Exploring the link between dietary inflammatory index and NAFLD through a structural equation modeling approach

Abstract

Nonalcoholic fatty liver disease (NAFLD) or metabolic dysfunction–associated steatotic liver disease (MASLD) is a significant global public health dilemma with wide-ranging social and economic implications. Diet and lifestyle modifications remain essential components of NAFLD management. The current study investigated the association between diet-related inflammation and NAFLD among 3110 Iranian adults participating in the Amol Cohort Study (AmolCS), employing the Structural Equation Modeling (SEM) approach.The inflammatory potential of the diet was quantified using an energy-adjusted dietary index (E-DII) score. Findings showed that in the total sample and separately in males, the E-DII score had a significant effect on NAFLD, with mediation through hypertension (β_standardized = 0.16, and 0.13, p < 0.001, respectively) and c-reactive protein (CRP) (β_standardized = 0.07, and 0.07, p < 0.001, respectively). In the total sample and separately in females, the E-DII score significantly affected NAFLD, with mediation through diabetes (β_standardized = 0.06, p < 0.001, and 0.07, p = 0.006, respectively). In full and both gender-specific models, dyslipidemia was a risk factor for NAFLD and partially mediated the effect of hypertension on NAFLD.

The current study concluded a mediated association between dietary inflammation and NAFLD through hypertension, CRP, diabetes, and dyslipidemia, suggesting further longitudinal studies, especially in high-risk populations. These findings underscore the complex interplay between diet, inflammation, and NAFLD in Iranian adults.

Supplementary Information

The online version contains supplementary material available at 10.1186/s41043-024-00721-1.

Introduction

Nonalcoholic fatty liver disease (NAFLD) or metabolic dysfunction-associated steatotic liver disease (MASLD) presents a significant global public health dilemma, estimated to affect a quarter of the adult population worldwide and leading to a considerable burden of disease with extensive social and economic consequences [1]. In Iran, about one-third of the adult population (33%) suffers from some degree of NAFLD [2]. This prevalence is higher than the global average, estimated at around 29.8% [3]. NAFLD is linked to other highly prevalent non-communicable diseases (NCDs), including type 2 diabetes mellitus, metabolic syndrome, and cardiovascular diseases [4, 5], with a great deal of overlap in the public health and healthcare system strategies required for managing and preventing these health conditions [6]. Diet and lifestyle changes and managing underlying metabolic risk factors are the cornerstones of treatment for all patients [7, 8].

Diet plays a critical role in regulating chronic low-grade inflammation, with the Dietary Inflammatory Index (DII) as a key tool to assess the inflammatory potential of dietary patterns [9, 10]. While specific compounds’ pro- and anti-inflammatory capacities are well-documented, little is known about the impact of whole dietary patterns and foods on NAFLD [11–15]. Diets rich in inflammatory foods, including saturated fats, refined sugars, and processed ingredients, activate pro-inflammatory pathways, increasing cytokines like TNF-α and IL-6, which contribute to hepatic insulin resistance, lipid accumulation, and oxidative stress, central to NAFLD’s pathogenesis [16, 17]. Prior research has shown that diets with higher inflammatory potential significantly increase the risk of NAFLD and are linked to various metabolic disturbances such as obesity, hypertension, dyslipidemia, and diabetes [10, 18–22]. These findings underscore the critical role of dietary inflammation in the pathogenesis of NAFLD and its mediating factors.

However, most research has assumed linear relationships between DII, cardio-metabolic profiles, and NAFLD, overlooking potential non-linear connections [10, 18, 19, 23, 24]. This study uses structural equation modeling (SEM) to explore the complex pathways [25] linking DII, NAFLD, and related traits, aiming to integrate multidimensional data into a cohesive framework. Understanding these interactions could help mitigate NAFLD’s global impact. The study hypothesizes that diet modulates inflammatory mechanisms influencing NAFLD and suggests that results may guide interventional strategies to prevent and treat diet-related diseases.

Based on the research background, a hypothetical model was proposed to estimate the association between the E-DII and NAFLD and verify the effects of mediators (Fig. 1). This study aims to test two hypotheses:1) E-DII has a significant direct association with NAFLD (H1) [16]; 2) the association between E-DII and NAFLD is mediated by BMI, HPTN, CRP, dyslipidemia, and diabetes (H2) [10, 18–22] (Fig. 1).

Fig. 1.

Proposed hypothesized model of the association between DII and NAFLD

Methods

Setting and study populations

The Amol Cohort Study (AmolCS) is a population-based cohort study in Amol to assess the prevalence and incidence rates of cardiovascular diseases (CVDs) and obesity-related metabolic disorders, including NAFLD. Amol, a medium-sized city near the Caspian Sea, lies approximately 200 km northeast of Tehran, Iran. AmolCS has been conducted in two phases (Phase 1 in 2009–2010 and Phase 2 in 2016–2017) using a multistage sampling design. Data were collected by standardized interviews, clinical examination, and tests of biological samples. The information in the second phase, which included Food Frequency Questionnaire (FFQ) data, was used in the present population-based cross-sectional study. A total of 5147 adults aged ≥ 18 were enrolled in the second phase of AmolCS (2016–2017). The ethical committee of the Iran University of Medical Sciences (IUMS) approved all procedures conducted (Approval number: IR.IUMS.REC.1401.950). Signed informed consent was acquired from all involved in the study. Further details on the project are available in the previous studies [26, 27].

We excluded participants based on the following criteria: a history of elevated alcohol intake (≥ 30 g/d for males and ≥ 20 g/d for females), viral hepatitis, hepatotoxic drugs use, significant weight changes in the three months preceding the abdominal sonography used for diagnosing NAFLD (n 486); pregnant or lactating women (n = 153); under-or over-reporting the energy intake (total energy intake of < 800 or > 4200 kcal/d for males, and < 600 or > 3500 kcal/d for females) [28] (n = 228); a CRP level of 10 mg/l or more, which may indicate infection or medication use [29] (n = 110); missing nutrition data (n = 249); missing abdominal ultrasonography data (n = 166); and missing other variables (186). As a result, we obtained a final dataset of 3,110 participants (1,398 females and 1,712 males) with complete and evaluable data. The present investigation adhered to the principles outlined in the Declaration of Helsinki, with protocols concerning human participants being approved by the ethical committee of the Iran University of Medical Sciences (IUMS) under reference number No. IR.IUMS.REC.1401.950. All participants involved in the study provided written consent before their participation.

Dietary assessment

Dietary intake was assessed using a validated 168-item semi-quantitative food-frequency questionnaire (FFQ) documented in Phase 2. Reported portion sizes and consumption frequencies were converted into daily intakes, expressed in grams, using standard household measures [30]. Nutrient and energy calculations were based on the United States Department of Agriculture (USDA) Food Composition Table (FCT), supplemented by the Iranian FCT for traditional foods not included in the USDA database [31].

Calculation of the energy-adjusted dietary inflammatory index (E-DII™)

The Dietary Inflammatory Index (DII^®) was employed to assess the inflammatory potential of individuals’ dietary patterns across a spectrum from highly anti-inflammatory (described by the most negative score) to highly pro-inflammatory (defined by the most positive score). Previous literature has extensively elaborated on the DII methodology [32]. In the AmolCS, values for 32 food parameters were derived through a validated 168-item semi-quantitative FFQ [33]. Nutrient intake values were then calculated and standardized against a global reference database to create Z-scores. These standardized values were transformed into centered percentile scores multiplied by their respective inflammatory weights. These values were derived from scientific literature linking specific nutrients or food components to inflammatory biomarkers like CRP or IL-6. The sum of these weighted scores yielded the DII score for each individual [32]. In the current study, to minimize the variation induced by metabolic efficiency, physical activity levels, and body size [28], E-DII by the residual method was used. E-DII scores were determined through a density approach by adjusting DII for 1000 kcal intake [34].

Laboratory measurements

Venous blood samples were collected from participants the day before the interview, following a 12-hour fasting period, to quantify biochemical parameters. High-density lipoprotein (HDL), total cholesterol, and triglycerides were assessed by Biochemical analysis. Serum aminotransferases (AST, ALT, GGT, and ALP levels) were also quantified. The calculation of serum cholesterol levels for low-density lipoprotein (LDL) was derived from the Friedewald formula [35]. Serum concentrations of (non-hs) CRP were evaluated using the auto-analyzer and diagnostic kits. All biochemical assessments were conducted in a single laboratory facility by a pathologist blinded to the clinical findings of study participants. Standard laboratory procedures were followed, utilizing the BS-200 automatic analyzer (Mindray, China). Approximately 10% of blood samples were randomly selected and reassessed at the national reference laboratory to evaluate the accuracy of the laboratory tests. The spectrum of coefficients of variation observed across all laboratory measurements was between 1.7% and 3.8%. Pars Azmoun Pharmaceuticals, Tehran, provided the testing kits.

Definition of NAFLD diagnosed by ultrasonography

NAFLD was defined as fat accumulation in the liver, identified through imaging or liver biopsy after ruling out other evident causes of liver damage, such as excessive alcohol consumption, hepatotoxic medications, toxins, viral infections, or genetic liver disorders [36]. The current study used hepatorenal contrast (bright liver) and vascular blurring to define ultrasonographic fatty liver [37]. A solitary sonologist, unaware of the clinical condition of the individuals, performed all ultrasound scans using a 3.5 MHz convex probe (Esaote SpA, Genova, Italy). Criteria such as the obscuration of portal or hepatic veins and a notable increase in hepatic echogenicity were employed to confirm fatty liver.

Definition of diabetes

DM type 2 was defined by a history of DM or taking anti-hyperglycemic drugs (oral agents, insulin) or an FBS ≥ 125 mg/dl with no history of DM type I [38].

Measurement of other variables

A thorough anthropometric examination included height, weight, waist, and hip circumference were measured using World Health Organization guidelines, and body mass index (BMI) was calculated accordingly [39]. Smoking was defined as no smoking and past/current smoking. Physical activity was evaluated using a reliable International Physical Activity Questionnaire (IPAQ) [40]. The metabolic equivalent (MET) of physical activity energy expenditure was measured according to how long and strenuous each participant’s workout was. Following the operating procedure, blood pressure was measured utilizing a manual sphygmomanometer after 5 min of rest based on Korotkoff noise, and the average blood pressure derived from two measurement readings was used [41]. The monitors underwent calibration, and the cuffs were fitted correctly. Blood pressure was evaluated according to the recommendations in the 7th report issued by the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure [42, 43].

Statistical analysis

The data’s normality assessment was conducted using the Kolmogorov–Smirnov test. Mean ± SD represented continuous-distributed variables. We used frequency or percentage for categorical variables and the χ² test to compare groups. A structural equation modeling (SEM) approach was employed to investigate the direct and indirect association of the dietary inflammatory index (DII) with NAFLD. The fundamental theoretical framework of SEM is depicted in Fig. 1, with NAFLD as the endogenous variable and DII as the exogenous variable. Hypertension (SBP and DBP combined), BMI, dyslipidemia (LDL, HDL, TG, and Cholesterol combined), CRP, age, and physical activity were considered key confounding and mediating variables interconnected with NAFLD. Confirmatory factor analysis (CFA) was utilized to validate the dyslipidemia measurement model by examining the relationship between the observed variable and its latent construct.

Model fitness was approached through the application of standard criteria encompassed the requirement of root-mean-square error of approximation (RMSEA) values lower than 0.08 and comparative fit index (CFI) values, goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), tucket lewis index (TLI) surpassing 0.90, and standardized root mean square residual (SRMSR) values lower than 0.05. Moreover, the maximum-likelihood method was employed to gauge the adequacy of the model, along with the presentation of standardized path coefficients and their respective statistical significance [44–46]. Within the realm of Structural Equation Modeling (SEM), the utilization of standardized estimates becomes imperative due to the disparate units in which each variable is gauged, thereby enhancing the interpretability and comparability of outcomes. The statistical data analysis was conducted using SPSS 24.0 and IBM SPSS Amos 24.0 (version 24.0; Amos Development Corp., Meadville, PA, USA) Statistical Software Packages. SPSS 24.0 facilitated a descriptive statistical analysis of the dataset, while SPSS Amos 24.0 was instrumental in examining structural equation modeling (SEM). All statistical tests followed a two-tailed approach, with statistical significance attributed to p-values below 0.05.

Results

Participant characteristics

A total of 3110 participants had evaluable data for these analyses, including 1398 females (44.95%). There were 1452 (46.68%) participants with NAFLD. The characteristics of these participants overall, and stratified by sex and NAFLD, are shown in Table 1. Compared with females, males were older (P < 0.001), more active (P < 0.001), smokers (P < 0.001), and were more likely to have a family history of diabetes (P = 0.01) and hypertension (P < 0.001). Furthermore, male participants presented statistically significant higher levels of WC, WHR, TG, SBP, DBP, and liver enzymes (AST, ALT, GGT) compared with females (all P < 0.001). For the other anthropometric, cardiometabolic, and liver parameters evaluated (Table 1), values were significantly greater in females than males (all P < 0.001 except P = 0.003 for SBP). The range of E-DII scores was − 4.13 to 4.45 (− 4.08 to 4.33 and − 4.13 to 4.45 in females and males, respectively), and correspondingly, their mean (SD) of dietary energy density (DED) was 1.46 (0.41) and 1.56 (0.51), respectively.

Table 1.

Baseline characteristics of the study adult participants by gender and NAFLD (n = 3110), AmolCS, 2016–2017

Characteristic	Overall	Women (n = 1398)	Men (n = 1712)		Non NAFLD (n = 1658)	NAFLD (n = 1452)
	N (%)	N (%)	N (%)	P value	N (%)	N (%)	P value
Family history of diseases
Family history of Diabetes	1122(36.1)	536(38.3)	586(34.2)	0.01	572(34.5)	550(37.9)	0.02
Family history of CVDs	217(7.0)	104(7.4)	113(6.6)	0.20	113(6. 8)	104(7.2)	0.37
Family history of HPTN	1395(44.9)	744(53.2)	651(38.0)	< 0.001	660(39.8)	735(50.6)	< 0.001
Residual areas
Rural	1366(43.9)	506(36.2)	860(50.2)	< 0.001	725(56.3)	641(44.1)	0.42
Urban	1744(56.1)	892(63.8)	852(49.8)		933(56.3)	811(55.9)
Comorbidities
Diabetes	469(15.1)	276(19.7)	193(11.3)	< 0.001	168(10.1)	301(20.7)	< 0.001
Metabolic syndrome	865(27.8)	511(36.6)	354(20.7)	< 0.001	232(14.0)	633(43.6)	< 0.001
Heart disease	142(4.6)	50(3.6)	92(5.4)	0.01	77(4.6)	65(4.5)	0.44
Lifestyle factors
Lowering serum glucose agent’s user	181(5.8)	90(6.4)	91(5.3)	0.10	70(4.2)	111(7.6)	< 0.001
Lowering serum lipid agent’s user	366(11.8)	197(14.1)	169(9.9)	0.001	167(10.1)	199(13.7)	0.001
Lowering hypertension agent’s user	574(18.5)	297(21.2)	277(16.2)	< 0.001	266(16.0)	308(21.2)	< 0.001
Past/Current smoking	443(14.2)	8(0.6)	435(25.4)	< 0.001	259(15.6)	184(12.7)	0.01
Alcohol drinker (≥ 40 g)	178(5.7)	5(0.4)	178(5.7)	< 0.001	85(5.1)	93(6.4)	0.16
Physical activity (MET-h/d)
Low	1069(34.4)	525(37.6)	544(31.8)	< 0.001	538(32.4)	531(36.6)	0.03
Moderate	1139(36.6)	534(38.2)	605(35.3)		615(37.1)	524(36.1)
high	902(29.0)	339(24.2)	563(32.9)		505(30.5)	397(27.3)
	Mean (SD)	Mean (SD)	Mean (SD)	P value	Mean (SD)	Mean (SD)	P value
Age (years)	46.97 ± 14.61	45.83 ± 14.03	47.91 ± 15.01	< 0.001	45.58 ± 16.11	48.56 ± 12.50	< 0.001
Anthropometric parameters
Body mass index (BMI, kg/m2)	28.04 ± 5.05	29.55 ± 5.37	26.81 ± 4.41	< 0.001	25.89 ± 4.45	30.50 ± 4.55	< 0.001
Waist circumference (WC, cm)	88.82 ± 11.34	87.70 ± 12.14	89.57 ± 10.57	< 0.001	83.58 ± 10.12	94.79 ± 9.56	< 0.001
Waist to height (WHtR) ratio	0.54 ± 0.07	0.56 ± 0.08	0.52 ± 0.06	< 0.001	0.51 ± 0.06	0.58 ± 0.06	< 0.001
Waist to hip (WHR) ratio	0.87 ± 0.08	0.84 ± 0.08	0.89 ± 0.07	< 0.001	0.85 ± 0.08	0.90 ± 0.07	< 0.001
Cardiometabolic parameters
Serum triglycerides (TG, mg/dl)	134.86 ± 91.21	129.07 ± 89.21	139.57 ± 92.57	0.001	112.52 ± 66.72	160.38 ± 107.33	< 0.001
Serum total cholesterol (T-C, mg/dl)	180.70 ± 40.28	183.41 ± 41.86	178.48 ± 38.81	0.001	175.85 ± 38.01	186.23 ± 41.01	< 0.001
Serum high-density lipoprotein (HDL, mg/dl)	43.67 ± 11.69	46.04 ± 11.85	41.74 ± 11.20	< 0.001	44.72 ± 11.75	42.47 ± 11.51	< 0.001
Serum low-density lipoprotein (LDL, mg/dl)	99.30 ± 26.29	99.65 ± 26.73	99.01 ± 25.94	0.49	97.60 ± 26.48	101.24 ± 25.95	< 0.001
Systolic blood pressure (SBP, mmHg)	114.86 ± 19.27	113.71 ± 20.64	115.80 ± 18.03	0.003	111.21 ± 18.36	119.03 ± 19.44	< 0.001
Diastolic blood pressure (DBP, mmHg)	71.69 ± 11.85	70.91 ± 12.33	72.32 ± 11.42	0.001	69.02 ± 11.14	74.72 ± 11.92	< 0.001
Fasting blood glucose (FBS, mg/dl)	106.23 ± 35.39	108.98 ± 40.73	103.98 ± 30.18	< 0.001	101.21 ± 31.33	111.96 ± 38.75	< 0.001
Serum C-reactive protein (CRP, mg/dl)	3.51 ± 6.78	4.36 ± 7.42	2.83 ± 6.12	< 0.001	2.94 ± 6.30	4.17 ± 7.24	< 0.001
Liver parameters
Alanine transaminase (ALT, mg/dl)	24.01 ± 17.58	19.87 ± 14.04	27.40 ± 19.37	< 0.001	19.87 ± 13.05	28.74 ± 20.64	< 0.001
Aspartate transaminase (AST, mg/dl)	21.52 ± 9.19	19.50 ± 8.29	23.17 ± 9.55	< 0.001	20.52 ± 8.69	22.67 ± 9.59	< 0.001
Gamma-glutamyl-transferase (GGT, mg/dl)	27.19 ± 19.24	24.27 ± 18.74	29.57 ± 19.31	< 0.001	23.98 ± 18.23	30.85 ± 19.70	< 0.001
Alkaline phosphatase (AP, mg/dl)	197.88 ± 59.58	196.33 ± 67.53	199.15 ± 52.18	0.18	193.08 ± 61.33	203. 36 ± 57.04	< 0.001
Dietary parameters
energy-adjusted dietary index (E-DII) score	-0.79 ± 1.18	-0.83 ± 1.17	-0.76 ± 1.18	0.13	-0.75 ± 1.18	-0.83 ± 1.17	0.06
Dietary energy density (DED)	1.52 ± 0.47	1.46 ± 0.41	1.56 ± 0.51	< 0.001	1.52 ± 0.49	1.51 ± 0.45	0.45

Data are Mean ± SD (all such values) unless indicated. Significant at P < 0.05 for independent t-test for continuous variables and chi-square test for categorical variables

Confirmatory factor analysis

The fit indices for CFA used to develop the constructs of dyslipidemia (combination of LDL, HDL, TG, and Cholesterol) had acceptable fit to the data (χ²/df = 12.144, p < 0.001, RMSEA = 0.060, SRMR = 0.038, GFI = 0.981, AGFI = 0.952, CFI = 0.960) (see supplementary Fig. 1). LDL explained a higher proportion of the variance in the dyslipidemia factor.

Association between E-DII and NAFLD

The SEM output of the Model in the three sub-samples (overall population, males, and females) is presented in Fig. 2, which depicts the relationships between the E-DII and NAFLD while considering the main confounders and mediators.

Fig. 2.

Final structural model of the relationship between energy adjusted dietary inflammatory index (E-DII) and NAFLD. The full-model fit indices: χ2/df = 12.144, p < 0.001, GFI = 0.981, AGFI = 0.952, CFI = 0.960, TLI = 0.920, RMSEA = 0.060, SRMR = 0.038. Note: The values on the paths represent standardized regression coefficients. The upper-faced numbers refer to the full sample, whereas the numerical values below refer to male and female sub-samples. Arrows in bold represent statistically significant associations (*P < 0.05; ** P < 0.01; *** P < 0.001). Square dot arrows refer to statistically significant associations in both total sample and males sub-sample, whereas, dash arrows refer to statistically significant associations in both the total sample and the female sub-sample

E-DII had no significant direct effect on NAFLD in all three models (overall model: β_standardized = 0.024, P = 0.234, 95% CI [0.034, 0.15]; males model: β_standardized = 0.049, P = 0.069, 95% CI [0.020, 0.069]; females model: β_standardized = -0.017, P = 0.582, 95% CI [-0.039, 0.023]). Therefore, hypothesis 1 of the present study was not supported.

However, hypothesis 2 was supported in the current study. E-DII score had a significant effect on NAFLD, which was mediated through hypertension (in both sexes: of E-DII to hypertension: ( β_standardized = 0.07, P = 0.04; and of hypertension to NAFLD: β_standardized = 0.16, p < 0.001; in males: of E-DII to hypertension: β_standardized = 0.08, P = 0.004; of hypertension to NAFLD: β_standardized = 0.13, p < 0.001), CRP (in both sexes: of E-DII to CRP: β_standardized = 0.07, p < 0.001; of CRP to NAFLD: β_standardized = 0.07, p < 0.001; in males: of E-DII to CRP: β_standardized = 0.08, P = 0.008; and of CRP to NAFLD: β_standardized = 0.07, p = 0.003), and diabetes (in both sexes: of E-DII to diabetes: β_standardized = 0.06, p < 0.001; of diabetes to NAFLD: β_standardized = 0.09, p < 0.001; in females: of E-DII to diabetes: β_standardized = 0.07, P = 0.006; and of diabetes to NAFLD: β_standardized = 0.08, p = 0.002). For females, results for both the hypertension and CRP were null. The effect of E-DII on diabetes was also ineffective in men.

In full and both gender-specific models, dyslipidemia was a risk factor for NAFLD (in both sexes combined: β_standardized = 0.11, p < 0.001; in males: β_standardized = 0.13, p < 0.001, and in females: β_standardized = 0.06, p = 0.034), and partially mediated the effect of hypertension on NAFLD (in both sexes combined: β_standardized = 0.09, p < 0.001; in males: β_standardized = 0.09, p < 0.001, and in females: β_standardized = 0.08, p = 0.014).

The goodness-of-fit metrics presented evidence that the models (full sample, males and females) perfectly fit the data: full model, χ²/df = 12.144, p < 0.001, RMSEA = 0.060, SRMR = 0.038, GFI = 0.981, AGFI = 0.952, CFI = 0.960, and TLI = 0.920. Males model, χ²/df = 7.053, p < 0.001, RMSEA = 0.057, SRMR = 0.043, GFI = 0.976, AGFI = 0.942, CFI = 0.958, TLI = 0.915. Females model, χ²/df = 7.905, p < 0.001, RMSEA = 0.066, SRMR = 0.043, GFI = 0.975, AGFI = 0.940, CFI = 0.951, TLI = 0.910.

Discussion

In this study, no direct association was observed between E-DII and NAFLD. To further understand the association between them, we proposed a hypothesis to evaluate whether some cardio-metabolic and inflammatory factors play a role as mediators. The results of our study supported the mediation hypothesis that E-DII through hypertension, CRP, dyslipidemia, and diabetes is associated with NAFLD evaluated by ultrasound.

The link between DII and NAFLD has been confirmed by growing evidence from both observational and interventional studies suggesting that anti-inflammatory diets might have hepatoprotective effects that improve NAFLD severity [47–49]. Zhang et al., in the analysis based on nationwide survey data including 10,052 US adults, found a significant association between inflammatory potential diet and Fatty Liver Index (FLI) by showing that individuals in the highest DII quartile had a 52% increased likelihood of fatty liver than those in the lowest DII quartile (OR 1.52, 95% CI 1.27–1.83, P _trend <0.001) [12].

In our study, adherence to a pro-inflammatory diet was associated with a greater intake of energy, carbohydrates, saturated fat, MUFA, cholesterol, and total fat, as well as a low intake of fiber, vitamins (except for Vitamin B-12), and minerals. Furthermore, a pro-inflammatory diet was also associated with low intakes of caffeine, garlic, pepper, onion, and green/black tea (see supplementary Table 1).

Polyphenol-rich foods such as coffee, tea, fruits, and vegetables have been shown to have anti-inflammatory effects [34]. A long-term diet that prioritizes plant-based foods, low-fat dairy, and fish while avoiding processed meats and sugar-sweetened beverages can prevent intestinal inflammatory processes via the gut microbiome [50]. Furthermore, minerals such as magnesium can regulate enzymes involved in liver lipid metabolism and may improve chronic metabolic disorders by reducing low-grade inflammation [51, 52]. An elevated risk of NAFLD has already been linked to a lack of adherence to these recommendations [11–13, 49].

The present study is the first to explore the association between E-DII and NAFLD using the SEM approach. Structural equation modeling has evolved to allow for non-linear relations and multiple pathways between DII and NAFLD [25]. The study found that the E-DII affects NAFLD through hypertension, CRP, and dyslipidemia in the total sample and males. Additionally, diabetes mediated the relationship between E-DII and NAFLD in the total sample and among females. Evidence from the studies has suggested the correlation between hypertension and hepatic steatosis [53, 54] and confirmed that the increase in hypertension could predict the occurrence and development of NAFLD [55, 56]. Another study conducted by Zhang et al. revealed that the level of DII was significantly positively correlated with SBP and DBP in US adults [57], which is consistent with our findings. Studies indicated that hypertension contributes to oxidative stress, systemic inflammation, and endothelial dysfunction, which promote hepatic insulin resistance and fat accumulation [58]. Hypertension can also exacerbate liver fibrosis by increasing renin-angiotensin system activity, leading to profibrotic signaling [59, 60]. This pathway highlights the role of hypertension as both a contributor to and a potential mediator in the progression of NAFLD.

The underlying mechanisms of the association between E-DII and NAFLD through hypertension, diabetes, and dyslipidemia might be explained by the fact that gut microbial metabolites like trimethylamine N-oxide and short-chain fatty acids act on downstream cellular targets contribute to hypertension [61–64]. Dietary contents affect the composition and function of the gut microbiota, and previous research has also confirmed the cardio-metabolic disease risk-reducing effects of anti-inflammatory dietary patterns by shaping the gut microbiome [63, 65, 66]. Disturbances in lipid metabolism, insulin resistance, and gut-derived endotoxins contribute to producing and releasing pro-inflammatory cytokines, inhibiting insulin receptor signaling and leading to steatosis and fibrosis [67, 68]. Evidence supports an inverse association between DII and lipid profiles [69, 70]. Pro-inflammatory diets may activate the NF-KB pathways, simulate TG production in the liver, and alter the balance of the lipid profile [71].

CRP’s role as a mediator is consistent with its ability to act as a regulator of nitric oxide production in the endothelium and coordinate the production and secretion of various cytokines, increasing the pro-inflammatory activity of different adipokines. MetS was associated with oxidative stress and chronic low-grade inflammation [72], and NAFLD is one criterion of MetS [73].

The association of E-DII with NAFLD through hypertension, CRP, and dyslipidemia was not observed in females, which means other determinants might be involved. Previous studies have emphasized that women are less affected by environmental stress (such as oxidative stress) than men [74]. Estrogen has antioxidant and anti-inflammatory properties in females, which may significantly contribute to this difference [75]. Therefore, eating habits with inflammatory potential may have less impact on the development of NAFLD in females. Our results emphasize the importance of gender-specific interventions to control NAFLD and also the pivotal role of anti-inflammatory diets in preventing hypertension, low-grade systemic inflammation, and dyslipidemia among high-risk individuals.

As far as we know, the present study is the first to investigate the relationship between DII and NAFLD through the SEM technique. One of its significant strengths is the temporal alignment of dietary data collection, blood sampling, and sonographic findings, which greatly enhances causal inference [76]. Despite its strengths and novelty, the present findings should be interpreted in the context of several limitations. First, the potential issue of reverse causation could arise due to the cross-sectional design employed in the study. The second limitation of the study is the single assessment of inflammatory markers (CRP) as risk factors. However, CRP is the most commonly measured biomarker of systemic inflammation in research examining the relationship between the dietary inflammatory index and various diseases [77]. This could be attributed to the extended half-life of CRP in the serum compared to other low-grade inflammatory biomarkers [78]. In addition, our study was limited to the population of the north of Iran. Finally, this study used liver ultrasonography, associated with lower accuracy than liver biopsy, to identify NAFLD, which may cause the misclassification. However, ultrasonography is considered a safe, accurate, and convenient tool for evaluating NAFLD in epidemiological surveys [79].

Conclusions

Our findings suggest that hypertension, CRP, and dyslipidemia mediate the association between dietary inflammatory potential and NAFLD. Therefore, a pro-inflammatory diet may exacerbate the effects of hypertension, CRP, and dyslipidemia and, as a result, may increase the risk of NAFLD. Altogether, these results demonstrate the importance of constantly assessing the dietary intake in patients with hypertension and dyslipidemia to improve diet quality and reduce its complication rate in the population. However, substantiating the efficacy of the DII requires further longitudinal studies, particularly among populations at high risk of chronic diseases.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

The research idea and structure were created by FZ, HA, and AD. NM and AD had full access to the data and ensured analysis accuracy. MM, EG, and BA contributed to gathering the data. The data was analyzed and interpreted by AD, SE, NM, JRH, and SP. The initial manuscript draft was written by AD, BA, and SE. All authors reviewed the manuscript for important content and approved the final version. HA is responsible for overseeing the entire paper as the guarantor.

Funding

The study was financed by the Iran University of Medical Sciences (IUMS) with grant number 1401-3-75-22501. Dr. Hébert received backing from grant U01 CA272977-01 from the United States National Cancer Institute. The funders were not involved in designing the study, collecting and analyzing data, interpreting results, or writing the report.

Data availability

The data gathered and analyzed in this study can be obtained from the corresponding author upon reasonable request. Interested researchers may reach out to the corresponding author, Dr. Hossein Ajdarkosh, email address: ajdarkosh.h@iums.ac.ir.

Declarations

Ethics approval and consent to participate

The investigation followed the guidelines specified in the Declaration of Helsinki and received approval from the ethical committee of the Iran University of Medical Sciences (IUMS) with the reference code IR.IUMS.REC.1401.950. Each individual involved in the study provided written consent after receiving comprehensive information regarding the study’s aims and procedures.

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.

Disclosure

Dr. James R. Hébert is the primary shareholder of Connecting Health Innovations LLC (CHI), which has secured authorization for his innovation, the dietary inflammatory index (DII^®), from the University of South Carolina. CHI plans to use this invention to create computer and smartphone applications for patient counseling and dietary intervention in clinical settings. CHI also holds the exclusive rights to the E-DII™. The content of this paper is not related to CHI’s work, and CHI has not influenced this project in any way.