Sex and age as risk factors for non-alcoholic fatty liver disease

The involvement of age and sex in the development of NAFLD has received increased attention in recent years yet the reasons why these are risk factors for NAFLD remain poorly understood. The prevalence of NAFLD is higher in men and is thought to increase into middle age and then decline after the age of 50–60 years^{(Reference Lonardo, Bellentani and Argo14)}. In contrast, pre-menopausal women appear to be relatively protected from NAFLD; however, this protective capacity is lost after the fifth decade of life when the prevalence of NAFLD is thought to be similar in both sexes^{(Reference Lonardo, Bellentani and Argo14,Reference Lonardo, Nascimbeni and Ballestri15)} . The incidence of NASH and cirrhosis is also thought to be greater in both men and women who are ≥50 years of age compared to younger age groups^{(Reference Sayiner, Koenig and Henry2)}. Recent meta-analysis suggests that whilst pre-menopausal women may have a lower risk of NAFLD, women ≥50 years of age may be at an increased risk of NAFLD progression, compared to men of a similar age^{(Reference Balakrishnan, Patel and Dunn-Valadez16)}. Specifically, among older age groups (≥50 years of age), the relative risk of NASH and advanced liver fibrosis was found to be 17 and 56 % higher respectively, in women compared to men^{(Reference Balakrishnan, Patel and Dunn-Valadez16)}. Conversely, the risk of NAFLD progression was not significantly different between men and women in populations with an average age of ≤50 years^{(Reference Balakrishnan, Patel and Dunn-Valadez16)}. Further work is required to elucidate potential mechanisms underlying the apparent increased risk of NAFLD progression in older women. For example, studies exploring sexual dimorphism in liver metabolism have recently linked hepatic actions of oestrogens to lipid metabolism and female reproductive functions^{(Reference Maggi and Della Torre17)}. Whether these or other sexually dimorphic metabolic or endocrine factors are important in NAFLD remains to be investigated^{(Reference Lefebvre and Staels18)}.

Advancing age also increases the risk of hepatic and extra-hepatic complications of NAFLD^{(Reference Lonardo, Bellentani and Argo14)}. Thus it is expected that older patients with NAFLD will have a higher likelihood of overall and disease-specific mortality^{(Reference Bertolotti, Lonardo and Mussi19,Reference Ong, Pitts and Younossi20)} . Whether the association between NAFLD and all-cause mortality is modified by sex is currently unclear. Previous studies suggest a worse outcome in men^{(Reference Ong, Pitts and Younossi20,Reference Bedogni, Miglioli and Masutti21)} , whilst others have found trends suggesting that NAFLD is associated with an increased risk of all-cause mortality in women but not men^{(Reference Liu, Zhong and Tan22)}. Thus, further large prospective cohort studies should explore whether the direction and magnitude of the association between NAFLD and mortality are modified by sex.

White adipose tissue mass and distribution in non-alcoholic fatty liver disease

A wealth of evidence indicates that obesity increases the risk of NAFLD^{(Reference Younossi23–Reference Younossi, Koenig and Abdelatif26)}. Obesity is defined as excess body fat and results from chronic overnutrition. For adults, it is most frequently classified as a weight for height index or BMI and includes underweight or ‘wasting’ (<18⋅5 kg/m2), overweight (≥25 kg/m²), obesity (≥30 kg/m²) and morbid obesity (≥40 kg/m²)^{(Reference Kelly, Yang and Chen27)}. In contrast, waist circumference provides a simpler anthropometric measurement to diagnose central obesity which is an important independent risk factor for NAFLD and an important component of the metabolic syndrome (MetS). As previously described^{(Reference Alberti, Eckel and Grundy28)}, MetS is defined as the presence of three or more of the following criteria; increased waist circumference, hypertriglyceridemia, reduced HDL-cholesterol, hypertension and hyperglycaemia. It is worth highlighting that neither BMI nor waist circumference is considered reliable indicators of adiposity per se since they do not provide an assessment of WAT mass nor volume⁽²⁹⁾. Nonetheless, BMI and waist circumference have proved to be extremely useful measures for population-based studies and firmly established the importance of obesity as a risk factor for NAFLD. Despite this, it is an oversimplification to consider NAFLD solely as a consequence of obesity given the growing evidence indicating that NAFLD can also occur in individuals with a non-obese BMI, or low WAT mass^{(Reference Ye, Zou and Yeo30,Reference Azzu, Vacca and Virtue31)} . It has been proposed that an increase in the accumulation of central WAT and a reduction in the functional capacity of WAT (particularly subcutaneous WAT (SAT)) to store excess energy as TAG are crucial factors that underpin the relationship between obesity, systemic metabolic disease and NAFLD^{(Reference Azzu, Vacca and Virtue31)}.

Studies utilising adipose tissue-targeted technologies coupled with histological assessment have suggested that the hypertrophic expansion of adipocytes within visceral WAT (VAT) rather than SAT is particularly associated with NAFLD. After approximately 4 years of follow up, a larger VAT area was found to be associated with a higher risk of incident NAFLD, whereas larger areas of SAT were associated with regression of NAFLD^{(Reference Kim, Chung and Kwak32)}. Moreover, several recent studies have demonstrated that increased VAT, as opposed to SAT, increases the risk of, and predicts advanced liver fibrosis in patients with NAFLD^{(Reference Eguchi, Eguchi and Mizuta33–Reference Yu, Kim and Kim35)}. Similarly, evidence also indicates that VAT accumulation is an independent risk factor for hepatocellular carcinoma (HCC) recurrence in patients with suspected NASH^{(Reference Ohki, Tateishi and Shiina36)}. Thus, this evidence supports a fundamental hypothesis that ‘the risk of developing metabolic disease associated with obesity is governed by the regional distribution of WAT within the individual, with the expansion of certain fat depots being more strongly associated with metabolic dysfunction than others’^{(Reference Jensen37)}. Collectively, it is likely that the distribution and capacity of SAT to effectively expand and store lipid, rather than the obesity per se, is a pivotal factor in the relationship between increased adiposity and NAFLD risk.

The distribution of WAT is known to differ significantly between sexes, changes with increasing age and has been hypothesised to be partly responsible for the increased prevalence of NAFLD in men and older age groups, particularly post-menopausal women (Fig. 1)^{(Reference De Carvalho, Justice and Freitas38–Reference Karastergiou, Smith and Greenberg40)}. Whilst the mechanisms regulating the distribution of WAT remain largely elusive, evidence indicates that ageing and male sex are associated with a restricted capacity to effectively expand so-called ‘metabolically protective’ SAT depots^{(Reference Mancuso and Bouchard41)}. Whilst pre-menopausal women typically have greater total adiposity, men tend to accumulate greater amounts of VAT with ageing and pre-menopausal women accumulate gluteal femoral SAT which is associated with a lower risk of metabolic disease and NAFLD^{(Reference Eaton and Sethi42)}. In both men and women, older age (i.e. post-menopausal women and men >50 years) is associated with a reduction in the capacity of SAT to expand and an increase in VAT^{(Reference Hughes, Roubenoff and Wood43–Reference Kuk, Saunders and Davidson45)}. The limited capacity of SAT to store TAG in men and with increasing age is likely to re-direct lipid accumulation ectopically in non-adipose tissues, including the liver, leading to lipotoxicity, a chronic local and systemic pro-inflammatory environment and eventually NAFLD development^{(Reference Godoy-Matos, Silva Júnior and Valerio46)}. The importance of effective SAT expansion can be seen in individuals with certain genetic or acquired lipodystrophies that are characterised by the complete or partial absence of SAT^{(Reference Polyzos, Perakakis and Mantzoros47)}. In spite of their often lean appearance, these individuals appear to exhibit much higher rates of NAFLD/NASH progression and other cardiometabolic complications than would be expected based on their BMI alone^{(Reference Azzu, Vacca and Virtue31,Reference Polyzos, Perakakis and Mantzoros47)} . Given this, it is likely that differences in WAT distribution between sexes and changes occurring with increasing age are both important in the increased risk of NAFLD associated with ageing and with male sex.

Fig. 1. Age-related changes in WAT distribution in men and women are associated with increased risk NAFLD, MetS, T2DM and CVD. Sex and age are key factors that modify the risk of NAFLD and NAFLD progression. NAFLD risk is lower in younger women compared to younger men whereas the risk of NAFLD is similar in older men and women (i.e. post-menopausal). Younger women have an increased capacity to preferentially expand gluteal femoral SAT consequently protecting them from NAFLD. Age-associated changes in WAT leads to the redistribution of WAT which is typically characterised by a marked reduction in SAT and increased central metabolically-unfavourable VAT which may partly explain the increased risk of NAFLD associated with ageing in both men and women. WAT distribution is different between men and women, is heavily influenced by ageing and is strongly associated with NAFLD risk. T2DM, type 2 diabetes; MetS, metabolic syndrome; SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue; NAFLD, non-alcoholic fatty liver disease; WAT, white adipose tissue.

Adipose tissue dysfunction and non-alcoholic fatty liver disease

WAT is composed of mature unilocular adipocyte fraction and a stromal vascular fraction, comprised of numerous cell types such as vascular, mesenchymal and immune cells. At a cellular level, WAT expansion can be mediated by an enlargement of individual adipocytes (hypertrophy), an increase in the number of adipocytes (hyperplasia) or a mixture of both. Adipocyte hypertrophy, rather than hyperplasia, is more closely associated with WAT dysfunction and metabolic disease^{(Reference Longo, Zatterale and Naderi48)}. Factors including hypoxia, low-grade chronic inflammation (i.e. metaflammation) and improper extracellular matrix remodelling are thought to limit adipocyte differentiation and the healthy expansion of adipose tissue (hyperplasia)^{(Reference Sethi and Vidal-Puig49,Reference Cawthorn, Heyd and Hegyi50)} . This limit can result in adipocyte hypertrophy, dysfunction, stress and eventually death^{(Reference Rutkowski, Stern and Scherer51,Reference Halberg, Khan and Trujillo52)} . In this context, WAT dysfunction refers to a reduction in the tissues ability to effectively sense and respond to dynamic changes in nutrient availability (i.e. metabolic inflexibility) and can coexist with adipose insulin resistance and metaflammation. Specifically, this dysfunction is thought to affect WAT metabolism and in particular its ability to handle lipids and increase the lipolytic rate of WAT due to a reduction in tissue insulin sensitivity, increasing the flux of non-esterified fatty acids (NEFA) to the liver and consequently increasing the risk of NAFLD^{(Reference Azzu, Vacca and Virtue31,Reference Blüher53–Reference Byrne55)} .

Accompanying the changes in the distribution of WAT, ageing is associated with a marked reduction in insulin, lipolytic and NEFA responsiveness in WAT. This metabolic inflexibility may underly the known association between ageing and increased risk of NAFLD^{(Reference Hughes, Roubenoff and Wood43–Reference Kuk, Saunders and Davidson45)}. The reduction in SAT with ageing in both men and women may in part be driven by a reduction in the adipogenic potential of progenitor cells and the accumulation of senescent adipocytes in aged WAT. Preadipocytes isolated from peripheral SAT in elderly individuals were found to have a reduced rate of replication compared to those isolated from younger individuals^{(Reference Caso, McNurlan and Mileva56)}. Additionally, ageing is associated with an accumulation of senescent adipocyte-derived stem cells within SAT which lack the ability to differentiate into adipocytes in response to metabolic stress, consequently affecting the tissue's capacity to store TAG^{(Reference Schipper, Marra and Zhang57)}. Through their senescence-associated secretory phenotype, senescent adipocyte progenitor cells within WAT are also likely to contribute to WAT inflammation and subsequent metabolic complications^{(Reference Palmer and Kirkland58,Reference Xu, Palmer and Ding59)} .

In addition to ageing, there are also sexually dimorphic differences in WAT function whereby WAT in females is generally more insulin-sensitive, more lipogenic and less susceptible to inflammation than WAT from males. This phenomenon is also strongly associated with differences in sex hormone concentrations^{(Reference Grove, Fried and Greenberg60,Reference Macotela, Boucher and Tran61)} . Menopause appears to associate with a preferential increase in VAT (rather than SAT) in both obese and non-obese women^{(Reference Phillips, Jing and Heymsfield62–Reference Leeners, Geary and Tobler65)}, further supporting a role for sex hormones, such as oestrogen, in regulating the beneficial distribution and function of WAT. Circulating concentrations of oestrogen decrease markedly after menopause which is thought to lead to the redistribution of lipids into VAT and the liver which, in combination with overnutrition, increases the risk of VAT accumulation and NAFLD in post-menopausal women^{(Reference DiStefano66)}. Pre-clinical studies utilising ovariectomised murine models also support a causative relationship between reduced oestrogen production, increased VAT mass and the development of NASH^{(Reference Eaton and Sethi42,Reference DiStefano66–Reference Rogers, Perfield and Strissel68)} . Whilst an in-depth discussion of the role of oestrogen within WAT is beyond the scope of this review (see other relevant reviews^{(Reference Eaton and Sethi42,Reference Lizcano and Guzmán69,Reference Cooke and Naaz70)} ), it is thought that the increased expression of oestrogen receptor α in the gluteal femoral SAT of premenopausal women promotes lipoprotein lipase activity and accumulation of TAG in adipocytes within this depot^{(Reference Frank, de Souza Santos and Palmer71)}. Thus, it is likely that differences in WAT function (partly driven by differences in sex hormone concentrations and the expression of functional target receptors) is an important factor underlying the observed differences in NAFLD risk between men and women. Furthermore, changes in WAT with ageing are likely to exacerbate WAT dysfunction associated with a state of chronic energy surplus and are likely to have an important role in the increased risk of NAFLD associated with older age.

Adipokines and non-alcoholic fatty liver disease

WAT is an endocrine tissue capable of secreting a wide range of adipokines which have various roles in the regulation of whole-body energy homeostasis and inter-organ communication^{(Reference Funcke and Scherer72)}. The aberrant production of these adipokines has been linked to multiple obesity-related metabolic diseases. Amongst these adipokines, leptin and adiponectin are predominately produced by adipocytes. In addition to its well-established role in regulating appetite and energy homeostasis^{(Reference Sethi and Vidal-Puig49,Reference Montague, Farooqi and Whitehead73)} , leptin exerts a dual action on hepatic function and NAFLD severity. Recent meta-analyses including an analysis of over thirty studies indicated that circulating concentrations of leptin are elevated in patients with NAFLD compared to healthy controls and supports a positive relationship between leptin and NAFLD^{(Reference Polyzos, Aronis and Kountouras74)}. As recently highlighted^{(Reference Polyzos, Kountouras and Mantzoros75,Reference Jiménez-Cortegana, García-Galey and Tami76)} , under normoleptinemia conditions, leptin is thought to suppress hepatic glucose production and hepatic lipogenesis thus providing an insulin-sensitising anti-steatotic effect. Conversely, in the context of chronic hyperleptinemia as is common in obesity, a state of leptin resistance can result, which may also contribute to the NASH phenotype. It is suggested that in the liver, high concentrations of leptin can increase the expression of matrix remodelling enzymes via interacting with leptin receptors on Kupffer and sinusoidal endothelial cells, in turn activating hepatic stellate cells, and possibly contributing to liver fibrosis^{(Reference Polyzos, Kountouras and Zavos77)}.

Sexual dimorphism has also been reported for leptin expression^{(Reference Saad, Damani and Gingerich78)}. Despite their lower risk of NAFLD, circulating concentrations of leptin are higher in pre-menopausal women compared to age-matched men and higher leptin levels are thought to be driven by both greater adiposity and an increased production rate of leptin per unit mass of WAT in women compared to men^{(Reference Castracane, Kraemer and Franken79)}. In both men and women, circulating concentrations of leptin are thought to gradually decline with ageing, with reductions being most noticeable in women compared to men whilst appearing to be independent of menopausal status^{(Reference Castracane, Kraemer and Franken79,Reference Isidori, Strollo and Morè80)} . Despite these findings, it is currently unknown whether differences in circulating concentrations of leptin between sexes and age groups have an impact on the risk of NAFLD.

Similar to leptin, a wealth of studies indicate that the circulating concentrations of adiponectin, the most systemically abundant adipokine, are altered in patients with NAFLD (as reviewed in^{(Reference Boutari and Mantzoros81)}). Adiponectin is a hepatoprotective adipokine that has well-established anti-inflammatory^{(Reference Maeda, Shimomura and Kishida82–Reference Tilg and Hotamisligil84)} and insulin-sensitising effects^{(Reference Yamauchi, Kamon and Minokoshi85)} both systemically and within the liver. Meta-analysis indicates that adiponectin concentrations are significantly lower in patients with NAFLD compared to healthy controls; furthermore, NASH is associated with lower adiponectin when compared to simple steatosis^{(Reference Polyzos, Toulis and Goulis86)}. Conversely, adiponectin concentrations are thought to increase in patients with NAFLD-cirrhosis potentially due to a reduction in the hepatic clearance of adiponectin and/or an increase in its production as a result of the tissue repair process associated with NAFLD-cirrhosis^{(Reference Sohara, Takagi and Kakizaki87–Reference Polyzos, Kountouras and Zavos89)}. Along with its well-established role in promoting hepatic insulin sensitivity^{(Reference Buechler, Wanninger and Neumeier90,Reference Polyzos, Kountouras and Zavos91)} , evidence indicates that adiponectin also has antifibrogenic effects via inhibiting the proliferation of hepatic stellate cells^{(Reference Adachi and Brenner92)}. Whilst the role of adiponectin in ageing remains uncertain, it is thought that circulating concentrations of adiponectin are paradoxically increased in older age and are positively associated with physical disability and mortality in elderly individuals^{(Reference Kizer, Arnold and Strotmeyer93)}. Furthermore, some evidence suggests that the association between adiponectin and ageing may be modified by sex^{(Reference Adamczak, Rzepka and Chudek94)}. In addition to leptin and adiponectin, a wealth of other studies have demonstrated that numerous other adipokines may be involved in the development and progression of NAFLD (Table 1). It should be noted that there is a substantial amount of conflicting evidence regarding the changes in circulating concentrations of other adipokines in the context of NAFLD and little is known about the potential pathological role of these adipokines in NAFLD (Table 1). Moreover, further studies are required to elucidate whether the effects of age and sex on adipokine production influences NAFLD risk.

Table 1. Changes in circulating concentrations of adipokines and their potential roles in NAFLD

HSC, hepatic stellate cells; NAFLD, non-alcoholic fatty liver disease; RBP-4, retinol binding protein-4.

In contrast to classic adipocyte-derived adipokines leptin and adiponectin, studies investigating changes in circulating concentrations of other adipokines in patients with NAFLD are largely inconsistent. Similarly, whilst the role of leptin and adiponectin in the development and progression of NAFLD remains somewhat debated, there is currently very little known about the potential roles of other adipokines on hepatic function and NAFLD. It should be noted that the expression of many adipokines (e.g. chemerin and RBP-4) is not restricted to WAT and also occurs within other tissues including the liver. Consequently, whilst changes in the secretion of WAT-derived adipokines may contribute to altered circulating concentrations, other sources (particularly hepatic) may also influence circulating concentrations, hepatic function and NAFLD.

WAT dysfunction and changes in adipokine secretion are also strongly associated with increased low-grade chronic inflammation in WAT (metaflammation); characterised by the infiltration of various leucocytes, an increase in the ratio of proinflammatory/anti-inflammatory macrophages and leucocytes and the increased presence of crown-like structures (dying adipocytes surrounded by pro-inflammatory macrophages)^{(Reference Reilly and Saltiel95)}. Consequently, metaflammation in WAT (particularly VAT inflammation) is associated with an increased expression of pro-inflammatory cytokines such as IL-6, IL-1β, TNF-α and monocyte chemoattractant protein-1^{(Reference Mohamed-Ali, Flower and Sethi96–Reference Sethi and Hotamisiligil98)}. Some of these have been shown to contribute to local insulin resistance, elevated fatty acid lipolysis, anti-adipogenesis and pro-inflammatory macrophage infiltration^{(Reference Sethi and Vidal-Puig49,Reference Cawthorn and Sethi97–Reference Ota101)} . This in turn can impact both metabolic and endocrine functions of WAT. Evidence from murine diet-induced obesity studies indicates that WAT inflammation and reductions in protective anti-inflammatory lipokines such as palmitoleic acid may be important in the development of NASH^{(Reference Duval, Thissen and Keshtkar102–Reference Souza, Teixeira and Biondo104)}.

By virtue of its anatomical links, via the portal vein, increased VAT inflammation is of particular importance in NAFLD/NASH since VAT-derived inflammatory cytokines (other adipokines, lipokines and metabolites (e.g. NEFA)) are initially transported to the liver and therefore may exacerbate NAFLD severity. Consequently, this may in turn increase the associated risk of T2DM and CVD^{(Reference Item and Konrad105,Reference Kabir, Catalano and Ananthnarayan106)} . Collectively, findings indicate that changes in the production of adipokines and increased WAT inflammation may contribute to NAFLD via modulating local and hepatic function, inducing insulin resistance and modulating the local and systemic pro-inflammatory conditions. Whether these changes in WAT function contribute to the increased risk of extra-hepatic diseases associated with NAFLD independently requires further investigation.

Intestinal dysfunction, dysbiosis and non-alcoholic fatty liver disease

Emerging evidence now suggests that changes in gut microbiota (GM) (i.e. dysbiosis) and intestinal function may exacerbate WAT dysfunction which may indirectly contribute to metabolic dysfunction and NAFLD^{(Reference Lundgren and Thaiss107)}. The gastrointestinal tract is the first point of contact for ingested nutrients where it has an integral role in nutrient breakdown and absorption, regulation of whole-body energy homeostasis and is an important host defence barrier. Occupying the gastrointestinal tract is an extensive number of microorganisms, collectively known as the GM which are thought to modulate local and distal tissue function via a range of complex mechanisms^{(Reference Valdes, Walter and Segal108,Reference Zmora, Suez and Elinav109)} . The microbial organisms occupying the gastrointestinal tract mainly include bacteria, archaea, fungi and viruses (predominantly bacteriophages); however, studies exploring the role of the GM in NAFLD have predominantly focused on bacteria^{(Reference Hu, Lin and Kong110,Reference Camarillo-Guerrero, Almeida and Rangel-Pineros111)} .

A plethora of studies have revealed that GM dysbiosis is associated with and is a contributing factor to NAFLD^{(Reference Canfora, Meex and Venema112–Reference Le Roy, Llopis and Lepage115)}. The dominating phyla within human GM are Bacteroidetes and Firmicutes with a significant inter-individual variation in the GM at lower taxonomical levels^{(Reference Bäckhed, Ley and Sonnenburg116,Reference Mouzaki, Comelli and Arendt117)} . Previous evidence indicates that the relative abundance of Bacteroidetes is lower in patients with NASH compared to those with hepatic steatosis and healthy controls^{(Reference Mouzaki, Comelli and Arendt117)}. More recently, Bacteroides abundance was found to be significantly increased in patients with NASH and the abundance of Ruminococcus was increased in patients with liver fibrosis^{(Reference Boursier, Mueller and Barret118)}. As recently reviewed^{(Reference Aron-Wisnewsky, Vigliotti and Witjes113)}, this shifting in GM in relation to NAFLD severity is supported by numerous other studies. Indeed, the presence of bacteria belonging to the Proteobacteria phylum was increased significantly in patients with ≥F3 when compared to patients with F0–F2 liver fibrosis (Table 2)^{(Reference Loomba, Seguritan and Li119)}. Emerging evidence also supports a strong link between the GM and NAFLD-cirrhosis indicating that the composition of the GM may be a useful tool for the identification and staging of NAFLD. Utilising a unique twin and family study design, one study identified a specific GM signature that had a robust diagnostic accuracy, with an area under the receiver operating characteristic of 0⋅92, for the detection of NAFLD-cirrhosis^{(Reference Caussy, Tripathi and Humphrey120)}. Further work demonstrated the robustness and potential universal applicability of this microbiome signature of NAFLD-cirrhosis in two independent cohorts across geographically and culturally distinct populations^{(Reference Oh, Kim and Caussy121)}. However, given the impact of host genetics and environmental factors on the composition of GM^{(Reference Li, Peng and Zhou122)}, it is unlikely that a single GM signature will be able to distinguish between NAFLD phenotypes at an individual level.

Table 2. Histological definitions of liver fibrosis stages and corresponding liver-biopsy validated liver VCTE cut-off values

VCTE, vibration-controlled transient elastography; kPa, kilopascal; PPV, positive predictive value; NPV, negative predictive value; NASH, non-alcoholic steatohepatitis.

Liver fibrosis stages and corresponding histological definitions are based on the NASH clinical scoring network scoring system^{(Reference Kleiner, Brunt and Van Natta201)}. Liver VCTE cut-off values are based on the findings from a recent large validation study^{(Reference Eddowes, Sasso and Allison202)}. The liver VCTE threshold of 8⋅2 kPa was found to have a: sensitivity of 0⋅71 (0⋅64–0⋅77), specificity of 0⋅70 (0⋅62–0⋅77), PPV of 0⋅78 (0⋅71–0⋅83) and NPV of 0⋅61 (0⋅54–0⋅69) for the identification of ≥F2 liver fibrosis. For the prediction of ≥F3 liver fibrosis, 9⋅7 kPa was found to have a sensitivity of 0⋅71 (0⋅62–0⋅78), specificity of 0⋅75 (0⋅69–0⋅80), PPV of 0⋅63 (0⋅55–0⋅71) and NPV of 0⋅81 (0⋅74–0⋅85). For the prediction of ≥F4 fibrosis, 13⋅6 kPa was found to have a sensitivity of 0⋅85 (0⋅69–0⋅95), specificity of 0⋅79 (0⋅74–0⋅83), PPV of 0⋅29 (0⋅24–0⋅57) and NPV of 0⋅98 (0⋅95–0⋅99).

Miele et al. were the first to identify that patients with NAFLD generally have increased intestinal permeability and alterations in intestinal tight junction integrity (observed as a reduction in zonula occludens-1 within intestinal crypt cells), compared to healthy subjects^{(Reference Miele, Valenza and La Torre123)}. Recent meta-analysis found that 39⋅1 % of NAFLD patients had evidence of increased intestinal permeability compared to 6⋅8 % of healthy controls (OR 5⋅08, 95 % CI 1⋅98, 13⋅05)^{(Reference Luther, Garber and Khalili124)}. Furthermore, subgroup analysis indicated that there was a higher incidence of increased intestinal permeability in patients with NASH compared to patients with simple steatosis^{(Reference Luther, Garber and Khalili124)}. It is generally well-accepted that the increased intestinal permeability commonly seen in NAFLD facilitates the translocation of GM-derived metabolites and bacterial products (such as lipopolysaccharides (LPS) and ethanol) which may in turn contribute to metaflammation and the pathogenesis of NAFLD^{(Reference Kolodziejczyk, Zheng and Shibolet125)}.

In addition to altered GM and intestinal permeability, the abundance of GM-dependent metabolites is thought to be altered in NAFLD, many of which may be detected in stool samples and may offer a tool for the assessment of disease severity. For example, work comparing the abundance of distinct stool metabolites in patients with NAFLD-cirrhosis v. healthy subjects revealed 17 metabolites which, in combination, were able to accurately detect the presence of NAFLD-cirrhosis (AUROC 0⋅91, 95 % CI 0⋅89, 0⋅93)^{(Reference Oh, Kim and Caussy121)}. Thus, evidence is accumulating to suggest that accumulation of certain microbial species, changes in intestinal function and increased intestinal permeability are likely to contribute not only to the pathogenesis of NAFLD but also to increased liver disease severity. Further studies are required to elucidate the potential role of non-bacterial species within the GM on the development and progression of NAFLD.

Intestinal dysfunction, dysbiosis and links with white adipose tissue function in non-alcoholic fatty liver disease

Associated with WAT dysfunction are changes in intestinal function and GM dysbiosis, which have also been proposed to be key factors contributing to NAFLD. Receiving about 70 % of its blood supply from intestinal vascularisation, the liver is constantly exposed to the metabolic products, toxins and nutrients produced by the GM^{(Reference Ridlon, Kang and Hylemon126)}. It has been suggested that when in a dysbiotic state, GM may contribute to the development and progression of NAFLD via a range of pathways; including changes in dietary energy harvest^{(Reference Bäckhed, Manchester and Semenkovich127,Reference Bäckhed, Ding and Wang128)} , alterations in SCFA production (particularly butyrate)^{(Reference Rau, Rehman and Dittrich129,Reference Zhou, Pan and Xin130)} , increased bacterial LPS translocation^{(Reference Kolodziejczyk, Zheng and Shibolet125,Reference Cani, Amar and Iglesias131)} , alternations in bile acid profiles^{(Reference Ferslew, Xie and Johnston132)} and increased endogenous ethanol production^{(Reference Zhu, Baker and Gill133)}. Indeed, the potential effects of these factors on hepatic function and NAFLD have been discussed in various recent reviews^{(Reference Hu, Lin and Kong110,Reference Canfora, Meex and Venema112–Reference Jiang, Zheng and Zhang114,Reference Kolodziejczyk, Zheng and Shibolet125,Reference Grabherr, Grander and Effenberger134)} ; furthermore, alterations in appetite-regulating gut hormones are also likely to have an important role in the development and progression of NAFLD, as recently reviewed^{(Reference Mells and Anania135–Reference Koukias, Buzzetti and Tsochatzis137)}.

Disruptions in intestinal permeability associated with obesity and NAFLD are likely to be accompanied by a reduction in the integrity of intestinal tight junctions^{(Reference Miele, Valenza and La Torre123,Reference Durkin, Childs and Calder138)} . Increased intestinal permeability in the presence of GM dysbiosis is thought to facilitate the translocation of bacterial products including pro-inflammatory endotoxins such as LPS. Circulating concentrations of LPS were found to be significantly higher in patients with NAFLD compared to healthy controls^{(Reference Harte, da Silva and Creely139,Reference Kitabatake, Tanaka and Fujimori140)} and have been shown to be positively associated with the expression of pro-inflammatory genes within both VAT and SAT in individuals with obesity^{(Reference Clemente-Postigo, Oliva-Olivera and Coin-Aragüez141)}. This is supported by evidence from pre-clinical murine studies indicating that increased LPS may directly contribute to WAT inflammation and increase the release of WAT-derived pro-inflammatory cytokines^{(Reference Chang, Sia and Chang142)}. Accompanying these findings, various other studies have proposed additional mechanisms by which changes in intestinal function and GM dysbiosis may impact NAFLD development both directly and in-directly via detrimentally impacting WAT function (Table 3 and Fig. 2).

Fig. 2. NAFLD is associated with changes in gut microbiota-derived factors that can alter hepatic and WAT function Changes in GM in NAFLD result in alterations in the production of various metabolites/factors that are thought to contribute to NAFLD both directly (i.e. by directly impacting hepatic function) and indirectly through detrimentally influencing WAT function. As highlighted on the left, intestinal eubiosis and healthy gut function (such as that typically found in young individuals) promotes intestinal barrier integrity and homeostasis whilst restricting the production and dissemination of metabolically detrimental factors (such as LPS and endogenous ethanol) into circulation, the liver and WAT. Conversely, as highlighted on the right, intestinal dysbiosis (such as that often associated with older age) leads to alterations in various GM-derived factors/metabolites that impair the function of tight junction-associated proteins located within the intestinal epithelium. Consequently, these changes are thought to contribute to an increased risk of NAFLD both directly (via inducing hepatic mitochondrial function, inflammation and steatosis) and indirectly through detrimentally impacting WAT function (impairing WAT expansion, metabolic flexibility and increasing the production of pro-inflammatory cytokines). The increased production of inflammatory cytokines is thought to lead to a state of chronic low-grade inflammation which is likely to further disrupt the function of tight junction-associated proteins, thus forming a vicious cycle of worsening metabolic dysfunction and NAFLD disease severity. GM, gut microbiota; LPS, lipopolysaccharide; TMAO, trimethylamine N-oxide; NAFLD, non-alcoholic fatty liver disease; WAT, white adipose tissue.

Table 3. Changes in GM-derived factors/metabolites in NAFLD and their proposed effects in WAT and the liver

LPS, lipopolysaccharide; HSCs, hepatic stellate cells; TMAO, trimethylamine N-oxide; NAFLD, non-alcoholic fatty liver disease; WAT, white adipose tissue.

Changes in the production of various GM-derived metabolites/factors may contribute to the development of NAFLD both directly (i.e. via a direct action within the liver) and indirectly via affecting the function of WAT. Whilst evidence of altered circulating concentrations of certain factors (namely LPS, endogenous ethanol and TMAO) is well-reported in patients with NAFLD, changes in circulating concentrations of other factors (particularly SCFA) require further investigation. Furthermore, more research is required to elucidate the potential contribution of endogenously produced ethanol on WAT dysfunction in the context of NAFLD.

Evidence also suggests that the composition of the GM and intestinal function can differ between sexes and such differences may partly explain differences in the risk of metabolic disease between sexes^{(Reference Kim, Unno and Kim143–Reference Sheng, Jena and Liu145)}. Similarly, changes in GM composition and intestinal function are strongly associated with ageing and are likely to contribute to the increased risk of NAFLD associated with older age both directly and indirectly via exacerbating WAT dysfunction^{(Reference Rea, O'Sullivan and Shanahan146–Reference Nagpal, Mainali and Ahmadi148)}. Similar to obesity, ageing is also associated with disruptions in intestinal permeability subsequently facilitating the translocation of bacterial products such as LPS which are known to contribute to both hepatic and WAT dysfunction (Fig. 2)^{(Reference Tran and Greenwood-Van Meerveld149,Reference Man, Bertelli and Rentini150)} . Collectively, existing studies demonstrate the existence of a gut-WAT axis which, in addition to the well-established gut-liver axis, may indirectly contribute to NAFLD pathogenesis. Furthermore, differences in GM composition and intestinal function between men and women and with ageing may contribute to both hepatic and WAT dysfunction and subsequently drive the development of NAFLD.

Non-alcoholic fatty liver disease and extra-hepatic complications