Non-alcoholic fatty liver disease (NAFLD) and its association to cardiovascular disease: A comprehensive meta-analysis

Abstract

Background

Non-alcoholic fatty liver disease (NAFLD) affects up to nearly a third of the Western population and has been inconsistently associated with cardiovascular diseases (CVDs). Therefore, we conducted a comprehensive meta-analysis to quantify the correlation of NAFLD with all major vascular diseases, acute coronary syndrome (ACS), subclinical atherosclerosis and endothelial dysfunction.

Methods

We searched PubMed and Embase for studies looking at the relationship between NAFLD and cardiovascular diseases published through September 2023. The parameters we used to assess cardiovascular diseases include acute coronary syndrome, brachial flow-mediated dilatation (FMD), serum asymmetric dimethylarginine (ADMA), carotid intima-media thickness (CIMT), and carotid stenosis (>50%). Data from these studies were then collected and meta-analysis was performed using the random effects model. RevMan v5.4 was used for statistical analysis.

Results

We interrogated a total of 114 publications which met our inclusion criteria. NAFLD patients showed statistically significant reduction in FMD% [MD: −4.83 (95% CI: −5.84 to 3.81, p < .00001)] and increased serum ADMA [MD: 0.08 (95% CI: 0.05–0.11, p < .00001)]. Mean CIMT was also increased in NAFLD patients [MD 0.13 (95% CI: 0.12–0.14, p < .00001)]. NAFLD showed a higher prevalence of pathological CIMT [MD: 0.11 (95% CI: 0.10–0.12, p < .00001)] and increased carotid plaques [OR: 2.08 (95% CI: 1.52–2.86, p < .00001)]. Furthermore, we demonstrated statistically significant increase in cardiovascular diseases among NAFLD patients compared to controls [OR: 1.92 (95% CI: 1.53–2.41, p < .00001)].

Conclusion

NAFLD is a strong predictor for endothelial dysfunction, subclinical atherosclerosis and cardiovascular disease. Further studies are required to determine whether incidental findings of fatty liver on abdominal ultrasonography should prompt the need for detailed assessment of other CVD risk factors.

Background

Non-alcoholic fatty liver disease (NAFLD) is the leading cause of chronic liver disease worldwide, afflicting up to 30% of the Western population. ¹ It is strongly associated with metabolic syndrome^1,2 and encompasses a wide morphological spectrum of liver lesions, in the absence of excessive alcohol consumption, from steatosis to steatohepatitis and cirrhosis.^3,4 Further progression can increase the risk of hepatocellular carcinoma (HCC). In the past decade, numerous findings have demonstrated that the mortality associated with NAFLD patients is largely attributed to cardiovascular diseases (CVD), sparking discussions regarding its independent or synergistic role with other risk factors in driving extrahepatic diseases.^5–7

Atherosclerosis is recognised as the underlying pathophysiological mechanism that promotes the development of cardiovascular diseases in NAFLD patients. ⁸ Among the key atherogenic mechanistic risk factors are endothelial dysfunction, arterial stiffness, and subclinical atherosclerosis, often defined by the thickening of intima-media, an early hallmark of atherosclerosis.^9,10 This relationship between NAFLD and cardiovascular disease can be assessed by investigating endothelial function, which includes markers such as carotid intima-media thickness (CIMT), flow-mediated dilation (FMD), and asymmetric dimethylarginine (ADMA). CIMT and FMD are two key non-invasive markers for evaluating endothelial dysfunction. ¹¹ ADMA, an endogenous analogue of L-arginine, which competes with nitric oxide (NO) synthesis, is closely associated with endothelial dysfunction. ¹² Robust evidence supporting the early occurrence of endothelial dysfunction in the progression of atherosclerosis strengthens the potential of these markers to provide invaluable insights into the vascular health of individuals with NAFLD.^13,14 Additionally, the presence of carotid plaques and acute coronary syndrome events serve as significant indicators of cardiovascular risk in this context. While several case-controlled studies have investigated these measurements, it is worth noting that many of these studies have been constrained by their limited sample sizes and inconclusive findings.

We conducted a comprehensive assessment of the association between NAFLD and major forms of vascular disease, including cardiovascular disease, subclinical atherosclerosis, and endothelial dysfunction in all ethnic groups. To the best of our knowledge, this meta-analysis represents the largest and most extensive effort to date in shedding light on the intricate relationship between NAFLD and vascular health.

Methods

Our search strategy was conducted in accordance with MOOSE criteria. We searched PubMed and Embase for relevant studies published through to September 2023. All languages were included. We also searched ClinicalTrials.gov for any ongoing or unpublished trials that could potentially provide additional data to the analysis.

Searched terms were ‘NAFLD’ or ‘nonalcoholic fatty liver disease’ or ‘nonalcoholic steatohepatitis’ and ‘carotid plaque’ or ‘carotid stenosis’ and ‘coronary heart disease’ or ‘myocardial infarction’ or ‘silent MI’ or ‘unstable angina’ or ‘heart attack’ and ‘ADMA’ or ‘asymmetric dimethylarginine’ and ‘flow mediated vasodilation’ or ‘brachial flow mediated dilatation’ or ‘FMD’ and ‘mCIMT’ or ‘carotid intima media thickness’ or ‘CIMT’. The above terms were combined with the Boolean operators ‘AND’ and ‘OR’, and a manual search was performed where appropriate on selected publications. The searched results were also filtered for human studies only. Cited bibliographies were also extracted. For duplicate publications identified, the one with larger sample size was included. Foreign language publications were translated. Review articles were searched for relevant references.

To determine the risk of cardiovascular outcomes in patients with NAFLD, the studies that looked at ACS (MI, silent MI, unstable angina) and carotid stenosis (stenosis of >50% in at least one coronary artery) were selected.

To evaluate endothelial dysfunction, we selected studies where data for brachial FMD or serum AMDA could be extracted.

Two independent reviewers were assigned to assess the articles prior to collecting the data. Any discrepancies were resolved in a joint discussion facilitated by a third reviewer.

Inclusion and exclusion criteria

All observational and prospective case–control studies comparing NAFLD and non-NAFLD (control) group were included, and the selection process was carried out by two separate reviewers. An article was considered relevant if NAFLD diagnosis was established either by liver biopsy, abdominal ultrasonography, CT scan or MRI with or without the presence of elevated liver enzymes. To evaluate endothelial dysfunction, we selected studies where data for brachial FMD or serum AMDA could be extracted. The assessment of subclinical atherosclerosis was based on studies providing ultrasonographic measured data of mCIMT and/or carotid plaques. We only included studies that presented mCIMT or FMD as mean ± SD. To determine the risk of cardiovascular outcomes in patients with NAFLD, studies looking at carotid stenosis of more than 50% and those with ACS as an end point were selected. Studies not meeting the above criteria were excluded.

Statistical analysis

Analysis was undertaken using RevMan v5.3. Results are reported as odds ratio (OR) or mean difference (MD) with 95% confidence intervals using random-effects model. We assessed heterogeneity using P_HET and I² statistical methods where I values of 25%, 50% and 75% were considered as low, moderate and substantial heterogeneity, respectively. ¹⁵ Publication bias was tested using Funnel plots.

Results

NAFLD and endothelial dysfunction

NAFLD and FMD%

We identified 14 studies evaluating FMD% in 1156 cases and 1158 controls.^{11,13,16–27} There was significant impairment of FMD% in NAFLD cases compared to controls with [MD −4.83 (95% CI: −5.84 to 3.81, p < .00001)] (Figure 1) There was significant heterogeneity between these studies (P_HET < .00001, I² = 95%). Excluding seven of the outlying studies,^{11,13,17,20,23,25,27} eliminated the heterogeneity (P_HET = .09, I² = 45%) and significant association remained [MD −6.71 (−7.95 to 5.47), p < .00001] (Figure 1). There was a low probability of publication bias as demonstrated by a symmetrical funnel plot.

Figure 1.

Random effects meta-analysis on the risk of FMD associated with NAFLD. Forest plot of comparison showing the relationship of FMD% in patients with NAFLD and without NAFLD. FMD, flow-mediated dilation. NAFLD, non-alcoholic fatty liver disease.

ADMA as a predictor of endothelial dysfunction

Five studies (289 cases and 185 controls) investigated serum ADMA as a potential marker of endothelial dysfunction in NAFLD.^{12,14,17,18,28} There was a statistically significant association between NAFLD and increased ADMA levels [MD 0.08 (95% CI: 0.05–0.11, p < .00001)] (Figure 2) with no evidence of heterogeneity (P_HET = .90, I² = 0%) and low probability of publication bias with a symmetrical funnel plot.

Figure 2.

Random effects meta-analysis on the risk of AMDA associated with NAFLD. Forest plot of comparison showing the relationship of AMDA in patients with NAFLD and without NAFLD. AMDA, serum asymmetric dimethylarginine. NAFLD, non-alcoholic fatty liver disease.

NAFLD and subclinical atherosclerosis

Association between NAFLD and CIMT

CIMT was the most extensively investigated measure of subclinical atherosclerosis with 83 studies (21,458 cases and 32,606 controls).^{11–13,15,16,19–22,25–27,29–100} NAFLD was associated with higher CIMT [MD 0.10 (95% CI: 0.09–0.11, p < .00001)] (Figure 3). There was significant heterogeneity between these studies (P_HET < 0.00001, I² = 97%). Excluding 15 outlying studies,^{11,31,43,44,50,55,58,62,63,71,76–78,84,93} did not eliminate heterogeneity (P_HET = .00001, I² = 96%). Funnel plot symmetry suggested low probability of publication bias.

Figure 3.

Random effects meta-analysis on the risk of CIMT associated with NAFLD. Forest plot of comparison showing the relationship of CIMT in patients with NAFLD and without NAFLD. CIMT, carotid intima media thickness. NAFLD, non-alcoholic fatty liver disease.

NAFLD and cardiovascular outcomes

Association between NAFLD and ACS

Twenty-five studies (593,635 cases and 4,915,788 controls) investigated ACS as an endpoint.^{8,10,14,31,101–121} There was significantly increased risk of ACS in patients with NAFLD [OR: 1.95 (95% CI: 1.49–2.55, p < .00001)] (Figure 4). There was significant heterogeneity between the studies (P_HET < .00001, I² = 97%). Excluding 10 outlying studies^{10,102–104,108,111,113,114,118,119} eliminated heterogeneity (P_HET < .00001, I² = 90%). After exclusions, a significant association remained [OR: 2.55 (1.82–3.58, p < .00001)]. There was a low probability of publication bias as demonstrated by a symmetrical funnel plot.

Figure 4.

Random effects meta-analysis on the risk of CHD associated with NAFLD. Forest plot of comparison showing the relationship of CHD in patients with NAFLD and without NAFLD. CHD, coronary heart disease. NAFLD, non-alcoholic fatty liver disease.

Association between NAFLD and carotid stenosis (>50%)

Twelve studies (2646 cases and 2540 controls) investigating carotid plaque as an endpoint were included.^{14,31,55,102,118,122–126} There was significantly increased risk of carotid plaque formation in patients with NAFLD [OR: 2.08 (95% CI: 1.52–2.86, p < .00001)] (Figure 5). There was substantial heterogeneity between the studies (P_HET < .0005, I² = 67%). Excluding two studies,^102,125 eliminated heterogeneity (P_HET = .11, I² = 38%). After exclusions a significant association remained [OR: 2.50 (1.93–3.24, p < .00001)]. There was a low probability of publication bias as demonstrated by a symmetrical funnel plot.

Figure 5.

Random effects meta-analysis on the risk of coronary plaque associated with NAFLD. Forest plot of comparison showing the relationship of coronary plaque in patients with NAFLD and without NAFLD. NAFLD, non-alcoholic fatty liver disease.

Discussion

Affecting about one-third of the western population, NAFLD is the commonest cause of chronic liver disease. It is commonly associated with cardiovascular and metabolic disease but many of these studies have suffered from small sample sizes. Although such individual studies shed little knowledge on any putative relationship, the totality of the body of work should provide a more reliable assessment. The presented meta-analysis robustly shows that NAFLD is strongly associated with endothelial dysfunction, atherosclerosis and acute coronary syndrome. This work is, to the best of our knowledge, the most comprehensive study conducted in this field to-date.

Several studies have evaluated the association between non-alcoholic liver disease and the risk of endothelial dysfunction, subclinical atherosclerosis, and CVD, but many of these previous studies have been small and inconsistent in their conclusions. The results from the studies evaluating endothelial dysfunction have shown more consistent results as the reporting of FMDA% and ADMA were more homogenous compared to CVD. Conversely, studies evaluating CVD reported more conflicting results which could be partially explained by variation of NAFLD definition as well as the diagnostic method of cardiovascular disease. There are two prior reviews, one narrative review published in 2010 by Ghouri et al. and one meta-analysis published in 2016 by Targher et al. both suggesting that CVD was unlikely to be caused by NAFLD diagnosis alone.^127,128

Our work adds to the body of evidence for an association between NAFLD and atherosclerosis but does not assist us in demonstrating a causal relationship. One way to establish a convincing causal relationship would be by demonstrating a reversal of atherosclerosis following treatment for NAFLD using a prospective study model. Notwithstanding the fact that we are not aware of such a study, providing a robust association provides confidence in pursuing the need for establishing causality.

We excluded studies with a poor study design not meeting accepted criteria or those that lacked an appropriate control group. Overall, our results indicate a strong correlation between NAFLD and supported by other smaller meta-analyses.^127,129The strengths of our study include a large number of participants with a diverse age group, all ethnicities, multiple languages and clinical background, including obesity and diabetes, being more representative of the actual population.

Endothelial cells that line blood vessels play a crucial role in maintaining vascular health by regulating blood flow, preventing clot formation, and promoting vasodilation. When endothelial function is compromised, it can lead to various cardiovascular diseases. ADMA (asymmetric dimethylarginine) is an endogenous molecule that inhibits the production of the potent vasodilator nitric oxide (NO) in endothelial cells. Ordinarily, NO helps maintain healthy blood vessels. Elevated levels of ADMA are associated with endothelial dysfunction and reduced NO availability. In the context of metabolic disorders such as NAFLD, endothelial dysfunction and altered ADMA levels are observed. NAFLD frequently coexists with type 2 diabetes mellitus and contributes to the development of atherosclerosis. Studies have shown that NAFLD patients with diabetes exhibit impaired FMD. Specifically, patients with NAFLD are more likely to have FMD ≤7% compared to those without NAFLD. ¹³⁰ Hypertension is also associated with an increased risk of impaired FMD. Interestingly, females appear to be protective against endothelial dysfunction. The mechanisms linking NAFLD, endothelial dysfunction, and ADMA involve insulin resistance, pro-inflammatory cytokines, and disturbed lipid metabolism. ¹³¹ Addressing NAFLD through screening and management is important in mitigating the risk of endothelial dysfunction and subsequent cardiovascular complications in individuals with metabolic disorders.

However, as with all meta-analyses several limitations should be taken into consideration when interpreting our results. The reader should note that by their very nature meta-analysis does not undertake an original new dataset but rather seek to bring together existing data. As such, the accuracy of our result is dependent on the quality and quantity of the publications which we have identified. Our specific definitions of CVD may result in an underestimation of the actual number of cardiovascular complications which may be broader. Our studies that looked at CVD as an outcome also included prevalence of stroke, chronic kidney disease and peripheral artery disease, which are excluded in our review, but could reflect poor cardiovascular health. Additionally, due to the nature of observational studies, it was not possible to assess the progression of participants over time. Despite attempts to exclude potentially biased studies, heterogeneity persisted in some outcomes despite our efforts to iteratively exclude it. These inevitable differences come from the study designs, participant characteristics, and diagnostic tools for NAFLD. Ultrasound was the most commonly used diagnostic tool which has been shown to have reduced sensitivity in milder hepatosteatosis.^132,133 Nevertheless, such imaging techniques are the most pragmatic and commonly used rather than the ‘gold standard’ of liver biopsy which is both invasive and tends to be clinically avoided. Although our population comprises a diverse ethnicity, which aids the strength of our study, to further quantify the relationship between NAFLD and cardiovascular disease, a more detailed analysis of each sub-group is needed, a stratification not provided in most studies. Further, we were unable to undertake the role of potential confounding factors for NAFLD such as central obesity, diabetes, medications (e.g. methotrexate, amiodarone, tamoxifen), liver biopsy size, presence of autoimmune diseases etc). ¹³⁴ As such we temper our conclusions for the practicing clinician. Finally, despite funnel plots appearing broadly symmetrical there is little doubt that publication bias exists within our meta-analysis, as is often the case with such studies. Investigators tend to actively submit for publication work that have positive results, whereas negative studies tend to be lower prioritised. Indeed, journals suffer the same bias by actively encouraging submission of positive results. While we have attempted to mitigate this in our comprehensive study by searching multiple databases and including the largest dataset to-date, publication bias cannot be eliminated.

Our results demonstrated that NAFLD is independently associated with multiple forms of cardiovascular disease. Further studies are required to determine whether patients presenting with such liver disease should be screened for other underlying vascular diseases.