Prediction of liver fibrosis severity in alcoholic liver disease by human microfibrillar‐associated protein 4

Abstract Background Alcoholic liver disease (ALD) is a public health concern that is the cause of half of all cirrhosis‐related deaths. Early detection of fibrosis, ideally in the precirrhotic stage, is a key strategy for improving ALD outcomes and for preventing progression to cirrhosis. Previous studies identified the blood‐borne marker human microfibrillar‐associated protein 4 (MFAP4) as a biomarker for detection of hepatitis C virus (HCV)‐related fibrosis. Aim: To evaluate the diagnostic accuracy of MFAP4 to detect ALD‐induced fibrosis. Method We performed a prospective, liver biopsy‐controlled study involving 266 patients with prior or current alcohol overuse. Patients were split into a training and a validation cohort. Results MFAP4 was present in fibrotic hepatic tissue and serum MFAP4 levels increased with fibrosis grade. The area under the receiver operating characteristic curve (AUROC) for detection of cirrhosis was 0.91 (95% CI 0.85‐0.96) in the training cohort and 0.91 (95% CI 0.79‐1.00) in the validation cohort. For detection of advanced fibrosis, the AUROC was 0.88 (95% CI 0.81‐0.94) in the training cohort and 0.92 (95% CI 0.83‐1.00) in the validation cohort. The diagnostic accuracy did not differ between MFAP4 and the enhanced liver fibrosis (ELF) test or transient elastography (TE) in an intention‐to‐diagnose analysis. MFAP4 did not predict hepatic decompensation in a time‐to‐decompensation analysis in a subgroup of patients with cirrhosis. Conclusion MFAP4 is a novel biomarker that can detect ALD‐related fibrosis with high accuracy.


| INTRODUC TI ON
Mortality attributable to cirrhosis has been rising in the United States since 2009 and increasingly affects young people caused by the development of end-stage alcoholic liver disease (ALD). 1,2 The recommended strategy to address the burden of ALD is early detection and alcohol abstinence as these actions can improve histological findings, decrease the likelihood of progression to cirrhosis and improve survival rates. [3][4][5][6][7] Additionally, early detection of ALD enables the implementation of evidence-based prophylaxis against liver-related complications and reinforces disease awareness that can positively modify drinking behaviours. 8 Unfortunately, the majority of ALD patients are diagnosed in an advanced fibrotic stage for which the prognosis is poor even if abstinence is achieved and treatment guidelines are followed. 9 Thus, patients should ideally be identified before their disease progresses to advanced fibrosis, as this point marks the threshold for severely increased rates of liver-related mortality. 6 Such early diagnosis should be achievable as nearly half of ALD patients have interacted with the healthcare system for an alcohol-related purpose prior to developing alcoholic cirrhosis. 10,11 The current gold standard reference for determining the level of hepatic fibrosis and exclusion of co-existing liver disease is a liver biopsy. 12 However, because of the invasiveness and low tolerability by patients, this procedure is not suitable as a screening tool to detect early stage ALD. 13 Recent advances in non-invasive testing using transient elastography (TE) and serum markers, such as the enhanced liver fibrosis test (ELF), may address some of these shortcomings. 14,15 Both ELF and TE have high diagnostic accuracy for the detection of ALD-related fibrosis, but both approaches are costly and frequently unavailable outside tertiary hospitals. Hence, there is an unfulfilled need for new biomarkers to detect fibrosis among patients with suspected ALD. Likewise, there are no tools available to identify patients who are most at risk for disease progression. Such tools would allow optimal allocation of healthcare resources and tailoring of treatment to individuals who are most in need.
The human microfibrillar-associated protein 4 (MFAP4) is ubiquitously distributed in the extracellular matrix in the human body. 16 Although its biological properties are not fully understood, MFAP4 is involved in orchestrating extracellular matrix remodelling during tissue repair. [17][18][19][20] MFAP4 was originally identified as a candidate biomarker of liver fibrosis from proteome analyses of microdissected cirrhotic septa isolated from patients infected with hepatitis C virus (HCV). 21 Subsequently, two studies confirmed the applicability of MFAP4 for the detection of HCV-induced liver fibrosis. 22,23 Owing to its reported diagnostic accuracy and robustness against variations in sampling handling and storage, serum MFAP4 has the potential to translate into a clinically meaningful diagnostic tool. 24 However, the diagnostic accuracy of serum MFAP4 to detect ALD-induced fibrosis has not been previously evaluated, nor has it been compared to that of other widely used non-invasive techniques such as TE or the ELF test.
In this study, we aimed to evaluate whether MFAP4 is upregulated during different stages of ALD-induced fibrosis and to compare the diagnostic accuracy of serum MFAP4 to that of TE and ELF. We also evaluated the prognostic potential of MFAP4 among patients with ALD-induced cirrhosis.

| ME THODS
We performed a prospective, biopsy-controlled, single-centre study with an internal validation cohort. Blood samples from 50 healthy gender-and aged-matched participants were used to determine the  All authors had access to the study data and reviewed and approved the final manuscript.

| Study population
This study included 266 patients with prior or current alcohol overuse, defined as more than 24 g and 36 g per day for women and men, respectively, for more than 1 year. Additional inclusion criteria were age 18-75 years old and informed consent to undergo a liver biopsy. Patients were recruited consecutively from two municipal alcohol rehabilitation centres and from three outpatient hospital liver clinics in the Region of Southern Denmark.

K E Y W O R D S
biomarker, cirrhosis, extracellular matrix protein, liver biopsy, non-invasive testing

Key points
There is a lack of new accurate diagnostic tools to detect alcoholic liver disease in an early reversible stage. In this study, we prove that the protein MFAP4 is present in the liver and blood in patients with alcohol overuse and document that it can be used to precisely assess the severity of scar tissue in the liver.
All participants consented after receiving verbal and written information. All patients were at significant risk for ALD that justified performance of a liver biopsy. The criteria were revised in January 2016 at which time patients younger than 30 years old and a liver stiffness below 6.0 kPa by TE were excluded based on our previous findings that patients who met these criteria did not have severe fibrosis. 26 Exclusion criteria were decompensated liver disease with clear clinical signs of cirrhosis, severe alcoholic hepatitis (defined by clinical criteria in the form of new onset of icterus and impairment in liver function in a patient with excessive alcohol overuse), debilitating disease with an expected survival of less than 1 year, concurrent liver disease, hepatic congestion or inability to comply with the study protocol. The healthy control cohort included 50 participants. The inclusion criterion for the control cohort was age 18-75 years old. Exclusion criteria were ongoing alcohol intake above 60 g/wk, daily alcohol intake or binge drinking habits.
Individuals with prior alcohol abuse, body mass index (BMI) above 28, concurrent liver disease, comorbidity or daily intake of any medication other than mild pain relievers (over-the-counter nonsteroidal anti-inflammatory drugs or paracetamol) were also excluded from the control cohort. All investigations were performed on the same day at Odense University Hospital (OUH), according to standard operating procedures after an overnight fast.

| Histological and immunohistochemical studies
Liver biopsies were performed percutaneously with a 17-G Menghini suction needle (Hepafix, Braun, Germany). We considered biopsies to be of adequate quality if they were > 10 mm long and contained > 5 portal tracts or if a regeneration nodule was present. A single experienced pathologist evaluated the biopsies according to the Kleiner fibrosis stage and non-alcoholic fatty liver disease activity score (NAS-CRN). 27 According to the Kleiner fibrosis stage, F0 is no fibrosis, F1 is perisinusoidal or periportal fibrosis only, F2 is perisinusoidal fibrosis in combination with portal or periportal fibrosis, F3 is bridging fibrosis and F4 is cirrhosis. We classified ≥ F3 as advanced fibrosis. The NAS-CRN is a semiquantitative score of steatosis (0- A tissue microarray with different types of normal tissues, such as skin, tonsil, lung, spleen, prostate, testis, uterus, kidney, gallbladder, normal liver and normal pancreas, was used as control. The walls of vessels in these tissues were used as positive control. Hepatic expression of MFAP4 was semiquantitatively evaluated in a subgroup of 116 patients. For this study, we designed a 6-ordered score de-

| Non-invasive liver evaluation
An experienced nurse operator (>500 scans) performed liver stiffness measurements by TE using a FibroScan 502 Touch (Echosens) according to standard procedures. We measured the commercially available ELF test (Siemens Healthcare Diagnostics, Inc) on an Advia Centaur XP according to the manufacturer's instructions (Siemens Healthcare Diagnostics, Inc). Covariation estimated from the assay control using three different concentrations ranged from 2.5% to 2.9% for tissue inhibitor of metalloproteinase-1, 4.1% to 6.1% for hyaluronic acid and 4.0% to 5.4% for N-terminal pro-peptide.

| Quantification of serum MFAP4 levels
Detection of serum MFAP4 levels was performed using an AlphaLISA technique as previously described. 16 The experiments were performed in duplicate and sample covariance < 10% was acceptable.

| Statistical analysis of MFAP4 in serum and liver biopsies
Summary statistics were used to describe patient characteristics. Kruskal-Wallis test and a subsequent Dunn's test were used to test for differences in serum MFAP4 levels between fibrotic stages. We investigated which factors influenced serum MFAP4 by performing a robust multivariate linear regression analysis using stepwise elimination of insignificant factors. Serum MFAP4 levels were non-normally distributed, and data were transformed using the natural logarithm prior to analysis. An ordinal logistic regression analysis was performed to identify variables that independently predict MFAP4 expression stage in liver biopsies. Variables

| MFAP4 serum levels in the cohorts
The mean serum level for MFAP4 in the total patient cohort with prior or current alcohol overuse was 59.5 (± 47.0) U/L compared to 27.7 (± 8.8) U/L in the healthy control group that had no history of liver disease or alcohol overuse. The boxplot for serum MFAP4 in the total cohort showed that the serum level increased in relation to the Kleiner fibrosis stage as seen in Figure 1.

| MFAP4 expression score in liver tissue
To corroborate our finding of the close linkage between serum

| Accuracy of MFAP4 for detection of advanced liver fibrosis and cirrhosis
In To optimize stratification of patients according to MFAP4 serum level, we determined rule-in and rule-out cut-offs for advanced fibrosis and cirrhosis by setting specificity (to rule-in) or sensitivity   Figure 3 and Supporting Information).
Both risk prediction plots and calibration plots are shown in Figure 4.
The calibration of serum MFAP4 was good for both advanced fibrosis

| Intention-to-diagnose analysis
To consider the impact of non-evaluable results, we performed an in-

TA B L E 2 (Continued)
or Brier score, was similar to that of TE and ELF in the intention-todiagnose analysis.

| Relationship between MFAP4 and time-todecompensation in cirrhosis
Cirrhotic patients with high MFAP4 expression in hepatic tissue had significantly larger spleens and tended to have lower platelet counts compared to cirrhotic patients who had low MFAP4 expression (Supporting Information). As these findings potentially reflect differences in the portal pressure between the groups, we subse-  in cirrhosis develop as part of a decompensating event or as acuteon-chronic liver failure. 28 These events are frequently triggered by inflammation rather than progression of fibrosis. 29 This dynamic natural history of cirrhosis likely cannot be reflected by a single fibrosis marker and may explain why MFAP4 failed as a predictor of decompensation in our study. 28 Ballooning was independently associated with both hepatic expression and serum level of MFAP4, suggesting a potential role for ballooning in the upregulation of MFAP4 expression in fibrotic tissue and subsequent release to the blood stream. Consistent with these findings, we identified ballooning as a risk factor for false classification of cirrhosis.

| D ISCUSS I ON
Although our study was specifically designed to evaluate the diagnostic accuracy of serum MFAP4, we do note some strengths and limitations. Our cohort included 266 patients who had varying degrees of alcohol consumption that covered the full fibrotic spectrum of ALD from no fibrosis to fully developed cirrhosis. By including patients recruited from primary care, we increased the generalizability of our results. Likewise, we chose to include patients with obesity or features of metabolic syndrome as these conditions often co-exist with alcohol overuse and reflect daily clinical practice. Patients who had obvious ascites and large varices were excluded as these patients rarely require further diagnostic workup in the form of a liver biopsy to secure a diagnosis of cirrhosis. No patients in the cohort suffered from severe pulmonary or cardiovascular diseases and there was only one case of (chronic) pancreatitis. Such conditions may affect MFAP4 serum levels and could potentially impact the diagnostic accuracy of MFAP4 in a clinical setting. 16,30,31 The study cohort was split into a training and a validation cohort by date as this leads to temporal validation. 32 MFAP4 maintained a high diagnostic accuracy, despite differences in clinical phenotypes and disease prevalence between the cohorts. However, differences in disease prevalence leading to spectrum bias should still be considered when generalizing results. 26 Further, the prognostic analyses should be interpreted with some caution as the follow-up time was relatively short for patients with early stage cirrhosis. However, since more than half of the patients met the primary endpoint during the follow-up period, the high event rate compensated for the short follow-up period and justified the time-to-decompensation analysis.
In conclusion, serum MFAP4 is a novel, highly accurate marker for assessing ALD-induced fibrosis with similar diagnostic accuracy as TE and ELF test.

ACK N OWLED G EM ENT
The specialist nurses at Odense University Hospital outpatient liver clinic contributed immensely to this study: Trine Møller, Charlotte