Continuous Safety Monitoring in Large Phase I Cancer Clinical Trials with Multiple Expansion Cohorts

Biostatistics and Bioinformatics Seminar

Masha Kocherginsky, PhD, Northwestern University and Theodore Karrison, PhD, University of Chicago

Traditional phase I oncology clinical trial designs have morphed into multiple, parallel phase I trials that incorporate concurrent expansion cohorts across multiple diseases or disease subtypes. These cohorts are used to provide additional safety data, as well as preliminary efficacy data. As a result, some of these Phase I trials have become quite large. For example, a recent study [Cancer Letter, Oct 7, 2016] looked at PD-1 drug development, and identified 35 phase I trials with ≥200 patients. In such trials, dose escalation is typically done using a standard phase I design, e.g. “3+3”, to determine the maximum tolerated dose (MTD), which is followed by enrollment of multiple expansion cohorts in different disease types or biomarker-defined subgroups. Safety is typically monitored only within each expansion cohort, and the results are not jointly evaluated in real time across the multiple expansion cohorts.

Here we propose to use a continuous safety monitoring approach based on the sequential probability ratio test (SPRT) and discuss its application to adverse event (AE) monitoring for large Phase I trials with multiple concurrent expansion cohorts. Trial designs based on this approach consist of a continuous safety monitoring boundary with a rejection number for each additional AE, and the rejection boundaries can be calculated using existing software. Based on our simulation studies and evaluation of the operating characteristics, we found that this is a practical approach which maintains Type I error rate and results in smaller average sample size than the current approach.