Document Type



Various studies have investigated potential predictors of age at natural menopause but have produced inconsistent results. The relationship between age at natural menopause and socioeconomic, reproductive, and health behavioral factors was evaluated using longitudinal data from 5961 Australian female twins, aged 17 to 88 years at the time of study. The sample consisted of women voluntarily enrolled in the Australian Twin Registry. Failure-time analysis was the principal statistical method used to handle censored observations. Kaplan-Meier estimates showed the overall median age at natural menopause to be 51 years (95% confidence interval, 50-51). Median age at menopause was earlier for women with earlier birth year, women with late age of menarche, women who had no children, or women who were smokers. Differences in age at menopause between social, occupational, and educational groups were statistically significant (Mantel-Cox test, p < 0.001) for education, major occupational classification, combined income, and selfrated social class, with higher age at menopause for higher levels of each variable. A Cox proportional hazards model was used to estimate the odds ratio of the occurrence of natural menopause among different subgroups, adjusted to reflect simultaneous effects of all other significant covariates. This large study provided clear trends of association in predictors relating to age at menopause. These trends may help to resolve uncertainties and conflicting results identified in studies of comparable white samples. The nature of the twin data also sets a solid background for future analyses of genetic and environmental variance components using statistical modeling or related methods.