Outcome definition and index validity in cross-national comparisons of diet-attributable metabolic burden
Year: 2026
Type: preprint
Abstract: Background In three national surveys, no cross-national ranking of diet-attributable metabolic burden was identified: the apparent ordering moved with the definition of the outcome and with the validity of the dietary index. Both were tested. Subjects/Methods We harmonised individual-participant data from national surveys in the USA (NHANES 2017–2018), Korea (KNHANES 2015–2021) and Mexico (ENSANUT 2018–2019); 56,365 adults in total, recovered Mexican blood-pressure data from the raw files, and defined metabolic syndrome (MetS) with a common five-component ATP III rule. We estimated survey-weighted sex-specific associations of the energy-adjusted Dietary Inflammatory Index (E-DII) with MetS, country-specific population-attributable fractions (PAFs) on a common counterfactual scale, and the E-DII’s correlation with log high-sensitivity C-reactive protein (hsCRP). A quantitative bias analysis examined whether validity differences could reverse the country ordering. Results The E-DII–MetS association was larger in women (USA OR 1.23, 95% CI 0.98–1.55; Korea 1.078, 1.028–1.132; Mexico 1.074, 0.992–1.163; interaction P=0.080) but disappeared without blood pressure in the outcome (pooled 1.065, 0.982–1.155; P=0.13). Female PAFs were 15.0%, 7.6% and 3.8%; no pairwise difference was distinguishable (all P≥0.23). E-DII validity was 0.137 in US women versus 0.040 in Korean women (P=2.0×10−4); the validity ratio was 0.29, correction reversed the observed ordering, and its interval straddled the reversal threshold. Conclusions Neither the direction nor the magnitude of cross-national differences in diet-attributable MetS burden was identified; the ordering was bounded by outcome definition and by index validity. Cross-national burden comparisons require local validation of both.
Source: medRxiv
Authors: Shangze Huang, Xinyan Wang
Language: English
Topic: Nutritional Studies and Diet
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7171647656