Rodent and nonrodent toxicology studies are currently expected to support Phase I trials of antineoplastic drugs in the United States. To determine the predictive value of these studies, we initiated a project to compare preclinical and clinical toxicity data within various drug classes. The first class analyzed was the platinum anticancer drugs. Twelve platinum analogues that had both preclinical (mice, rats and/or dogs) and clinical data from matching drug administration schedules were identified. The rodent LD10 (the dose that causes lethality in 10% of treated animals) or dog toxic dose high (a dose that when doubled causes lethality in dogs) correlated well with the human maximally tolerated dose on a mg/m2 basis. For every platinum analogue investigated, one-third the rodent LD10 or one-third the dog toxic dose high in mg/m2 gave a starting dose and a first escalation dose that did not exceed the clinical maximally tolerated dose. The dose-limiting toxicities in patients were previously observed in 7 of 7, 7 of 8, and 9 of 11 mouse, rat, and dog studies, respectively. Our data indicate that mice, rats, and dogs all had value in predicting a safe starting dose and the qualitative toxicities in humans for platinum anticancer compounds. The efficiency of Phase 1 trials could have been improved without sacrificing patient safety by allowing higher starting doses for this drug class than conventionally permitted.