A multivariate analysis of prognostic factors including tumor response after endocrine therapy in advanced breast cancer patients.

Although clinicians want to know how tumor response affects the survival of advanced cancer patients, the direct comparison of response with survival is severely discouraged because of biases essentially involved. In 159 patients with advanced breast cancer treated with adreno-oophorectomy, we analyzed the survival of the patients using the Kaplan and Meier method with the landmark method (landmark time of 3 months after treatment). A multivariate analysis with the Cox proportional hazard model for survival with explanatory variables including response of patients at the landmark time (3 months after therapy) was performed in order to find probable prognostic factors. By the Kaplan and Meier curves with the landmark method, response category showed a significant difference in the survival of the patients (log rank test; p < 0.0001). In the Cox model in the patients of landmark time of 3 months after therapy, we found that response was the most powerful factor for survival out of 10 variables (p = 0.0001), and the dominant site of metastasis significantly and independently modified the survival length (p = 0.001). ER status was indifferent in the analysis; however, when the response category was not included, ER was shown to be most influential. We propose that a combination of the Cox model with the landmark method would be a rational and useful approach to estimate probable prognostic factors including treatment outcome variables such as response for the survival analysis in advanced cancer patients.