The aim of this study was to determine the value of basal FSH as a predictor of assisted reproduction outcome in women >or=35 years undergoing ovarian stimulation with gonadotrophin-releasing hormone (GnRH) antagonist. A retrospective clinical study was carried out on 83 infertile women, 35-45 years old, divided into three groups according to their day 3 FSH concentration (group A = FSH <or=10 mIU/ml, group B = FSH >10 and <15 mIU/ml, group C = FSH >15 mIU/ml). Patients underwent ovarian stimulation with a GnRH-antagonist protocol. Group A women had significantly higher basal inhibin B concentrations (P < 0.001), lower cancellation rate (P < 0.001), required a significantly lower dosage of recombinant FSH (P < 0.0001) and had significantly higher oestradiol concentration under stimulation compared with the other groups (P < 0.0001). Oocyte and embryo numbers were comparable in all groups, although groups B and C had more low quality embryos compared with group A. The number of metaphase II oocytes and embryos was related to patients' ovarian reserve markers only in group C. Pregnancy and delivery rates were 35 and 22.5% in group A, 22.2 and 16.6% in group B and 5 and 0% in group C. It is concluded that a basal FSH cut-off of 10 mIU/ml seems predictive of ovarian reserve, while basal FSH cut-off of 15 mIU/ml seems predictive of pregnancy potential and probably of oocyte quality.