Abstract
OBJECTIVE: To investigate whether dual-energy x-ray absorptiometry (DXA) estimates of adiposity improve risk prediction for cardiometabolic diseases over traditional surrogates, body mass index (BMI), waist circumference (WC), and waist-to-hip ratio (WHR) in older women. PATIENTS AND METHODS: We analyzed up to 9744 postmenopausal women aged 50 to 79 years participating in the Women's Health Initiative who underwent a DXA scan and were free of cardiovascular disease and diabetes at baseline (October 1993 to December 1998) and followed through September 2015. Baseline BMI, WC, WHR, and DXA-derived percent total-body and trunk fat (%TrF) were incorporated into multivariable Cox proportional hazards models to estimate the risk of incident diabetes, atherosclerosis-related cardiovascular diseases (ASCVDs), heart failure, and death. Concordance probability estimates assessed the relative discriminatory value between pairs of adiposity measures. RESULTS: A total of 1327 diabetes cases, 1266 atherosclerotic cardiovascular disease (ASCVD) cases, 292 heart failure cases, and 1811 deaths from any cause accrued during a median follow-up of up to 17.2 years. The largest hazard ratio observed per 1 standard deviation increase of an adiposity measure was for %TrF and diabetes (1.77; 95% CI, 1.66-1.88) followed by %TrF and broadly defined ASCVD (1.22; 95% CI, 1.15-1.30). These hazard ratios remained significant for both diabetes (1.47; 95% CI, 1.37-1.57) and ASCVD (1.22; 95% CI, 1.14-1.31) even after adjusting for the best traditional surrogate measure of adiposity, WC. Percentage of trunk fat was also the only adiposity measure to demonstrate statistically significant improved concordance probability estimates over BMI, WC, and WHR for diabetes and ASCVD (all P<0.05). CONCLUSION: DXA-derived estimates of abdominal adiposity in postmenopausal women may allow for substantially improved risk prediction of diabetes over standard clinical risk models. Larger DXA studies with complete lipid biomarker profiles and clinical trials are needed before firm conclusions can be made.