Randomized discontinuation design: Application to cytostatic antineoplastic agents

Purpose: Propose a phase II study design to evaluate the activity of a putative cytostatic agent, acknowledging heterogeneity of tumor growth rates in the population of patients. Methods: In the setting of renal cell carcinoma, some patients' tumors will grow slowly naturally. An appropriate design has to distinguish antiproliferative activity attributable to the novel agent from indolent disease. We propose a randomized discontinuation design that initially treats all patients with the study agent (stage 1) and then randomizes in a double-blind fashion to continuing therapy or placebo only those patients whose disease is stable (stage 2). This design allows the investigators to determine if apparent slow tumor growth is attributable to the drug or to selection of patients with naturally slow-growing tumors. Results: By selecting a more homogeneous population, the randomized portion of the study requires fewer patients than would a study randomizing all patients at entry. The design also avoids potential confounding because of heterogeneous tumor growth. Because the two randomly assigned treatment groups each comprise patients with apparently slow growing tumors, any difference between the groups in disease progression after randomization is more likely a result of the study drug and less likely a result of imbalance with respect to tumor growth rates. Stopping rules during the initial open-label stage and the subsequent randomized trial stage allow one to reduce the overall sample size. Expected average tumor growth rate is an important consideration when deciding the duration of follow-up for the two stages. Conclusion: The randomized discontinuation design is a feasible alternative phase II study design for determining activity of possibly cytostatic anticancer agents.