The PubMed, EMBASE, Cochrane Library, and Web of Science databases were systematically searched from their inception to 22 May 2022. There were no language restrictions, and the search strategy combined the use of medical subject headings (MeSH) and keywords. The search terms in the study were as follows: (“Non-alcoholic Fatty Liver Disease ” OR “Non-alcoholic Fatty Liver Disease” OR “NAFLD” OR “Non-alcoholic Fatty Liver*” OR “Non-alcoholic Steatohepatiti*”) AND (“Dementia” OR “Alzheimer's disease” OR “Cognitive decline” OR “Cognitive impairment” OR “Cognitive disorder” OR “Cognitive dysfunction”). The complete search strategy is presented in Additional File 1. Additionally, we conducted manual searches of reference lists and related reviews of included studies for other relevant articles.

Trials were included according to the following eligibility criteria: (1) cohort or case–control study designs; (2) reports of dementia or cognitive impairment diagnosis as the outcome, with the diagnostic criteria of NAFLD being hepatic steatosis in addition to more than 1 of (i) overweight or obesity status, (ii) type 2 diabetes mellitus (T2DM), and (iii) metabolic dysregulation; and (3) investigations of the association of NAFLD with the risk of dementia or cognitive impairment. If multiple data were reported from the same study, we included the study with the longest follow-up or the largest number of participants.

The following literature was excluded: (1) conference abstracts or study protocols, (2) duplicate publications, and (3) studies with no outcomes or inadequate data.

The literature included in this study was independently screened by 2 reviewers (WLP and ZZY) based on eligibility and exclusion criteria. First, duplicate articles and those deemed to not be relevant to this study based on titles and abstracts were excluded. Thereafter, potential articles were identified, downloaded, and read in full to ensure that data from the article could be effectively included. In case of disagreement concerning the inclusion of an article, the 2 researchers discussed with a third reviewer to reach a consensus.

According to guidelines on data extraction for systematic reviews and meta-analyses (Taylor et al., 2021), the 2 above-mentioned reviewers independently extracted and summarized the following information from each of the included studies: first author, year of publication, country, study type, sample size, study period, number of follow-up years, age of participants, gender, diagnosis of dementia or cognitive impairment, dementia type, and adjustment for confounders.

We assessed the quality of the included articles using the Newcastle–Ottawa Scale (NOS) for quality assessment (Stang, 2010), focusing on 3 aspects: selection, comparability, and outcome. The score ranged from 0 to 9 points for cohort studies and case–control studies. According to the NOS, the higher the score is, the better the quality of the included studies. We considered the NOS scores of 0–3, 4–6, and ≥7 as indicative of low, medium, and high quality, respectively. The quality of a cross-sectional study was assessed using the American Agency for Health Care Research and Quality (AHRQ) methodology checklist. The checklist contains 11 items, with the risk of bias scored as 1 for “Yes” and 0 for “Unclear” or “No” risk. After recording the total score, the articles were divided into 3 grades: “low” (0–3 points), “medium” (4–7 points), and “high” (8–11 points).

To assess the association between NAFLD and the risk of dementia or cognitive impairment, we extracted the adjusted odds ratio (OR) and 95% CI from each study. The extracted adjusted OR and 95% CI were statistically analyzed using Stata v. 16.0 software (Stata Corp, College Station, TX, USA). The heterogeneity of the included studies was assessed by the chi-square test and I². When P was>0.10 for the χ2 test or when I² was <50%, statistical heterogeneity was considered non-significant, and the fixed-effects model was subsequently adopted. If significant heterogeneity was indicated from I² > 50% or P < 0.1, the random-effects model was employed. Robustness of the results with respect to the overall effect of statistics and sources of heterogeneity were verified through subgroup analysis and sensitivity analysis. Publication bias was confirmed by visual inspection of funnel plots, and the Egger regression test was used to statistically detect publication bias for asymmetry.

The systematic search of studies published before 22 May 2022, identified 1,095 results. Of the 1,095 articles, 227 were duplicate articles and were excluded. After screening the title and abstract, another 852 articles were excluded. After perusal of the full text of the remaining of the articles, 9 articles were excluded. Of these 9 articles, 8 failed to provide relevant data and outcomes and 1 article provided incomplete data. Ultimately, 7 studies (Labenz et al., 2021; Lampignano et al., 2021; Shang et al., 2021; Jeong et al., 2022; Kim et al., 2022; Liu et al., 2022; Yu et al., 2022) were included in this study. The search selection process is presented in Figure 1.

Seven studies were included in this review. Five studies (Labenz et al., 2021; Shang et al., 2021; Jeong et al., 2022; Kim et al., 2022; Liu et al., 2022) were cohort studies and 2 studies (Lampignano et al., 2021; Yu et al., 2022) were case–control studies. The 7 studies were published from 2020 to 2022 and involved a total of 891,562 individuals, with sample sizes ranging from 1542 to 608,994 individuals per study. The average follow-up time ranged from 3 to 9.5 years. There were 2 studies (Jeong et al., 2022; Kim et al., 2022) from Korea, 1 (Yu et al., 2022) from the United States, 1 (Liu et al., 2022) from China, 1 (Shang et al., 2021) from Switzerland, 1 (Labenz et al., 2021) from Germany, and 1 from (Lampignano et al., 2021) Italy. The results of 5 studies (Labenz et al., 2021; Lampignano et al., 2021; Shang et al., 2021; Jeong et al., 2022; Kim et al., 2022) described outcome measures of dementia, and the 2 other studies (Liu et al., 2022; Yu et al., 2022) involved cognitive impairment. The main characteristics of the included studies are shown in Table 1.

The quality of included studies was assessed by the NOS scale, and the results are shown in Table 1. The average score across the 7 studies was 7.8, indicating that the overall quality was high. Among them, 1 article (Jeong et al., 2022) scored as high as 9 points, and the remaining studies scored 7 points and 8 points. According to the AHRQ criteria, the scores for the 2 cross-sectional studies (Lampignano et al., 2021; Yu et al., 2022) were 6 and 7, respectively, and both were classified as being of moderate quality.

Two studies (Liu et al., 2022; Yu et al., 2022) assessed the association between NAFLD and risk of cognitive impairment. As shown in Figure 2, data analysis showed a strong link between NAFLD and cognitive impairment (OR = 1.44; 95% CI: 1.17–1.78; I² = 0%; P = 0.001). Through sensitivity analysis, the pooled-effect size was not reversed by individual studies, which means that the outcomes demonstrated robustness (Figure 3).

As shown in Figures 4, 5 studies (Labenz et al., 2021; Lampignano et al., 2021; Shang et al., 2021; Jeong et al., 2022; Kim et al., 2022; Liu et al., 2022; Yu et al., 2022) described the association of all-cause dementia with NAFLD; however, results from the data showed that the relationship between NAFLD and an increased risk of all-cause dementia was not obvious (OR = 1.03; 95% CI: 0.97–1.09; I² = 84.7%; P = 0.341). Since Alzheimer Disease (AD) and vascular dementia (VD) belong to the category of dementia, we extracted the data of AD and VD separately for reanalysis. Two of five studies (Labenz et al., 2021; Jeong et al., 2022) described the association between NAFLD and VD, with the data demonstrating a significant relationship (OR = 0.88; 95% CI: 0.79–0.98; I² = 0.0%; P = 0.020; Figure 2). Three studies (Labenz et al., 2021; Shang et al., 2021; Jeong et al., 2022) evaluated the risk of AD in NAFLD and found that NAFLD was not associated with an increased risk of NAFLD (OR = 0.95; 95% CI: 0.83–1.09; I² = 61.0%; P = 0.489; Figure 2). Through sensitivity analysis, the pooled-effect size was not reversed by individual studies, which means that the outcomes demonstrated robustness (Figure 5).

In a subgroup analysis of gender, 4 studies (Labenz et al., 2021; Shang et al., 2021; Kim et al., 2022; Liu et al., 2022) described the likelihood of dementia or cognitive impairment in females with NAFLD (OR = 1.09; 95% CI: 0.93–1.27; I² = 83.2%; P = 0.273; Table 2) and 3 studies (Labenz et al., 2021; Kim et al., 2022; Liu et al., 2022) described the likelihood of dementia or cognitive impairment in males with NAFLD (OR = 1.01; 95% CI: 0.95–1.07; I² = 31.0%; P = 0.765; Table 2). The data demonstrated that there was no gender difference in the chances of developing dementia in patients with NAFLD.

In the subgroup analysis of study type, 2 studies (Lampignano et al., 2021; Yu et al., 2022) evaluated the risk of dementia or cognitive impairment using a cross-sectional study design and found that NAFLD was associated with an enhanced risk of dementia or cognitive impairment (OR = 1.49; 95% CI: 1.19–1.88; I² = 0.0%; P = 0.001; Table 2). The other 5 included studies (Labenz et al., 2021; Shang et al., 2021; Jeong et al., 2022; Kim et al., 2022; Liu et al., 2022) showed that other retrospective cohorts of NAFLD were not associated with an increased risk of dementia or cognitive impairment (OR = 1.03; 95% CI: 0.97–1.09; I² = 84.3%; P = 0.294; Table 2).

Through the visual inspection of the funnel plot, there was no evidence of a significant publication bias in the outcomes for NAFLD and the risk of dementia or cognitive impairment (Figure 6). Egger regression test (P = 0.952) also indicated no publication bias in our meta-analysis.

In this systematic review and meta-analysis of observational studies, a total of 7 population-based studies provided a comprehensive evaluation of the relationship between NAFLD and the risk of dementia or cognitive impairment in approximately 891,562 individuals. We found a significant increase in the risk of cognitive impairment among individuals with NAFLD, with an overall 1.44-fold increase in risk compared to healthy controls. In contrast, the association between NAFLD and VD was unexpected. Patients with NAFLD had a 0.88-times risk of developing VD; hence, patients with NAFLD were less likely to develop VD. Subgroup analyses showed that gender was not a significant factor in determining whether patients with NAFLD were at a high risk for dementia or cognitive impairment. In the subgroup analysis of study type, NAFLD was associated with a higher risk of dementia or cognitive impairment in cross-sectional studies compared with retrospective cohorts.

We found a statistically significant difference between NAFLD and cognitive impairment after undertaking separate statistical calculations for dementia and cognitive impairment although this result was based on just 2 observational studies. However, because of the low heterogeneity of the statistical results, this may increase statistical efficiency. The conclusions drawn from the results of the data can also be verified from the results of previous research. A review (George et al., 2022) of 11 observational studies showed NAFLD to be associated with low cognitive ability across several domains, such as “general cognition,” “mental speed, attention, and psychomotor speed,” and “ideas, abstraction, figural creations, and mental flexibility.” In the study by Sang et al. (Seo et al., 2016), 874 (19.5%) of 4,472 patients with confirmed NAFLD were diagnosed with cognitive impairment. However, the study failed to provide enough data to calculate the increased risk of cognitive impairment in relation to NAFLD. In addition, a prospective cross-sectional study (Filipović et al., 2018) suggested that a cognitive deficit was more pronounced in persons with NAFLD, and patients suffering from NAFLD had about a 4-times higher risk of having cognitive impairment as assessed by the Montreal Cognitive Assessment test. The study of Filipović et al. (2018) shows that the volume of gray and white matter in NAFLD patients with lower Montreal Cognitive Assessment (MoCA) scores may be significantly reduced. However, this study did not include data on the incidence of cognitive impairment due to NAFLD; hence, this result was not included in our meta-analysis. In contrast, more recent studies were added to the current analysis, thus providing robust evidence on the association between NAFLD and the risk of cognitive impairment.

After grouping by type of dementia, we found that NAFLD patients had a lower risk of VD than did patients with AD. The result of our meta-analysis showed that patients with NAFLD have a reduced risk of developing VD; however, converse, results were reported in a recent study by Wang et al. (2022), in which an overall 2.62-fold increase in the risk of VD among individuals with NAFLD was observed. Owing to the small number of the study populations included in this review and the fact that the control group of the study was not compared with healthy people, we were unable to further investigate the effect of NAFLD on dementia risk.

Subgroup analyses showed that there was no statistical interaction of NAFLD with male or female gender with respect to the likelihood of dementia or cognitive impairment, which is largely consistent with the results suggested by previous studies (Wang et al., 2022). In the subgroup analysis of study type, NAFLD was associated with an increased risk of dementia or cognitive dysfunction in cross-sectional studies compared to retrospective cohort studies. We speculate that this may be related to the following reasons: there were only 2 cross-sectional papers in the study, which might have reduced statistical efficiency, and the smaller sample size relative to the case–control study might influence the results.

The pathophysiological mechanisms of the association between NAFLD and dementia or cognitive impairment remain largely unclear. Nevertheless, some researchers continue to speculate upon the relationship between NAFLD and dementia or cognitive impairment. Systemic neuroinflammation has been considered to be one of the important causes of cognitive impairment and dementia in NAFLD. Liver-induced macrophages (Haukeland et al., 2006; Tosello-Trampont et al., 2012) infiltrated by lipid accumulation secrete cytokines and chemokines into the systemic circulation (Tosello-Trampont et al., 2012; Dixon et al., 2013; Kratz et al., 2014; Narayanan et al., 2016), but it is unclear how inflammation spreads across the blood–brain barrier to the brain (Viscogliosi et al., 2013; Miller and Spencer, 2014). In addition, cytokine-activated receptors on peripheral endothelial cells promote the release of proinflammatory factors in the central nervous system (Cai and Liu, 2012), thereby causing an inflammatory response in the brain to disrupt brain function. Second, studies have demonstrated that the dysfunction of the gut–liver–brain axis in patients with liver cirrhosis is associated with hepatic encephalopathy (Bajaj, 2014; Ahluwalia et al., 2016), which has led to an interest in determining whether or not cognitive impairment in patients with NAFLD is associated with altered gut microbiota. In the early stages of NAFLD, gut microbiota homeostasis is disrupted; for example, by small intestinal bacterial overgrowth and low-abundant bacteria (Bauer et al., 2002; Qin et al., 2014; Bajaj et al., 2015). These microbial signatures are the drivers of “leaky gut” (Miele et al., 2009; Li et al., 2013). Gut microbes that enter the systemic circulation, such as endotoxins, ammonia, and bacterial DNA, can initiate and propagate systemic and neuroinflammation leading to cognitive impairment or dementia (Bajaj, 2014; Liu et al., 2020). In addition, studies have found that vascular complications in patients with NAFLD (Targher et al., 2016), such as atherosclerosis and cerebrovascular dysfunction, may contribute to dementia or cognitive impairment (De La Torre, 2012; Long et al., 2015). The upregulation of cerebral blood volume was found to be significantly impaired by oxygenated hemoglobin concentration in patients with NAFLD as measured on a cognitively challenging verbal fluency task (Takahashi et al., 2017). Notably, in addition to the need for more research to further understand the pathological mechanism of this new metabolic encephalopathy caused by NAFLD, future research should pay increased attention to the accurate diagnosis of the disease and its impact on daily life, so as to formulate a better treatment plan for improving patients' quality of life.

Our meta-analysis drew on existing evidence to summarize the association between a history of NAFLD and the risk of dementia or cognitive decline. Albeit, our results have some limitations. Due to the small amount of experimental data (acquired from all of the studies) and the methodological heterogeneity of the included studies, the accuracy of our results might have been affected. The diagnostic criteria for NAFLD and dementia or cognitive impairment are inconsistent, which might have biased the results. In addition, caution should be taken concerning the generalizability of the included studies' results. Since these studies were conducted in the United States, China, Italy, Korea, and Switzerland, regional bias may be present.