Evaluating Temporal Consistency and Dietary-Recall Sensitivity of the Association Between the Dietary Inflammatory Index and Cardiovascular Disease: Cross-Sectional and Mortality Analyses of NHANES 2005–2018

Background: /Objectives: Associations between the Dietary Inflammatory Index (DII) and cardiovascular disease (CVD) in the National Health and Nutrition Examination Survey (NHANES) have usually relied on selected cycles, short-term recalls, and prevalent disease. This study examined the association’s consistency across survey periods, its dependence on recall measurement error, and reverse causation. Methods: Adults aged ≥20 years in NHANES 2005–2018 with two 24 h recalls (n = 30,579) were analyzed with survey-weighted logistic regression of self-reported CVD on a 27-parameter DII. Pre-specified analyses were an equivalence test of 2005–2010 versus 2011–2018, regression calibration using the replicate recall, restrictions for post-diagnosis dietary change, and survey-weighted Cox models for mortality through 2019 in 27,239 adults free of CVD. Results: Each 1-unit higher DII was associated with higher odds of CVD (odds ratio [OR] 1.09, 95% CI 1.03–1.15). Period-specific ORs were 1.09 and 1.08, but equivalence within ±10% was not established (interaction-model ratio of ORs per SD 0.94, 90% CI 0.85–1.04). After calibration, ORs per unit of usual DII were 1.17 (1.07–1.27) for the two-day mean and 1.20 (1.10–1.31) for day one. Excluding special diets or restricting to diagnoses within 5 or 10 years did not attenuate the association. Hazard ratios were 1.09 (0.98–1.21) for CVD mortality and 1.09 (1.04–1.14) for all-cause mortality. Post hoc, cycle-specific estimates showed no evidence of heterogeneity, and each DII unit was associated with 13% higher serum C-reactive protein. Conclusions: The DII–CVD association was modest, similar across survey periods and recall strategies, and larger after correction for recall error; restriction analyses did not indicate that post-diagnosis dietary change created it, and the CVD mortality estimate was concordant but imprecise. Formal equivalence across periods, a CVD mortality association, and causation were not established.