OBJECTIVE The clinical records of the node-positive breast cancer patients treated at our department were reviewed, to evaluate if there is a correlation between the ratio of involved axillary lymph nodes and the overall and cause specific survival. METHODS From 1984 until July 2001, 2073 files from patients with an invasive breast carcinoma were submitted to retrospective analyses. In 810 cases, a node positive status was diagnosed. All pT-stages were included. The total number of dissected nodes (pNtot) and the number of involved nodes (pN+) were available for 741 patients. The ratio of nodal involvement (pN+%) was categorized into three groups, pN+%< or =10% (n = 212) between 11 and 50% (n = 346) and between 51 and 100% (n = 183). RESULTS The actuarial overall survival (OS) at 5 and 10 years was, respectively, 78.2 and 59.1%. Cause specific survival (CSS) rates were, respectively, 83.6 and 69.1%. In univariate analyses, age (P = 0.01), grade (P = 0.02), pT-stage (P < 0.0001), chemotherapy (P = 0.0002), the number of involved nodes < or =3 versus >3 (pN+) (P < 0.0001) and ratio pN+% (P < 0.0001) were associated significantly with overall survival. A multivariate analysis using the Cox proportional hazards model found that pN+% was the most significant prognostic factor; pN+lost significance when pN+% was taken into account. CONCLUSIONS The percentage of positive lymph nodes in an axillary lymph node dissection appears to be an important prognostic factor for survival. The nodes ratio improved on the absolute numbers of involved axillary lymph nodes for assessment of prognosis.