Clinical results with irinotecan (CPT-11 [Camptosar]) and other camptothecin derivatives in various cancers, although encouraging, have fallen short of the expectations predicted by preclinical models. One proposed explanation for this is that preclinical xenograft models do not predict for the sensitivity of human cancer. In this article, we describe the results of several studies suggesting that this explanation is incorrect. Instead, our results indicate that the discrepancy between clinical response rates and findings in preclinical models may be due to a failure to incorporate the principles learned from preclinical studies into the design of clinical trials. Our analysis suggests that if differences in host tolerance are taken into account, the xenograft models are quite accurate predictors of clinical response. Moreover, application of the principles derived from preclinical models to the design of clinical trials may significantly enhance clinical response rates. Thus, the camptothecin analogs provide a paradigm for better integrated, pharmacokinetically driven, preclinical and clinical development of new drugs.