In vitro tests of tumor cell drug sensitivity have been suggested as a means of selecting more appropriate clinical strategies against human cancer and improving preclinical drug development. The lack of biotransformation in these in vitro assays precludes the meaningful assessment of several major chemotherapeutic agents, including dacarbazine and cyclophosphamide. In vivo drug exposure of tumor cells in agar diffusion chambers placed in mice offers a possible solution to problems of drug biotransformation and pharmacokinetics. We have prospectively used this system as an assay for sensitivity of clonogenic human melanoma cells. Tumor cells were tested fresh, cryopreserved, and/or cultured in vitro before or after clinical use of dacarbazine, semustine, and mitolactol in 41 patient-drug combinations in which a clinical correlation could be made. Tumor cell drug sensitivity in the assay using fresh or cryopreserved tumor cells was highly correlated with clinical response and resistance with clinical nonresponse. Cultured melanoma cells exhibited enhanced plating efficiency in comparison to both fresh and cryopreserved cells of the same tumor. Cultured cells also showed increased drug sensitivity which did not correlate with drug sensitivity of the same fresh or cryopreserved tumor or with clinical response. Tumor cell drug sensitivity assays carried out in vivo provide a possible basis for preclinical evaluation of drugs which are unsuitable for in vitro testing.