During the 16 years of follow-up from 1988 through 2004, a total of 2,771 cases of incident CHD were confirmed among the 42,351 men and 2,359 cases were confirmed among the 76,703 women. The distributions of selected CHD risk factors by BMI category at baseline are shown in .
Baseline characteristics, by category of body mass index in 1986, 42,351 men in the Health Professionals Follow-up Study and 76,890 women in the Nurses’ Health Study
Compared to men with BMI 18.5 to 22.9 kg/m2, men with a BMI 23.0 to 24.9 kg/m2 had an age-adjusted relative risk (RR) of CHD of 1.11 (95 percent confidence interval (CI) 0.97–1.26) (). The age-adjusted RR increased further with excess weight, to 2.28 (95% CI 1.96–2.66) for BMI of ≥ 30.0 kg/m2. After adjustment for potential confounding factors, the multivariate adjusted RRs were only modestly attenuated.
Relative Risk of CHD, by category of body mass index among men, the Health Professionals Follow-up Study*, and women, the Nurses’ Health Study#
Among women, we used the same BMI category cutoffs and referent group. Women with a BMI 23.0 to 24.9 kg/m2 had an age-adjusted RR of CHD of 1.10 (95% CI 0.97–1.25). The age-adjusted RR increased with excess weight to 2.44 (95% CI 2.17–2.74) with BMI of ≥ 30.0 kg/m2. After adjustment for potential confounding factors, multivariate adjusted RRs were slightly stronger.
In order to assess the effect of updating BMI and covariates on the effect estimates, we re-ran the multivariate models in each cohort using BMI updated with each (2 year) period of follow-up, though still lagged by 2 years. Compared to the estimates derived from our primary approach (in ), the RRs in men were slightly attenuated, with the RR associated with BMI 23.0 to 24.9 of 1.04 (95% CI 0.91–1.18), BMI 25.0 to 26.9 of 1.37 (95% CI 1.21–1.55), with BMI 27.0 to 29.9 kg/m2 RR of 1.56 (95% CI 1.37–1.77) and with BMI of ≥ 30.0 kg/m2 RR of 1.77 (95% CI 1.53–2.05). A similar pattern of modest attenuation was found in women, in models containing updated BMI and covariates, with RR associated with BMI 23.0 to 24.9 of 1.14 (95% CI 0.99–1.30), with BMI 25.0 to 26.9 of 1.34 (95% CI 1.16–1.54), with BMI 27.0 to 29.9 kg/m2 RR of 1.53 (95% CI 1.34–1.75) and with BMI of ≥ 30.0 kg/m2 RR of 2.09 (95% CI 1.85–2.37).
shows the RRs of CHD by BMI category across comorbid conditions, adjusted for potential confounding factors, among the men. The common reference category for all comparisons is men with a BMI of 18.5 to 22.9 kg/m2 and no associated comorbid conditions. Among men without either hypercholesterolemia, hypertension, or diabetes, we found a RR of CHD of 1.08 (95%CI 0.92–1.28) among those with a BMI of 23.0 to 24.9 kg/m2 and a RR 1.94 (95% CI 1.57–2.40) among those with BMI 30 kg/m2 and above. For men with hypercholesterolemia only, the RR was 1.28 (95% CI 0.85–1.91) among those with a BMI of 23.0 to 24.9 kg/m2 and 2.17 (95% CI 1.23–3.83) among those with BMI 30 kg/m2 and above.
Relative Risk of CHD by BMI, hypercholesterolemia, hypertension, both, or diabetes, the Health Professionals Follow-up Study
shows the corresponding relative risks among the women. For those without either hypercholesterolemia, hypertension, or diabetes, we found a RR of 1.08 (95% CI 0.93–1.26) for women with a BMI 23.0 to 24.9 kg/m2, and among those with BMI 30 kg/m2 and above, a RR of 2.12 (95% CI 1.81–2.47). Among those women with both hypercholesterolemia and hypertension, but not diabetes, the RR was 2.37 (95% CI 1.43–3.94) for BMI 23.0 to 24.9 kg/m2 and 3.90 (95% CI 2.66–5.71) for BMI of ≥ 30.0 kg/m2. Among women with diabetes, the RRs were 5.35 (95% CI 3.25–8.80) for BMI 23.0 to 24.9 kg/m2 and 9.51 (95% CI 7.72–11.72) for BMI of ≥ 30.0 kg/m2.
Relative Risk of CHD by BMI, hypercholesterolemia, hypertension, both, or diabetes, the Nurses Health Study
To evaluate the incremental risk of CHD associated with each level of excess weight within the subgroups of men and women with comorbid health conditions we ran multivariate models in each subgroup separately, with results presented in (men) and (women). We found a positive trend associated with increasing BMI among the men without these health conditions as well as among men with hypertension alone, with both hypercholesterolemia and hypertension, and with diabetes. The tests for trend were not significant for men with hypercholesterolemia only. Among the women, we found a significant positive trend of increasing CHD risk with increasing BMI in women without comorbid conditions, with hypercholesterolemia alone and with hypertension alone, as shown in . Trend tests were not significant in subgroups of women with both hypercholesterolemia and hypertension in combination and in women with diabetes.
Relative Risk of CHD, by category of body mass index and associated comorbid conditions among men, the Health Professionals Follow-up Study*
Relative Risk of CHD, by category of body mass index and associated comorbid conditions among women, the Nurses’ Health Study#
To address possible residual confounding by cigarette smoking, we conducted sub-analyses restricted to those men and women who had never smoked. Among men, with respect to reference BMI of 18.5 to 22.9 kg/m2, the RRs were for each category very similar to those of the full cohort, with a RR of 1.21 (95% CI 0.98–1.49 for BMI 23.0 to 24.9 kg/m2, 1.51 (95% CI 1.23–1.86) for BMI 25.0 to 26.9 kg/m2, 2.01 (95% CI 1.62–2.49) for BMI 27.0 to 29.9 kg/m2, and 2.33 (95% CI 1.81–3.01) for BMI > 30 kg/m2. In analyses restricted to women who had never smoked, the RRs for each BMI category were somewhat stronger than those derived from the full cohort, 1.48 (95% CI 1.16–1.88) for BMI 23.0 to 24.9 kg/m2, 1.80 (95% CI 1.41–2.31) for BMI 25.0 to 26.9 kg/m2, 2.30 (95%CI 1.81–2.90) for BMI 27.0 to 29.9 kg/m2, and 3.26 (95% CI 2.60–4.07) for BMI > 30 kg/m2.
The estimates of PARF for participants with diabetes necessarily included any additional effects of hypercholesterolemia and/or hypertension as well, since individuals with diabetes were all categorized as such, regardless of the presence or absences of these other comorbid conditions. In sub-analyses where the category of diabetes was split between those with or without an associated comorbid condition, we did not find meaningful modification of the BMI-CHD effect estimates (data not shown).
Simpler, more parsimonious models of CHD risk by BMI category and comorbid conditions were developed, with covariate terms adjusting for age and smoking only. This step was necessitated by the lack of directly comparable covariate information in the NHANES data. Comparisons of estimates from these models to those from the full multivariate models showed acceptably minimal (<10%) changes in RRs in all categories in both men and women. Accordingly, the RRs derived from these simpler models in the HPFS and NHS were applied to prevalence estimates of the joint distributions of the same categories of age, smoking, BMI, and comorbid conditions, in the US population to produce PARF estimates.
In PARF analyses based on BMI alone (without consideration of presence or absence of hypercholesterolemia, hypertension, or diabetes), adjusting for age and smoking, the reference group was participants in the BMI category 18.5 to 22.9 kg/m2. Among men, the PARF associated with higher BMI categories using this reference was 38.7%. Among women, the PARF was 43.5%.
PARF analyses were also conducted including the full affects of the associated comorbid conditions of hypercholesterolemia, hypertension, and diabetes, with the reference group in this case being participants in the BMI category of 18.5 to 22.9 kg/m2 and without any of the associated comorbidities. The PARF for men in these joint effects analyses was 60.4%, while that for women was 64.7%.