Cardiometabolic Disease Risk Assessment Predicts Incident Diabetes in a Large and Diverse Population

Background: Risk scores for predicting type-2 diabetes mellitus (T2DM) have been developed, but few have included adequate African-Americans.

Objective: To derive and validate scoring systems to predict future diabetes in populations including caucasian- and African-Americans, weighted by sex and race.

Design, Setting, and Participants: We estimated 9.5-year risk of T2DM in participants with valid follow-up data taken from 12,1321 people in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study, and externally validated it in 9,712 participants from the Atherosclerosis Risk in Communities study.

Methods: Cardiometabolic disease scores were calculated for each sex-race pair using quantitative clinical data. Incidence diabetes was defined as fasting glucose ≥ 126 mg/dL, non-fasting glucose ≥ 200mg/dL, currently on medication for T2DM, or self-report. Covariates included: sex, race, and age; waist circumference; blood pressure; HDL cholesterol; fasting triglycerides or on lipid-lowering medication. We assigned a simple integer score for each risk factor component using logistic regression coefficients for each sex-race pair. Discrimination was assessed with area under the receiver operating characteristic curves (AROCs).

Results: Diabetes incidence rates at a median follow-up of 9.5 years were 18.3% and 18.7% in African-American men and women, and 12.3% and 9.2% in caucasian-American men and women. Weighted covariates differed more by race than sex (p = 0.001). The AROCs ranged from 0.72 for black men to 0.78 for white women. Externally validated models performed well; AROC was 0.75 (black men) to 0.83 (white women).

Conclusion: The weighted CMDS score has high model discrimination using readily available clinical data, and can be used to quantify T2DM risk based on race- and sex- specific data; thus, diabetes prevention can be targeted to high risk individuals in a manner that optimizes outcomes, benefit/risk ratio, and cost-effectiveness.