Modified-Weight Healthy Eating Index-2015 Scores Are More Strongly Associated With Mortality Risk Than Standard Scores

Abstract

Background: Various diet quality scores are consistently and similarly associated with mortality risk. Emerging evidence suggests that individual diet quality components are differentially associated with mortality risk, but it is unclear whether or not modified weights reflective of relative component differences would strengthen mortality associations.

Objective: This study examined whether Healthy Eating Index 2015 (HEI-2015) scores with modified (vs standard) component weights are differentially associated with mortality risk.

Design: This was a longitudinal analysis of the National Health and Nutrition Examination Survey III (1988-94) with 23 years of mortality follow-up. The HEI-2015 and modified-weight scores were calculated from one 24-hour recall. The a priori Key Facets HEI was a subset score equivalently weighting fruits, vegetables, whole grains, and seafood and plant proteins. In the least absolute shrinkage and selection operator regression-weighted HEI, components were assigned weights reflecting relative mortality risk associations.

Participants/setting: Analyses included 10,789 US adults (aged 20 years and older) who were not pregnant and without prior diabetes, cancer, cardiovascular disease, or chronic kidney disease diagnoses.

Main outcome measures: All-cause and cardiovascular disease mortality risk were the primary outcomes examined.

Statistical analyses performed: Three energy-adjusted HEI scores were assigned to quintiles; covariate-adjusted sex-stratified Cox models with age as the timescale assessed associations between and trends across quintiles of HEI scores and all-cause and cardiovascular disease mortality risk.

Results: Modified-weight HEI scores were associated with 23% to 38% reduced all-cause mortality risk in the highest vs lowest quintile, whereas comparisons were only significant for women (Key Facets P = 0.02 and least absolute shrinkage and selection operator regression-weighted P = 0.001; for men P = 0.06 on both scores), trends across quintiles of modified-weight scores were significant for men and women. The HEI-2015 was not significantly associated with all-cause mortality risk, and none of the scores were associated with cardiovascular disease mortality risk.

Conclusions: Only modified-weight HEI scores were associated with reduced all-cause mortality risk. Findings suggest modified diet quality weighting schemes warrant further examination to determine their replicability.

Keywords: Diet quality components; Healthy Eating Index; Machine learning; Mortality; National Health and Nutrition Examination and Survey.