Four hundred and eleven evaluable patients with suboptimal (greater than 1 cm residual) stage III and IV and recurrent ovarian cancer after surgical exploration and tumor debulking were prospectively randomized to receive cyclophosphamide, doxorubicin, and cisplatin (CAP) with or without bacillus Calmette-Guerin (BCG). Therapy was planned for eight courses with the BCG to be given by the sacrification technique on Days 8 and 15 of each course. The addition of BCG did not improve response rate, progression-free interval (PFI), or survival. In a multivariate analysis prognostic factors significantly favorable for survival include nonmeasurable disease and young age. Those patients having tumor with a mucinous histology had poorer survival and PFI than patients with tumors composed of other cell types.