Abstract 24: Childhood Risk Factor Patterns Predict Adult Type 2 Diabetes Better Than Single Childhood Risk Factor Measures: The International Childhood Cardiovascular Cohort (i3C) Consortium
Background: Childhood risk factors predict adult onset type 2 diabetes mellitus (T2DM). We developed a novel statistical approach using risk factor patterns to improve prediction.
Methods: The cohort included 5269 individuals from the Bogalusa Heart Study, Cardiovascular Risk in Young Finns Study, Childhood Determinants of Adult Health Study, Childhood Cohort Studies in Minneapolis and Cincinnati, and the Muscatine Study. All had 2+ childhood measures of risk factors between ages 3 and 20 in 1970s-90s and reported whether they had T2DM between ages 21 and 63 in 2011-16. We established sex-specific polynomial age patterns to represent standard growth curves for each risk factor and computed the residual from this standard for each observation within ages 3-20. Residuals were regressed on age within each person to get a personal intercept and slope, and then 9 patterns were formed as intercept tertiles crossed with slope tertiles within each intercept tertile. T2DM associations with risk factors were assessed by using either the first childhood measurement or childhood patterns of body mass index (BMI), systolic blood pressure (SBP), and height (HT), with covariates age at first visit, sex, race, and study.
Results: Mean age at first measure was 9.2 (SD 3.3) y, 41% male, 85% white, 12% black, with 2-19 repeats of childhood risk factor measures. Risk factor patterns tended to diverge from each other by mid-childhood and measured adult risk factor values were well predicted, diverging widely across the childhood patterns (not shown). After mean follow-up of 29.0 (SD 8.1) y, 282 (5.4%) participants self-reported T2DM. T2DM was predicted by the model including first measures of BMI (positive), SBP (positive), and HT (inverse), but the corresponding pattern model improved prediction probability by 55% (Table).
Conclusion: These findings show the value of using multiple risk factor measurements during childhood and the potential for developing risk factor pattern charts to be used as clinical standards for predicting adult T2DM risk.
Author Disclosures: T. Hu: None. A. Sinaiko: None. E. Urbina: None. O. Raitakari: None. J. Woo: None. L. Bazzano: None. J. Steinberger: None. T. Burns: None. M. Juonala: None. A. Venn: None. S. Daniels: None. R. Prineas: None. T. Dwyer: None. C. Magnussen: None. D. Jacobs: None.
- © 2017 by American Heart Association, Inc.