OBJECTIVETo analyze the value of ovarian reserve markers for predicting ovarian response in women undergoing in vitro fertilization-embryo transfer.

METHODSAccording to the ovarian response, 331 patients undergoing oocyte retrieval cycles were divided into of normal, poor, and high response groups. Serum anti-Mvllerian hormone (AMH) was determined using AMH ELISA kit on day 3 of the menstrual cycle, antral follicle count (AFC) was measured using vaginal ultrasound, and basal serum follicle-stimulating hormone (FSH), luteinizing hormone (LH) and estradiol (E(2)) levels were detected using chemiluminescence method.

RESULTSSerum AMH and FSH levels, FSH/LH ratio, AFC, and the patients age, but not the basal E(2) level (P>0.05), were correlated with the number of oocytes collected (×1000/ampules of Gn) (P<0.001). AFC and serum AMH were the strongest single predictors for low ovarian response, with the areas under curve (AUC) of 0.855 (0.787-0.924) and 0.832 (0.764-0.900) (P<0.05), and cutoff values of ≤9 and ≤1.88 ng/ml, respectively. AFC was the strongest single predictor for high ovarian response, with an AUC of 0.787 (0.728-0.847) and the cutoff value of ≥15. Logistic regression model found that the combination of AFC, serum AMH and FSH improved the predictive power for poor ovarian response, but not for high ovarian response.

CONCLUSIONAFC, serum AMH, FSH, FSH/LH, and age are all predictors of ovarian response, among which AFC is the strongest single predictor. A multivariable model can improve the predictive power for low ovarian response but not for high ovarian response.