The Dietary Inflammatory Index Is Associated With Diabetes Severity

Dana E. King, MD, MS; Jun Xiang, MS, MA


J Am Board Fam Med. 2019;32(6):801-806. 

In This Article


Study Population

The present study was a retrospective cross-sectional study using data from the continuous National Health and Nutrition Examination Survey (NHANES 2013 to 2014). The NHANES is a series of complex and multistage surveys, conducted by the National Center for Health Statistics (NCHS), designed to assess the health and nutritional status of the noninstitutionalized US population. Since 1999, the continuous NHANES[13] collected demographic, socioeconomic, dietary, and health-related information through 2 components, an in-home interview and a medical examination, on selected participants in 2-year cycles. Informed consents were obtained from all participants and the protocol for conducting the NHANES survey was approved by the NCHS Research Ethics Review Board. Details on survey design and response rate can be found on the NHANES Web site.[13] Analyses for this study were limited to adults ≥20 years of age (the customary classification cutoff in the NHANES) with nonmissing information for variables of interest. The NHANES uses ≥20 years as the cutoff for adults, and we have used it to be consistent with many previous NHANES studies. The focus of the study was adults with diabetes because the role of inflammation in diabetes, while well established in adults, is not as well established in children.[14] West Virginia University Institutional Review Board approved this study to be exempt.

Definition of Nondiabetes, Prediabetes, Diabetes, and Severe Diabetes

To define diabetes status of a participant, we followed the guideline from the American Diabetes Association using measured HgbA1c as a diagnostic criterion: without diabetes treatments, participants with HgbA1c less than 5.7%, between 5.7% to 6.4%, or 6.5% or greater would be categorized as having no diabetes, prediabetes, or diabetes respectively; 9% or greater HgbA1c would be defined as having severe diabetes.[15] We also added those people who answered positively to the question, "Were you told by a doctor that you have diabetes?" to identify additional individuals with diabetes.

The DII and NHANES 2013 to 2014

The DII is a tool, created to examine the inflammatory potential of individuals' diets. A description of the design and development of the original DII can be found elsewhere.[1]

The current study incorporated the latest version of DII, which represents an improved scoring algorithm based on extensive review of the literature and a world food consumption data from several countries.[16] Briefly, a total of 45 food parameters (types of food and nutrients) derived from dietary data were assigned inflammatory effect scores based on the research findings from 1943 selected articles, examining the role of the food parameters on the 6 established inflammatory biomarkers (IL-1β, IL-4, IL-6, IL-10, TNF-a, and C-reactive protein), published from 1950 to 2010. World food consumption data, based on 11 diverse populations around the world, was used to generate a mean and standard deviation for each food parameter. An individual's diet was then linked to the world food database as a z-score, calculated by subtracting the "standard global mean" and dividing its standard deviation. This z-score was then converted to a centered percentile score to minimize the risk of "right skewing." The product of the centered percentile score and the respective article generated inflammatory effect score for each food parameter was then summed to create an overall DII score for an individual. A total DII score could be positive or negative. Higher positive DII scores indicate more proinflammatory diets and more negative scores imply more anti-inflammatory diets.

In this study, we utilized a total of 28 out of the 45 food parameters, for which we had dietary intake data available from the 2 24-hour dietary recalls data in the NHANES 2013 to 2014 to calculate DII scores. These parameters include total calories, total fat, saturated fat, monounsaturated fat, polyunsaturated fat, omega-3 fatty acids, omega-6 fatty acids, protein, carbohydrate, fiber, alcohol, cholesterol, niacin, thiamin, vitamin A, vitamin B2, vitamin B6, vitamin B12, vitamin C, vitamin D, vitamin E, iron, magnesium, selenium, zinc, folic acid, β carotene, and caffeine. We calculated the total DII scores per 1000 calories of food consumed to control for the effect of different amounts of total energy intakes.

Population Covariates

We extracted population characteristics including age, gender, race, BMI (body mass index), physical activity, smoking status, alcohol use, and socioeconomic status (education level, health insurance status) as potential covariates. Age was divided into 3 groups: 20 to 44 years, 45 to 64 years, and 65 years and older. We examined race in 4 race groups of non-Hispanic white, non-Hispanic Black, Hispanic, and Other race, as they are categorized in the NHANES.[13] There were 4 BMI categories combined as underweight (<18.5 kg/m2), normal (18.5 to 24.9 kg/m2), overweight (25 to 29.9 kg/m2), and obese (≥30 kg/m2) based on the Center for Disease Control and Prevention breakdown. Participants' education level was grouped into 2 categories of "<High school" and "≥High school." Health insurance status was defined as "Yes" for having health insurance and "No" for not having health insurance. Two levels of physical activity were defined as "≥150" or "<150" minutes moderate-intense recreational physical activity per week. Smoking status was coded as "smoke" for current smokers and "not smoke" for current nonsmokers. For alcohol use, the cutoff for "alcohol use" and "no alcohol use" was at least 12 alcohol drinks in the past year.

Statistical Analysis

All analyses in this study were conducted using SAS (version 9.4, 2013, SAS Institute Inc. Cary, NC). To account for the complex survey design (including oversampling, survey nonresponse, and poststratification), we incorporated 2-year sampling weights and SAS survey analysis procedures following NHANES survey methods and analytic guidelines.[13]

Population characteristics of the study sample were compared across diabetes status using χ2 test. Regression analyses were performed to determine the differences in DII scores between diabetes status and between the severity levels of diabetes. To examination the relationship between diabetes and DII scores adjusted for all covariates included age, sex, race, socioeconomic status, BMI categories, alcohol use, smoking status, and physical activity, we estimated multivariable adjusted odds ratios (ORs) using logistic regression models. There were no missing values for DII, and only 108 for HgbA1c. Missing values were addressed by the assumption of missing at random. All tests were 2 tailed, and P-values less than .05 were considered statistically significant.