The PubMed (1951), EMBASE (1966), the Cochrane Library (2000), and Web of Science (1900) were systematically searched to July 22, 2021 using the medical subject headings (MeSH) and related text words. Supplementary Table 1 presents the detailed search strategy. The reference lists of the original studies included in our analysis were also searched as well as those listed in the published review and meta-analyses (19–23).

This study included original publications to evaluate the association between HP infection and the risk of NAFLD. We excluded the followings studies: (1) no information of the country for included population; (2) animal experiments, case reports, case series, reviews, practice guidelines, commentaries, and editorials; (3) unavailable data; (4) non-English or non-Chinese language publications. If more than one publication on the same study population was available, only the most recent publication was included. Two reviewers (Xiajing Chu, Zhiyuan Ma) independently screened the titles and abstracts, selected relevant full texts, and assessed them for eligibility. Non-conformity was resolved by discussion.

Using a predefined data collection form, two reviewers (Xiajing Chu, Zhiyuan Ma) independently extracted the following data: the first author, year of publication, country of participants, study design, publication type, sample size, participants setting, time of the study, age of participants, methods used for identification, diagnosis of the HP infection and NAFLD, adjusted in the multivariate analysis or non-adjusted effect estimates with the 95% CI. If effect estimates were not provided, the odds ratio (OR) and 95% CI were calculated. Any discrepancies between the reviewers were resolved by discussion, or by resort to a third reviewer (Xiang Yan) if consensus could not be reached.

Two independent reviewers (Zhiyuan Ma, Xiang Yan) assessed the risk of bias based on the Newcastle-Ottawa Scale (NOS) (31) for the cohort study and case-control study and the Agency for Healthcare Research and Quality (AHRQ) for the cross-sectional study (32). In the case of disagreement, a third investigator (Xiajing Chu) was consulted. The included cohort studies and case-control studies were rated as “low quality” (0–3 points), “moderate quality” (4–6 points) or “high quality” (7–9 points) based on their overall score on the NOS. The included cross-sectional studies were regarded as high quality (8–11 points), moderate quality (4–6 points), and low quality (< 4 points) based on their overall score on the AHRQ.

This meta-analysis was performed using R software v3.6.1. The odds ratio (OR), hazard ratio (HR) and 95% CI were used to explore the relationship between HP infection and NAFLD. The extent of heterogeneity was interpreted based on the total percentage of variation between the relevant studies, as measured by the I² statistical parameter. The heterogeneity was categorized as low if I² was 0–25%, moderate if I² was 25–50%, and high if I² was more than 50% (33). Additionally, Cochrane's Q-test was used to assess the presence of heterogeneity. The P-value by the Cochrane's Q-test > 0.05 indicated no significant heterogeneity among the included studies (22). When the heterogeneity was low, a fixed-effects model was used. Otherwise, a random-effects model was used (19). If necessary, multiple reported analyses per outcome were combined using fixed-effects models, so that each study contributed at most one effect size for each outcome.

Previous studies found a difference in the association of the HP infection with the risk of NAFLD between the Asian and non-Asian populations. Therefore, we analyzed Asian and non-Asian populations, respectively. The Asian population was defined as the population from Asian countries, such as China, India, Iran, Japan, and South Korea. The Non-Asian population was defined as the population from a non-Asian country.

Subgroup and sensitivity analyses were performed to explore the sources of heterogeneity. Subgroup analysis was conducted based on the different diagnosis methods of NAFLD and HP. The common NAFLD methods included liver biopsy, ultrasound, hepatic steatosis index (HIS), NAFLD liver fat score (NAFLD-LFS), and fatty liver index (FLI). HP methods included invasive tests, serology, breath test (UBT), and Stool antigen tests. Furthermore, subgroup analysis was also based on the degree of covariate adjustment, the study design, and gender. Because these factors may have an influence on the overall association between NAFLD and HP based on previous studies (19–23).

Additionally, sensitivity analyses were performed to determine the influence of individual studies on the overall estimates by serially excluding each study.

Funnel plots and the Egger test were used to assess potential publication bias. If the number of included studies were < 10, testing for publication bias was not performed (34).

For this study, 612 potentially relevant studies were retrieved using the predefined search strategy, and another 28 studies were retrieved through other sources. Among these articles, 345 were duplicate publications. A total of 181 studies were excluded by screening the titles and abstracts. Of the remaining 114 potential eligible articles, 89 studies were excluded by carefully examining the abstracts or full texts. Finally, 25 studies met the inclusion criteria and were included in our meta-analysis. Two studies reported the HR (35, 36), and 23 studies reported the OR (3, 25–29, 37–53). In these 23 studies, one study reported the OR values based on different levels of NAFLD (mild, moderate and severe) (50); another study reported the OR values based on the different white blood cell (WBC) count (52). Therefore, 25 studies were included in this meta-analysis. Figure 1 shows the detailed flowchart of the selection process of eligible studies.

Twenty-five studies involving 117,458 participants were included (Table 1). This review contained two cohort studies (35, 36), 17 cross-sectional studies (3, 25–27, 29, 37–39, 41–47, 49, 52), and six case-control studies (28, 40, 48, 50, 51, 53). These studies were published between 2013 and 2021, three of which were conference abstracts (43, 47, 54). The participants from 17 studies were from Asia, and eight studies involved participants who were not from Asia. Regarding the HP test method, two studies used multiple methods. Of the remaining studies, three used only invasive tests, 10 used the breath test, nine used the serology, and one used stool Ag tests. Additionally, most studies (17 studies) used ultrasound to diagnose NAFLD.

In this study, quality assessments were conducted on full texts. Supplementary Table 2 presents the quality assessment results according to NOS for the cohort study and the case-control study. The scores of included studies ranged from six to nine (mean 7.75), and the comparability had a low score for most included studies. Supplementary Table 3 presents the results of AHRQ for the cross-sectional study. The scores of included studies ranged from one to 15 (mean 6.53). In only one study, the handling of missing data in the analysis was explained (49), and in only one study, the patient response rates and completeness of the data collector were summarized (41).

In an overall pooled analysis, HP-positive populations had a higher risk of NAFLD than HP-negative populations (OR = 1.30) both in Asian and non-Asian countries. Regarding The Asian population, the meta-analysis using random-effects model showed HP infection was associated with a risk of 1.30 of developing NAFLD (OR = 1.30, 95% CI: 1.13–1.49, P < 0.01; I² = 94.30%, P < 0.01; Figure 2). The HP infection was also associated with an increase risk of NAFLD in non-Asian population (OR = 1.42, 95% CI: 1.04–1.94, P = 0.03; I² = 44.90%, P = 0.09; Figure 2). No significant difference in the association was found between non-Asia and Asia populations (P = 0.61). Besides, two studies reported the effect size of HR (35, 36), and the pooled HR supported the association between HP infection and risk of NAFLD (HR = 1.13, 95% CI: 1.04–1.23) (Supplementary Figure 1).

Sensitivity analysis was performed by removing one study at a time, and the P-value confirmed the stability of the results for the Asian population. However, the results of the non-Asian population were not stable enough (Supplementary Figures 1–3). Analysis of the funnel plot of the OR for publication bias suggested the absence of bias because of plot symmetry (Supplementary Figure 4). Furthermore, the Egger test showed no publication bias (P = 0.17).

This meta-analysis included 25 studies (two control studies, six case-control studies, and 17 cross-sectional studies) involving 107,306 participants from 11 countries. The data studies showed a positive association between HP and the risk of development of NAFLD both for Asian and non-Asian populations. Similar results were observed in most subgroups of the Asian populations, except for the male group. However, the association for the non-Asian population was not stable enough.

In several studies, the underlying mechanism involved in HP infection and the risk of NAFLD development were explained. HP infection is known to cause chronic low-grade systemic inflammation by increasing the levels of proinflammatory cytokines (55). Additionally, HP influences the development of NAFLD through hormonal effects (13). Moreover, HP-induced bacterial translocation in chronic liver disease can be detected by human β-defensin-1(40).

A prospective multicenter pilot cohort study (35) showed that after eradication therapy of HP infection, there was a significant reduction in levels of C-reactive protein, leptin, insulin resistance, NAFLD-LFS, TNF-α, and IL-6. After a 24-month follow-up, the incidence rate of NAFLD in patients with eradication therapy was five times lower compared to that in untreated patients. Additionally, a randomized controlled trial (56) reported a significant improvement in insulin resistance after 24 weeks of successful eradication of HP. Thus, the eradication of HP had an advantageous effect on metabolic diseases, such as NAFLD. These findings were consistent with our results. Although the forest plot of the OR both for Asian and non-Asian populations showed a moderate or high heterogeneity, the sensitivity analysis and publication bias substantiated the robustness of our results for the Asian population. NAFLD diagnosis methods and study design contributed to the heterogeneity of the association for the Asian population.

Subgroup analysis based on different country populations showed positive associations between HP infection and the risk of NAFLD in China, India, and Greece. Considering the small size effect, we thought that the positive association in China was more robust. This observation was consistent with the data presented in previous studies (24, 29). The pathogenesis of NAFLD is known to include adipose tissue-derived hormones, nutritional factors, genetic and epigenetic factors, gut microbiota, and insulin resistance (57, 58). Because some information on these pathogenesis factors and metabolic risk factors (e.g., body weight, hypertension, hyperglycemia, and dyslipidemia) was lacking, subgroup analysis based on these factors was not conducted. Further studies are required to verify these results by well-designed human studies, considering the complex interactions with confounding factors, such as environmental and genetic susceptibility factors (19). Additionally, we found that females with HP infection had a higher risk of NAFLD in the Asian population. Although the underlying mechanism of action is unclear, many studies have shown that gender is a factor influencing the risk of an infection (36, 41, 48). Besides, the observed differences in associations are most likely due to social determinants of health, but not decided by biological or genetic factor only. The difference on associations between Asian and non-Asian population were possibly caused by adoption of dietary and lifestyle habits, crowded living conditions, poor sanitation, and lack of access to care (24, 59). However, the small sample size in our included populations might be a confounding variable.

Five studies (19–23) assessed the association between HP infection and NAFLD, and all studies concluded that the presence of a significant relationship between HP and NAFLD, and one study noted a 36% increased risk of NAFLD in patients with HP infection (20). HP infection indeed showed a positive association with NAFLD for the Asian population. These findings were similar to the findings of our study. Compared with previous meta-analyses, this study used a larger sample size to explore the association between HP infection and NAFLD in the non-Asian population than in a previous Meta-analysis. In our review, we included additional studies, including cross-sectional, case-control, and cohort studies. The comprehensive subgroup analysis, which was based on the region of the population, the different diagnosis methods of NAFLD and HP, the degree of covariate adjustment, and the study design, confirmed the robustness of our results.

A specific subgroup analysis was performed based on the country of the population, the different diagnosis methods used for NAFLD and HP, the degree of covariate adjustment, and the study design. These methods supported our conclusions. Our study had several limitations. First, the included studies had a small sample size of fewer than 500 participants, which might affect the quality of evidence. Second, the bias of the included retrospective studies might affect the quality of evidence. Additionally, the Asian and Non-Asian populations were defined as a population from the Asian or Non-Asian countries, respectively. The designation may not be reliable enough, but there is no more precise approach due to lack of data. Finally, subgroup analyses were performed based on potential confounding variables, but subgroup analyses by several important pathogenesis factors, HP test ranges, and metabolic risk factors (e.g., body weight, hypertension, hyperglycemia, and dyslipidemia) could not be conducted because of lacking information. Further prospective studies are required to perform an in-depth analysis of the heterogeneity.