There is accumulating evidence that immunohistochemical staining for p53 can identify patients with endometrial carcinoma who have an adverse outcome, but the interpretation of existing data is complicated by differences between studies in the way that p53 immunohistochemistry results have been assessed. In this study, we sought to determine the appropriate cut-off level for stratification of patients with endometrial carcinoma into high- and low-risk groups, based on p53 immunohistochemical staining. A total of 200 cases of endometrial carcinoma treated by hysterectomy were retrieved from the archives of the Department of Pathology, Vancouver General Hospital, from the period 1983 to 1998. Follow-up information was available for all cases. Slides were reviewed and the diagnosis confirmed, tumors graded according to FIGO grading system, and tumor cell type assessed. A tissue microarray consisting of duplicate 0.6-mm cores of tumor was constructed and immunostained for p53. Immunoreactivity for p53 was scored by counting the number of positively stained tumor cell nuclei and expressing this as a percentage of the total number of tumor cell nuclei counted (p53 index). Kaplan-Meier survival curves were constructed and compared by calculation of log-rank statistic, and multivariate analysis was performed by Cox regression modeling. The distribution of p53 index results was bimodal, with most cases having a very low or very high p53 index. The peaks of the bimodal distribution were clearly separated using a p53 index of > or =50%. Immunoreactivity was a significant adverse prognostic indicator of disease-specific survival (p<0.0001 by univariate analysis). Patients with strongly p53 immunoreactive tumors (p53 index >or =50%) had a significantly worse outcome than patients with weakly immunoreactive (p53 index > or =5% and <50%) or p53-negative (p53 index <5%) tumors (p = 0.0001). There was no significant difference between the outcomes for patients in the latter two groups. By multivariate analysis, p53 overexpression was a significant prognostic indicator independent of patient age and tumor stage (p = 0.008) but was not independent when the analysis was extended to include FIGO grade and tumor cell type. p53 immunostaining was of prognostic significance in the subset of patients with endometrioid carcinomas (p = 0.02), but not in patients with clear cell or papillary serous carcinomas. Using a p53 index of > or =50% as a cut-off between positive and negative p53 staining, immunohistochemical staining for p53 is a prognostic indicator in patients with endometrial carcinoma of endometrioid type. p53 immunostaining was not found to be of prognostic significance independent of tumor cell type and grade.