One- and two-stage designs for phase II window studies.
The primary goal of phase II window studies in cancer is testing new single agents or combination therapies for newly diagnosed patients who are less likely to have multiple-drug-resistant tumours than the typical phase II patients who have failed one or more chemotherapeutic regimens. In addition, by utilizing newly diagnosed patients, one can assess the responsiveness in a population more representative of where the phase II agent might be applied in a true phase III setting. In general, the outcome of a patient on a phase II window study can be categorized as response, stable disease, or early disease progression. Phase II window studies sometimes require early stopping rules for both insufficient response rate and excessive early progression rates. In this paper, one-stage and two-stage designs for phase II window studies are developed, requiring the monitoring of two rates. It is shown that the power function is monotone in response rate and early disease progression rate. Consequently, the significance level and power are easy to compute. Computational procedures are described and examples are provided. The proposed method can be applied to other studies with categorical outcomes.