Stratified multivariate analysis of prognostic markers in breast cancer: a preliminary report

Published results using prognostic markers in breast cancer have been very confusing to oncologists, surgeons, and pathologists alike. As a result, there is a wide variation in opinion among oncologists about the utility of these markers in clinical practice. This study was undertaken to determine the utility of stratified multivariate survival analysis in integrating the commonly used prognostic factors into a user-friendly prognostic scheme, and its implication for treatment decision making. 300 women with invasive ductal carcinoma of the breast who were followed-up for 28-112 months (median 72 months) were entered in the study. Patients with distant metastases, those with bilateral or multifocal tumors, and special types of carcinoma were excluded. Variables included in the stratified multivariate survival analysis were estrogen receptor (ER) status, tumor size, nodal status, histological grade, number of mitotic figures per ten high power fields (MF/10HPF), and type of initial therapy. Data was subjected to Kaplan-Meier survival analysis and the log rank test for statistical significance at different steps of the analysis. Cut-off values for ER that produced a significant difference in survival varied from 9 fmol/mg protein to as high as 76 fmol/mg protein in different patient groups, and MF/10HPF varied from 6 to 21. Patients were stratified into different groups that enabled better evaluation of treatment outcome. Patients could also be combined into three groups with significantly different survival rates (p < 0.0001). Stratified multivariate survival analysis show that prognostic markers a) are interdependent, and their cut-off values vary depending on other tumor characteristics, and b) if used in a systematic way, they can be used to guide treatment decisions.