The ARIC predictive model reliably predicted risk of type II diabetes in Asian populations

Table 2 Comparison of published and locally-estimated multivariate predictive functions

Bold and italicized fonts: Estimates from local models whose 95% confidence intervals do not overlap with the published estimates, i.e. we assumed that the published estimates were the true values in the hypothesis testing
*Categorical variables are used for the Framingham model with the following rule: fasting glucose level: (1 = 100 - 125 mg/dL, 0 ≤ 100 mg/dL), overweight (1 = BMI ≥ 25 & BMI < 30, 0 = otherwise), obese (1 = BMI ≥ 30, 0 = otherwise), low high density lipoprotein (HDL) (1 = HDL < 40 mg/dL in men or < 50 mg/dL in women, 0 = otherwise), high triglyceride level (1 ≥ 150 mg/dL, 0 = otherwise), elevated blood pressure (systolic blood pressure ≥ 130 or diastolic blood pressure ≥ 85 mm Hg, 0 = otherwise)For the SAHS model, the 95% confidence intervals of the coefficients estimated from the NHS-92 cohort included the published SAHS coefficient for all but four risk factors: age, gender, FPG at baseline and BMI; the coefficients for age and gender were greater for the published SAHS model than the locally-estimated coefficients. In contrast, the locally estimated coefficients for FPG and BMI were larger than the published coefficients.

ISSN: 1471-2288