Waist-corrected BMI predicts incident diabetes mellitus in a population-based observational cohort study

Abstract

Introduction: Waist-corrected body mass index (wBMI), which combines BMI and waist circumference (WC) measurements, has proven superior to either measure alone for predicting obesity but has not yet been applied to the prediction of diabetes mellitus (DM).

Methods: Over a 5-year period, 305,499 subjects were eligible for this study based on citizen health check-ups in the Tacheng Area of northwest China. Diagnosis of DM was defined as the end point.

Results: After exclusion, a total of 111,851 subjects were included in the training cohort and 47,906 in the validation cohort. Participants of both sexes with wBMI in the upper quartiles had significantly higher incidence of DM than those with wBMI in the lower quartiles (log-rank χ^{2 =} 236, p< 0.001 for men; log-rank χ^{2 =} 304, p< 0.001 for women). After adjusting for multiple variables, WC, BMI, wBMI, and waist-to-height ratio (WHtR) were all independent predictors for diabetes. In men, the adjusted hazard ratios (HRs) of wBMI for diabetes for the second, third, and fourth quartiles were 1.297 [95% CI: 1.157, 1.455], 1.664 [95% CI: 1.493, 1.853], and 2.132 [95% CI: 1.921, 2.366], respectively, when compared with the first quartile. In women, they were 1.357 [95% CI: 1.191, 1.546], 1.715 [95% CI: 1.517, 1.939], and 2.262 [95% CI: 2.010, 2.545], respectively. Compared with WC, BMI, and WHtR, wBMI had the highest C-index in both men (0.679, 95% CI: 0.670, 0.688) and women (0.730, 95% CI: 0.722, 0.739). Finally, a nomogram was constructed to predict incident DM based on wBMI and other variables. In conclusion, wBMI had the strongest predictive capacity for incident DM when compared with WC, BMI, and WHtR, especially in women.

Discussion: This study provides a reference for advanced investigation of wBMI on DM and other metabolic diseases in the future.

Keywords: body mass index; diabetes mellitus; waist circumference; waist-corrected body mass index; waist-to-height ratio.

Figure 1

Kaplan–Meier survival curves of non-diabetes by wBMI quartiles in both sexes. Incident diabetes mellitus risk increases with wBMI quartiles in (A) men and (B) women. Log-rank χ^2 = 236, p< 0.001 for men; log-rank χ^2 = 304, p< 0.001 for women. wBMI, waist-corrected body mass index.

Figure 2

Forest plots of multivariable Cox regression results. After regression, quartiles of wBMI, age, DM family history, FPG, HBP status, education years, and exercise frequency were independent predictors for diabetes in men (left) and women (right). wBMI, waist-corrected body mass index; DM, diabetes mellitus; FPG, fasting plasma glucose; HBP, hypertension. *p< 0.05; **p< 0.01; ***p< 0.001.

Figure 3

Concordance indexes of quartiles of wBMI, BMI, WC, and WHtR for both sexes. Compared to the other body composition measures, wBMI had the highest concordance in both men (left) and women (right). *p<0.05, ***p<0.001 vs. wBMI. qwBMI, quartile of waist-corrected body mass index; qWC, quartile of waist circumference; qBMI, quartile of body mass index; qWHtR, quartile of waist-to-height ratio.

Figure 4

A nomogram for prognostic prediction of non-DM using qwBMI in men (left) and women (right). A male patient aged 60 years (40 points) with wBMI of 26.0 (15 points), no DM family history (0 points), a history of hypertension (10 points), education<9 years (3 points), FPG 6.0 mmol/L (78 points), and who seldom exercises (8 points) would have 154 total points. A line is drawn downward from the total points axis to the survival axes to determine probability of 2‐year non-DM (approximately 90%) and 4-year non-DM (approximately 70%). DM, diabetes mellitus; qwBMI, quartile of waist-corrected body mass index; FPG, fasting plasma glucose; HBP, hypertension.

Figure 5

Calibration curves for the nomogram of both sexes. (A, B) Calibration curves of 2‐ and 4‐year non-DM probability for male subjects in the training cohort. (C, D) Calibration curves of 2‐ and 4‐year non-DM probability for female subjects in the training cohort. (E, F) Calibration curves of 2‐ and 4‐year non-DM probability for male subjects in the validation cohort. (G, H) Calibration curves of 2‐ and 4‐year non-DM probability for female subjects in the validation cohort. The green line indicates the ideal reference line where predicted probabilities would match observed survival rates. Red dots represent the performance of the nomogram and were calculated by bootstrapping. The closer the solid red line is to the green line, the more accurate the model’s predictions. DM, diabetes mellitus.