High incidence of type 2 diabetes in a population with normal range body mass index and individual prediction nomogram in Vietnam.

Aims: The study aimed at determining 5-year incidence and prediction nomogram for new-onset type 2 diabetes (T2D) in a middle-aged population in Vietnam.

Methods: A population-based prospective study was designed to collect socio-economic, anthropometric, lifestyle and clinical data. Five-year T2D incidence was estimated and adjusted for age and sex. Hazard ratio (HR) for T2D was investigated using discrete-time proportional hazards model. T2D prediction model entering the most significant risk factors was developed using the multivariable logistic-regression algorithm. The corresponding prediction nomogram was constructed and checked for discrimination, calibration and clinical usefulness.

Results: The age- and sex-adjusted incidence was 21.0 cases (95% CI: 12.2-40.0) per 1000 person-years in people with mean BMI of 22.2 (95% CI: 21.9-22.7 kg/m ). The HRs (95% CI) for T2D were 1.14 (1.05-1.23) per 10 mmHg systolic blood pressure, 1.05 (1.03-1.08) per 1 cm waist circumference, 1.40 (1.13-1.73) per 1 mmol/L fasting blood glucose, 1.77 (1.15-2.71) per sleeping time (<6 h/day vs 6-7 h/day) and 2.12 (1.25-3.61) per residence (urban vs rural). The prediction nomogram for new-onset T2D had a good discrimination (area under curve: 0.711, 95% CI: 0.666-0.755) and fit calibration (mean absolute error: 0.009). For the predicted probability thresholds between 0.03 and 0.36, the nomogram showed a positive net benefit, without increasing the number of false positives.

Conclusion: This study highlighted an alarmingly high incidence of T2D in a middle-aged population with a normal range BMI in Vietnam. The individual prediction nomogram with decision curve analysis for new-onset T2D would be valuable for early detection, intervention and treatment of the condition.