Fahimeh Ramezani Tehrani1, Mohammad Ali Mansournia, Masoud Solaymani-Dodaran, Ewout Steyerberg, Fereidoun Azizi. 1. 1Reproductive Endocrinology Research Center 2Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran 3Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran 4Minimally Invasive Surgery Research Center, Iran University of Medical Sciences, Tehran, Iran 5Division of Epidemiology and Public Health, University of Nottingham, Nottingham, UK 6Center for Medical Decision Making and Department of Public Health, Erasmus MC, Rotterdam, the Netherlands.
Abstract
OBJECTIVE: This study aimed to improve existing prediction models for age at menopause. METHODS: We identified all reproductive aged women with regular menstrual cycles who met our eligibility criteria (n = 1,015) in the Tehran Lipid and Glucose Study-an ongoing population-based cohort study initiated in 1998. Participants were examined every 3 years and their reproductive histories were recorded. Blood levels of antimüllerian hormone (AMH) were measured at the time of recruitment. Age at menopause was estimated based on serum concentrations of AMH using flexible parametric survival models. The optimum model was selected according to Akaike Information Criteria and the realness of the range of predicted median menopause age. RESULTS: We followed study participants for a median of 9.8 years during which 277 women reached menopause and found that a spline-based proportional odds model including age-specific AMH percentiles as the covariate performed well in terms of statistical criteria and provided the most clinically relevant and realistic predictions. The range of predicted median age at menopause for this model was 47.1 to 55.9 years. For those who reached menopause, the median of the absolute mean difference between actual and predicted age at menopause was 1.9 years (interquartile range 2.9). CONCLUSIONS: The model including the age-specific AMH percentiles as the covariate and using proportional odds as its covariate metrics meets all the statistical criteria for the best model and provides the most clinically relevant and realistic predictions for age at menopause for reproductive-aged women.
OBJECTIVE: This study aimed to improve existing prediction models for age at menopause. METHODS: We identified all reproductive aged women with regular menstrual cycles who met our eligibility criteria (n = 1,015) in the Tehran Lipid and Glucose Study-an ongoing population-based cohort study initiated in 1998. Participants were examined every 3 years and their reproductive histories were recorded. Blood levels of antimüllerian hormone (AMH) were measured at the time of recruitment. Age at menopause was estimated based on serum concentrations of AMH using flexible parametric survival models. The optimum model was selected according to Akaike Information Criteria and the realness of the range of predicted median menopause age. RESULTS: We followed study participants for a median of 9.8 years during which 277 women reached menopause and found that a spline-based proportional odds model including age-specific AMH percentiles as the covariate performed well in terms of statistical criteria and provided the most clinically relevant and realistic predictions. The range of predicted median age at menopause for this model was 47.1 to 55.9 years. For those who reached menopause, the median of the absolute mean difference between actual and predicted age at menopause was 1.9 years (interquartile range 2.9). CONCLUSIONS: The model including the age-specific AMH percentiles as the covariate and using proportional odds as its covariate metrics meets all the statistical criteria for the best model and provides the most clinically relevant and realistic predictions for age at menopause for reproductive-aged women.
Authors: A C de Kat; Y T van der Schouw; M J C Eijkemans; G C Herber-Gast; J A Visser; W M M Verschuren; F J M Broekmans Journal: BMC Med Date: 2016-10-03 Impact factor: 8.775