Statistical considerations of chemoprevention clinical trials in prostate cancer.

The design and analysis of chemoprevention trials in prostate cancer is more complicated than that of an ordinary treatment efficacy trial. The following issues must be taken into account when designing prospective randomized chemoprevention trials: (1) The choice of the primary endpoint: biopsy-proven prostate cancer or mortality from prostate cancer? The choice of the endpoint has a major impact on the design of the trial, its overall duration, and the number of patients required. (2) The sample size calculation which depends on the event rate in the control group, the size of the difference to be detected, the size of the type I and type II errors, the non-adherence and crossover rates, and the duration of patient entry and follow-up. (3) The possible confounding between the intervention and ascertainment of the endpoint. What is the effect of the intervention on the sensitivity and specificity of prostate screening tests? What are its implications on trial design? (4) The effect of 'noncompliance' on intent-to-treat analyses: nonadherence or treatment refusal, crossovers, loss to follow-up, biopsy refusal, and intercurrent deaths. (5) The assessment of side effects, both acute and long term.