Pregnancy loss and prevalent cardiometabolic multimorbidity: a two-stage individual participant data analysis of CHARLS, ELSA, and HRS

Abstract

Cardiovascular disease remains the leading cause of death worldwide, and cardiometabolic multimorbidity (CMM) is associated with substantially higher mortality and reduced life expectancy. Pregnancy loss is common and has been linked to later cardiovascular risk, but multinational evidence using harmonized individual participant data and CMM as the outcome remains limited. To examine the cross-sectional association between self-reported pregnancy loss and prevalent cardiometabolic multimorbidity (CMM) among middle-aged and older women in three nationally representative ageing cohorts, and to evaluate the consistency of findings across countries. Cross-sectional analysis using the last valid observation for each woman as the individual-level main analysis, with person-wave analyses as sensitivity analyses. The study did not model incident CMM or CMM progression. Country-specific estimates were combined using a two-stage individual participant data approach. The China Health and Retirement Longitudinal Study (CHARLS, 2011–2018), the English Longitudinal Study of Ageing (ELSA, 2002–2018), and the Health and Retirement Study (HRS, 2017/2019 Life History Mail Survey). Women aged 45 years or older with available pregnancy history and cardiometabolic outcome data (CHARLS: n = 7,514; ELSA: n = 3,773; HRS: n = 1,630; total N = 12,917). Self-reported history of pregnancy loss. In CHARLS and HRS, pregnancy loss corresponded to miscarriage or stillbirth; in ELSA, the available variable combined miscarriage, stillbirth, and induced abortion and could not be disaggregated. CMM, defined as the coexistence of at least two of five self-reported physician-diagnosed conditions: hypertension, diabetes, dyslipidemia, heart disease, and stroke. The prevalence of pregnancy loss was 25.2% in CHARLS, 24.9% in ELSA, and 35.3% in HRS; the individual-level prevalence of CMM was 30.1%, 46.6%, and 36.9%, respectively. In the fully adjusted individual-level model (M3), pregnancy loss was associated with higher odds of prevalent CMM in CHARLS (OR = 1.45, 95% CI: 1.20–1.75, p < 0.001), but not in ELSA (OR = 1.08, 95% CI: 0.90–1.29, p = 0.424) or HRS (OR = 1.24, 95% CI: 0.71–2.17, p = 0.455). The DerSimonian-Laird random-effects pooled OR was 1.24 (95% CI: 0.99–1.56, p = 0.062; HKSJ 95% CI: 0.82–1.89). In the person-wave M3 sensitivity analysis, the pooled OR was 1.15 (95% CI: 0.98–1.34, p = 0.083; HKSJ 95% CI: 0.77–1.72). In parity-adjusted person-wave analyses restricted to ELSA and HRS, the pooled OR was 1.02 (95% CI: 0.81–1.29, p = 0.849). The individual-level main analysis suggested that the association between pregnancy loss and prevalent CMM was primarily observed in CHARLS, whereas no statistically significant association was observed in ELSA or HRS. The three-country pooled estimate did not reach statistical significance and was attenuated after parity adjustment in ELSA and HRS. Given the cross-sectional design, the heterogeneous exposure definition in ELSA, sample reduction in HRS, and sensitivity of estimates to analytic choices, the findings should be interpreted as exploratory evidence rather than evidence of a temporal or causal association.