Statistical analysis
The baseline characteristics in terms of demographics, lifestyle, health conditions and female specific factors were presented as mean (standard deviation, SD) and numbers (percentage). General linear models and a chi-square test was used to compare the differences for baseline characteristics by each parity. Also, we used general linear model to calculate the estimated values of FI, ΔKDM-biological age and HD corrected for multiple covariates for each pairty. A Cox proportional hazards model was performed to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for all-cause premature mortality using the “1 production” group as a reference, with the follow-up time as the time scale. The dose-response relationship was fexibly modeled by the restricted cubic spline (RCS) to explore the potential nonlinear correlation between parity and the hazard of all-cause premature mortality.
We performed stratified analyses by following factors: Born year (<1980 or ≥1980), race/ethnicity (White, Mixed, Asian or Asian British, Black or Black British, Chinese or Other ethnic group), Townsend deprivation index (<median or ≥median), BMI (<25, 25–29.9, or ≥30 kg/m2 ), smoking (never, ever, current), Alcohol intake frequency (<once/mouth,≥once/mouth), hypertension (no or yes), diabetes (no or yes), hypertension (no or yes), asthma (no or yes), emphysema and chronic bronchitis (no or yes). To evaluate interactions between the number of live births and these factors, multiplicative interaction was assessed by adding interaction terms to the Cox models.
Three sensitivity analyses were performed. The first analysis excluded the participants whoes follow-up duration was less than 1 or 2 years, to check if the severe illness would affect the results. The second analysis evaluated the participants with additional adjusting for covariates dietary factors (including fresh fruit intake, dried fruit intake, oily fish intake, salt added to food, cereal intake, processed meat intake,mineral and other dietary supplements), to examinte whether dietary factorshad had effect on the relationship. The third analysis further adjusted covariate biochemical indicators (including albumin, triglyceride, glucose, LDL, cholesterol, total bilirubin) for participants, to remove the impact of certain biochemical indicators unrelated to production on outcomes.
All statistical analyses were conducted by R 4.2.2, and p-values < 0.05 were considered statistically significant.