Administration of chemotherapy to premenopausal women shortens their reproductive lifespan by depleting nonrenewable oocytes. Preservation of fertility is a priority for many such women, and identification of women at risk of infertility is therefore important. However, age is the only patient characteristic currently recognized to be predictive of long-term ovarian function after chemotherapy. Our objective was to assess markers of ovarian reserve and age as long-term predictors of ovarian function after chemotherapy. We conducted a prospective, longitudinal study at a university hospital and research institute. Patients included women who were premenopausal at the time of diagnosis of early breast cancer. Ovarian function was assessed at 5 yr follow-up in relation to pretreatment hormonal and ultrasound markers of ovarian reserve. Forty-two women received (neo-)adjuvant chemotherapy. Continuing menses 4-5 yr after diagnosis closely reflected ovarian activity as assessed by a range of serum markers, including estradiol, inhibin B, FSH, and anti-müllerian hormone (AMH). Pretreatment serum AMH, FSH, antral follicle count, and age predicted late ovarian activity by univariate analysis. However, only AMH was predictive in a multivariate logistic regression (odds ratio = 13.0; 95% confidence interval = 2.5-66.7); 0.71 ng/ml gave peak likelihood ratio of 7.0 with 54% sensitivity and 92% specificity. Bone mineral density fell over the 4-5 yr after diagnosis with greater loss in women with lower ovarian activity. Higher pretreatment AMH was associated with lower bone mineral density at both lumbar spine and hip at 5 yr (P < 0.02). Measurement of AMH at cancer diagnosis predicts long-term ovarian function after chemotherapy. Use of this in clinical practice may allow better prediction of chemotherapy-related risk to future fertility.