Abstract P343: Using Semi-Parametric Models to Examine the Association Between Changes in Animal-Source and Fat Calories Over 18 Years With the Distribution of Hba1c in Chinese Adults
Background: Most researchers assume a homogeneous association between diet behaviors and cardiovascular disease (CVD) risk factors like diabetes, despite the fact that individuals differ in susceptibility and unobserved characteristics.
Objective: To assess whether distributional changes in dietary intake over time relate to current diabetes and whether initial diet status or changes in diet are comparatively more relevant for diabetes risk.
Methods: With population-based data from 18,987 Chinese adults ages 18-75y seen up to 7 times between 1991 and 2009 in the China Health and Nutrition Survey, we derived diet components from three consecutive 24-hour recalls (linked to individualized recipes and Chinese nutrition facts panels) at each survey (% energy from fat [%fat] and % energy from animal-source foods [%animal-source]) in relation to HbA1c measured via fasting blood in 2009. We used sex-stratified, longitudinal growth mixture models with random effects and a skew-T distribution to model individual trajectories separately for each dietary component over 18 years, controlling for time-varying urbanization, age, income, and caloric intake. Then we used the intercepts and slopes from the diet trajectories to predict the distribution of HbA1c in 2009 in separate models.
Results: Using longitudinal data and adjusting for time-varying covariates, the shape of the distributions of %animal-source calories and %fat calories became flatter and more positively skewed over 18 years of modernization, with an increasing proportion of the population at the higher ends of fat and animal-source food distributions over time. The slope of %fat calories over time was strongly associated with HbA1c (men: p<0.001; women: p=0.016), whereas both the intercept (men & women: p<0.0001) and slope (men: p=0.004; women: p=0.038) of %animal-source calories over 18 years strongly predicted mean HbA1c in 2009.
Conclusions: Our novel, semi-parametric growth mixture models revealed that targeting %fat calories may be particularly relevant for diabetes prevention because there was a strong association between change in %fat and HbA1c, irrespective of initial %fat calorie levels. Addressing animal-source food consumption remains important for diabetes prevention.
Author Disclosures: S.M. Attard: None. A. Howard: None. A.H. Herring: None. B.M. Popkin: None. P. Gordon-Larsen: None.
This research has received full or partial funding support from the American Heart Association, Mid-Atlantic Affiliate (Maryland, North Carolina, South Carolina, Virginia & Washington, DC).
- © 2015 by American Heart Association, Inc.