The prognostic values of p53 and of its downstream mediator p21WAF1/Cip1 in patients receiving adjuvant chemotherapy for epithelial ovarian cancer have not been clearly established. Tumor extracts from a series of 120 patients treated postsurgically with cisplatin or carboplatin alone or together with other chemotherapeutics for primary ovarian carcinoma were assayed both for p53 protein by an immunofluorometric assay developed by us and for p21 protein by a commercially available immunoassay. Relative risks (RRs) for cancer relapse and death after 24 months of follow-up were determined by multivariate Cox regression analysis. Disease-free (DFS) and overall survival (OS) probabilities were also examined by the Kaplan-Meier method and log-rank tests. All other procedures were similarly nonparametric and based on two-sided tests of significance. Concentrations of p53 were elevated in patients with advanced stage disease (P = 0.02) or poorly differentiated (P = 0.03), suboptimally debulked tumors (P = 0.02), as well as in patients who failed to respond to chemotherapy (P = 0.03), as assessed by computed tomography scanning, serum CA125 determination, and second-look laparotomy. Statistically significant associations between concentrations of p53 and p21 were not found, nor were relationships demonstrated between concentrations of p21 and other clinicopathological variables or treatment response. Univariate analysis showed that p53 concentrations above the median indicated significantly higher risks for relapse (P = 0.04) and death (P < 0.01) and showed trends for increasing risks for relapse (P = 0.04) and death (P < 0.01) when p53 was considered as a four-level categorical variable. Multivariate analyses adjusted for age, stage, grade, and residual tumor size confirmed these observations (RR = 1.50; P = 0.05 for DFS and RR = 1.92; P = 0.03 for OS) for median-dichotomized p53, but the trends were of borderline significance (P = 0.09 for DFS and P = 0.07 for OS). In contrast, p21 positivity was not a significant predictor of favorable outcome in univariate survival analysis, and use of a three-level variable combining positivity or negativity status for both p53 and p21 did not yield greater separation of patients into risk groups (P = 0.07 for DFS and P = 0.06 for OS) than the use of p53 alone. Assessment of p53 expression may be an independent indicator of poor prognosis in ovarian cancer patients treated with adjuvant chemotherapy. The prognostic value of p21 expression, however, could not be demonstrated in our series of ovarian cancer patients.