Reproductive History and Age at Menarche in Relation to Reproductive Lifespan and Natural Menopause: A population-based analysis from NHANES 1999-March 2020

Abstract

Background: To evaluate the associations of reproductive history(gravidity, parity, pregnancy loss) and age at menarche (AAM) with age at natural menopause (ANM) and reproductive lifespan (RL = ANM - AAM), and to assess effect modification and nonlinearity using a design-based, multiple-imputation framework. Methods: We analyzed data from in NHANES 1999-March 2020 using stratified, clustered, survey-weighted methods with five imputations pooled by Rubins rules. We included women aged [&ge;]60 years with self-reported AAM, ANM, gravidity and parity, excluding surgical or induced menopause. Primary outcomes were RL and ANM. Main analyses used linear models under two strategies: independent effects and mutual adjustment. Nonlinearity was assessed with restricted cubic splines (RCS) via joint Wald tests and AIC; pregnancy loss (approximately 65% zeros) was additionally evaluated using two-part models. Prespecified effect modifiers were income, race/ethnicity, and smoking; exploratory modifiers underwent BH-FDR control. Results: The analytic sample included 3,167 participants. AAM showed a robust linear inverse association with RL (fully adjusted {beta} = -0.871 years per later menarche year; 95% CI -1.011 to -0.731; FDR p<0.001) and a borderline positive association with ANM ({beta} = 0.129; 95% CI -0.011 to 0.269; p=0.07). Overall, parity and pregnancy loss were not significantly associated with RL or ANM across adjustment tiers. Two-part models indicated no transition effect and no dose-response among pregnancy-loss >0. Smoking modified the parity[-&gt;] RL association (prespecified joint Wald p=0.026). Income displayed a nonlinear association with RL, with larger gains from low to middle income. Discussion: In nationally representative data using rigorous design-based and multiple-imputation analyses, AAM was a stable determinant of RL with an approximately linear inverse association. In contrast, independent effects of gravidity, parity, and pregnancy loss on ANM/RL were limited. Lifestyle factors(smoking) and socioeconomic context modified specific associations. Confirmation in longitudinal cohorts with biomarker endpoints is needed to establish causality.