Does anti-Müllerian hormone predict menopause in the general population? Results of a prospective ongoing cohort study.

STUDY QUESTION: Do ovarian reserve tests (ORTs) predict age at natural menopause (ANM) in a cohort of healthy women with a regular menstrual cycle? SUMMARY ANSWER: Of the ORTs researched, anti-Müllerian hormone (AMH) alone predicts age at menopause. However, its predictive value decreased with increasing age of the woman, prediction intervals were broad and extreme ages at menopause could not be predicted. WHAT IS KNOWN ALREADY: A fixed interval is hypothesized to exist between ANM and age at loss of natural fertility. Therefore, if it is possible to predict ANM, one could identify women destined for early menopause and thus at higher risk for age-related subfertility. Of ORTs researched in the prediction of ANM, AMH is the most promising one. STUDY DESIGN, STUDY SIZE AND DURATION: A long-term, extended follow-up study was conducted, results of the first follow-up round were previously published. Two hundred and sixty-five normo-ovulatory women (21-46 years) were included between 1992 and 2001, 49 women (18.5%) could not be reached in the current follow-up round. PARTICIPANTS, SETTING,
METHODS: Two hundred and sixty-five healthy normo-ovulatory women were included, recruited in an Academic hospital. We measured baseline AMH, follicle-stimulating hormone and the antral follicle count (AFC). At follow-up (2009 and 2013), menopausal status was determined via questionnaires. Cox regression analysis calculated time to menopause (TTM) using age and ORT. A check of (non-) proportionality of the predictive effect of AMH was performed. A Weibull survival model was used in order to predict individual ANM. MAIN RESULTS AND THE ROLE OF CHANCE: In total, 155 women were available for analyses. Eighty-one women (37.5%) had become post-menopausal during follow-up. Univariable Cox regression analysis demonstrated age and ORTs to be significantly correlated with TTM. Multivariable Cox regression analysis, adjusting for baseline age and smoking; however, demonstrated AMH alone to be an independent predictor of TTM (Hazard Ratio 0.70, 95% Confidence Interval 0.56-0.86, P-value <0.001). A (non-)proportionality analysis of AMH over time demonstrated AMH's predictive effect to decline over time. LIMITATIONS, REASON FOR CAUTION: The observed predictive effect of AMH became less strong with increasing age of the woman. Individual AMH-based age at menopause predictions did not cover the full range of menopausal ages, but did reduce the variation around the predicted ANM from 20 to 10.1 years. WIDER IMPLICATIONS OF THE
FINDINGS: Age-specific AMH levels are predictive for ANM. Unlike in our previous publication however, a declining AMH effect with increasing age was observed. This declining AMH effect is in line with recent long-term follow-up data published by others. Moreover, the accompanying predictive inaccuracy observed in individual age at menopause predictions based on AMH, makes this marker currently unsuitable for use in clinical practice. STUDY FUNDING/COMPETING INTERESTS: No external funds were used for this study. M.D., M.J.C.E, S.L.B., G.J.S. and I.A.J.R. have nothing to declare. J.S.E.L. has received fees and grant support from the following companies (in alphabetical order): Ferring, Merck-Serono, MSD, Organon, Serono and Schering Plough. F.J.M.B. receives monetary compensation: member of the external advisory board for Merck Serono, the Netherlands; consultancy work for Gedeon Richter, Belgium; educational activities for Ferring BV, the Netherlands; strategic cooperation with Roche on automated AMH assay development.