Abstract Context Several statistical models were introduced for the prediction of age at menopause using a single measurement of anti-müllerian hormone (AMH); however, individual prediction is challenging and needs to be improved. Objective The objective of this study was to determine whether multiple AMH measurements can improve the prediction of age at menopause. Design All eligible reproductive-age women (n = 959) were selected from the Tehran Lipid and Glucose Study. The serum concentration of AMH was measured at the time of recruitment and twice after that at an average of 6-year intervals. An accelerated failure-time model with Weibull distribution was used to predict age at menopause, using a single AMH value vs a model that included the annual AMH decline rate. The adequacy of these models was assessed using C statistics. Results The median follow-up period was 14 years, and 529 women reached menopause. Adding the annual decline rate to the model that included single AMH improved the model’s discrimination adequacy from 70% (95% CI: 67% to 71%) to 78% (95% CI: 75% to 80%) in terms of C statistics. The median of differences between actual and predicted age at menopause for the first model was –0.48 years and decreased to –0.21 in the model that included the decline rate. The predicted age at menopause for women with the same amount of age-specific AMH but an annual AMH decline rate of 95 percentiles was about one decade lower than in those with a decline rate of 5 percentiles. Conclusion Prediction of age at menopause could be improved by multiple AMH measurements; it will be useful in identifying women at risk of early menopause.