Statistical issues in studies of toxicity modifiers

The primary focus of oncology is the search for a cure. In pursuit of this goal, therapeutic regimens have increased in intensity, often causing increased normal tissue morbidity. Pharmacologic agents are being developed to ameliorate or prevent these toxicities. The stepwise investigation of these agents requires a different paradigm from standard drug development. Traditionally, phase I studies are designed to find the maximum tolerated dose of an experimental drug; however, in pharmacologic agents that aim to reduce toxicity, defining the maximum tolerated dose, may not be necessary. The focus of phase I studies in these agents is the establishment of the maximum effective dose. Phase II studies traditionally are designed to establish efficacy of an experimental regimen before advancing to a phase III study. To determine potential efficacy, predictable efficacy levels must be available in the literature for standard therapy. In this manner, the phase II study can be evaluated against the historical control to determine if a phase III study should be conducted. This generally is not possible in the context of toxicity modifiers. Because of variability in toxicity assessment, prior rates of toxicity may not be reliable when evaluated against a phase II study focusing on toxicity amelioration. Because the assessment of toxicity is subjective, retrospective comparisons may be biased and lead to the wrong conclusion. One method to correct for this bias is to conduct placebo-controlled, randomized phase II studies. Phase III studies of toxicity modifiers have toxicity amelioration or prevention as the primary objective. These studies should be designed to assess the subjective and objective components of toxicity, evaluate quality of life, and assess possible adverse effects on therapeutic outcomes. Design and analysis issues are discussed in this report.