Prothrombotic factors in histologically proven nonalcoholic fatty liver disease and nonalcoholic steatohepatitis


  • An Verrijken,

    Corresponding author
    1. Department of Endocrinology, Diabetology and Metabolism, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    • Address reprint requests to: Luc Van Gaal, M.D., Ph.D., Department of Endocrinology, Diabetology and Metabolism, Antwerp University Hospital, Wilrijkstraat 10, 2650 Edegem (Antwerp), Belgium. E-mail:; fax: +32-3-825.49.80.

    Search for more papers by this author
    • These authors contributed equally to this work.

  • Sven Francque,

    1. Department of Gastroenterology and Hepatology, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ilse Mertens,

    1. Department of Endocrinology, Diabetology and Metabolism, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    Search for more papers by this author
  • Janne Prawitt,

    1. Université Lille Nord de France, INSERM U1011, UDSL, Institut Pasteur de Lille, Lille, France
    Search for more papers by this author
  • Sandrine Caron,

    1. Université Lille Nord de France, INSERM U1011, UDSL, Institut Pasteur de Lille, Lille, France
    Search for more papers by this author
  • Guy Hubens,

    1. Department of Abdominal Surgery, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    Search for more papers by this author
  • Eric Van Marck,

    1. Department of Pathology, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    Search for more papers by this author
  • Bart Staels,

    1. Université Lille Nord de France, INSERM U1011, UDSL, Institut Pasteur de Lille, Lille, France
    Search for more papers by this author
  • Peter Michielsen,

    1. Department of Gastroenterology and Hepatology, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    Search for more papers by this author
  • Luc Van Gaal

    Corresponding author
    1. Department of Endocrinology, Diabetology and Metabolism, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
    • Address reprint requests to: Luc Van Gaal, M.D., Ph.D., Department of Endocrinology, Diabetology and Metabolism, Antwerp University Hospital, Wilrijkstraat 10, 2650 Edegem (Antwerp), Belgium. E-mail:; fax: +32-3-825.49.80.

    Search for more papers by this author

  • See Editorial on Page 16

    Potential conflict of interest: Nothing to report.

  • This work is part of the project “Hepatic and adipose tissue and functions in the metabolic syndrome” (HEPADIP), which is supported by the European Commission as an Integrated Project under the 6th Framework Program (contract LSHM-CT-2005-018734).


An independent role of nonalcoholic fatty liver disease (NAFLD) in the development of cardiovascular disease has been suggested, probably mediated through increased levels of prothrombotic factors. Therefore, we examined whether NAFLD is linked to a prothrombotic state, independently of metabolic risk factors in a large single-center cohort of overweight/obese patients. Patients presenting to the obesity clinic underwent a detailed metabolic and liver assessment, including an extensive panel of coagulation factors. If NAFLD was suspected, a liver biopsy was proposed. A series of 273 consecutive patients (65% female) with a liver biopsy were included (age, 44 ± 0.76 years; body mass index: 39.6 ± 0.40 kg/m2). Increase in fibrinogen, factor VIII, and von Willebrand factor and decrease in antithrombin III correlated with metabolic features, but not with liver histology. Levels of plasminogen activator inhibitor-1 (PAI-1) increased significantly with increasing severity of steatosis (P < 0.001), lobular inflammation (P < 0.001), ballooning (P = 0.002), and fibrosis (P < 0.001). Patients with nonalcoholic steatohepatitis had significantly higher PAI-1 values than those with normal liver (P < 0.001). In multiple regression, including anthropometric and metabolic parameters, steatosis remained an independent predictor of PAI-1 levels, explaining, together with fasting C-peptide and waist circumference, 21% of the variance in PAI-1. No consistent correlations with histology were found for the other coagulation factors. Conclusion: In obesity, NAFLD severity independently contributes to the increase in PAI-1 levels, whereas other coagulation factors are unaltered. This finding might, in part, explain the increased cardiovascular risk associated with NAFLD. (Hepatology 2014;58:121–129)


alanine aminotransferase


alkaline phosphatase


activated protein C resistance


activated partial thromboplastin time


aspartate aminotransferase


body mass index


computed tomography


cardiovascular disease


factor VII coagulant


factor VIII coagulant


factor XI coagulant


gamma-glutamyl transpeptidase


glycated hemoglobin


high-density lipoprotein cholesterol


insulin resistance estimation using homeostasis model assessment


high sensitive c-reactive protein


international diabetes federation


isolated impaired fasting glucose


isolated impaired glucose tolerance


low-density lipoprotein cholesterol


metabolic syndrome


magnetic resonance spectroscopy


nonalcoholic fatty liver disease


NASH activity score


nonalcoholic steatohepatitis


NASH Clinical Research Network


third adult treatment panel of the national cholesterol education program


plasminogen activatior inhibitor-1


plasma prothrombin


portal vein thrombosis


standard error of the mean




upper limit of normal




visceral adipose tissue


von Willebrand factor antigen.

Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver injury worldwide.[1] NAFLD is highly associated with obesity and the metabolic syndrome (MetS).[2] It might be accompanied by liver inflammation and signs of hepatocyte damage, leading to nonalcoholic steatohepatitis (NASH).[1] NASH can lead to progressive fibrosis and cirrhosis.[1]

In recent years, a possible role of NAFLD in the development of cardiovascular disease (CVD) has been suggested and this relationship seems to be independent from obesity and the MetS.[3, 4] However, the exact mechanisms linking NAFLD to CVD are still poorly understood. Different pathways could be involved, one of which is overproduction of prothrombotic factors in patients with fatty liver.[5]

Prothrombotic factors have also been identified in patients with cirrhosis and portal vein thrombosis (PVT)[6] and might be implicated in the pathogenesis of several liver diseases.[7] Recently, Villa et al. reported a lower rate of complications in patients with cirrhosis treated with low-molecular-weight heparins, suggesting again the role of local thrombotic events in the progression of liver disease.[8]

Different human studies have assessed the relationship between liver steatosis and prothrombotic factors. Although most of these studies report several alterations in different coagulation factors, results are inconsistent. This might, in part, be explained by methodological issues. Several studies used elevated serum liver tests,[9, 10] ultrasonography,[9, 11] or magnetic resonance spectroscopy (MRS)[12, 13] for the diagnosis of liver dysfunction. Only a few studies used data obtained from liver biopsy.[11, 14-16] Low patient numbers and retrospectivity might further account for the discrepant results.

The aim of the present study was therefore to assess the relation between an extended set of coagulation factors and the histological subtypes and severity grades of NAFLD in patients with obesity, independent of metabolic risk factors (including visceral fat measurement by computed tomography [CT]). Therefore, NAFLD and coagulation status were assessed in a large group of thoroughly characterized, consecutively recruited subjects who were overweight or obese, using liver biopsy as the gold standard for diagnosis and staging of NAFLD/NASH.[17]

Patients and Methods

Patient Selection

Patients visiting the obesity clinic of the Antwerp University Hospital (Antwerp, Belgium) were prospectively recruited. Patients were predominantly Caucasian, and at enrollment, none were involved in a weight management program. Every patient underwent a metabolic work-up combined with a liver-specific program (see the Supporting Materials), both approved by the ethics committee of the Antwerp University Hospital and requiring written informed consent. Because diabetes has a specific clinical picture, patients already known to have diabetes were not included. Patients with significant alcohol consumption (self-reported; >20 g/day or intermittent, but important, drinking) were excluded.

The possibility of liver involvement was defined by abnormal liver tests (aspartate aminotransferase [AST] and/or alanine aminotransferase [ALT] and/or gamma-glutamyl transpeptidase [GGT] and/or alkaline phosphatase [ALP]) and/or liver ultrasound (US) abnormality (steatotic liver)[18] and/or abnormal aminopyrin breath test.[19] For ALT, three different cut-off levels were used: the upper limit of normal (ULN) set by the biochemistry laboratory (56 IU/L), the classical cut-off of 40 IU/L, and the limits proposed by Prati et al. (30 IU/L in men and 19 IU/L in women).[20] When one or more of these criteria were met, a liver biopsy was proposed. For patients going to surgery, a liver biopsy was proposed regardless of the preset criteria. Liver biopsy was performed (after additional informed consent) percutaneously (16-G Menghini) or perioperatively (14-G Tru-Cut).

Laboratory Analysis

The following coagulation factors were determined: thrombocytes; activated partial thromboplastin time (APTT); plasma prothrombin (PT); fibrinogen; factor VII coagulant (FVII:c); factor VIII coagulant (FVIII:c); von Willebrand factor antigen (vWF:Ag); factor XI coagulant (FXI:c); antithrombin III; protein C; activated protein C resistance (APC-R); platelet function; and plasminogen activator inhibitor-1 (PAI-1). Methodological aspects are listed in the Supporting Materials.

Liver Biopsy

Hematoxylin and eosin stain, Sirius Red stain, reticulin stain, and Perl's iron stain were routinely performed on all biopsies, which were reanalyzed by one experienced pathologist (E.V.M.) blinded for clinical data.

The diagnosis of NASH was made according to Brunt et al.,[21] and the different features were scored according to the NASH Clinical Research Network (NASH-CRN) scoring system.[22] Patients were divided in two groups according to the presence or absence of NASH, as defined by Brunt et al.,[21, 23] requiring the combined presence of some degree of steatosis, ballooning, and lobular inflammation to make the diagnosis of NASH (detailed information in the Supporting Materials).

In addition, the NASH activity score (NAS) was calculated as the sum of the scores for steatosis, ballooning and lobular inflammation.[22] Since the publication of the NASH-CRN scoring system, the definition of NASH by a NAS ≥5 has been widely adopted. However, this definition is slightly different from the former definition of Brunt et al.[23] and its use outside the setting of interventional studies was recently questioned.[21] Therefore, we used both definitions. Furthermore, the NAS-based diagnostic classification therefore refers to three categories: NAS <3 (no NASH); NAS 3 or 4 (indeterminate NASH); or NAS ≥5 (definite NASH).


The presence of the MetS was evaluated according to the third adult treatment panel of the National Cholesterol Education Program (NCEP-ATP III)[24] and International Diabetes Federation (IDF) criteria.[25]

PAI-1 Gene Expression by RNA Extraction and Real-Time Quantitative Polymerase Chain Reaction

In patients in whom a spare liver biopsy sample was available, RNA was isolated by guanidinium thiocyanate/phenol/chloroform extraction[26] (see the Supporting Materials).

Statistical Analyses

Statistical calculations were carried out using IBM SPSS Statistics 19.0 (SPSS, Inc., Chicago, IL). Values are expressed as mean ± standard error of the mean (SEM). Normality of variables was verified with Kolmogorov-Smirnov's test. Because most values were not normally distributed, and were not normalized after log or square root transformation, nonparametric tests were used. Spearman's rank correlations were calculated. Differences in continuous variables were tested with Mann-Whitney's U test or Kruskal-Wallis' test, as appropriate. Linear regression with studentized residuals was used to test whether differences were independent of confounding factors. Stepwise multiple regression was used to determine the influence of different variables on PAI-1 levels. Results were considered significant if P < 0.05.


Main Characteristics

Between August 2006 and January 2011, 577 patients were screened. Patient selection from screening to liver biopsy is shown in Fig. 1.

Figure 1.

Patient selection from screening to liver biopsy.

Subject Characteristics

We included 273 patients (177 women and 96 men). Subject characteristics are shown in Table 1. Eight patients (2.9%) had overweight (body mass index [BMI] ≥25 to <30 kg/m2), 150 (50.0%) had a BMI ≥30.0 to <40 kg/m2, and 115 (42.1%) had a BMI ≥40 kg/m2. Men had significantly higher PAI-1 levels, compared to women (3.5 ± 0.31 versus 2.7 ± 0.20 ng/mL; P = 0.005). Regarding glucose tolerance status, we identified 5 subjects with isolated impaired fasting glucose (I-IFG), 80 patients with isolated impaired glucose tolerance (I-IGT), and 7 with combined IFG/IGT. Diabetes mellitus was newly diagnosed in 25 patients.

Table 1. Clinical Characteristics of the Study Population
  1. The study population was comprised of 177 female and 96 male patients.

Age, years440.761874
BMI, kg/m239.60.4027.969.1
Waist, cm119.00.7989.0162.0
Fat mass, %48.60.4826.865.3
VAT, cm22145.5560622
Subcutaneous adipose tissue, cm261110.411221059
Fasting glucose, mg/dL861.4961387
Fasting insulin, mU/L17.70.730.391.8
Fasting C-peptide, nmol/L1.150.020.482.57
hs-CRP, mg/dL0.850.060.048.98
AST, U/L331.0215142
ALT, U/L471.4716144
GGT, U/L421.987299
Total cholesterol, mg/dL2052.63100407
HDL-C, mg/dL480.8425104
Triglycerides, mg/dL1544.0945594
PAI-1, ng/mL3.000.170.519.2
NAS, median (range)4.0 (0-8)   
Steatosis grade, n (%)    
0 (<5)68 (24.9)   
1 (5-33)89 (32.6)   
2 (33-66)64 (23.4)   
3 (>66)52 (19.0)   
Lobular inflammation, n (%)    
0 (no foci)89 (32.6)   
1 (<2 foci per 200× field)109 (39.9)   
2 (2-4 foci per 200× field)52 (19.0)   
3 (>4 foci per 200× field)23 (8.4)   
Ballooning, n (%)    
0 (none)84 (30.8)   
1 (few balloon cells)100 (36.6)   
2 (many cells/prominent ballooning)89 (32.6)   
NASH diagnosis based on NAS, n (%)    
No NASH (NAS<3)103 37.7)   
Indeterminate (NAS = 3-4)75 (27.5)   
Definite NASH (NAS ≥5)95 (34.8)   
NASH diagnosis based on Brunt et al.[22, 24] (%)153 (56.0)   
Fibrosis stage, n (%)    
0 (no fibrosis)171 (62.6)   
1 (perisinusoidal or periportal)46 (16.8)   
2 (perisnusoidal or portal/periportal)34 (12.5)   
3 (bridging fibrosis)21 (7.7)   
4 (cirrhosis)1 (0.4)   


Mean biopsy length was 16.0 ± 0.47 mm (range, 2-45), and the mean number of portal tracts was 8.5 ± 0.30 (range, 1-25). Distribution, according to grade of steatosis (Fig. 1A), NAS (Fig. 1B), and stage of fibrosis (Fig. 1C), is shown in Fig. 1. Based on the definition by Brunt et al., 153 of 273 (56.1%) had NASH.[23] Based on the definition by Kleiner et al., 95 of 273 (33.5%) had NASH.[22] One patient had NASH based on Kleiner et al., but not by Brunt et al., and 59 had NASH according to Brunt et al., but with a NAS score <5. The latter 59 all had a borderline or possible NASH (NAS 3 or 4) according to Kleiner et al.[22]

Association With Metabolic Features

Spearman's rank correlations between coagulation factors and anthropometric variables, liver tests, high sensitive c-reactive protein (hs-CRP), and parameters of lipid and glucose metabolism are listed in the Supporting Materials.

Association With Liver Histology

Kruskal-Wallis' analyses were performed to compare prothrombotic factors according to the different histological features of the NASH-CRN scoring system (Table 2). Only PAI-1 levels were systematically different (P < 0.05). For this reason, further statistical analyses were performed with PAI-1. Levels of PAI-1 increased significantly with increasing severity of steatosis (P < 0.001), lobular inflammation (P < 0.001), ballooning (P = 0.002), and fibrosis (P < 0.001; Fig. 2A-D).

Table 2. Differences in Prothrombotic Factors Across the Different Histological Stages
 SteatosisInflammationBallooningNAS ScoreFibrose
  1. Prothrombotic variables were compared using Kruskal-Wallis' test; significant if P < 0.05.

  2. Abbreviations: INR, international normalized ratio; PFA, platelet function analyzer.

Thrombocytes, ×109/L0.8880.5290.0300.1130.492
APTT, seconds0.9240.0800.5580.1440.490
PT, %0.1070.6030.0900.6020.527
PT, INR0.2480.7110.1240.7150.587
Fibrinogen, mg/dL0.2820.1040.0080.0280.420
FVII:c, %0.1830.5950.7990.4020.957
FVIII:c, %0.2150.3410.3780.4920.114
vWF:Ag, %0.3190.7000.2640.7630.081
FXI:c, %0.2940.1400.0020.1850.845
Antithrombin III, %0.0230.0450.5010.0830.051
Protein C, %0.0670.2390.0030.0430.584
APC resistance0.5590.3690.5120.0940.404
PFA_epi, seconds0.4090.0470.4200.3080.191
PFA_adp, seconds0.0680.9000.8090.2400.107
PAI-1, ng/mL<0.001<0.0010.002<0.001<0.001
Figure 2.

PAI-1 levels according to the different histological features. Kruskal-Wallis' analysis shows increasing PAI-1 levels with increasing severity of steatosis (A), lobular inflammation (B), ballooning (C), and fibrosis (D). PAI-1 levels also increased significantly with NAS score (E) and the different diagnostic categories of NAS (F). All P < 0.01.

Mann-Whitney's analyses comparing patients with NASH, as defined by the Brunt et al. criteria,[21] to patients without NASH, showed significantly higher levels of PAI-1 in patients with NASH (3.52 ± 0.24 versus 2.33 ± 0.21 ng/mL; P < 0.001). PAI-1 levels also increased with the NAS score (P < 0.001; Fig. 2E) and the different diagnostic categories of NAS (P < 0.001; Fig. 2F; Kruskal-Wallis' analyses).

Linear regression with studentized residuals was used to test the confounding influence of BMI, visceral adipose tissue (VAT), and C-peptide. Correcting for these three variables did not influence the results on steatosis (P = 0.001), ballooning (P = 0.045), or inflammation (P = 0.018), but did influence the results on fibrosis (P = 0.240). Correcting for BMI, VAT, and C-peptide did not influence the differences in PAI-1 levels between patients with NASH and without NASH, according to Brunt et al. (P = 0.01), and did not influence the results on NAS score (P = 0.042) or NAS-based diagnostic classification (P = 0.008). Correcting for BMI, VAT, and insulin resistance estimation using homeostasis model assessment (HOMA-IR) did not influence the results on steatosis (P = 0.002), ballooning (P = 0.007), or inflammation (P = 0.009), whereas results for fibrosis became borderline significant (P = 0.050). Correcting for BMI, VAT, and HOMA-IR did not influence the observed differences in PAI-1 levels between patients with or without NASH according to the Brunt Score (P = 0.001), according to NAS (P = 0.020) or NAS-based diagnostic classification (P = 0.001).

Spearman's Rank Correlation Between Prothrombotic Factors, Anthropometric and Metabolic Factors, and Liver Tests

Spearman's rank correlations between prothrombotic factors and different anthropometric, metabolic, and hepatic variables were calculated (see the Supporting Materials). PAI-1 was the only prothrombotic factor that correlated with all anthropometric variables (P < 0.001), all liver tests (P < 0.05), and with high-density lipoprotein cholesterol (HDL-C; r = −0.33; P < 0.001) and triglycerides (TGs; r = 0.21; P = 0.001). PAI-1 levels were not related to total cholesterol levels (r = 0.04; P = 0.473) or low-density lipoprotein cholesterol (LDL-C) levels (r = 0.10; P = 0.088).

Looking at the relationship between PAI-1 and parameters of glucose metabolism, the strongest relationship was found between PAI-1 and C-peptide (r = 0.43; P< 0.001). Of all prothrombotic factors studied, PAI-1 was the only factor to be related to HOMA-IR levels (P < 0.001). No correlation was found with hs-CRP (P < 0.05).

Mann-Whitney's Test Comparing PAI-1 Levels in Patients With and Without the MetS

Based on NCEP-ATPIII criteria, PAI-1 levels were significantly higher in subjects with the MetS (n = 133), compared to subjects without the MetS (n = 130; 3.5 ± 0.25 versus 2.5 ± 0.21 ng/mL; P = 0.001). Based on IDF criteria, we also found higher levels of PAI-1 in patients with the MetS (3.5 ± 0.24 versus 2.5 ± 0.23; p < 0.001; see the Supporting Materials).

Stepwise Multiple Regression Analysis

To determine the independent determinants of PAI-1, we performed stepwise multiple regression analysis, including variables that were significantly related to PAI-1 (r ≥ 0.20; P < 0.05). Variables calculated from another variable were not included. First, we performed a multiple stepwise regression analysis with PAI-1 as the dependent variable and liver histology parameters (steatosis, ballooning, inflammation, and fibrosis) as independent variables. We found steatosis to be the most important determinant of PAI-1 levels and fibrosis to be a second independent determinant, together explaining 12% of the variance in PAI-1 levels. In a next step, we performed multiple regression analysis with variables concerning the MetS (present or absent, number of components of the MetS), anthropometry (BMI, waist, and VAT), liver tests (AST, ALT, and GGT), lipids (HDL-C and TG), and parameters of glucose metabolism (HOMA-IR, glycated hemoglobin [HbA1c], and C-peptide or glucose, insulin, C-peptide, and HbA1c).

Variables that were independent determinants of PAI-1 in these subanalyses were included in a combined multiple regression analysis (Table 3). In this analysis, we found C-peptide to be the most important determinant of PAI-1 levels. Other independent determinants were steatosis, waist circumference, and GGT, together explaining 21% of the variance in PAI-1 levels. Including gender into the analysis did not influence the results.

Table 3. Stepwise Multiple Regression Analysisa
 BSEBR2 ModelP Value
  1. Abrreviations: r2, adjusted R square of stepwise multiple regression analysis; B, regression coefficient; SEB, Standard error of B.

  2. a

    With PAI-1 as dependent variable and steatosis, fibrosis, waist, AST, GGT, HDL-C, fasting C-peptide, HbA1c and number of components of the metabolic syndrome as independent variables.

Fasting C-peptide, nmol/L1.3980.4650.140.003
Waist, cm0.0360.0140.200.009
GGT, U/L0.0120.0050.210.012

Liver PAI-1 Gene Expression in Association With Liver Histology

To further study the role of NASH in the elevation of PAI-1 levels, we analyzed PAI-1 gene expression in a subgroup of 105 of the 273 patients included in the study in whom a spare sample was available for further analysis. Comparing PAI-1 gene expression in patients without NASH (n = 40), borderline NASH (n = 29), or NASH (n =3 6), according to the definition of Kleiner et al.,[22] PAI-1 expression was significantly higher in patients with NASH, compared to those with borderline or no NASH (7.5 ± 0.69 versus 5.5 ± 0.19 versus 5.2 ± 0.24; Kruskal-Wallis' nonparametric independent samples test; P < 0.001; Fig. 3).

Figure 3.

PAI-1 gene expression. PAI-1 gene expression is higher in patients with NASH (n = 36), compared to patients without NASH (n = 40;*P = 0.001) and patients with borderline NASH (n = 29;**P = 0.007; Kruskal-Wallis' analysis).


The aim of the present study was to investigate whether NAFLD and its different subtypes in obese subjects were associated with a prothrombotic state and whether this link existed independently of metabolic risk factors. Therefore, we studied a large group of overweight or obese subjects, using liver biopsy as the gold standard for the definition and classification of NAFLD. From an extended panel of prothrombotic factors, only PAI-1 showed a significant increase in PAI-1 levels throughout the different stages of the histological features of NAFLD, with the highest levels in patients with more severe steatohepatitis. No consistent differences were found in other prothrombotic factors. Multiple regression showed C-peptide, steatosis, waist circumference, and levels of GGT, a marker of liver damage, as independent determinants of PAI-1 levels, together explaining 21% of its variance.

Obesity and the MetS are known to be associated with an increased risk of thromboembolic events.[27] An increase in PAI-1, fibrinogen, FVIII, and vWF and a decrease in antithrombin III are most frequently reported. Also, in our large series of overweight and obese patients, the increase in PAI-1, fibrinogen, FVIII, and vWF and decrease in antithrombin III in relation to metabolic factors is confirmed, whereas the other coagulation factors were not significantly altered.

Because the liver is an important source of the majority of coagulation factors and NAFLD and NASH are closely linked to the features of the MetS, the specific contribution of NAFLD and NASH to the observed alterations needs to be assessed. Furthermore, NAFLD has been identified as a cardiovascular risk factor independent of frequently associated metabolic risk factors. Therefore, it can be hypothesized that alterations in the hepatic production of coagulant factors might, in part, explain both the prothrombotic state in obesity and the role of NAFLD in increasing the risk of cardiovascular events. Furthermore, prothrombotic factors have been implicated in the pathogenesis of cirrhosis and in PVT.[6] Recently, anticoagulant therapy has shown to beneficially influence the natural history of cirrhosis.[8] Therefore, prothrombotic factors may play an important role in disease progression in several liver diseases.[28]

A first conclusion of our study is that most of the alterations observed in obesity are unrelated to liver status. The correlations between alterations in fibrinogen or vWF and liver histology are insignificant when corrected for metabolic factors, such as BMI or VAT, which implies that the observed alterations are related to the common underlying metabolic features, but not to a specific contribution, of the fatty or inflamed liver. For many other factors studied, no consistent correlation with liver histology was observed, except for PAI-1. We can therefore conclude that in NAFLD and NASH, at least in obese patients, most of the prothrombotic factors are unaltered.

Few studies have assessed a large panel of coagulation factors in patients with NAFLD. Kotronen et al.[13] found an increase in FVIII, FIX, FXI, and XII in a retrospective analysis of 54 patients with NAFLD, mostly diagnosed by MRS, compared to 44 controls. The differences remained significant after adjustment for metabolic factors. Assy et al.[29] found an increase in protein C in relation to the degree of fibrosis in 15 patients with biopsy-proven NAFLD and 15 with NAS,H compared to chronic hepatitis C or healthy individuals. There was no correction for metabolic factors. Papatheodoridis et al.[30] studied 60 biopsy-proven NAFLD patients, compared them to 90 chronic viral hepatitis patients, and found an increase in fibrinogen, protein S, and protein C in NAFLD. There was no correction for metabolic factors. They also showed a relation between NAFLD and the number of thrombotic risk factors, with the highest numbers in patients with NASH and fibrosis. Targher et al.[31] studied 100 apparently healthy volunteers and found no differences between 35 NAFLD patients (based on US) and the no-NAFLD group after correction for VAT. Hence, most of the studies have low patient numbers or lack histological diagnosis.

In contrast to these studies, our study consecutively recruited a large series of patients with a wide range of BMI and other metabolic factors in whom there was no a priori suspicion of liver disease. Liver biopsy was performed in all patients who underwent bariatric surgery, and in patients not going to surgery, liver biopsy was proposed if there was any suspicion of NAFLD. This resulted in a large series of histologically characterized patients presenting the whole spectrum from normal liver to NASH and cirrhosis. The patients who ultimately appeared to have a histologically normal liver were considered to be an internal control group, avoiding the need to compose a matched control group (in which histological data are usually lacking). These methodological considerations strengthen the conclusion that, besides PAI-1, the liver is not contributing to the prothrombotic state associated with obesity. It might be worthwhile, in future studies, to include another type of control group consisting of nonobese patients, such as liver donor patients or patients undergoing elective abdominal surgery.

A second conclusion of our study is that NAFLD is independently contributing to the increase in PAI-1, observed in obese patients, and that the increase in PAI-1 is related to histological severity, especially steatosis and NASH.

Literature data on PAI-1 are mostly, but not all,[32] consistent with an increase in PAI-1 in relation to NAFLD. In the study of Targher et al.,[31] the link between NAFLD and PAI-1 disappeared after correction for VAT, suggesting that not the liver, but VAT was the main contributor to thrombophilia in the MetS. Also, in the study of Sookoian et al.,[16] studying 113 biopsy-proven NAFLD patients, compared to 102 age- and gender-matched controls without biopsy, and of Cigolini et al.[9] (studying 31 NAFLD patients, based on US, with 32 controls) differences in PAI-1 disappeared upon correction for metabolic parameters. Not all studies corrected the associations for metabolic features. In the studies by Kotronen et al.,[13] Barbato et al.[33] and Bruckert et al.,[34] the association of increased PAI-1 levels with NAFLD (based on MRS, US, or liver enzymes, respectively) remained significant after adjusting for different metabolic parameters. By contrast, Espino et al.[32] found no differences in PAI-1 serum levels according to the presence or absence of NAFLD. In the latter study, histology was, however, only available in a small group of 50 morbidly obese patients and comparison was only made between patients with NAFLD versus patients without NAFLD.

As outlined before, some of the methodological shortcomings of other reported studies were avoided in our study. Therefore, we estimate that we can reliably conclude that, in the case of obesity, the liver independently contributes to the observed increase in PAI-1 and that this increase is correlated with the severity of steatosis and steatohepatitis, with the highest values in the most severe NASH.

Although the grade of steatosis does seem to have a higher effect on PAI-1 levels, compared to inflammation, our data clearly show increasing levels of PAI-1 with increasing grades of inflammation or severity of steatohepatitis, as expressed by the NAS. We also found highly significantly increased levels of PAI-1 in patients with NASH, compared to patients with simple steatosis. Furthermore, PAI-1 gene expression was also significantly higher in patients with NASH, compared to those with simple steatosis, further substantiating the role of steatohepatitis. This is probably not restricted to NASH, because PAI-1, being an acute-phase reactant, has shown to be increased in different types of both acute and chronic hepatic inflammation.[35]

Previous studies have also shown a link between the prothrombotic state and hepatic fibrosis.[36] In multiple regression analysis entering only histological features, we found fibrosis to be independently correlated with PAI-1 levels. However, steatosis was the most important determinant, showing that even patients with simple steatosis had increased levels of PAI-1. When metabolic factors were included in multiple regression, fibrosis was no longer an independent determinant of PAI-1. Our population comprises low numbers of patients with advanced fibrosis, probably resulting in insufficient statistical power to show an independent correlation between fibrosis and PAI-1.

Different mechanisms could explain the link between NAFLD/NASH and increased PAI-1 levels. One of the possible mechanisms is that increased liver fat could directly stimulate hepatocytes to secrete PAI-1. Sookoian et al.[16] found higher circulating levels of PAI-1 in patients with NAFLD and also higher scores of PAI-1 expression in the liver, compared to controls. In our series, we confirm the higher expression of PAI-1 in livers of patients with NASH. Patients with normal livers or patients with simple steatosis have lower levels, again pointing toward an effect of the inflammatory component of the disease in the increase in PAI-1 levels.

The true clinical significance of the increase in PAI-1 can, of course, not be assessed by a cross-sectional analysis. In plasma, PAI-1 inhibits degradation of fibrin clots, contributing to an increased atherothrombotic risk. Large epidemiological studies have shown elevated plasma levels of PAI-1 as a predictor of myocardial infarction.[37] PAI-1 levels are related to the severity of vessel wall damage and are good predictors of subsequent development of a first acute myocardial infarction.[38] One can hypothesize that the increase in PAI-1 levels by the steatotic and inflamed liver, at least in part, explains the observed link between NAFLD and increased cardiovascular risk. Furthermore, local thrombotic events might contribute to fibrogenesis and liver disease progression. Antithrombotic treatment has recently been proven to beneficially affect disease evolution in cirrhosis. However, large longitudinal studies are needed to assess the role of PAI-1 in NAFLD-associated cardiovascular risk and fibrosis progression and the potential therapeutic implications of these findings.

The association found between a prothrombotic status and NAFLD, independent of obesity, might be of relevance for the treatment paradigm of NAFLD and NASH. As more effective treatments become available in the future, it might be considered to treat patients with NAFLD and NASH, regardless of the degree of fibrosis, in an attempt to reduce the associated cardiovascular risks.[39]

In conclusion, our results show that NAFLD and NASH independently contribute to the prothrombotic state in obesity by an increase in PAI-1, whereas other prothrombotic factors are unaffected by liver status.