AUTHOR=Wang Xue , Jin Lei , Mao Yun-dong , Shi Juan-zi , Huang Rui , Jiang Yue-ning , Zhang Cui-lian , Liang Xiao-yan 

TITLE=Evaluation of Ovarian Reserve Tests and Age in the Prediction of Poor Ovarian Response to Controlled Ovarian Stimulation—A Real-World Data Analysis of 89,002 Patients

JOURNAL=Frontiers in Endocrinology

VOLUME=Volume 12 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2021.702061

DOI=10.3389/fendo.2021.702061

ISSN=1664-2392

ABSTRACT=Aims: We aimed to explore the value of ovarian reserve tests (ORTs) for predicting poor ovary response (POR) and if an age cut-off improved this forecasting to facilitate clinical decision-making for women undergoing in vitro fertilization (IVF).
Methods：A retrospective cohort study was conducted on poor ovary response(POR) patients using real world data from five large reproductive centers  in China. A total of 89,002 women with infertility undergoing their first traditional ovarian stimulation cycle for in vitro fertilization from January 2013 to December 2019 were included. The receiver operating characteristic (ROC) curve was performed to estimate the prediction value of POR by the following ORTs: anti-Mullerian hormone (AMH), antral follicle count (AFC) and basal FSH, as well as patient age.
Results: In this retrospective cohort, the frequency of POR in the first IVF cycle was 14.8% . Age, AFC, AMH, bFSH were used as predicting factors for POR, of which AMH and AFC were the best indicators when using a single factor for prediction (AUC 0.862, AUC 0.842). The predictive values of the multivariate model included age and AMH (AUC 0.865), age and AFC (AUC 0.850), age and all three ORTs (AUC 0.873). Compared with using a single factor alone, the combinations of ORTs and female age can increase the predictive value of POR. Adding age to single AMH model improved the prediction accuracy compared with AMH alone (AUC 0.865 vs. 0.862), but did not significantly. The AFC with age model significantly improved the prediction accuracy of the single AFC model (AUC 0.846 vs. 0.837). To reach 90% specificity for POR prediction, the cutoff point for age was 38 years old with a sensitivity of 40.7%, 5 for AFC with a sensitivity of 55.9%, and 1.18ng/ml for AMH with a sensitivity of 63.3%. 
Conclusion: AFC and AMH demonstrated a high accuracy when using ROC regression to predict POR. When testing is reliable, AMH can be used alone to forecast POR. When AFC is used as prediction parameter, age is suggested to be considered as well.  AFC≤ 5 and AMH≤ 1.18ng/ml should be recommended to predict POR more accurately in IVF/ICSI patients.