Comparing four ovarian reserve markers--associations with ovarian response and live births after assisted reproduction.

Abstract

We compared the ability of four different ovarian reserve tests (ORTs) to predict live births per started in vitro fertilization-intracytoplasmic sperm injection (IVF-ICSI) cycle, and poor and excessive response to controlled ovarian hyperstimulation. This was a cohort study in a private infertility center in collaboration with Uppsala University, comprising 1230 IVF-ICSI cycles in 892 consecutive women between April 2008 and June 2011. Anti-Müllerian hormone (AMH) levels, antral follicle counts (AFC), combinations of basal levels of follicle-stimulating hormone and luteinizing hormone, and menstrual cycle lengths were analyzed for correlation and treatment outcome prediction in age-adjusted statistical models. Stepwise multivariable generalized estimating equation analyses were carried out in a sub-group with complete data on all four ORTs (620 cycles in 443 women). Odds ratios and c-statistics were calculated in the largest available set of data for each significant variable. Primary outcomes were live birth rate per started cycle and poor and excessive ovarian response to controlled ovarian hyper-stimulation (defined by the ovarian sensitivity index). All ORTs correlated significantly with each other, with the strongest correlation between AFC and AMH (r = 0.71, p < 0.0001). Univariately, AMH and age equivalently predicted live birth (c-statistic 0.61), and together they provided a significantly better model (c-statistic 0.64). For prediction of poor and excessive response the best model included AMH, AFC and age (c-statistic 0.89). AMH improves the ability to estimate live birth rates after assisted reproduction compared with female age alone. AMH, AFC and age together constituted the best model for prediction of ovarian response.