Association between human tumor colony-forming assay results and response of an individual patient's tumor to chemotherapy.

An in vitro tumor colony-forming assay was utilized to measure the sensitivity of 800 individual patients' tumors to a variety of antineoplastic agents. Thirty-six separate histologic types of cancer were represented. Only 199 of the 800 patients' tumors (25 percent) both formed colonies in vitro and had enough cells in the biopsy or fluid specimen to perform drug sensitivity assays. In 123 instances the drug tested in vitro against the tumor was also used clinically to treat the patient. The clinician caring for the patient did not know the results of the in vitro test. When analyzed in a retrospective manner, the probability of a positive prediction from the assay, given the patient responded clinically, was 0.88. The probability for a negative prediction of the assay given the patient did not respond, was 0.94. Associations of in vitro and in vivo results in the 123 correlations were highly significant (p less than 0.001). We conclude that, as now constituted, the human tumor colony-forming assay can provide useful sensitivity information for only about 25 percent of the general oncology patients. Secondly, a prospective clinical trial of the assay is needed to insure that the assay is indeed predictive of which drug will produce a patient response and that it is not merely an indicator that a particular patient's tumor is highly responsive in vivo.