Age at treatment and the total cumulative dose of cyclophosphamide are established predictors of chemotherapy-induced ovarian failure in premenopausal women with breast cancer. In a prospective trial of 49 women receiving adjuvant chemotherapy, we sought to identify whether other factors including family history of osteoporosis, lifestyle factors, reproductive hormones, bone turnover markers, or bone mineral density (BMD) of the hip or total spine measured before (baseline) and 6 months after chemotherapy increased the risk of ovarian failure. Univariate logistic regression analyses were used to identify the variables related to the risk of ovarian failure, and a multivariate logistic regression model was used to find the best of predictors of ovarian failure among the variables. In the multivariate model a higher total spine BMD at baseline (OR=6.0, p=0.02, 95% CI 1.4-27.3), and a 10% change in estradiol (E2) from baseline to six months (OR=0.80, p=0.02, 95% CI 0.67-0.97) were predictive of ovarian failure adjusted for age. In particular, as total spine BMD at baseline increases by 0.1 g/cm2 the odds of developing chemotherapy-induced ovarian failure increases by 6-fold. The multivariate model provides a good fit to the data (GOF p-value=0.8852), and provides excellent discrimination between those women who will and will not develop ovarian failure (Area under ROC=0.9487). If confirmed in larger data sets, using baseline spinal BMD to identify women who are at higher risk of developing ovarian failure independent of their age would be of value.