Results
Cross-sectional Data at Study Entry
The following baseline variables were cross-tabulated with the categories of parity: age, cigarette smoking, any alcohol intake, body mass index (BMI), serum lipids and lipoproteins, diabetes, hypertension, peak expiratory flow, prior CHD or stroke, atrial fibrillation, depression score, physical activities of daily living and self-rated health. Only age, hypertension, diabetes and BMI showed possible associations with parity and the data are presented in Table 1. With the exception of women with 6+ children, the mean age was lower in those at higher parity. Average BMI rose with increasing parity, as did the prevalence of diabetes and hypertension in women with 4+ children.
Prediction of BMI, diabetes and hypertension by parity was explored in separate multiple logistic models controlling for age and other variables and the findings are presented in Table 1. The odds ratios only reached significance for hypertension in women with 6+ children.
All-cause Mortality in Women
All-cause mortality rate declined with increasing parity (Table 2). It was highest in nulliparous women, progressively falling until women had 3+children. All-cause mortality findings in proportional hazards models in relation to parity are presented in Table 2. In a model controlling only for age, hazard ratios were lower at high degrees of parity but did not reach statistical significance. In a model which then included other significant predictors of mortality, but excluded hypertension, diabetes and BMI (the 'confounder model'), the hazard ratios for all-cause mortality fell progressively with increasing parity beyond 1 child, reaching statistical significance in those with 6+ children, where the risk of death was 40% lower than in nulliparous women (test of trend for parity P < 0.004).
The prediction of all-cause mortality by parity was not materially changed if the 'confounder model' now included hypertension or diabetes or BMI (separate models in Table 2). A final model incorporating all variables showed little change from the 'confounder model', even with the inclusion of hypertension, diabetes and BMI (Table 2). Test of trend for parity in this final model was highly significant (P < 0.002).
The relationship of parity to all-cause mortality in women is demonstrated in the hazard curves in Figure 1. These curves demonstrate a broad hierarchy of reducing mortality with increasing parity in the final multivariate model which controls for the contribution of other predictors and potential confounders.
(Enlarge Image)
Figure 1.
Hazard curves for all-causes mortality by parity in the full proportional hazards model.
The final proportional hazards model included the following variables in addition to parity (hazard ratio and 95% CI): age, 1.10 (1.08–1.11); BMI, 0.99 (0.97–1.00); any alcohol intake, 0.79 (0.67–0.93); current cigarette smoking, 1.76 (1.36–2.27); low peak expiratory flow, 2.10 (1.66–2.67); increased physical disability, 1.65 (1.32–2.08); poor self-rated health, 1.26 (1.01–1.56); diabetes, 2.00 (1.54–2.59); hypertension, 1.49 (1.22–1.81) and atrial fibrillation, 2.23 (1.46–3.40). Additional variables listed above in the cross-sectional analysis did not approach statistical significance and were excluded from all-cause mortality models.
Cause-specific Mortality in Women
Cause-specific mortality rates are presented in Table 3 for the predominant groupings. The pattern of decline in all-cause mortality with parity was accompanied by parallel reductions in deaths from cancer and respiratory conditions and 'other' causes of death. In multivariate models employing the final list used in Table 2, this was statistically significant only in a consistent manner for deaths from other causes. CHD mortality was generally increased in all parous women, statistically so in many groups, who had an increase in the range 60–111% compared with nulliparous women.
Mortality in Men
Mortality and 'parity' were similarly evaluated in men in the full proportional hazards model. There were 704 deaths (57% of all men) and there was modest evidence of reduced all-cause mortality with increasing number of children. But this finding did not approach statistical significance (hazard ratio and 95% CI): childless, 1.00; 1 child, 0.88 (0.64–1.21); 2 children, 0.92 (0.70–1.20); 3 children, 0.90 (0.69–1.17); 4 children, 0.95 (0.72–1.27); 5 children, 0.87 (0.61–1.24); 6+ children, 0.83 (0.60–1.17) (test of trend for parity not significant P < 0.5). There were 222 CHD deaths (18% of all men), but there was no significant association with number of children, certainly no suggestion of the excess CHD mortality seen in women (data not shown).