Supplementary material

Supplementary Table 1. Characteristics of the KORA models (regression coefficients of multivariable logistic models)

Base model (model 1) / Clinical model (model 2) / Clinical model (model 3)
Intercept / -12.6453 / -21.2526 / -22.9422
Sex (male) / 0.5235 / 0.2073 / -0.3897
Age (years) / 0.0825 / 0.0946 / 0.1028
BMI (kg/m2) / 0.1353 / 0.1115 / 0.0892
Parental diabetes (yes) / 0.8482 / 0.8063 / 0.7876
Former smoker (yes) / 0.3912 / 0.4121 / 0.4682
Current smoker (yes) / 0.9140 / 0.9862 / 1.2013
Hypertension (yes) a / 0.7632 / 0.5569 / 0.4350
Fasting glucose (mg ⁄ dl) / - / 0.0865 / 0.0828
Serum uric acid (mg ⁄ dl) / - / - / 0.4231

a Hypertension was defined as Systolic ⁄ diastolic blood pressure ≥140 ⁄ 90 mmHg or use of medication for hypertension .

Supplementary Table 2. Baseline characteristics of individuals in KORA and PREVEND cohorts (aged ≥55 yr)

Characteristics / KORA (n = 873) / PREVEND (n=2050) / p value a
Sex (males) (%) / 50.0 / 55.0 / 0.01
Age (years) / 63.14 (5.4) / 64.2 (5.6) / <0.001
BMI (kg/m2) / 28.1 (4.0) / 27.2 (3.8) / <0.001
Parental diabetes (%) / 22.7 / 15.3 / <0.001
Smoking status
Never smoker (%) / 52.0 / 26.2
Former smoker (%) / 35.9 / 46.3 / <0.001
Current smoker (%) / 12.1 / 27.5
Hypertension (%) / 49.3 / 54.6 / 0.01
Fasting glucose (mmol/l) / 5.48 (0.52) / 4.97 (0.63) / <0.001
Serum uric acid (µmol/l) / 333.09 (81.5) / 326.0 (80.7) / 0.03

Data are percent for categorical variables and mean±SD for continuous variables.

a Comparison between KORA and PREVEND cohorts in each groups by incident diabetes: χ2 tests for categorical variables; t tests for continuous variables.

4

Supplementary Method Part 1

The formula of the KORA base model is:

To do logistic recalibration, logistic regression models were fitted with the KORA linear predictors as the only covariate in the PREVEND data set:

Supplementary Method Part 2

Where model 2cases and model 1cases is the mean predicted risk of model 2 and model 1, respectively, for those who developed diabetes, and the rest of values denote the corresponding risks for those who did not develop diabetes, the integrated discrimination improvement (IDI) is calculated as follow:

4