The authors assayed 209 stage I and II mammary carcinomas for the estrogen receptor (ER) with an immunocytochemical assay (ICA) and quantitated the nuclear stain with the SAMBA 4000 Cell Image Analysis System (Imaging Products International, Inc., Chantilly, VA). The cases had been followed for 54-214 months (average, 64 months). The results were correlated with the patients' overall and disease-free survival times. Three of the parameters obtained from the quantitative analysis were evaluated: labeling index, mean optical density, and quick score. Statistical analysis was performed with the Kaplan-Meier product limit estimator for quantitated values and Cox regression for risk of mortality and disease progression. Mean optical density of the nuclei at a cut-off value of 10 produced the strongest association between quantitative ERICA and overall and disease-free survival times in discriminating high- and low-risk groups (P = .016 and P = .018, respectively). When the mean optical density results were categorized into ranges of values of potential biologic significance according to overall survival times, patients with mean optical density greater than 15 were found to be at lowest risk; patients with values between 5 and 15 had a statistically significant intermediate risk; and the group with values less than 5 had the worst outcome (P = .018). The probability of disease-free survival also was statistically significant (P = .038) among the three groups with cut-off points of less than 10, 10-35, and greater than 35. These results showed that quantified optical density has better discriminating power for risk prediction than that obtained by the biochemical assay for ER at a cut-off value of 10 or 20 fmol/mg. Estrogen receptor ICA has been adequately shown to be as or more accurate than ligand-binding assays. Our results corroborate those studies and support the utility of ICA assays for ER by showing that quantitation performed by image analysis is an objective and reproducible method yielding clinical prognostic information of higher reliability than that given by the dextran-coated charcoal ER assay.