Combined modalities are currently used for cancer therapy, although their mechanisms of activity remain incompletely deciphered. The design of new drug combinations suffers from our inability to anticipate accurately their efficacy or toxicity. They can be evaluated in vivo, using human tumors grafted into immunodeficient mice, as we did here with combined protocols used in the clinical setting. Xenografts of small cell lung carcinoma (SCLC) from eight patients were used to test the tumor sensitivity to etoposide (VP16; 12-16 mg/kg/days, days 1, 2, and 3), cisplatin (CDDP; 6-9 mg/kg/day, day 1) and ifosfamide (IFO; 90-210 mg/kg/day, days 1, 2, and 3) as single agents and to evaluate the efficacy of the two-drug or three-drug combinations. Five xenografts came from untreated patients (SCLC-61, SCLC-6, SCLC-10, SCLC-41, and SCLC-96) and three after treatment (SCLC-74, SCLC-101, and SCLC-108). p53 was inactivated in all of them. Tumor growth inhibition, growth delay, and the survival rate of tumor-bearing mice reflected individual SCLC chemosensitivity. As single agents, IFO inhibited tumor growth in a dose-dependent manner, whereas CDDP and VP16 had little or no effect. Both CDDP and IFO potentiated VP16, inducing complete regressions in the most sensitive SCLCs; VP16-IFO was more effective than VP16-CDDP, with complete regressions in six versus three of the eight tumors tested, respectively. CDDP-IFO was less effective than VP16-IFO, with three of eight SCLCs giving complete regressions. The three-drug combination led to modest improvement over the best two-drug combination but only for sensitive SCLCs. Because drug-responses distinguished two classes of SCLCs, as sensitive or refractory, MDR1, glutathione S-transferase pi, lung-related multidrug resistance protein, multidrug resistance protein, and topoisomerase IIalpha mRNA expression was studied by semiquantitative reverse transcription. There was no correlation with SCLC sensitivity; topoisomerase IIalpha and multidrug resistance protein was expressed in all cases, lung-related multidrug resistance protein and glutathione S-transferase pie in seven of eight, and MDR1 gene in four of eight. In conclusion, these SCLC xenografts displayed a pattern of chemotherapy response close to that observed in patients. This model confirmed that in two-drug combinations, each component potentiated the effects of the other, with VP16-IFO tending to be the best two-drug combination, both of which were more effective than VP16-CDDP and better tolerated than CDDP-IFO. The addition of a third agent gave a modest, if any, therapeutic benefit in the responders but none in refractory SCLCs. There was no correlation between the extent of response and resistance markers.