Purpose To assess the antitumor activity of vitespen (autologous, tumor- derived heat shock protein gp96 peptide complexes) by determining whether patients with stage IV melanoma treated with vitespen experienced longer overall survival than patients treated with physician's choice. Patients and Methods Patients (N = 322) were randomly assigned 2:1 to receive vitespen or physician's choice (PC) of a treatment containing one or more of the following: dacarbazine, temozolomide, interleukin-2, or complete tumor resection. This open-label trial was conducted at 71 centers worldwide. Patients were monitored for safety and overall survival. Results Therapy with vitespen is devoid of significant toxicity. Patients randomly assigned to the vitespen arm received variable number of injections (range, 0 to 87; median, 6) in part because of the autologous nature of vitespen therapy. Intention-to-treat analysis showed that overall survival in the vitespen arm is statistically indistinguishable from that in the PC arm. Exploratory landmark analyses show that patients in the M1a and M1b substages receiving a larger number of vitespen immunizations survived longer than those receiving fewer such treatments. Such difference was not detected for substage M1c patients. Conclusion These results are consistent with the immunologic mechanism of action of vitespen, indicating delayed onset of clinical activity after exposure to the vaccine. The results suggest patients with M1a and M1b disease who are able to receive 10 or more doses of vitespen as the candidate population for a confirmatory study.