Abstract
BACKGROUND AND AIMS: Most epidemiological studies ignore long-term burden, gain and variability in body weight in assessing cardiometabolic disease risk. We investigated the associations of body mass index (BMI) trajectories measured by general practitioners with incident type 2 diabetes (T2D) and coronary artery disease (CAD).</p>
METHODS: We used electronic healthcare data from 111,615 European-ancestry participants from UK Biobank (57.1 (SD 7.8) years, 59.6 % women) with at least three BMI measurements (median trajectory period: 14.9 [interquartile range 9.5, 20.1] years). We calculated six variables capturing different long-term aspects, including i.e. burden (long-term average, area under the curve), gain (slope) and variability (standard deviation, average of the [absolute] consecutive BMI differences). The variables were used in principal component (PC) analyses and k-means clustering. Newly-derived dimensions and subgroups were used as exposures in cox-proportional hazard models.</p>
RESULTS: The BMI-trajectory indices were captured in two PCs reflecting BMI burden and BMI gain. The BMI-burden PC associated with higher T2D (hazard ratio [95 % confidence interval] per SD higher PC: 1.57 [1.55,1.60]) and CAD (1.17 [1.15,1.19]) risks, while weak or no associations were observed with the BMI-gain PC (T2D: 1.03 [1.01,1.05]; CAD: 1.01 [0.98,1.03]). Participants with the highest BMI burden, compared to those with lowest BMI burden without significant gain, had highest T2D (6.96 [6.41,7.55]) and CAD (1.57 [1.45,1.69]) risks. Both methods to capture BMI burden, gain and variability showed superior model fit compared to a single baseline BMI assessment.</p>
CONCLUSIONS: Long-term high BMI burden, irrespective of BMI gain, was a risk factor for cardiometabolic disease.</p>