The aim of this study was to investigate whether there is an association between the presence of neuroendocrine elements in relapsed prostate cancer and sensitivity to estramustine/etoposide and carboplatin or cisplatin. Thirty patients with progressive metastatic castrate prostate cancer were selected on the basis of clinical criteria for treatment with cytotoxic chemotherapy. The criteria included a tumor biopsy specimen taken during relapse showing neuroendocrine features based on morphology alone (carcinoid elements, small cell tumor) or by immunohistochemistry (detection of chromogranin A, neuron-specific enolase or synaptophysin). Patients were treated with cis- or carboplatin, estramustine (orally) and etoposide (orally or intravenously). Remission of radiographically visualized lesions, decline of prostate-specific antigen (PSA) or death owing to any cause constituted (separately reported) the endpoints. Tumor remission was found in about half of the patients, determined either by changes in measurable lesions or by a 50% decline in serum PSA. Neuroendocrine elements--irrespective of how they were identified--were not predictive of tumor remission or survival. Regression of measurable lesions by > 50% was seen in 4/9 (44%) cases of small cell carcinoma, 6/13 (46%) of poorly differentiated carcinoma, 7/13 (54%) of tumors with one marker immunohistochemically detected and 3/7 (43%) of tumors without any staining. It is concluded that response to chemotherapy was not predicted solely on the basis of the presence or absence of neuroendocrine elements in a relapsed tumor specimen. The results support the use of cytotoxic drugs in the relapsed setting and definitive trials are ongoing to prove any benefit to survival.