Development and Validation of an Empirical Dietary Inflammatory Index

Abstract

Background: Knowledge on specific biological pathways mediating disease occurrence (e.g., inflammation) may be utilized to construct hypotheses-driven dietary patterns that take advantage of current evidence on disease-related hypotheses and the statistical methods of a posteriori patterns.

Objective: We developed and validated an empirical dietary inflammatory index (EDII) based on food groups.

Methods: We entered 39 pre-defined food groups in reduced rank regression models followed by stepwise linear regression analyses in the Nurses' Health Study (NHS, n = 5230) to identify a dietary pattern most predictive of 3 plasma inflammatory markers: interleukin-6 (IL-6), C-reactive protein (CRP), and tumor necrosis factor α receptor 2 (TNFαR2). We evaluated the construct validity of the EDII in 2 independent samples from NHS-II (n = 1002) and Health Professionals Follow-up Study (HPFS, n = 2632) using multivariable-adjusted linear regression models to examine how well the EDII predicted concentrations of IL-6, CRP, TNFαR2, adiponectin, and an overall inflammatory marker score combining all biomarkers.

Results: The EDII is the weighted sum of 18 food groups; 9 are anti-inflammatory and 9 proinflammatory. In NHS-II and HPFS, the EDII significantly predicted concentrations of all biomarkers. For example, the relative concentrations comparing extreme EDII quintiles in NHS-II were: adiponectin, 0.88 (95% CI, 0.80, 0.96), P-trend = 0.003; and CRP, 1.52 (95% CI, 1.18, 1.97), P-trend = 0.002. Corresponding associations in HPFS were: 0.87 (95% CI, 0.82, 0.92), P-trend < 0.0001; and 1.23 (95% CI, 1.09, 1.40), P-trend = 0.002.

Conclusion: The EDII represents, to our knowledge, a novel, hypothesis-driven, empirically derived dietary pattern that assesses diet quality based on its inflammatory potential. Its strong construct validity in independent samples of women and men indicates its usefulness in assessing the inflammatory potential of whole diets. Additionally, the EDII may be calculated in a standardized and reproducible manner across different populations thus circumventing a major limitation of dietary patterns derived from the same study in which they are applied.

Keywords: dietary inflammatory potential; dietary patterns; hypothesis-driven; inflammation; inflammatory markers.

FIGURE 1

Adjusted mean (95% CI) plasma inflammatory marker concentrations in quintiles of the EDII in regular users (A) and nonusers (B) of aspirin/NSAIDs (Nurses’ Health Study; n = 5230; 1986–1990). Values are mean concentrations of biomarkers adjusted for age at blood draw, physical activity, smoking status, BMI, menopausal status, postmenopausal hormone use, case-control status, batch effects for biomarker measurements, and an inflammation-related chronic disease comorbidity score. Chronic diseases and conditions included in the score were hypercholesterolemia, cancer, diabetes, high blood pressure, heart disease, and rheumatoid or other arthritis. All tests were 2-sided and all 95% CIs were statistically significant (i.e., did not include 1). All biomarker concentrations were back-transformed (e^x), and all P-trends < 0.0001. P values for the interaction of EDII and aspirin/NSAIDs were as follows: IL-6 = 0.11; CRP = 0.36; TNFαR2 = 0.21. Sample sizes in EDII quintiles were as follows—nonusers of aspirin/NSAIDs: Q1 = 668, Q2 = 669, Q3 = 669, Q4 = 669, and Q5 = 669; regular aspirin/NSAID users: Q1 = 377, Q2 = 377, Q3 = 378, Q4 = 377, and Q5 = 377. CRP, C-reactive protein; EDII, empirical dietary inflammatory index; NSAID, nonsteroidal anti-inflammatory drug; Q, quintile; TNFαR2, TNF-α receptor 2.

FIGURE 2

Adjusted mean (95% CI) plasma inflammatory marker concentrations in quintiles of the EDII in normal-weight women (BMI <25 kg/m²) (A) and overweight/obese women (BMI ≥25 kg/m²) (B) from the Nurses’ Health Study (n = 5230; 1986–1990). Values are mean concentrations of biomarkers, adjusted for age at blood draw, physical activity, smoking status, aspirin/NSAID use, menopausal status, postmenopausal hormone use, case-control status, batch effects for biomarker measurements, and an inflammation-related chronic disease comorbidity score. Chronic diseases and conditions included in the score were hypercholesterolemia, cancer, diabetes, high blood pressure, heart disease, and rheumatoid or other arthritis. All tests were 2-sided and all 95% CIs were statistically significant (i.e., did not include 1). All biomarker concentrations were back-transformed (e^x), and all P-trends < 0.0001. P values for interaction of EDII and aspirin/NSAIDs were as follows: CRP = 0.13, IL-6 = 0.12, and TNFαR2 = 0.43. Sample sizes in EDII quintiles were as follows—normal-weight women: Q1 = 544, Q2 = 545, Q3 = 544, Q4 = 545, and Q5 = 544; overweight/obese women: Q1 = 501, Q2 = 502, Q3 = 502, Q4 = 502, and Q5 = 501. CRP, C-reactive protein, EDII, empirical dietary inflammatory index; NSAID, nonsteroidal anti-inflammatory drug; Q, quintile; TNFαR2, TNF-α receptor 2.