This study was performed following the protocol published in PROSPERO (registration ID: CRD42022316369) and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria (22). Two reviewers independently searched six databases: PubMed, Cochrane Library, EMBASE, CNKI, Wanfang, and CBM. Supplementary retrieval in the Clinical Trials Database (clinicaltrials.gov) and the Grey Literature Database (opengrey.eu) was also performed. The search was not limited by language and included publications up to April 8, 2022. Using the combination of subject terms and free-text words, the search query was formed as follows: (“Non-alcoholic Fatty Liver Disease” (MeSH) OR “NAFLD” OR “Nonalcoholic Steatohepatitis” OR “NASH” OR “Metabolic Associated Fatty Liver Disease” OR “MAFLD” OR “metabolic associated steatohepatitis” OR “MASH”) AND (“ITLN1 protein, human” (Supplementary Concept) OR “omentin” OR “intestinal lactoferrin receptor” OR “hIntL protein”). This query was slightly modified according to the different databases. A specific strategy is provided in Supplementary Table 1 . If the full text or essential data of the included articles were not available, the corresponding authors were contacted via email for relevant information.

Two reviewers (QZ and SHC) independently assessed the eligibility of all studies, and any contradiction was resolved via discussion with a third reviewer (YNK). Studies were included if (1) patients were diagnosed with MAFLD, based on radiologically- proven hepatic steatosis and one of the three following conditions: overweight/obesity, diabetes mellitus, or metabolic dysregulation (e.g., HOMA-IR score ≥2.5); (2) the control group comprised healthy participants without MAFLD; (3) both groups included adults (age ≥18 years); and (4) outcomes included circulating serum or plasma omentin levels. Observational research cohort and case-control studies were included.

Studies were excluded if (1) patients had other causes of liver disease (e.g., alcoholic fatty liver disease or viral or autoimmune hepatitis) or severe disease; (2) omentin levels were not detected or were not derived from serum or plasma; (3) there were patient overlaps (i.e., more than one study conducted on the same patient cohort or duplicate publications); or (4) full-text or critical data of the articles was not obtained. Reviews, editorials, case reports, letters to the editor, hypotheses, studies on animals or cell lines, and abstracts from conferences were excluded.

Two reviewers (QZ and SHC) independently extracted the relevant data and evaluated the quality of the included literature. If conflicts arose during the process, a third reviewer (YNK) adjudicated the disagreement between the two reviewers. The following data were extracted from the included studies: (1) basic characteristics of the studies (name of the first author, publication year, country where studies were conducted, and design), (2) criteria applied to diagnose and classify MAFLD, (3) methods used to detect circulating omentin levels, (4) characteristics of patient/control cohort (number, sex, age, body mass index [BMI], co-existing disorders including obesity and T2DM), (5) circulating omentin levels in serum or plasma, (6) biochemical measurements (aspartate aminotransferase [AST], alanine aminotransferase [ALT], total cholesterol [TC], total triglyceride [TG], low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], C-reactive protein, interleukin-6, fasting serum insulin [FINS], fasting blood glucose [FBG]), and (7) calculation of homeostasis model assessment of insulin resistance [HOMA-IR].

The Newcastle–Ottawa Scale (NOS) is a useful tool for performing a semi-quantitative assessment of the quality of observational studies in a meta-analysis. The NOS contains eight items based on three dimensions: selection of cases, comparability of groups, and evaluation of exposures. Each item corresponds to one score, except that the item related to comparability corresponds to two scores. The quality of the studies is rated from 0 (very poor) to 9 (high) (23). The Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) scale was used to assess the certainty of these studies (https://gdt.gradepro.org).

Meta-analysis was performed using Stata/SE 15.1 (Stata Corporation, TX, USA). Omentin levels across each group were calculated as the mean ± standard deviation (SD). For omentin levels expressed as the median (first quartile, third quartile), the mean and SD were calculated using standardized formulas (24, 25). The standardized mean difference (SMD) was used to calculate the effect size in all studies. Heterogeneity among different studies was examined using the I² test, Cochran’s Q-test, and Galbraith’s test. A fixed-effects model was used for low heterogeneity (P≥0.1) and a random-effects model for high heterogeneity (P<0.1) (26). Subgroup and meta-regression analyses were conducted to explore the sources of heterogeneity. The subgroups were established based on the basic characteristics of the participants and physical indicators. Univariate meta-regression analysis was performed. The SMD of the omentin levels served as the dependent variable. The subgroup analysis variables and other biological indicators were added as explanatory covariates. Variables that met the significance level (P ≤ 0.05) in the univariate analysis were entered into the multivariate meta-regression analysis. Additionally, a sensitivity analysis was conducted to assess the stability of the outcomes by sequentially excluding studies. Egger’s test and funnel plots were used to examine publication bias (27).

The entire literature search and selection process are presented in a PRISMA flow diagram ( Figure 1 ). After retrieving articles from eight databases, 681 records were obtained. From the 681 records, 65 duplicates were removed, and the remaining studies were screened. Ultimately, 12 articles were included in the meta-analysis. The 12 articles were published between 2011 and 2021, reporting data on 1624 individuals (927 cases and 697 controls). Eight studies were conducted in China, one in Germany, one in Turkey, one in Poland, and one in Iran. The eight studies conducted in China were published in Chinese, and the remaining four were written in English. All articles were case-control studies that measured circulating omentin levels using enzyme-linked immunosorbent assays (ELISA). Seven of the twelve studies indicated that they evaluated omentin-1 levels, whereas the other five studies did not mention the specific subtype of omentin. The percentage of male subjects ranged from 31.71% to 100% in the patient groups and 38.76% to 100% in the control groups. The average ages of the patient and control groups were 51.51 and 51.26 years, respectively. BMI ranged from 26.1 to 41.0 kg/m² in MAFLD cases and 21.35 to 27.4 kg/m² in healthy controls. Supplementary specific information regarding the characteristics of the included studies can be found in Table 1 .

The quality of the 12 included studies was evaluated using the NOS scale, with an average score of 5.58, which indicated that the overall risk of bias was moderate ( Table 2 ). Nine studies were scored 6 or above, and the remaining studies were scored 4. Regarding the articles with a low score (<5), the studies conducted by Liu et al. and Wang (36) did not include community controls and lacked specific selection criteria for healthy controls (31, 36). The studies by Waluga et al. also did not include community controls and failed to meet the standard of representativeness (34). According to the GRADE scale (https://gdt.gradepro.org), the certainty of the present meta-analysis was very low (see Supplementary Table 2 ). Because all studies were observational, the certainty started at a low level. In addition, a low NOS score and significant heterogeneity led to a serious risk of bias and inconsistency, which combined with unmatched confounding factors between groups, accounted for a degradation of certainty.

This meta-analysis examined 12 studies consisting of 14 comparisons of omentin levels between MAFLD and control groups ( Figure 2 ) (28–39). Among these, the articles by Li et al. and Montazerifar et al. presented two statistical results (Li. et al.: MAFLD vs. controls and MAFLD+T2DM vs. controls; Montazerifar et al.: MAFLD vs. controls and MAFLD+abdominal obesity vs. abdominal obesity). The omentin levels were significantly lower in the MAFLD groups than in the control groups (SMD=-0.950 [-1.724, -0.177]). The random-effects model was applied because of its high heterogeneity (P<0.001, I^{2 =} 97.8%). Obvious heterogeneity according to Galbraith’s test is also vividly exhibited in Figure 3 , showing that most of the studies were not within a reasonable range. Therefore, subgroup analysis and meta-regression were necessary to further explore the sources of heterogeneity.

Subgroup analysis was implemented by region, sex, age, BMI and FBG ( Figure 4 ). As for subgroup analysis by region, in East Asia, omentin levels of patients with MAFLD were significantly lower than those of controls with an SMD of -1.69 [-2.63, -0.75]. In contrast, in the Middle East subgroup, patients with MAFLD had significantly higher omentin levels than healthy individuals with an SMD of 0.30 [0.02, 0.58]. There was no significant difference in omentin levels between patients with MAFLD and controls in the European subgroup, with an SMD of 0.78 [-1.01, 2.56]. It is worth noting that the heterogeneity in the East Asian and European groups were still high (P<0.001), and only that in the Middle East group decreased (P=0.327).

Regarding subgroup analysis by BMI. MAFLD patients with BMI between 23 and 27.5 kg/m² had significantly lower omentin levels than controls (SMD=-1.63 [-2.92, -0.34]), whereas MAFLD patients with BMI ≥27.5 kg/m² showed no significant difference compared with controls (SMD=0.28 [-0.21, 0.77]). Furthermore, patients with BMI between 23 and 27.5 kg/m² had relatively lower omentin levels than those with BMI ≥27.5 kg/m².

In addition, we performed subgroup analysis based on FBG. The cutoff values were based on Standards of Medical Care in Diabetes (40) (i.e., FBG ≥7 mmol/L for a diagnosis of diabetes and FBG of 5.6-6.9 mmol/L for impaired fasting glucose). In the subgroup with FBG <5.6 mmol/L, omentin levels were not significantly different between MAFLD patients and controls (SMD=0.04 [-0.31,0.40]). However, patients with FBG of 5.6-6.9 mmol/L or ≥7 mmol/L had significantly lower omentin levels than healthy controls (SMD=-1.77[-2.14, -1.40] and -2.30 [-3.25, -1.36], respectively). Intragroup heterogeneity was also reduced (FBG <5.6 mmol/L: I^{2 =} 67.50%, P=0.015; FBG ≥7 mmol/L: I^{2 =} 90.50%, P<0.001; 5.6 mmol/L≤ FBG<7 mmol/L: I² <0.001%, P=0.739). Other results of subgroup analysis by sex and age are listed in Table 3 .

Initially, 10 factors (region, sex, age, BMI, HOMA-IR, TC, TG, LDL-C, HDL-C, and FBG) were included in the univariate meta-regression. The results demonstrated that both region and FBG level exerted significant effects on the outcome. Region and FBG levels accounted for 39.22% and 78.47% of the between-study variance, respectively. Moreover, patients in Europe and the Middle East had higher omentin levels than those in East Asia. Patients with higher FBG levels had lower circulating omentin levels (Europe vs. East Asia: coefficient=2.467, P=0.024; Middle East vs. East Asia: coefficient=1.855, P=0.041; FBG: coefficient=-0.578, P=0.001). Region and FBG levels were then included in the multivariate meta-regression. The results showed that FBG level could better explain the heterogeneity, along with an inverse association with the SMD of circulating omentin levels (coefficient=-0.538, P=0.009) ( Table 4 ).

No study exerted a significant influence on the overall outcomes, which indicated benign stability ( Figure 5 ). Egger’s test indicated a small possibility of publication bias (P=0.919). No obvious asymmetry was observed in the funnel plot, suggesting a low possibility of publication bias ( Figure 6 ).

This meta-analysis investigated the association between circulating omentin levels and MAFLD. Subgroup analysis and meta-regression revealed that FBG levels might be a source of heterogeneity. The outcomes were reasonable and consistent, as previously discussed. Importantly, our study provides useful evidence for the clinical application of omentin as a biomarker to facilitate diagnosis or as a supplement to treatment.

This meta-analysis had certain limitations. First, the number of included studies and participants was small. Omentin is a novel adipokine that was identified in 2006, and relevant studies have been limited so far. Further studies are encouraged to investigate the association between omentin levels and patients with MAFLD, and an updated meta-analysis should be performed in the future. Second, the absolute value of omentin varied significantly from 2.85 to 460 ng/mL, although the units were uniform. Therefore, SMD and a random-effects model were chosen to pool the outcomes. Because all studies used ELISA to measure omentin levels, the distinctions between them might be derived from different experimental kits of different manufacturers. Third, the heterogeneity was high. We have conducted the subgroup analysis and meta-regression to explore the source of heterogeneity. Although we conducted subgroup analysis using as many variables as possible, they were still limited. The results demonstrated that FBG might be a possible source of heterogeneity, but several considered indicators (i.e., ALT, AST, FINS, and HbA1c) were not included in the meta-regression due to an insufficient number of studies that reported those indicators. Some of those indicators were essential to the association between omentin and MAFLD. Therefore, we hope that relevant articles in the future will pay more attention to multiple serum biomarkers. Furthermore, in order to fully identify the source of heterogeneity, different severities of MAFLD are another important factor to be considered. Since classification of the severity of MAFLD was rarely seen in the included studies, we could not perform a subgroup analysis based on disease severity.