To investigate the contribution of serum levels of testosterone (TS) and sex hormone binding globulin (SHBG) in association with body mass index (BMI) as a surrogate marker of obesity, to the predictive capability of tumor size (T), lymph node (N) and estrogen receptor (ER) status and proliferative activity (TLI). We investigated 120 women with primary breast cancer and median follow-up of 138 months. Serum levels of TS and SHBG and patient's BMI were evaluated before surgery. The contribution of TS, SHBG, their ratio (TS/SHBG) and BMI to the predictive capability of tumor-specific biomarkers was investigated by Harrell's c statistic. TS alone did not affect prognosis, whereas SHBG was protective in postmenopausal patients, in which BMI was associated with a progressive increase in the relapse-specific hazard ratio (HR). When in combination, TS, SHBG and BMI, affected prognosis in different ways depending on menopausal status. The best predictive capability (c = 0.78) was observed in postmenopausal patients when at the basic model (N + TLI) were added TS, BMI, TS * BMI interaction, with or without SHBG. In premenopause subgroup, the best predictive capability (c = 0.67) was provided by the basic model (N + TLI) plus TS and SHBG or their ratio, BMI and TS * BMI or TS/SHBG * BMI interaction. Patient-associated features such as BMI and serum levels of TS and SHBG can improve the predictive capability of consolidate tumor-specific biomarkers in both pre- and postmenopause, thus providing a relevant contribution to the decision-making process.