Evaluating the association of serum ferritin and hepatic iron with disease severity in non‐alcoholic fatty liver disease

Hyperferritinemia, with or without increased hepatic iron, represents a common finding in non‐alcoholic fatty liver disease (NAFLD). However, it is unclear whether it reflects hepatic inflammation or true iron‐overload and, in case the latter is confirmed, whether this influences disease progression. We therefore explored the association between serum ferritin, degree and pattern of hepatic iron deposition and liver disease severity in patients with NAFLD.


| INTRODUC TI ON
Non-alcoholic fatty liver disease (NAFLD) is increasingly prevalent in Western countries, affecting approximately 30% of unselected population. 1 It is usually diagnosed in patients with metabolic syndrome (MetS), but also in a small percentage of subjects with normal weight (7%). 2 NAFLD encompasses a wide spectrum of histological conditions, from simple steatosis to steatohepatitis or NASH (characterized by ballooning and lobular inflammation) with subsequent development of fibrosis that can lead to cirrhosis. Despite its high prevalence, only a small proportion of subjects with NAFLD develops NASH and faces a higher risk of liver disease progression. 3 A high serum ferritin (SF) is a common finding in NAFLD, involving up to 30% of affected subjects, and it is often the biochemical abnormality that leads to medical attention. However, it is still unclear whether it simply reflects hepatic inflammation or represents true hepatic iron-overload. 4,5 In this context, SF has been proposed as a marker of both NASH and liver fibrosis with contrasting results [6][7][8] and has been incorporated in panels for liver fibrosis assessment. 9 More recently, SF has been proposed as an independent predictor of long-term mortality in NAFLD. 10 Mild-to-moderate hepatic iron accumulation is encountered in a number of liver diseases including chronic hepatitis C, alcoholic and non-alcoholic fatty liver disease and cirrhosis, 11 but the underlying mechanism is unclear. Also, although it is known that iron excess can damage the liver by inducing oxidative stress and lipid peroxidation, 12 in these settings the effect of iron accumulation is not completely understood.
Moreover, the relationship between SF levels, the pattern of iron deposition (hepatocellular [HC], reticuloendothelial [RES] or mixed) 13 and liver disease stage in NAFLD has not been elucidated.
The aim of this study was therefore to assess the presence and pattern of hepatic iron accumulation in patients with biopsy-proven NAFLD, and to examine whether such a finding is associated with more severe/progressive liver disease. Furthermore, we analysed the role of serum ferritin as predictor of liver disease severity.

| Patients population
We retrospectively evaluated all consecutive outpatients

Key points
There are contrasting results on the association of serum ferritin and presence or pattern of hepatic iron deposition with disease severity in people with fatty liver. Therefore, we reviewed liver biopsies of NAFLD patients in order to provide a conclusive answer to the above question. We showed that iron deposition both in hepatocytes and macrophages (mixed pattern) was associated with non-alcoholic steatohepatitis, which is the progressive form of fatty liver. Serum ferritin alone could not be used as a marker of liver scarring (fibrosis).

| Iron status
Serum iron, SF and total iron binding capacity (TIBC) and/or transferrin saturation (TSAT) were determined using automated biochemical methods. HFE genetic test had been performed in patients with hyperferritinemia (above 200 μg/L in females and 300 μg/L in males or menopause females) and an abnormal transferrin saturation (TSAT > 45%) and in selected patients with hyperferritinemia and normal TSAT according to the physicians' discretion, by PCR-based techniques.

| Histological assessment
All liver biopsy specimens were obtained by percutaneous or trans-jugular route, with a median length of 19 mm (6-58 mm).
Specific criteria for liver biopsy were: length of at least 10 mm, comprising at least six portal tracts or less if cirrhotic or considered of sufficient quality for a diagnosis and staging by the pathologist. We included 6 biopsies <10 mm (6, 7, 8, 9, 9 and 9 mm respectively) presenting simple steatosis since the histopathologist was satisfied with the representativity of the sample and the results of the analysis were not changed by their removal (data not shown). Liver sections were routinely stained with haematoxylin/ eosin, silver reticulin, blue aniline or Sirius red for collagen, Perls' Prussian blue for iron. Liver biopsies were centrally reviewed by a single pathologist in each centre. NAFLD lesions were scored according to the NASH Clinical Research Network (CRN) NAS scoring system. 17 NASH was diagnosed in the presence of the combination of any degree of hepatic steatosis, hepatocellular ballooning and lobular inflammation. 18,19 Hepatic fibrosis was staged on a 5-point scale (0 = absence of fibrosis, 1 = zone 3 perisinusoidal/ perivenular fibrosis, 2 = zone 3 and periportal fibrosis, 3 = septal/ bridging fibrosis, 4 = cirrhosis 17 ). Significant and advanced fibrosis was defined as stages ≥F2 and ≥F3, respectively. Advanced fibrosis was chosen as one of the main variables of interest based on previous studies demonstrating that it is associated with long-term clinical outcomes and increased mortality in NAFLD patients. 20,21 Inter-observer agreement regarding histological evaluation by the 2 pathologists was tested on a set of 30 slides by weighted Cohen's kappa, with a resulting k value for fibrosis of 0.76, meaning excellent agreement.The presence of iron was assessed both in hepatocytes and reticuloendothelial cells and the degree of liver siderosis was classified according to a modified Scheuer's system (Table S1 13,22 ).

| Ethical approval
Blood tests and liver biopsy were performed as part of the standard or routine care. Both centres had the approval from the local ethical committee to use registered parameters and liver biopsies for studies. The study was carried out in accordance with the principles of the Helsinki Declaration.

| Statistical analysis
Continuous data were presented as mean ± standard deviation if parametric or median and interquartile range if nonparametric. Categorical data were presented as number and percentage.
Comparisons between frequencies or percentages were performed by using the chi-square test or the Fisher's exact test. Betweengroup comparisons of continuous variables were performed using the Student's t test or Analysis of Variance for normally distributed variables, and the Mann-Whitney or Kruskal-Wallis tests for nonnormally distributed variables.
Multiple logistic regression analysis, stepwise approach, was used to examine the relationship between serum ferritin, presence and pattern of hepatic iron deposition and presence of NASH or fibrosis. All the variables that were associated with NASH at the univariate analysis with a statistical significance corresponding to a P value up to 0.1 were included in the multivariate analysis. Similarly, logistic regression analysis was performed to find variables associated to advanced fibrosis. A 2-sided P value <0.05 was considered significant. All analyses were performed using IBM SPSS (22.0, IBM, New York, USA).

| Characteristics of the NAFLD population
Of the 477 patients initially considered, 2 were subsequently excluded because of suboptimal biopsy sample, 3 because of missing clinical and/or biochemical data and 4 because of compound C282Y/ H63D heterozygosity at the HFE gene test. The demographic, clinical, biochemical and histological details of the 468 patients included in the study are shown in Table 1: the mean age was 47 years, 76% of patients were of Caucasian ethnicity and 38% were females. The mean BMI was 30.4 kg/m 2 ; a history of IFG was observed in 19% patients and DM in 29% of patients, of which 65% were on hypoglycemic agents, mainly metformin. There was a high percentage of patients affected by dyslipidemia (68%), of which only 25% were on statin treatment. Hypertension had already been diagnosed in 32% of subjects, and 62% were treated mainly with renin-angiotensin system inhibitors. Female subjects were more likely to have diabetes and hypertension (45% vs 29%, P < 0.001, and 39% vs 27%, P = 0.009, respectively). None of the included patients had radiological or histological evidence of HCC at the time of liver biopsy.
Histological criteria for NASH were fulfilled in 247 (53%) patients, while advanced fibrosis was prevalent in 89 (19%) patients (81 of which having NASH and 8 with likely 'burnt-out NASH'). The British cohort had a higher prevalence of obesity (mean BMI 31 vs 29, P = 0.003), hypertension (37% vs 24%, P = 0.003) and dyslipidemia than the Italian cohort (77% vs 55%, P < 0.0001) and this was reflected by a higher prevalence of patients who fulfilled the criteria for MetS (37% vs 20%, P < 0.0001). On the other hand, a higher proportion of patients with NASH (83% vs 33%, P < 0.0001) and hepatic iron deposition (36% vs 17%, P < 0.0001) was found in the Italian cohort. No difference in the proportion of patients with advanced fibrosis was found when comparing the 2 cohorts (18% vs 20%, P = 0.53).
The characteristics of the 247 patients with NASH are shown in
Clinical and laboratory data of subjects according to the presence of stainable hepatic iron and to the pattern of iron deposition are shown in Table 3: subjects with stainable hepatic iron were more likely to be male and have a lower BMI, had increasingly higher levels of SF (particularly Mixed>RES>HC, P < 0.0001), serum iron (23.4 μmol/L vs 17 μmol/L, P < 0.0001) and TSAT (37% vs 27%, P = 0.04), with a higher proportion of patients with hyperferritinemia (60% vs 15%, P < 0.0001) and TSAT > 45% (27% vs 10%, P = 0.003, data not shown). In addition, they had significantly higher levels of ALT (P = 0.007).

| SF, iron deposition and liver disease severity
Hyperferritinemia was found in 122 (26%) patients, with no significant difference between the 2 cohorts (34% vs 28%, P = 0.17). When looking at mild alcohol consumption, an increasing trend in SF values was found for increasing alcohol intake; however, the prevalence of hyperferritinemia was not different between abstinent vs mild drinkers (P = 0.1), even when considering subclasses of alcohol intake (P = 0.097, Table SS).
Looking at the single components of histological NASH, a positive association was found between SF and increasing degree of steatosis (P = 0.008) but not with hepatocellular ballooning or lobular inflammation and overall SF did not associate with diagnosis of NASH (Table 2). Interestingly, SF showed a peculiar pattern throughout fibrosis stages, increasing from F0-F1 to F3, and subsequently decreasing in cirrhosis (Figure 1). When comparing the inter-group SF distribution, it was significantly higher for F3 compared to F0-F1 (P = 0.024), with an average increase of 14% from F0-F1 to F2 and of 61% from F2 to F3 and a decrease of 43% from F3 to F4.
On the other hand, patients with a mixed pattern of iron deposition were more likely to have NASH if compared to patients with other patterns of iron deposition or to patients without hepatic iron.
No difference was seen between patients with and without hepatic stainable iron when considering the individual components of histological NASH, NAS > 3 or > 5 (data not shown).
No association was found between presence of stainable hepatic iron and fibrosis (mild, significant, advanced).

| D ISCUSS I ON
In this study, we showed that a mixed pattern of hepatic iron deposition is associated with the presence of NASH, while ferritin levels increase with worsening of fibrosis up to a pre-cirrhotic stage but are not independent predictors of advanced fibrosis.
Stainable hepatic iron was found in 25% of patients, in line with previous studies reporting a prevalence ranging from 14% to 50%. [23][24][25] The degree of iron accumulation was rarely above a mild to moderate degree and never above grade 3: this can be considered a typical characteristic of NAFLD patients with stainable hepatic iron, and would be one of the reasons explaining why many studies of phlebotomy in NAFLD have failed to show a benefit or improvement in liver histology. 26,27 A mixed pattern of iron deposition was the most common, accounting for 43% of patients with stainable hepatic iron and this is consistent with data from the literature. 5 When addressing the potential effect of mild-moderate alcohol intake on SF, a previous observational study on more than 12 000 patients showed that consumption of 2 or more alcoholic drinks per day was associated with significant risk of iron overload, whereas consumption of a lesser amount was associated with reduced risk of iron deficiency. 33 In our cohort, SF increased with increasing degree of alcohol intake, but only patients who drank more than 10 g/day of alcohol showed significantly higher SF when compared to total abstinent and lower alcohol intake subgroups. Moreover, there was no association between alcohol intake and hepatic iron deposition, NASH or fibrosis, suggesting that in NAFLD patients, the effect of mild alcohol consumption on iron metabolism and NASH is negligible, and different pathways are involved.
Few studies have explored the role of diet and its influence on SF and iron levels: although there are not many solid data, the consumption of red meat or insufficient amount of vegetables, common in MetS patients, seems to be associated with higher SF levels. 34 Unfortunately, we did not have such information from our cohort.
Considering our results, and the fact that ferritin levels in NAFLD patients tend to be higher than in haemochromatosis patients with the same amount of hepatic iron, 35