Disease definition and spectrum

Primary steatosis describes the ectopic deposition of fat within the liver that occurs in the absence of inciting factors such as alcohol, drugs, genetic and endocrine disorders [1]. Nonalcoholic fatty liver disease (NAFLD) is an umbrella definition embracing primary steatosis, nonalcoholic steatohepatitis (NASH), NASH-cirrhosis and NASH-hepatocellular carcinoma (HCC), namely the entire spectrum of alcohol-like liver lesions in individuals who, while not drinking of excessive amounts of alcohol, do exhibit dysmetabolic features [2, 3]. Of concern, paralleling the global epidemic of diabesity, NAFLD is rapidly emerging as a globally prevailing cause of chronic liver disease eventually culminating in progressive liver fibrosis, cirrhosis, and liver failure in a proportion of individuals [4]. Irrespective of whether primary steatosis will lead to hepatic damage under the form of concurrent inflammatory changes (NASH) and abnormal extracellular matrix deposition (NASH-fibrosis, NASH-cirrhosis, and NASH-HCC). NAFLD and the metabolic syndrome (MetS), with which it has a dynamic bi-directional relationship, are typically associated with severe extra-hepatic manifestations and complications, such as cardio-renal diseases, metabolic decompensation, and certain forms of extra-hepatic cancers [5–7].

NAFLD heterogeneity

Such a remarkable hepatic and extra-hepatic phenotypic changeability seemingly results from disease heterogeneity as affected by variable genetic background, epigenomic mechanisms, metabolic factors, and environmental cofactors [8, 9]. Importantly, this heterogeneity of clinical phenotypes raises the key question of predicting the odds of disease progression in any given individual patient, aimed at developing effective prevention and management strategies [10]. While the best approach to accomplish this goal remains undefined both in primary care and in specialist practice, ruling out advanced fibrosis with validated noninvasive scores [e.g., either NAFLD fibrosis score (NFS) or fibrosis 4 (FIB-4) index] presently holds on as the leading paradigm [11]. However, although other simple methods for risk stratification, such as evaluation of gamma-glutamyl transferase (GGT) and fatty liver index score (which also includes GGT assessment) are available [12, 13], it is anticipated that multi-stratified techniques to categorize NAFLD patients more accurately will be adopted in the near future [14].

Rationale and aim

Alongside with increasingly advocated personalized and precision medicine approaches, risk stratification has become more and more investigated in the hepatological arena aimed at identifying those primary care patients needing hepatological consultation; as well as those who may benefit most of personalized follow-up schedules and more intensive treatment approaches given that they are prone to the highest risk of severe complications [11, 15]. Additionally, the identification of homogeneous patient populations is of key importance in correctly describing subsets of enrollees in randomized clinical trials (RCTs) of novel drug treatments therefore possibly decreasing the presently disappointing results of RCTs [16].

Within the rapidly evolving scenario described above, the present narrative review article will address the heterogenous pathogenesis of NAFLD and its variable clinical course, as well as the background information for analyzing those available tools useful for predicting the risks of progressive fibrosis, cirrhosis, HCC, metabolic decompensation, cardio-nephro-vascular outcomes, and extra-hepatic cancer. With this backset, my review specifically emphasizes surrogate indices of fibrosis given that the role of liver biopsy out of clinical trials has become increasingly limited [17].

Liver-related outcomes

Several seminal studies have depicted the natural course of NAFLD [24]. Here I report on two of among the most recently published. In 2016, Vilar-Gomez et al. [25] followed for an average of 5.5 years 458 patients with biopsy-proven NAFLD with either fibrosis stage 3 (F3) bridging fibrosis (n = 159), or compensated cirrhosis (299 patients with Child-Turcotte-Pugh class A cirrhosis) recruited in a multi-national study involving Spain, Australia, Hong Kong, and Cuba. Data have shown that the risk of either hepatic decompensation or HCC were higher among individuals with cirrhosis than among patients with F3 fibrosis; conversely, those with F3 fibrosis compared to patients with cirrhosis, had a higher risk of incident vascular events and extra-hepatic cancers. Finally, the cumulative incidence of extra-hepatic malignancies was higher in patients with F3 fibrosis than among those with cirrhosis. Death or liver transplantation, hepatic decompensation, and HCC were strongly associated with baseline cirrhosis and mild steatosis. Details on fibrosis staging adopted in NAFLD/NASH arena may be retrieved elsewhere [26, 27]. In short, in both principal classification systems, i.e., the NAFLD activity score (NAS, NASH Clinical Research Network) and steatosis, activity, and fibrosis (SAF)/fatty liver inhibition of progression (FLIP) algorithm, F3 identifies bridging fibrosis, namely an advanced form of pre-cirrhotic disease [28]. This seminal study was first in demonstrating that NAFLD-cirrhosis is associated with liver-related clinical complication, whereas those with bridging fibrosis are at risk principally of extra-hepatic cancers and vascular events [25].

In 2021, expanding these findings further, Sanyal et al. [29] by following for a median of 4 years 1,773 adults with biopsy-proven NAFLD found that, after adjusting for confounding factors, the stage of fibrosis predicted the odds of any incident event of hepatic decompensation and increased all-cause mortality. In agreement with other reports, this study also found that, compared to those with absent-to-early fibrosis stages (F0 to F2) individuals with fibrosis stage F4 exhibited a higher risk of incident type 2 diabetes (T2D) and chronic kidney disease (CKD) [29]. However, in this study, incident cardiac events and extra-hepatic cancers did not vary across fibrosis stages [29].

As to the risk of primary liver cancer, data suggest that a dysmetabolic milieu, such as seen in NAFLD, is a definite risk factor/co-factor for HCC and an increasingly acknowledged risk factor for cholangiocarcinoma as well [30]. In agreement, a recent meta-analysis by Yi et al. [31] found that cholangiocarcinoma was strongly associated with NAFLD [odds ratio (OR) 2.05 1.53–2.75].

In NAFLD, differently from chronic liver disease owing to other etiologies, HCC risk is not restricted to those exhibiting NAFLD-cirrhosis. Indeed, also non-cirrhotic NAFLD, namely uncomplicated steatosis and NASH, may be precursor lesions of HCC although, among these patients, the incidence of HCC fails to reach those thresholds which justify screening campaigns [32].

The pathomechanisms underlying the association of metabolic fatty liver syndromes with primary liver cancer have recently been reviewed [30, 33].

Extra-hepatic outcomes

Extra-hepatic outcomes may be classified as cardiovascular, metabolic and (extra-hepatic) cancer [34, 35].

Risk stratification in primary care. When should a NAFLD patient be referred to specialist hepatological practice?

An outstanding panel of experts from USA, Europe, Australia, and Asia, have issued the following criteria to follow to identify those NAFLD patients who should be referred from a primary care pathway to hepatological practice [11]. According to these authorities, sequential steps should be taken. It is important to identify those who have multiple metabolic risk factors; T2D; or imaging-proven steatosis or raised transaminases. Such people should undergo further laboratory testing including FIB-4 evaluation, and those with indetermined risk, be submitted to liver stiffness measurement (LSM). Those patients deemed to be at a high risk, defined by LSM > 12 kPa, should undergo hepatological evaluation [11]. Liver stiffness may be measured with different imaging techniques which are based either on magnetic resonance (magnetic resonance elastography) [42] or on ultrasonography. The latter include transient elastography [43], shear wave elastography [44], acoustic radiation force impulse techniques, and strain elastography technique [45].

Risk stratification of advanced fibrosis, cirrhosis and HCC among those with NAFLD

Several studies have addressed risk stratification with reference to variable liver outcomes (Table 1).

These studies are commented as follows. Dongiovanni et al. [46] conducted a methodologically robust study combining data of at-risk probands as well as people from the general population, in which a Mendelian randomization analysis using risk alleles in PNPLA3, TM6SF2, GCKR and MBOAT7, and a polygenic risk score for intra-hepatic fat were utilized. Data support the notion that, probably, it is hepatic steatosis per se that causes incident liver fibrosis, irrespective of inflammatory changes, and that genetic polymorphisms associated with increased intra-hepatic fat content increase the risk of hepatic fibrosis, to the degree predicted by their steatogenic effects [46]. This study indirectly suggests that severity of steatosis and its duration might be utilized for stratifying the risk of fibrosing chronic liver disease, at least among those carriers of “at risk” alleles in PNPLA3, TM6SF2, GCKR and MBOAT7.

Interestingly, although biomarkers are available to identify those fibrotic NAFLD forms, measurement of liver stiffness appears to be more accurate than non-invasive, blood-based scores in ruling out advanced fibrosis. Illustrating this notion, Labenz et al. [47] found that measurement of liver stiffness was able to identify advanced fibrosis more accurately than NFS, FIB-4, and APRI.

Ioannou et al. [48] in their study specifically addressing those variables that might predict the risk of incident HCC among those with NAFLD-cirrhosis found that four out of seven predictors, selected out of 25 potential variables, accounted for most of the observed HCC risk in NAFLD-cirrhosis [48]. This study portends an improved benefit of prediction models risk-based screening strategy over the screen-all approach [48]. Moreover, these simple models are available as web-based tools (www.hccrisk.com) and can be used in individual patients or by healthcare systems or, also, to select high-risk patients to enroll in RCTs.

More recently, De Vincentis et al. [49] reported that, compared to validated clinical and biochemical parameters, the combination of common genetic variants improved the ability to predict severe liver disease both in the setting of general population and in at high risk NAFLD subjects, particularly among those individuals exhibiting metabolic risk factors. This study encourages the utilization of combined PRS-HFC with non-invasive clinical fibrosis scores to identify those individuals who, being at risk of incident severe liver disease, are most at need of aggressive treatment approaches as well as intensive follow-up schedules.

Fujiwara et al. [50] adopted a complex study design. In short, these authors first defined the signature in their derivation set for recurrent HCC. Next, this signature was validated in tissue validation set 1 and tissue validation set 2. Finally, PLS-NAFLD was translated to a serum-protein panel, PLSec-NAFLD, using their computational algorithm, TexSeC, which was externally validated in serum validation set. These authors discovered a 133-gene signature, PLS-NAFLD which was able to predict incident HCC over a 15-year follow-up. High-risk PLS-NAFLD was associated with specific immune cell phenotypes in fibrotic portal tracts along with impaired metabolic regulators. PLS-NAFLD was after validated in independent cohorts of patients with NAFLD who were either HCC naive or HCC. PLS-NAFLD was bioinformatically translated into a four-protein secretome signature, PLSec-NAFLD, which was validated in an independent cohort of HCC-naive patients with NAFLD and cirrhosis. Combination of PLSec-NAFLD with a previously defined index (the etiology-agnostic PLSec-AFP) further improved HCC risk stratification. Therefore, PLS-NAFLD can be used for RCTs and also as a surrogate end-point in studies of HCC chemoprevention [49]. Despite its high academic standard, findings from this study seems to have scarce chances to enter clinical arena by now.

Jambulingam et al. [51] reported that those individuals exhibiting a fast fibrosis progression could accurately be identified through a specific combination of metabolites and lipoproteins which performed better than noninvasive markers. Clearly, those 14 metabolites, that at multivariate analysis were significantly associated with fibrosis progression, are not universally available at hospital laboratories and this study, presently, seems to have high academic importance more than true clinical relevance.

Chen et al. [52] reported that prognostication of liver outcomes may be improved with PNPLA3 genotyping which suggests a more liberal use of this test in clinical practice and in research.

A recent study has followed an innovative approach to predict mild liver fibrosis through analysis of BA serum profile with ultra-performance liquid chromatography coupled with tandem mass spectrometry. To this end Liu et al. [53] found that the combination of serum BAs with widely available clinical and laboratory parameters identified mild fibrosis more accurately than other validated scores of common use [52]. Although this study elegantly shows that assessment of secondary BA levels combined with anthropometric and hepato-metabolic biomarkers accurately identifies mild fibrosis non-invasively, the requirement of sophisticated laboratory machinery hampers a universal diffusion of this strategy.

The technical details of those studies commented above are summarized in Table 1.

Stratification of risk of extra-hepatic cancers among those with NAFLD

Hong et al. [54] speculated that nonalcoholic fatty pancreas (NAFPD) and NAFLD might be risk factors for breast cancer. By evaluating 961 breast cancer patients and 1,006 non-cancer controls, these authors found that, at multivariate analysis, NAFLD, NAFPD, and serum uric acid were independently associated with breast cancer [54]. On these grounds, the authors established a risk assessment model whose rising scores were associated with sharply increasing OR of breast cancer indicating that incident breast cancer may accurately be predicted with the proposed scoring system [54].

Cardiovascular risk assessment in NAFLD

Conceptually, it is uncertain whether it is correct to risk-stratify NAFLD patients utilizing the same risk scores that are validated in the general population. Predictably, such algorithms tend to underscore cardiovascular risk in NAFLD given that they fail to specifically address multiple pro-atherogenic pathogenic NAFLD features, such as subclinical inflammation, insulin resistance, hypertriglyceridemia and amount of intrahepatic fat content [55, 56]. In agreement with this prediction, Wu et al. [57] in their study retrospectively evaluating 10,453 individuals (3,519 with and 6,934 without NAFLD) followed for 116 months and exhibiting 957 clinical and 752 subclinical CVD events, have shown that ultrasonographic assessment of steatosis severity, non-invasive liver fibrosis scores and apolipoprotein profiles predicted the 10-year risk of CVD more accurately than the conventional CVD risk scores [57].

More recently, it is becoming increasingly clear that metabolomic signature plays a key role in precision medicine approaches in NAFLD arena [58]. For example, Martínez-Arranz et al. [59] by evaluating metabolome serum profile in 1,154 individuals with biopsy-proven NAFLD, and from four mouse models of NAFLD with impaired very low-density lipoprotein (VLDL)-triglyceride (TG) secretion, and one with normal VLDL-TG secretion, were able to identify three different metabolic subtypes: A, B, and C, occurring with variable prevalence (47%, 27%, and 26%, respectively). While the percent occurrence of NASH and fibrosis was comparable across these subtypes, serum concentrations and rate of secretion of VLDL-TG were lower among subtype A than subtypes B and C, and so was the 10-year high risk of CVD, measured with the Framingham risk score, and the frequency of patatin-like phospholipase domain-containing protein 3 NAFLD risk allele. Collectively, data demonstrate that metabolomic signatures, mirroring known CVD and genetic risk factors, offer clinically relevant information to stratify the risk of CVD in NAFLD with such metabolomic features [59]. Finally, it should not be neglected that cardiovascular risk assessment is of vital importance in determining the entire natural history of NAFLD and that biomarkers of fibrosis predict cardiovascular risk in these patients [60, 61].