Efficacy of three COS protocols and predictability of AMH and AFC in women with discordant ovarian reserve markers: a retrospective study on 19,239 patients

BACKGROUND: Recent studies have consistently shown that AFC and serum AMH are good predictors of ovarian response and have shown strong correlations. However, it is not unusual for reproductive medicine specialists to encounter discordance between them. This is the first study to investigate the efficacies of the different COS protocols when the AFC and AMH levels are discordant. Based on the association between COS protocols and pregnancy outcomes, we attempt to explain the controversial results and clarify the predictive value of AMH and AFC in this context. METHODS: 19,239 patients undergoing their first fresh in vitro fertilization (IVF)/intracytoplasmic sperm injection (ICSI) cycles with GnRH antagonist protocols, GnRH-a long protocols or GnRH-a ultra-long protocols between January 1, 2016, and December 31, 2019, were enrolled and then divided into four groups in accordance with the boundaries for the AFC and serum AMH level provided by the Poseidon Classification. Our study was divided into two parts. Firstly, we retrospectively compared the effects of the three COS protocols in patients with discordant AMH and AFC. Multivariate logistic regression models were conducted in a forward manner to exclude the influence of confounding factors. Afterward, to increase comparability between Group 2 (low AMH and normal AFC) and Group 3 (normal AMH and low AFC), propensity score matching (PSM) analysis was performed based on age, BMI, the number of embryos transferred, and COS protocol. IVF intermediate and reproductive outcomes were compared between Group 2 and Group 3. RESULTS: For people with low AMH and normal AFC (Group 2), the number of total oocytes, clinical pregnancy rate (CPR), live birth rate (LBR) and cumulative live birth rate (CLBR) were significantly higher in GnRH-a ultra-long protocol compared with GnRH antagonist protocol. In multivariate logistic regression models, significant associations of COS protocol with fresh LBR and CPR were found after adjusting for age, BMI, AFC, AMH and the number of embryos transferred. Whereas, in patients with normal AMH and low AFC (Group 3), the number of total oocytes, CLBR, LBR and CPR were highest in the long GnRH-a protocol although there was no statistically significant difference. After PSM, the results showed that although oocytes yield and available embryos in patients with normal AMH and low AFC were significantly higher, there was no significant difference in reproductive outcomes between Group 2 and Group 3. CONCLUSIONS: We found that women with normal AFC and low AMH may benefit from the GnRH-a ultra-long protocol. Nevertheless, for women with normal AMH and low AFC, the long GnRH-a protocol seems to be associated with better clinical outcomes. Furthermore, after eliminating the confounding factors including the COS protocol, we found that AMH can only predict the number of oocytes but not the quality of oocytes when there was discordance between AFC and AMH. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s13048-021-00863-4.