Objective: The aim of this study was to compare the ability of body mass index (BMI), waist circumference (WC), waist-to-height ratio (WHtR) and percent body fat (PBF) to predict incident diabetes.
Methods: Areas under receiver operating characteristic curves (AUCs) for these parameters were calculated for a Japanese population consisting of 1,704 men and 1,016 women to determine incident diabetes and hazard ratios (HRs) for each one SD increase in these parameters, and also when they were dichotomously defined.
Results: During 4 years of follow-up (mean 3.4 years), 54 men (3.2%) and 19 women (1.9%), of whom 69 (94.5%) were prediabetes at baseline, developed diabetes. The AUCs [95% confidence intervals (CIs)] for BMI, WC, WHtR and PBF were 0.69 (0.61–0.76), 0.68 (0.61–0.76), 0.70 (0.63–0.78) and 0.65 (0.59–0.71), respectively, in men and 0.74 (0.61–0.86), 0.75 (0.61–0.88), 0.76 (0.64–0.88) and 0.77 (0.65–0.88), respectively, in women. After adjusting for fasting glucose, hemoglobin A1c and other confounders, only BMI in men was significant [HR (95% CI): 1.49 (1.07–2.07), p=0.018] among continuous parameters, while only WC and PBF in women were significant [HRs (95% CIs): 5.22 (1.24–21.91), p=0.024 and 5.67 (1.33–24.20), p=0.019, respectively] among dichotomous parameters.
Conclusion: The abilities of these anthropometric parameters to predict incident diabetes were almost equal and limited when both fasting glucose and hemoglobin A1c were measured.
View full abstract