Early stopping designs based on progression-free survival at an early time point in the initial cohort.

We introduce a new study design in which patients are evaluated early in their treatment for disease progression. Our design is appropriate when lack of progression, both early and late, is the criterion for treatment success. An initial cohort of n(1) patients is followed until the last one has been evaluated. If enough of these patients are progression free (PF) at an early time point, t(1) after arrival, a second cohort is recruited until n(2) total patients are evaluable for PF survival at the final time t(2). Otherwise, the trial is terminated for futility both early in time and with a minimal number of patients. Patients in the initial cohort who are PF at t(1) continue on study and are again evaluated at t(2). The design permits early stopping for rapid progression of disease, an indication of futility both for cytotoxic and newer non-cytotoxic targeted therapies. The design tests the composite hypothesis of a probability p(1) of being PF at t(1) and p(2) of being PF at t(2) given PF at t(1). Power and type I error are maintained at design point levels over a wide range of parameters p(1) and p(2). No distributional assumptions are needed other than the binomial, so the design provides rigorous power analysis for this type of study. Tables of optimal designs are supplied for a broad range of requirements. Copyright (c) 2007 John Wiley & Sons, Ltd.