Can modest escalations of dose be detected as increased tumor control?

Clinically defined groups of tumors are usually characterized by shallow dose-response curves, and this results from heterogeneity among individual dose-response curves, each of which is very likely quite steep. A review of published results for human tumors indicates that a 10% escalation of dose to tumors controlled at the 50% level, where changes in outcome are most likely to be detected, will be detectable in a population of unselected patients only in sizable clinical trials (130-300 patients per dose level). With a few exceptions, a dose escalation of 20% will be detectable in much smaller trials (50-130 patients per dose level). Therefore, clinical trials of improved treatment modalities will be confounded by patient heterogeneity, and modest improvements may go undetected in all but the largest trials. Mathematical modeling was used to study the effect on the steepness of the dose-response curve of selecting patients on the basis of the radiosensitivity measure SF2 (surviving fraction at 2 Gy). If SF2 is a faithful predictor of response in a group of tumors, then heterogeneity could be reduced by excluding the patients with the most sensitive (controlled with near certainty) and most resistant (recurring with near certainty) tumors. The resulting "stochastic fraction" (tumors for which treatment outcome is probabilistic) would be characterized by a steep dose response, and the number of patients required to demonstrate the effect of dose escalation would be substantially reduced (by about 50%).