Data of five phase II clinical trials on iproplatin and carboplatin, conducted by the ECTG, have been pooled in order to evaluate the extent of toxicities of these compounds. One hundred and seventy patients treated with iproplatin and 65 patients treated with carboplatin were evaluable. Most of them (81%) had been previously treated with chemotherapy. Doses ranged from 180 to 300 mg/m2 every 4 weeks for iproplatin, and from 350 to 450 mg/m2 every 5 weeks for carboplatin, according to the initial status of the patient. WHO criteria were used to grade toxic effects. Weekly blood counts were performed, and lowest observed counts were analysed by non-parametric methods. Censored data were analysed by actuarial methods. Thrombocytopenia was the dose-limiting toxicity and was dose related. Leucopenia was less severe. The risk of thrombocytopenia varied largely amongst patients, and could be predicted from the initial platelet count, the initial creatinine level and prior therapy with alkylating agents. The cumulative risk increased with the total dose, but with a decreasing hazard rate, and without additional delay to platelet recovery. Nausea, vomiting and diarrhoea were the most frequently observed non-haematological side-effects, and were more severe with iproplatin than with carboplatin. Peripheral neuropathy was observed in some cases, but could be due to prior treatments. Renal toxicity did not cause major problems. Our results confirm the findings of the phase I trials: thrombocytopenia is dose-limiting for both drugs, and renal side-effects are negligible. The risk model of thrombocytopenia, consistent with Egorin's model for carboplatin, could serve as a basis for dose adjustment. The feasibility of the scheme could be insufficient for prolonged treatment.