Meta-analysis of the association between dietary inflammation index and C-reactive protein level

Abstract

Background:

There have been various clinical studies on the effect of dietary inflammatory index (DII) on circulating inflammatory biomarkers, but the findings from these are contradictory. The aim of the present study was to clarify any association.

Methods:

The PubMed, Embase, Web of Science and Cochrane Library database were searched for relevant studies from inception February 2021. There were no language restrictions. Two investigators independently selected eligible studies. Measures of association were pooled by using an inverse-variance weighted random-effects model. The heterogeneity among studies was examined using the I² index. Publication bias, sensitivity and subgroup analyses were also performed.

Results:

A total of 13 cross-sectional studies were identified, involving 54,813 participants. The adjusted pooled OR of C-reactive protein (CRP) levels for the highest (the most pro-inflammatory diet) versus lowest (the most anti-inflammatory diet) DII categories was 1.25 (95% CI: 1.18–1.32; I² = 59.4%, P = .002). Subgroup analyses suggested the main source of study heterogeneity was the geographic area (Asia, Europe, or USA) and CRP levels (>3 mg/L or others). This finding was remarkably robust in the sensitivity analysis.

Conclusion:

The meta-analysis suggests that more pro-inflammatory DII scores were positively associated with CRP, the DII scores can be useful to assess the diet inflammatory properties and its association with low-grade inflammation.

1. Introduction

Low-grade chronic systemic inflammation, has a complex and multifocal etiology, and as indicated by the continuous presence of serum inflammatory mediators such as C-reactive protein (CRP), fibrinogen, and various interleukins.^[1] Elevated levels of CRP were associated with an increased risk of cancer and as a predictor of all-cause mortality in many studies.^[2–4] Diet is well known to play a key role in regulating chronic inflammation^[5] that is involved in the onset and progression of most chronic diseases.^[6] Many studies have indicated that dietary components modulate inflammatory status, the consumption of a variety of vegetables is inversely associated with lower CRP,^[7] the Mediterranean diet rich in fruits and vegetables, has been associated with lower inflammation levels,^[8] On the other hand, The Western dietary pattern high intakes of red and processed meat, refined grains and high in fat, has been associated with increased markers of inflammation.^[9] The Dietary Inflammatory Index (DII) was developed to measure the diet inflammatory power, scoring individuals’ diets from anti- to a pro-inflammatory.^[10] The DII has been used in several studies to predict mortality and diseases, such as cancer^[11,12], cardiovascular disease^[13] and metabolic syndrome.^[14,15] Previously, the DII has been shown to predict levels of inflammatory markers, such as CRP, interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α).^[16–18]

Recently, many studies have assessed the relationship between the DII and CRP levels. However, the results remain inconsistent and controversial. Elevated DII has been shown to be associated with CRP levels in some,^[19,20] but not all^[21,22] studies. These conflicting findings might be due to different geographic area or dietary assessment tools and so on. Therefore, we conducted this meta-analysis to systematically assess the association between DII and CRP levels.

2. Materials and methods

2.1. Ethics statement

As all analyses were based on previously published studies, and no ethical approval or patient consent was required.

2.2. Search strategy

PubMed, Embase, Cochrane Library and Web of Science are going to be searched for studies published up to February 2021, and with using the keywords of (“C-reactive protein” OR “C-reactive protein” OR CRP) AND (inflammatory potential of diet OR DII OR anti-inflammatory diet OR pro-inflammatory diet).

Reference lists of reviews are also manually searched. No language restrictions.

2.3. Study selection

Studies meeting the following inclusion and exclusion criteria were applied.

Inclusion: all observational studies (cross-sectional, case–control and prospective) that assessed the association between dietary inflammation index and CRP level; those provided the multivariable-adjusted RR, hazard ratio (HR), or odds ratio (OR) with corresponding 95%confidence intervals (CI) of CRP level.

Exclusion: studies that did not investigate the association between DII and CRP level; reviews, case-reports, protocols, short-communications, personal opinions, letters, posters, conference abstracts, and laboratory research (in vivo and in vitro studies); adolescents.

2.4. Data extraction and quality assessment

From each article, the following data was extracted in standard format: first author surname, publication year, country of the study origin, study participants, study design, sample sizes, gender, age range or mean age, dietary assessment tool, DII types, CRP levels, most fully adjusted risk estimate, and adjustment for confounding factors in the statistical analysis.

Study selection, extraction of study characteristics and quality assessment were independently performed by 2 reviewers (Hua and Liang). The selection process was performed in 2 phases. Phase-1 2 blinded reviewers (Hua and Liang) screened the title and abstracts of all identified references. Phase-2, the same 2 reviewers applied the eligibility criteria to full-text articles, any disagreements were mutually discussed, and if necessary, a third reviewer was involved (Yang) to make a final decision. The methodological quality of the included studies was evaluated using a 9-star Newcastle-Ottawa Quality Assessment Scale (NOS).^[23] This scale judges a study quality based on selection, comparability, and ascertaining of outcome. A study achieving 7 or more stars was considered to be high quality.

2.5. Statistical analysis

The multivariate-adjusted risk estimates were selected if they were reported in the original publication. ORs and 95%CI were considered as the effect size for all studies. We pooled OR estimates for the highest versus the lowest DII score. The heterogeneity among studies was assessed using Cochrane Q and I-squared (I²) statistic, defining a significant heterogeneity as Cochrane Q < 0.10 and/or I² > 50%. The fixed-effects model was selected when there is no significant heterogeneity was observed; otherwise, the random-effects model was applied. Subgroup analyses were conducted by dietary assessment tool (food frequency questionnaire, 24-h dietary records or 7 days dietary records), geographic area (Asia, Europe, or USA), CRP levels (>3 mg/L or others) and DII types (DII, energy-adjusted DII or alternate-DII). Publication bias was evaluated using Egger and Begg tests with visual inspection of funnel plots.^[24,25] The number of missing studies and the effect that these studies might have had on the outcome was explored by using nonparametric rank-based data augmentation techniques (trim-and-fill procedure) developed by Duval and Tweedie^[26] A sensitivity analysis was conducted by removing individual studies each time to analyze the robustness of the pooling risk estimate. All statistical analyses were carried out in STATA version 12.0 (Stata Corp, College Station, TX).

3. Results

3.1. Search results

A total of 13 eligible studies^{[19–22,27–35]} from 2259 relevant articles were identified in this meta-analysis. There were 1582 articles left after duplicates removed. Through title and abstract scanning, 1543 articles were excluded because of the following reasons: irrelevant studies (n = 1501), conference abstract (n = 13), review articles (n = 23), RCT design (n = 6). We reviewed the full texts of the remaining articles and 26 articles were excluded because of the following reasons: incomplete data (n = 22), different markers (n = 2), adolescents (n = 1) and not reported adjusted OR (n = 1). Finally, 13 eligible studies were included in this meta-analysis. Flow chart of the study selection is presented in Figure 1.

Figure 1.

Flow chart of the study selection.

3.2. Studies characteristics

The detailed characteristics of the 13 studies are showed in Table 1. All of these studies were cross-sectional studies and published from 2014 to 2020, including 54813 individuals in total. Three studies were conducted in Asia,^[28,30,33] 3 studies in Europe^[19,22,27]and 7 studies in USA.^{[20,21,29,31,32,34,35]} NOS of 13 studies ranged from 4 to 8 stars and a mean score was 6.38, suggesting moderate methodological quality.

Table 1.

Characteristics of studies included in the meta-analysis.

Author/yr	Country	Study participants	Sample size,n	% Female	Mean or age range	Dietary assessment tool	DII types	OR (95%CI) multivariate	Adjustment confounders	CRP levels	NOS stars
Corley et al.[27]	United Kingdom	Older adults	928	51.5	70.00	FFQ	E-DII	1.12 (1.02–1.24)	age, sex, smoking, BMI, physical activity, and hypercholesterolemia	>3 mg/L	8
Julia et al.[19]	France	Health	226	NR	NR	24-h dietary records	DII	1.32 (0.89–1.95)	sex, baseline age, educational level, baseline smoking status, baseline physical activity, energy intake and number of dietary records available, BMI.	>3 mg/L	7
							A-DII	1.33 (0.94–1.88)
Na et al.[28]	Korean	Health	28086	70.0	40–79	24-h dietary records	DII	1.24 (1.07–1.43)	age, sex, BMI, smoking status, education level, blood pressure status, total calorie intake, use of oral contraceptive and physical activity lipid-lowering medication	>3 mg/L	6
Phillips et al.[29]	USA	Health	1992	NR	50–69	FFQ	E-DII	1.35 (1.05–1.72)	age, gender, BMI, physical activity, smoking status, alcohol consumption and use of anti-inflammatory and	NR	6
Shivappa et al.[22]	Belgium	Health	2487	52.0	35–55	FFQ	DII	1.03 (0.86–1.17)	energy, age, sex, BMI, smoking status, education level, use of non-steroidal anti-inflammatory drugs, blood pressure, use of oral contraceptives, anti-hypertensive therapy, lipid-lowering drugs and physical activity	>3 mg/L	7
Shin et al.[30]	Korean	Adults	3014	57.0	43.00	24-h dietary records	DII	1.70 (1.07–2.69)	age, sex, education, marital status, alcohol consumption, smoking status, BMI, high-density lipoprotein cholesterol, and physical activity	>2 mg/L	8
Shivappa et al.[20]	USA	Health	495	46.7	49.00	24-h dietary records	DII	1.47 (1.03–2.12)	MET (metabolic equivalents of task), gender, light season, race, marital status, serum total cholesterol, employment status, anti-inflammatory medication use, alcohol status and herbal supplement use	>3 mg/L	6
			559	47.8		7 d dietary records		1.61 (1.15–2.27)
Shivappa et al.[36]	USA	Adults	7215	52.5	49.18 ± 17.88	24-h dietary records	DII	1.53 (1.20–1.95)	age, sex, ethnicity, BMI, education, smoking, poverty index and physical activity	>3 mg/L	6
Shivappa et al.[32]	USA	Adults	5292	51.1	34.40 ± 21.78	24-h dietary records	E-DII	1.81 (1.42–2.31)	age, sex, ethnicity, BMI and poverty index	>3 mg/L	6
Suzuki et al.[33]	Japan	Health	1176	44.7	63.40	FFQ	DII	1.32 (1.01–2.52)	sex, age, smoking habits, drinking habits, history of hypertension, total energy intake, and BMI.	>1 mg/L	8
Tabung et al.[21]	USA	Postmenopausal women	2567	100.0	50–79	FFQ	DII	1.30 (0.97–1.67)	age, BMI, race, educational level, smoking status, physical activity, inflammation-related comorbidity, regular use of antidepressants, statins, and NSAIDs	>3 mg/L	7
Wirth et al.[35]	USA	African-American churchgoers	329	79.0	54.8 ± 11.4	FFQ	DII	3.17 (1.52–6.62)	gender, insurance, perceived health, age, and the Multigroup Ethnic Identification Measure	>3 mg/L	4
Wirth et al.[34]	USA	Police Officers	447	25.0	42.4 ± 8.5	FFQ	DII	1.57 (0.85–2.88)	age, education, and sleep quality	>3 mg/L	4

95%CI = 95% confidence internal, A-DII = alternate-DII, BMI = body mass index, CRP = C-reactive protein, DII = dietary inflammation index, E-DII = energy-adjusted DII, FFQ = food frequency questionnaire, NOS = Newcastle-Ottawa Quality Assessment Scale, NR = not reported, OR = odds ratio.

3.3. Association between DII and CRP level

The adjusted pooled OR of CRP level for the highest (the most pro-inflammatory diet) versus lowest (the most anti-inflammatory diet) DII categories was 1.25 (95% CI: 1.18–1.32) in a random effect model. Meanwhile, significant heterogeneity between studies was revealed (I² = 59.4%, P = .002) (Fig. 2).

Figure 2.

Forest plots of associations between DII and C-reactive protein level. DII = dietary inflammatory index.

3.4. Subgroup meta-analysis

To further elaborate on the significant study heterogeneity observed, subgroup analyses was performed, whereby studies were stratified based on: dietary assessment tool, geographic area, CRP levels and DII types (Table 2). Subgroup analysis suggested the source of potential study heterogeneity may be applicable to the geographic area and CRP levels.

Table 2.

Subgroup analysis.

Subgroup	Number of studies	Odds ratio (95%CI)	I2 (%)	P heterogeneity
Dietary assessment tool
FFQ	7	1.24 (1.07–1.43)	54.9	.039
24-h dietary records	7	1.44 (1.27–1.63)	28.6	.210
7 d dietary records	1	1.61 (1.15–2.26)	-	-
Geographic area
Asian	3	1.28 (1.12–1.46)	0.0	.437
Europe	4	1.11 (1.03–1.20)	0.0	.435
USA	8	1.54 (1.36–1.73)	13.7	.323
CRP levels
>3 mg/L	12	1.36 (1.20–1.54)	65.9	.001
Others	3	1.40 (1.15–1.71)	0.0	.661
DII types
DII	11	1.37 (1.20–1.56)	50.0	.029
E-DII	3	1.38 (1.03–1.84)	85.4	.001
A-DII	1	1.33 (0.94–1.88)	-	-

3.5. Sensitivity and publication bias analysis

The sensitivity analysis indicated that the pooled ORs were not obviously influenced by any single study (Fig. 3), suggesting that the results of this meta-analysis are stable.

Figure 3.

Sensitivity analysis was performed by omitting one study each time and recalculating the pooled OR estimates. OR = odds ratio.

In addition, Egger test revealed significant publication bias (P = .005). The funnel plot also indicated evidence of publication bias. After imputing 8 missing studies by using the trim-and-fill method, the recalculated pooled ORs were attenuated, but not substantially changed from the initial estimates (imputed OR (95% CI),1.36 (1.22–1.52); P = .001) (Fig. 4).

Figure 4.

Funnel plots of DII and C-reactive protein level using the trim-and-fill method. The circles alone are real studies and the circles enclosed in boxes are “filled” studies. DII = dietary inflammatory index.

4. Discussion

To our knowledge, the present study is the first meta-analysis summarizing the independent positive association between DII scores with CRP levels. In this meta-analysis, as a result of collecting multivariable logistic regression analysis to analyze the relationship between DII and CRP levels, it was found that the highest category of DII (the most pro-inflammatory diet) was independently associated with higher CRP levels when compared to the lowest category (the most anti-inflammatory diet).

Diet represents a complex set of exposures that often interact, and cumulative effects may modify both inflammatory responses and health outcomes.^[36] The DII was developed and refined to quantify the inflammatory potential of individual diets, and it was created based on the literature to assess each food having a positive or negative effect on inflammation,^[37] 45 food or nutrients and inflammation were reviewed and showing a positive association between the food parameters and pro-inflammatory cytokines (i.e., IL-1β, IL-6, TNF-α, and CRP) or a negative association with anti-inflammatory cytokines (IL-4 and IL-10).^[10,31] As mentioned before, the total score is dependent on the whole diet, rather than any single nutrient. The index is not dependent on population means or recommendations of intake; it is based on results published in the scientific literature. The index is not limited to micronutrients and macronutrients but also incorporates commonly consumed components of the diet including flavonoids, spices, and tea. Previous research indicate that an anti-inflammatory diet may protect individuals from an inflammatory response characterized by elevated levels of hs-CRP and thus indirectly against the development of cancer, and other inflammation-related chronic health conditions. Shivappa et al^[38] suggest that increasing intake of anti-inflammatory dietary factors, such as plant-based foods rich in fiber and phytochemicals, and reducing intake of pro-inflammatory factors, such as fried foods or processed foods rich in saturated fat or animal protein, may be a strategy for reducing risk of laryngeal cancer. These findings reinforce the idea that a diet rich in pro-inflammatory food parameters (sweets, butter and other animal fats, cholesterol, saturated fat), and relatively poor in anti-inflammatory food parameters (vegetables and fruits) may increase inflammation.

The strengths of the present study are as follows: comprehensive systematic literature search was performed to find all relevant studies, and different subgroup analyses were carried out to identify sources of heterogeneity. However, this study has some limitations which should be considered in interpreting the results. Firstly, DII was derived from various dietary assessment tools that were based on self-report using food frequency questionnaire, 24-h dietary records and 7 days dietary records, which carries an inherent degree of recall bias and can lead to a potential misclassification of the exposure. Secondly, publication bias was observed in Egger tests, but using the trim-and-fill method to include supposedly missing negative studies, a significant positive association still persisted. Third, substantial heterogeneity was observed, one possible explanation for this could be the differences in CRP levels; another reason could be the different geographic area. Fourth, a limited amount of research has incorporated healthy people without disease to analyze the relationship between the DII and CRP levels. Therefore, more studies are needed to identify the association between DII and CRP levels among diverse populations. Finally, the DII was developed based on results acquired from research on diverse inflammatory indices (IL-1B, IL-4, IL-6, IL-10, TNF-α and CRP), however, the present meta-analysis focused its analysis on the relationship between DII and CRP. Further research should also take into account the relationship between DII and other inflammatory indices.

5. Conclusions

In summary, the results of the present meta-analysis indicate that DII score was positively associated with CRP. Additionally, the more prospective studies are needed to clarify this relation.

Author contributions

Conceptualization: Fangying Yang.

Data curation: Rongyu Hua.

Formal analysis: Rongyu Hua, Guanmian Liang.

Investigation: Rongyu Hua.

Methodology: Rongyu Hua.

Project administration: Rongyu Hua, Guanmian Liang, Fangying Yang.

Resources: Guanmian Liang.

Software: Rongyu Hua, Guanmian Liang.

Supervision: Fangying Yang.

Validation: Guanmian Liang.

Writing – original draft: Rongyu Hua.