Nowadays, it is well known that CVD occurrence represents the most important factor of morbidity and mortality in patients with NAFLD/MASLD, in fact CVD events are higher than those liver-related [38]. Indeed, NAFLD/MASLD is accounted for being an independent CVD risk factor, even independent from smoke, obesity, diabetes, and metabolic syndrome (and its components, hypertension, dyslipidemia, visceral obesity and hyperglycemia) [39]. This peculiar characteristic is even more interesting today, as the pathogenetic relationship between non-alcoholic steatosis and the dysmetabolic state has been more forcefully established by the new definition of MASLD. A non-invasive assessment of the CVD risk for a MASLD patient should be the most appropriate approach, considering the global increasing burden of this condition, although this evaluation is always difficult. In the present study, we precisely aimed to evaluate the clinical risk of CVD occurrence by calculating it with a well-validated risk score (the 2008 version of the Framingham Hearth Risk score) and correlating it with a mostly non-invasive assessment (a precision medicine multi-OMIC approach) of MASLD severity. We believe that our interesting results clarify how and which MASLD patients should be attentioned for CVD risk. In fact, our overall MASLD cohort presents most of known CVD risk parameters (BMI, glycemia, prevalence of hypertension, diabetes and metabolic syndrome) significantly different, apart from those correlated with the liver disease itself (such as FIB-4 and LSM), but not FHRs itself, compared to age and sex-matched controls. At a first glance, these results look inconsistent with the supposed higher CVD risk in NAFLD patients, however, further analyzing the data we found out that only patients with low risk of advanced fibrosis (group A) and controls had similar FHRs. In fact, FHR scores were significantly different between controls and MASLD patients of group B, C and D (aggregated) (p: 0.024, Mann–Whitney U for independent samples), whereas there was no statistical difference between controls and group A (p: 0.451). Indeed, the group A patients had very likely simple steatosis, a metabolic condition very similar to matched healthy controls. Therefore, our finding seems to confirm, once again, that simple steatosis doesn’t confer any additional risk to its carrier.

Interestingly, analyzing FHRs among the four groups of MASLD, we observed a statistically significant increasing trend (as showed in Table 1 and Fig. 1; p: 0.004). These findings suggest that the non-invasive algorithm used to stratify MASLD patients based on fibrosis, was useful also to predict their CVD risk. In turn, this confirms previous reports that one of the strongest predictors of CVD risk in MASLD is the severity of the liver disease itself [40]. However, to better understand the relationship between CVD risk and MASLD disease we performed univariate and multivariate analyses with FHRs as the dependent variable. In the univariate analysis (Table 2), we observed well knows CVD risk factors associated with higher values of FHRs, such as age, sex and metabolic syndrome (and its components: glycaemia, hypertension, HDL cholesterol) and diabetes, apart from factors associated with liver disease (ALT, FIB-4 and LSM). Interestingly, the SNPs for NAFLD/MASLD risk were not correlated with FHRs. At the multivariate analysis, performed with all the significant variables at the univariate as independent factors, age, HDL, diabetes and FIB-4 were confirmed independently associated with FHRs (Table 3a). However, since FHRs is calculated from age, sex, total cholesterol, HDL cholesterol and blood pressure (other than smoking habits), there is a high risk of an incorporation bias on the statistical significance of these variables. Therefore, we performed another multivariate analysis, excluding Age and HDL cholesterol (Table 3b). Thus, only diabetes and LSM were significantly associated with higher FHRs, demonstrating, again, that liver disease severity (also if assessed in a non-invasive way) is the stronger driver of CVD risk, together with the presence of type 2 diabetes mellitus.

FHRs has the limitation to be precisely a risk score, calculated from clinical parameters, therefore there is no direct measurement of any pathophysiological impairment predisposing the patient to CVD. For this reason, we also performed, in a subgroup of patients, a direct measurement of endothelial dysfunction, which is the early vascular derangement that correlates with CVD [41]. Endothelial dysfunction, measured by peripheral artery tonometry was already present in 38.2% of the 110 patients who underwent the procedure. Overall, there was no differences between the classes of fibrosis risk, however there was a higher percentage (52.6% vs 39.3%) of altered endoPAT in group C (clinical cirrhosis) compared to all the other groups that did not reach the statistical significance due to the small sample size. Moreover, we found a linear correlation between endoPAT values and FHRs, as expected. Since the included patients were all free from active/previous CVD, these findings suggest that an already existing vascular derangement is, once again, correlated with the worst-case scenarios concerning liver disease severity.

Noteworthy, several interesting results emerged from the untargeted metabolomics analyses. First, the PLS-DA analysis, conducted to discover metabolic difference in FHRs, revealed no differences in metabolites profiles below the value of 40%. This is of particular significance, because it is reported that a FHRs > 30% is classified as high CVD risk. In our cohort, the number of patients with FHRs ≥ 40% was 65 out of 466 (13.95%), 21 out of 185 (11.35%) in group A, 18 out of 83 in group B (21.68%), 18 out of 86 in group C (20.93%) and 5 out of 26 in group D (19.24%). Also, 10 out of 73 controls (13.69%) had this FHR score. Once again, there was a statistically significant difference between group A and the other groups (A vs B p: 0.027; A vs C p: 0.037, Mantel–Haenszel), and no difference with controls (p: 0.602). Very interestingly, the metabolites classified as predicting an FHRs > 40% were in the ascorbate and aldarate metabolism, butanoate metabolism, pyruvate metabolism and glycolysis/gluconeogenesis. Ascorbate (vitamin C) is a well-known anti-oxidant, associated with CVD and cancer protection [42]. Butanoate metabolism derangement showed detrimental effects on CVD risk and endothelial dysfunction [43, 44]. In fact, it is well known that the decrease of butyrate-producing bacteria abundance in the Gut-microbiota leads to increased CVD risk [45]. Pyruvate metabolism impairment has been associated with increased fatty acid oxidation, mitochondrial dysfunction (typical of NAFLD/NASH), and reduced cardiac efficiency [46]. Finally, also glycolysis/gluconeogenesis intermediates have been demonstrated to contribute to metabolic classes separation between subjects with a high (> 40%) and a low FHRs. This finding is unsurprising, given the well-known association between diabetes/insulin resistance and CVD risk.

The metabolomics analysis on endothelial dysfunction, measured by EndoPAT, revealed an association between higher concentrations of butanoate, phenylalanine, and tyramine and normal EndoPAT values. In accordance with the forementioned supposed protective mechanisms against CVD, Butanoate concentrations are higher in patients without ED [43,44,45]. Also, higher phenylalanine levels in patients without preclinical ED is in line with previous reports which highlight that its concentrations are inversely correlated with carotid atherosclerosis and CVD risk [47]. As well, low levels of tyramine have been associated with cardiometabolic risk and inflammation in Metabolic Syndrome, therefore, finding it elevated in normal EndoPAT has a plausible explanation [48]. Conversely, thymine, serine, glucose, choline, deoxyglucose, arabinose and pipecolic acid were in higher concentrations in patients with ED (abnormal EndoPAT). Thymine was already demonstrated as a marker of coronary artery disease (CAD) in one metabolomic study [49]. Unsurprisingly, we observed high levels of glucose and deoxyglucose in patients with ED, in accordance with the already known association between hyperglycemic states and CVD risk. As well, high levels of Choline have been associated with major cardiovascular events in patients with myocardial infarction [50]. Pipecolic acid is an intermediate metabolite of the essential amino acid lysine whose degradation metabolites have been associated with the risk of type 2 diabetes and cardiovascular disease [51]. On the contrary, arabinose, a dietary pentose, has been associated with protective effects on diabetes, metabolic syndrome and dyslipidemia [52]. In the same way, high levels of the amino acid Serine have been inversely associated with the risk of peripheral artery disease in previous studies [53]. Therefore, our contradictory findings on these two metabolites need more explanations in future studies; perhaps, however, we could hypothesize a “metabolic attempt” to compensate the endothelial dysfunction in a preclinical stage.

Finally, it has to be noticed that in our cohort, the most studied SNPs for NAFLD/MASLD risk didn’t intercept any adjunctive cardiovascular risk. This could be mostly due to several reasons. First, these SNPs were identified mostly by inferring from the liver disease severity (inflammation and fibrosis), and most of them have specific activity in the processes of inflammation and fibrosis within the liver. Furthermore, even if some of them (i.e. TM6SF2) have been reported to exert also an effect on cardiometabolic risk (both detrimental and protective), it is not known already their penetrance in the general population, apart from NAFLD patients. Finally, our patients showed no statistically different SNPs prevalence among the liver disease severity groups, except for PNPLA3 dominant model that was more represented (as expected) in MASLD overall compared to controls, and in patients with higher risk of fibrosis (Group B) compared to those with lower risk (group A). Therefore, a possible explanation is that these SNPs are to be taken into account only partially for liver disease severity which, in turn, is itself the major marker for CVD risk.

However, a convincing theory postulated that “Metabolic NAFLD” could be different from “Genetic NAFLD” [54], therefore, we performed such an analysis in our population. Thus, we clustered MASLD patients in “Genetic” (subjects having at least one pathological mutation in any of the SNPs we analyzed) and “Non-genetic” (no mutations at all) MASLD. One more time we found that patients with “Genetic MASLD” had an increased risk of liver fibrosis, but not an adjunctive CVD risk (see supplementary Table 2).

Taken as a whole, we believe that our approach, if confirmed and validated in larger cohorts and in other populations, could be a valid strategy to correctly address the clinical risk of MASLD patients, both in terms of what type (e.g. CVD and/or liver-related) and what degree. Our method could possibly represent a cost-effective diagnostic tool for such a large population, which in turn may have an impactful effect in decreasing the costs of National Healthcare systems, both by improving disease progression prevention and diagnosis.

Finally, our present work has some limitations to discuss. First of all, the present data are to be considered cross-sectional. Therefore, no data are available at the time, on the real incidence of CV diseases in our population. However, the score we chose (FHR) has been widely demonstrated to be reliable in predicting CVD risk in longitudinal studies. Moreover, by study protocol, we didn’t collect liver biopsy of every single patient in the cohort, but only in 68 of them (14.59%). Therefore, it is likely that there is no absolute certainty on liver disease severity in each patient. However, our data showed a good discrimination between the classes in terms of both clinical and omics data, thus successfully demonstrating the effectiveness of this diagnostic algorithm, which is also recommended by guidelines. Furthermore, the effective presence of endothelial dysfunction by EndoPAT was measured only in a subgroup of patients. However, our data demonstrated that there was a significant linear correlation between PAT results and FHR score. Finally, it has to be mentioned that our study population had an average age of about 65 years (65.35 years for controls and 66.67 years for MASLD patients), therefore, it is to be considered a high CVD risk category, as universally agreed [55]. Therefore, even if our data demonstrated that, when accounting for the incorporation bias of the FHRs calculation, the presence of T2DM and higher liver stiffness values were the only variables that intercepted the higher CVD risk, it could be advisable to repeat such analyses in further longitudinal studies in younger populations to verify our findings in lower risk settings.