One-third of patients with node-negative breast carcinoma develop a local recurrence that will decisively influence their outcome. We conducted this quantitative ultrastructural study to determine the value of nucleolar features in predicting recurrence of node-negative breast carcinomas. Forty-three node-negative infiltrating duct carcinomas were examined by means of electron microscopy. The number of cells with three or more nucleoli (N3), percentage of nucleoli contacting the nuclear membrane (NC), patient age, tumor size, grading, treatment, follow-up time, development of recurrence, and status were recorded. There were no differences in patient age, tumor grade, or modality of surgical treatment. Significant differences were found among patients with and without recurrence for N3 ( P=0.02), NC ( P=0.01), and tumor size ( P=0.01). For N3, the positive predictive value (PPV) was 64.3, and the negative predictive value (NPV) was 72.4 (sensitivity=52.9%; specificity=80.8%; P=0.04). For NC, the PPV was 62.5, and the NPV was 74.1 (sensitivity=58.8%; specificity=76.9%; P=0.02). For tumor size, the PPV was 66.7, and the NPV was 64.9 (sensitivity=23.5%; specificity=92.3%; P not significant). Using logistic regression analysis, independent predictive value was shown for N3, NC, and tumor size. The number of nucleoli per cell and the number of cells with nucleoli in contact with the nuclear membrane have independent predictive value for the development of recurrence in lymph node-negative infiltrating duct carcinoma of the breast. Although electron microscopy is a highly accurate method, improving the precision of light microscopic techniques would allow more universal assessment of these nucleolar features.