Context: Anti-Mullerian hormone (AMH) has already been used for prediction of age at menopause with promising results.
Objective: We aimed to improve our previous prediction of age at menopause in a population-based cohort by including all eligible subjects and additional follow-up time.
Design and setting: All reproductive-aged women who met our eligibility criteria were selected from the Tehran Lipid and Glucose Study. The serum concentration of AMH was measured at the time of recruitment, and participant's date of menopause was recorded over a 10-year follow-up.
Subjects: A total of 1015 women, aged 20 to 50 years, with regular and predictable menstrual cycles at the initiation of the study were recruited.
Main outcome measure: The actual ages at menopause were compared with the predicted ones obtained from accelerated failure time model.
Results: We observed 277 occurrences of menopause. Median menopausal age was 50 years (range 30.1-58.2 years). The median (SD) of differences between the actual menopausal age and those predicted by our model was 0.5 (2.5) years. Model adequacy (measured by C-statistics) for correct prediction of age at menopause was 92%. The estimated ages at menopause and their 95% confidence intervals for a range of values of AMH and age were calculated and summarized in a table.
Conclusions: Using a model built on age and AMH, we can predict age at menopause many years earlier. This could provide opportunities for interventions in those who are at risk of early or late menopause.