Abstract Study question Which of AMH (measured with fully-automated assay) or AFC is the best predictive biomarker of ovarian response to exogenous gonadotropins in IVF/ICSI cycles? Summary answer In patients >35 years, AMH predicts the number of oocytes retrieved better than AFC and it has greater discriminating ability to detect suboptimal response. What is known already Individual variability in ovarian response to a starting dose of gonadotropins is a well-known aspect during controlled ovarian stimulation and many efforts have been made for obtaining the personalization of the treatment, identifying different biomarkers (age, basal FSH, BMI, AMH, AFC) that may predict the ovarian response. Among these biomarkers, AMH and AFC demonstrated the best performance in predicting ovarian response. Nevertheless, approximately one in five patients in clinical practice has a discordance between AMH and AFC; in these cases, the clinicians do not know which indicator should be chosen to individualize the protocol and the starting dose of gonadotropins. Study design, size, duration This prospective, observational, multivariate study compared fully-automated AMH with AFC in the prediction of ovarian response, defined as the number of oocytes retrieved and it involved 161 couples attending their first IVF/ICSI cycle between 2019 and 2022 (ClinicalTrials.gov NCT04168892). The patients underwent a GnRH antagonist protocol and received a HMG starting dose exclusively based on their age (150 IU if ≤ 35 years, 225 IU if > 35 years). Two physicians performed all the AFC determinations. Participants/materials, setting, methods The study was conducted at ANDROS Day Surgery Clinic, Palermo, Italy. Inclusion criteria were: female age 18-40 years, BMI 18-30 kg/m2, FSH ≤15 IU/l, normal menstrual cycles, normal uterine cavity, presence of both ovaries. Exclusion criteria were: PCOS, severe endometriosis, previous ovarian surgery, ovarian cysts, endocrinological diseases. Considering a power (1-β) of 90% and an alpha of 0.01, a total sample size of 160 couples was considered sufficient to verify the study hypothesis. Main results and the role of chance Three out of 161 patients were excluded after randomization. Eighty-eight out of 158 patients (55.7%) received 150 IU as starting dose (age=31.03±3.33; range 21-35 years); 70 patients (44.3%) received 225 IU (age=37.68±1.31, range 36-40). 9/158 cycles (5.6%) were suspended for hypo-response. The two groups were different in number of oocytes retrieved (p < 0.05), with higher number in ≤ 35 years group (M≤35=11.30±7.14, M>35=8.77±5.30). Univariate analyses revealed that the number of oocytes retrieved was correlated, in both groups, with AMH (ps < 0.01) and AFC (ps < 0.01), and in ≤ 35 years group with basal FSH (p = 0.03). No relationships were found with cause of infertility, BMI, smoking. Correlation coefficient (ρ) between AMH and AFC was 0.74 (p < 0.01) in ≤ 35 years group and 0.57 (p < 0.01) in > 35 years group. Multivariate Poisson regressions confirmed that AMH and AFC significantly predicted the number of oocytes retrieved in ≤ 35 years group (p < 0.01, Exp(B)=1.08; and p < 0.01, Exp(B)=1.03, respectively), while only AMH was a significant predictor in > 35 years group (p < 0.01, Exp(B)=1.18) (AFC: p = 0.34, Exp(B)=1.01). For prediction of suboptimal ovarian response (<8 oocytes retrieved), ROC analysis showed that AMH was not different from AFC in ≤ 35 years group (AUCAMH=0.811, AUCAFC=0.767; p = 0.331) but significantly better in > 35 years group (AUCAMH=0.858, AUCAFC=0.675; p < 0.001). Limitations, reasons for caution This is a single-center study; a multicenter trial could be useful for a further validation of the results. Moreover, many laboratories do not use automated assays for measuring AMH and this could make the results of this study hardly applicable in all contexts. Wider implications of the findings The results of this prospective trial highlight that, in patients >35 years, automatized AMH predicts the ovarian response better than AFC and it shows greater discriminating ability to detect suboptimal response. This could open new possibilities of individualizing treatment with AMH-based algorithms, especially in patients with the worst reproductive prognosis. Trial registration number NCT04168892
Read full abstract