We studied the effects of retinoids on the in vitro survival of melanoma colony-forming cells in biopsies obtained from ten patients with metastatic melanoma. The results indicate that specific retinoids reduce the ability of fresh human melanoma cells to form colonies in soft agar. The retinoids studied had differential effects on the survival of clonogenic melanoma cells, and these effects vary from patient to patient. The data provide support for the clinical trial of selected retinoids in micrometastatic and advanced melanoma.

A potential application of the human tumor stem cell colony assay is to guide Phase II clinical investigations by identifying classes of tumors (or individual patients) which are sensitive in vitro to a new antitumor compound. We have tested human tumor stem cells from 140 tumor biopsies representing 20 different tumor types for chemosensitivity to the Phase II drug 4'-(9-acridinylamino)methanesulfon-m-anisidide. In vitro sensitivity was defined as a reduction in the number of tumor colony-forming cells to 30% of the control or less after a 1-hr exposure to one-tenth of the pharmacologically achievable plasma concentration of 4'-(9-acridinylamino)methanesulfon-m-anisidide. In vitro sensitivity was found in 29 cases: non-Hodgkin's lymphoma (2 of 2); cervical carcinoma (1 of 1); sarcoma (3 of 6); neuroblastoma (1 of 2); acute myelogenous leukemia (6 of 16); chronic myelogenous leukemia (1 of 3); melanoma (8 of 34); uterine carcinoma (1 of 5); lung carcinoma (1 of 9); ovarian carcinoma (4 of 36); and breast carcinoma (1 of 11). Prospective in vitro-in vivo correlations in eight patients with various tumor types showed that three of three patients sensitive in vitro to 4'-(9-acridinylamino)methanesulfon-m-anisidide responded in vivo, while five of five patients resistant in vitro had no clinical response. The results provide support for further evaluation of the utility of the human tumor stem cell colony assay for targeting Phase II clinical trials.

The value of chromosomal analysis is well established in human hematologic neoplasms. In contrast, the relation between chromosomal abnormalities and clinical outcome in solid tumors in humans has received little study. We undertook this study to determine whether chromosomal abnormalities could provide information on the survival of patients with malignant melanoma. Chromosome-banding analysis was performed on tumor-biopsy samples from 62 patients with metastatic melanoma, and recurring cytogenetic abnormalities were correlated with survival. Patients with structural abnormalities of chromosome 7 or 11 had significantly shorter survival than patients without these abnormalities. We conclude that cytogenetic analysis may provide useful prognostic information about patients with metastatic melanoma.

The human tumor stem cell assay (HTSCA) has been used to study the in vitro sensitivity rates of anticancer drugs used in the treatment of 115 patients with previously untreated and relapsing ovarian cancer. The data from these studies have identified patterns of cross resistance and residual sensitivity between these agents, and have allowed the prospective selection of single agents possessing in vitro activity for the treatment of 32 patients with relapsing disease. cis-Platinum and vinblastine were the most active agents in vitro against ovarian TCFUs from both previously untreated and relapsing patients. Prior therapy with even one drug was associated with the acquisition of resistance to several classes of compounds (e.g., melphalan resistance was almost always associated with in vitro adriamycin resistance, P less than 0.001). A clinical trial yielding similar data would have required nearly 450 evaluable ovarian cancer patients. In 11 of 32 patients in vitro testing predicted sensitivity to single agents: eight of these had partial remissions for a predictive accuracy of 73%. In 33 instances the HTSCA had 100% accuracy in predicting the lack of clinical response. Thus, the HTSCA for advanced ovarian cancer appears to have a similar predictive accuracy rate to the estrogen receptor assay for predicting the response to hormonal therapy for disseminated breast cancer.

Ninety-seven evaluable patients with measurable, advanced, malignant melanoma were treated with recombinant alpha interferon in a cooperative phase II efficacy trial, whose primary objective was to estimate the response rate. Interferon (rIFN alpha-2a, Roferon-A) was injected subcutaneously daily for 70 days. Dose was escalated in four steps from three million units to 36 million units over ten days. Eight patients responded objectively and six patients (6%) had a complete response. The median duration of complete response was 11 months. Patients achieving complete response had only cutaneous, nodal, or pulmonary disease; some had extensive prior therapy; some could tolerate no more than three million units per day. Few patients could tolerate the target dose of 36 million units daily for 70 days. Limiting toxicity was primarily fatigue. Interferon in tolerable doses is effective in a small subset of patients with melanoma. Comparison of published trials of dacarbazine and recombinant alpha interferon indicates the two drugs have similar activity.

Objective methods have been developed to quantitate results of the in vitro human tumor stem cell assay, the degree of the association between the in vitro assay and clinical response as well as the likelihood of response. Methods considered to quantitate in vitro assay data included first-order kinetics of percent survival with drug concentration, minimal percent of tumor colony-forming unit survival at low drug concentrations, and area under the in vitro percent survival-drug concentration curve. Based upon experimental data, the percent tumor colony survival and the area under the curve (i.e., in vitro sensitivity indices) were concluded to better account than other methods for the commonly observed nonlinear shape of the in vitro curves. The two methods also yielded equivalent quantitative descriptions of the in vitro data. A logistic regression model was used for explicit quantitation of the relationship between the in vitro sensitivity index and predicted probability of clinical response. Very high association was observed between the predicted in vivo and actual clinical response for the cytotoxic drugs considered. Incorporation of other pharmacologic and patient prognostic factors into the quantitative methods is discussed and shown to improve their effectiveness.