Association between the number of live births and hazard of all-cause premature mortality
Table 2 shows the results of the multivariate Cox proportional hazard regression models between the number of live births and hazard of all-cause premature mortality. After adjustment for age, race, education, Townsend deprivation index, BMI, smoking statues, moderate drinking, moderrate activity, we found that the nulliparous women (HR0=1.17, 95% CI: 1.10-1.25) had higher risk of premature all-cause death, and the women with two, three and four childern had lower risk of premature all-cause death (HR2=0.86, 95% CI: 0.82-0.92, HR3=0.87, 95% CI: 0.82-0.93, HR4=0.91, 95% CI: 0.83-0.99), compared with the reference group. These results did not change appreciably after further adjustment for diabetes, hypertension, asthma, emphysema and chronic bronchitis, or female specific factors (use of oral contraceptive pills, history of stillbirth, spontaneous miscarriage or termination, age at menarche, age at menopause, and length of menstrual). The HR (95%CI) of all-cause premature mortality across groups were 1.17(1.09-1.25), 1.00(reference), 0.87(0.82-0.93), 0.88(0.82-0.94), 0.90(0.82-0.99) and 0.94(0.83-1.08), respectively (P-trend<0.001). The hazard of all-cause premature mortality was higher among childless women, comparied with those who had only one child, and women with two, three, or four children have lower all-cause premature mortality, and this is true across models 3 to 5.
The potential nonlinear association between all-cause premature mortality and the number of live births was intuitively described using restricted cubic spline (RCS) based on model5. As shown in Supplementary Figure 1, there was a non-linear relationship between the number of live births and all-cause premature mortality (P for non-linear<0.001), which proves the above conclusion. The nadir for all-cause premature death risk was estimated from RCS to be at two live births, which was significantly lower than one live birth. Also, the all-cause premature mortality of mothers with three or four children was lower than mothers who had one child. Moreover, women without child was signifcant increases in risk were observed for premature all-cause death than women with one child. Although the data for the women with five or more children was not significant, there is a clear upward trend in the RCS curve. More specifcally, we prefer to describe this nonlinear relationship with approximates U-shape where women with none, one, or four children have higher all-cause premature mortality than those with two or three children.
Sensitivity analyses did not affect the assessment of the relationship between parity and all-cause premature mortality risk (Supplementary Table 1). Similar results were observed after excluding participants with less than 1 and 2 years of follow-up, respectively. To avoid the possible influence of biochemical indicators on the results, we repeated our analyses by adjusting the covariate biochemical indicators additionally. The results were barely changed. The results of furthur adjusting for covariates dietary factors showed the conclusion remains alomost the same.