Abstract MP61: Body Mass Index and Mortality among Adults with Incident Type Diabetes: Results from Two Prospective US Cohort Studies

Abstract

Background: Recent evidence suggests that having normal weight at time of diagnosis of type 2 diabetes (T2D) is associated with an increased risk of premature death, compared to being overweight or obese; these studies were limited in their sample size and ability to adequately address residual confounding by smoking and reverse causation. Objective: To prospectively evaluate the association between body mass index (BMI) at T2D diagnosis and mortality in two large cohorts. Methods: Women and men with incident T2D from the Nurses’ Health Study (1978-2010; N=8,984) and Health Professionals Follow-up Study (1988-2010; N=2,443) were included if they were free of major chronic disease (cardiovascular disease [CVD], cancer) at T2D diagnosis. Participants’ self-reported body weight preceding diagnosis (mean 11 months) and height was used to calculate BMI (kg/m2). Cox proportional hazards models estimated the relative risk (HR) and 95% confidence interval (CI) for mortality across BMI categories. Multivariable models adjusted for age, smoking, baseline comorbidities (high blood pressure, cholesterol), and several other lifestyle factors. Fixed-effects meta-analyses were used to combine individual cohort estimates. Results: In all, 3,119 total deaths were observed over a follow-up of 36 years in women (18.7 deaths/1,000 person-years) and 26 years in men (25.2 deaths/1,000 person-years). A J-shaped association was observed across BMI categories (18.5-22.4, 22.5-24.9, 25.0-27.4, 27.5-29.9, 30.0-34.9, >35.0) and all-cause mortality (HR [CI] by category: 1.26[1.03, 1.55], 1[reference], 1.11[0.97, 1.28], 1.08[0.94, 1.25], 1.19[1.04, 1.37], 1.33[1.15, 1.55]). After stratifying by smoking status, a direct linear association was present among never smokers and a J-shaped relationship persisted among ever smokers. Excluding deaths in the first 4 years of follow-up and adjusting for BMI change prior to diagnosis further accentuated the linear relationship between BMI and all-cause mortality among never smokers (0.90[0.57, 1.43], 1.00 [reference], 1.19[0.91, 1.57], 1.20[0.91, 1.58], 1.27[0.97, 1.65], 1.50[1.14, 1.99]; p-trend<0.001). The association across BMI categories and CVD mortality was also linear among never smokers, and flat among ever smokers. In addition, among ever smokers, cancer mortality was highest among those with BMI 18.5-22.5. No clear trend was observed between BMI and mortality due to other causes. Excluding insulin users did not appreciably modify the associations. Conclusions: We found no evidence to support lower mortality rates among diabetics who were overweight or obese at diagnosis, compared to their normal-weight counterparts. In contrast, after accounting for confounding by smoking, we observed direct linear relationships between BMI and both all-cause and CVD mortality in our cohorts. Reducing other biases strengthened these relationships.