A wide-spectrum paired comparison of the properties of the Rolling 6 and 3+3 Phase I study designs

Abstract

The recently proposed Rolling 6 Phase I study design is an alternative to the standard 3+3 patient cohort design. It was proposed as a way to reduce the time necessary to complete a trial by reducing the downtime during which Phase I trials are suspended for evaluation of dose-limiting toxicity (DLT). We performed a paired comparison of the Rolling 6 and 3+3 designs via a large simulation study over a wide spectrum of conditions defined by inter-patient arrival time, number of doses, DLT dose response slope, DLT dose response shape, and starting dose. Our results in general confirm that, on average, trials will be completed more quickly with the Rolling 6, with negligible penalty in terms of patient safety during the trial or as a result of the selected MTD, although the Rolling 6 on average will require more patients and thus will incur higher research costs. While the reduction in average trial duration with the Rolling 6 can be notable in absolute terms, in relative terms the reduction may be modest, in contrast to the relative increase in numbers of patients treated, which can be substantial. For identical sequences of patients, the difference between the two designs in trial duration and numbers of patients treated is highly variable. Therefore it is not so predictable whether the Rolling 6 will result in more rapid completion of a single trial or of a small series of trials.