating the effect of a single infusion of anti-COVID-19 hyperimmune intravenous immunoglobulin (hIVIG), in outpatient adults with recently diagnosed SARS-CoV-2 infection, in both the overall cohort and in the subgroup of participants who had not received monoclonal antibodies or antiviral treatments, we propose a method termed the “correlation correction” to control the type I error at a predetermined rate while taking the estimated correlation into account, thus increasing efficiency. A closed form for the estimated correlation between test statistics is provided assuming maximum likelihood estimation. We evaluated this method through extensive simulation studies. Our findings indicate that the proposed method controls the type I error at the desired rate, improves power, and reduces the expected sample size compared to a Bonferroni correction.