Factors associated with axillary lymph node metastasis from breast carcinoma - Descriptive and predictive analyses

BACKGROUND. Although axillary lymph node metastasis is one of the most impo rtant prognostic determinants of breast carcinoma prognoses, the reasons wh y tumors vary in their capability to produce for axillary metastases remain unclear. METHODS. The authors used data from the nationwide Patient Care Evaluation (PCE) survey of the American College of Surgeons to evaluate the correlatio ns between patient/tumor characteristics and lymph node status, and to expl ore the use of these factors, which are all known prior to axillary dissect ion, in predicting lymph node status. The PCE data set contained 18,025 bre ast carcinoma cases diagnosed in 1990 after exclusion of women older than 7 9 years or with fewer than 6 lymph nodes examined. RESULTS. In a multivariate logistic regression model, larger tumor size, yo ung age, African American or Hispanic race, outer half tumor location, poor or moderate differentiation, aneuploidy, and infiltrating ductal histology were independently associated with a higher likelihood of one or more posi tive lymph nodes. Contrary to expectation, cases negative for estrogen rece ptor (ER) and progesterone receptor (PR) had a lower risk of positive lymph nodes when adjusted for other factors (odds ratio = 0.82; 95% confidence i nterval: 0.74-0.91) compared with cases positive for both receptors. This m odel accurately predicted lymph node status in 2 validation data sets (a 50 % random sample of 1990 PCE data and 1992 data from the National Cancer Dat a Base), but was less accurate in a third, older data set (1983 PCE data). However, the percentage of cases (1990 validation set) with predicted proba bilities less than 0.05 or greater than 0.95 were only 4.6% and <0.1%, resp ectively. CONCLUSIONS. The authors concluded that 1) most variation in axillary lymph node metastatic status can be explained by routinely available data, 2) ER and PR status may be involved in the mechanism of this behavior, and 3) th e difficulty of using prediction models to avert axillary dissection should not be underestimated. (C) 1999 American Cancer Society.