PURPOSE: Clinical features characteristic of small-cell prostate carcinoma (SCPC), "anaplastic," often emerge during the progression of prostate cancer. We sought to determine the efficacy of platinum-based chemotherapy in patients meeting at least one of seven prospectively defined "anaplastic" clinical criteria, including exclusive visceral or predominantly lytic bone metastases, bulky tumor masses, low prostate-specific antigen levels relative to tumor burden, or short response to androgen deprivation therapy. EXPERIMENTAL DESIGN: A 120-patient phase II trial of first-line carboplatin and docetaxel (CD) and second-line etoposide and cisplatin (EP) was designed to provide reliable clinical response estimates under a Bayesian probability model with early stopping rules in place for futility and toxicity. RESULTS: Seventy-four of 113 (65.4%) and 24 of 71 (33.8%) were progression free after four cycles of CD and EP, respectively. Median overall survival (OS) was 16 months [95% confidence interval (CI), 13.6-19.0 months]. Of the seven "anaplastic" criteria, bulky tumor mass was significantly associated with poor outcome. Lactic acid dehydrogenase strongly predicted for OS and rapid progression. Serum carcinoembryonic antigen (CEA) concentration strongly predicted OS but not rapid progression. Neuroendocrine markers did not predict outcome or response to therapy. CONCLUSION: Our findings support the hypothesis that patients with "anaplastic" prostate cancer are a recognizable subset characterized by a high response rate of short duration to platinum-containing chemotherapies, similar to SCPC. Our results suggest that CEA is useful for selecting therapy in men with castration-resistant prostate cancer and consolidative therapies to bulky high-grade tumor masses should be considered in this patient population.