Serum ferritin levels are associated with a distinct phenotype of chronic hepatitis C poorly responding to pegylated interferon-alpha and ribavirin therapy

Authors


  • Potential conflict of interest: Nothing to report.

  • This work was supported by the Swiss National Science Foundation (3100A0-122447 and 31003A-138484 [to D.M.], 32003B-127613 [to P.Y.B.], as well as 3347C0-108782 and 33CSC0-108782 [to the SCCS]), the Novartis Foundation (09C53; to D.M.), the Leenaards Foundation (to P.Y.B.), and the Santos-Suarez Foundation (to P.Y.B.). C.M.L. is the recipient of a Research Fellowship from the Deutsche Forschungsgemeinschaft (LA 2806/1-1).

Abstract

Elevated serum ferritin levels may reflect a systemic inflammatory state as well as increased iron storage, both of which may contribute to an unfavorable outcome of chronic hepatitis C (CHC). We therefore performed a comprehensive analysis of the role of serum ferritin and its genetic determinants in the pathogenesis and treatment of CHC. To this end, serum ferritin levels at baseline of therapy with pegylated interferon-alpha and ribavirin or before biopsy were correlated with clinical and histological features of chronic hepatitis C virus (HCV) infection, including necroinflammatory activity (N = 970), fibrosis (N = 980), steatosis (N = 886), and response to treatment (N = 876). The association between high serum ferritin levels (>median) and the endpoints was assessed by logistic regression. Moreover, a candidate gene as well as a genome-wide association study of serum ferritin were performed. We found that serum ferritin ≥ the sex-specific median was one of the strongest pretreatment predictors of treatment failure (univariate P < 0.0001, odds ratio [OR] = 0.45, 95% confidence interval [CI] = 0.34-0.60). This association remained highly significant in a multivariate analysis (P = 0.0002, OR = 0.35, 95% CI = 0.20-0.61), with an OR comparable to that of interleukin (IL)28B genotype. When patients with the unfavorable IL28B genotypes were stratified according to high versus low ferritin levels, SVR rates differed by >30% in both HCV genotype 1- and genotype 3–infected patients (P < 0.001). Serum ferritin levels were also independently associated with severe liver fibrosis (P < 0.0001, OR = 2.67, 95% CI = 1.68-4.25) and steatosis (P = 0.002, OR = 2.29, 95% CI = 1.35-3.91), but not with necroinflammatory activity (P = 0.3). Genetic variations had only a limited impact on serum ferritin levels. Conclusion: In patients with CHC, elevated serum ferritin levels are independently associated with advanced liver fibrosis, hepatic steatosis, and poor response to interferon-alpha-based therapy. (Hepatology 2012)

Chronic hepatitis C (CHC) is one of the most significant infectious diseases by being a leading cause of liver-related morbidity and liver transplantation worldwide.1, 2 Triple therapy comprising the hepatitis C virus (HCV) nonstructural protein 3-4A protease inhibitors, telaprevir or boceprevir, in combination with pegylated interferon-alpha (PEG-IFN-α) and ribavirin is becoming the new standard treatment for patients infected with the difficult-to-cure HCV genotype 1.3 Recently completed phase III clinical trials have demonstrated superior efficacy of such triple-therapy regimens in treatment-naïve and in treatment-experienced HCV genotype 1 patients, compared to PEG-IFN-α and ribavirin alone.4-8 However, such triple-therapy regimens are burdened with additional adverse effects and significant cost. In addition, the success of triple-therapy regimens still depends on the IFN sensitivity of a given patient, as, for example, the efficacy of telaprevir in previous null responders (<2 log10 reduction in HCV RNA after 12 weeks of PEG-IFN-α and ribavirin) remains limited (boceprevir was not evaluated in null responders), and the risk of resistance development decreases with an extended rapid virologic response (RVR) (i.e., undetectable HCV RNA at weeks 4 and 12 of triple therapy).6, 9 Therefore, despite enormous progress, there is still a need to establish algorithms (including, among others, on-treatment viral kinetics and interleukin [IL]28B genotype) to predict treatment outcome of IFN-α-based therapy (e.g., to identify patients who do not necessarily need or who may require only short durations of triple therapy).10-13

HCV interferes with the host's iron metabolism, and hepatic iron measures were correlated with the grade and stage, as well as with the treatment outcome, of CHC.14-18 Infection with HCV leads to iron accumulation in the liver and increased serum ferritin levels, which can be, at least partially, explained by down-regulation of hepcidin, a key regulator of iron homeostasis.17, 19 However, serum ferritin is also frequently elevated in inflammatory conditions. Excess iron in the liver promotes liver inflammation, oxidative stress, and mitochondrial dysfunction.20 Accordingly, repeated phlebotomy in patients with CHC has been shown to reduce necroinflammatory activity, and a recent study has reported a reduced risk of hepatocellular carcinoma development in the phlebotomy group, compared to the control group.21-23 However, the role of serum ferritin as an independent predictor of sustained virologic response (SVR) to IFN-α-based therapy remains controversial.18, 24-28

In the present study, we performed a comprehensive analysis of the relationship of serum ferritin levels and the natural course and treatment outcome of CHC in patients enrolled in the large, well characterized Swiss Hepatitis C Cohort Study (SCCS), including a genome-wide association study (GWAS) of determinants of serum ferritin levels in HCV-infected individuals.

Abbreviations

ALT, alanine aminotransferase; BMI, body mass index; CHC, chronic hepatitis C; CI, confidence interval; GWAS, genome-wide association study; HCV, hepatitis C virus; HIV, human immunodeficiency virus; IL, interleukin; OR, odds ratio; PEG-IFN-α, pegylated interferon-alpha; RVR, rapid virologic response; SCCS, Swiss Hepatitis C Cohort Study; SVR, sustained virologic response; SNP, single-nucleotide polymorphism.

Patients and Methods

Patients.

Patients were followed within the framework of the SCCS, a multicenter study performed at eight major Swiss hospitals and their local affiliated centers, including a total number of 3,648 patients with chronic or resolved HCV infection.29, 30 SCCS patients were selected for the present retrospective analysis if they met the following inclusion criteria: proven CHC, defined as anti-HCV seropositivity (using enzyme-linked immunosorbent assay and confirmed by recombinant immunoblotting assay) and HCV RNA detectable by a qualitative or quantitative assay; availability of serum ferritin measurements either at baseline before antiviral therapy or before liver biopsy (≤3 months); written informed consent for genetic testing; and availability of genomic DNA for genotyping. Patients with hemochromatosis were excluded from the present study. Assessment of treatment response was restricted to treatment-naïve patients who were treated during clinical practice conditions with either PEG-IFN-α2a or PEG-IFN-α2b in combination with weight-based ribavirin, with standard treatment durations (48 weeks for HCV genotypes 1 and 4 and 24 weeks for HCV genotypes 2 and 3), and who received ≥80% of the recommended dose of both agents during the first 12 weeks of therapy. SVR was defined as HCV RNA below the limit of detection in a sensitive assay ≥24 weeks after treatment completion, and all patients who failed to achieve SVR were classified as no SVR. Data on rapid, early, and end-of-treatment virologic response were not available. Demographic and clinical characteristics, including age, sex, HCV genotype, human immunodeficiency virus (HIV) coinfection, baseline HCV viral load, liver histology, serum ferritin levels, serum alanine aminotransferase (ALT) levels, HCV treatment, alcohol consumption, body mass index (BMI), and presence or absence of diabetes, were extracted from clinical databases. Serum ferritin levels were determined at local laboratories. Heavy alcohol intake was defined as consumption >40 g per day over a period of ≥5 years. Liver biopsies were evaluated by experienced local pathologists. Liver fibrosis was classified according to the METAVIR score. Necroinflammatory activity was stratified into two groups: absent to mild activity (METAVIR A0-A1) versus moderate to high activity (METAVIR A2-3). Steatosis was considered present when detected ≥5% of all hepatocytes. Quantitative intrahepatic iron determination was not systematically performed in the SCCS. The study was approved by local ethical committees.

GWAS and Genotyping of IL28B Single-Nucleotide Polymorphisms.

A GWAS of determinants of serum ferritin levels in CHC patients was performed. Patients were included in this analysis if they met the above-indicated inclusion criteria and if DNA samples were genotyped previously within the framework of a GWAS of spontaneous and treatment-induced clearance from HCV infection.30 Genotyping and genotype data analyses were described previously.30 The present GWAS focused on serum ferritin levels measured before treatment initiation and/or liver biopsy, as described above. The following phenotypes were defined: serum ferritin as continuous variable; serum ferritin < versus ≥ the sex-specific median; serum ferritin < versus ≥ the 90% percentile. Each of the more than 500,000 single-nucleotide polymorphisms (SNPs) was tested for associations among each phenotype by linear or logistic regression models, as appropriate. The covariates, age and sex, were included in the regression analyses. Bonferroni correction was used to adjust for multiple testing, applying a significance threshold of 5 x 10−8. IL28B rs8099917 genotyping was performed previously or performed as previously described.30, 31

Statistical Analysis.

Associations between serum ferritin < versus ≥ the sex-specific median with continuous (i.e., HCV viral load and serum ALT levels) and dichotomic variables (e.g., SVR versus no SVR, necroinflammatory activity [none-mild versus moderate-high], stage of liver fibrosis [METAVIR F0-F1 versus F2-F4], and presence or absence of steatosis) were assessed in linear and logistic regression models, respectively. After univariate analyses, multivariate analyses were performed for significant associations. Multivariate analyses were obtained by using backward selection, using a P value ≥0.15 for removal from the model. Only patients with complete data for the remaining covariates were included in multivariate analyses. Sex, age, and serum ferritin levels were forced into the model. Group differences (e.g., serum ferritin levels according to IL28B rs8099917 TT versus GT/GG) were assessed by means of χ2 contingency tables or Wilcoxon-Mann-Whitney U tests, as appropriate. P values <0.05 were considered to be significant.

Results

Patient Characteristics.

A total of 876 patients were included in the response to treatment study, 980 in the liver histology study, and 707 Caucasian patients in the GWAS of ferritin serum levels in chronic HCV infection. In the response to treatment group, 606 (69%) and 270 (31%) patients were treated with PEG-IFN-α2a and PEG-IFN-α2b, respectively, in combination with weight-based ribavirin. The overall SVR rates in patients infected with HCV genotypes 1, 2, 3, and 4 were 43%, 86%, 78%, and 47%, respectively. Demographic and clinical characteristics of patients are shown in Table 1. The sex-specific median of serum ferritin levels in the whole cohort was 85 μg/L in females and 203 μg/L in males. Ferritin levels above these values were considered as elevated in the following analyses.

Table 1. Baseline and Demographic Characteristics
CharacteristicsTreatment Response (N = 876)Liver Histology (N = 980)GWAS (N = 707)
  • A number of patients overlap between the treatment response, liver histology, and GWAS groups.

  • *

    Detection before treatment initiation in the treatment response group and before first liver biopsy in the liver histology group.

  • Alcohol consumption data were missing in 16 treatment and 21 biopsy patients.

  • Diabetes data were missing in 11 treatment and 77 biopsy patients.

  • §

    ALT data were missing in 278 treatment and 1 biopsy patients.

  • Steatosis data were missing in 309 treatment and 94 biopsy patients.

  • Fibrosis data were missing in 314 treatment patients.

  • #

    Activity data were missing in 316 treatment and 10 biopsy patients.

  • **

    HIV serostatus was missing in 155 treatment and 116 biopsy patients.

Male sex, n (%)569 (65)638 (65)450 (0.64)
Caucasian ethnicity815 (93)917 (94)707 (100)
Age (years), mean (range)*45 (19-75)44 (17-76)45 (21-75)
Alcohol (40 g/day ≥5 years), n (%)*,157 (18)182 (19)104 (16)
Diabetes, n (%)*,69 (8)66 (7)51 (7)
BMI (kg/m2), mean (range)26 (14-60)25 (14-42)24 (16-42)
ALT (U/L), mean (range)*,§102 (6-693)103 (4-2,082)104 (8-720)
Ferritin (μg/L), mean (range)*237 (3-2,183)240 (3-4,525)250 (3-3,358)
Steatosis (presence), n (%)417 (74)588 (66)373 (70)
Fibrosis stage, n (%)   
 0-1180 (32)707 (72)212 (49)
 2-4382 (68)273 (28)336 (61)
Activity#   
 None to mild421 (75)780 (80)437 (81)
 Moderate to severe139 (25)190 (20)105 (19)
HCV genotype, n (%)   
 1383 (44)528 (54)343 (49)
 298 (11)82 (8)83 (12)
 3322 (37)269 (27)222 (31)
 473 (8)101 (10)59 (8)
HCV RNA (log10), mean (range)*5.96 (1-8)5.88 (1-8)6.00 (1-8)
HIV coinfection, n (%)*,**35 (5)61 (6)40 (7)

Serum Ferritin Levels and Response to Treatment of CHC.

Elevated serum ferritin (ferritin >sex-specific median) was among the strongest pretreatment predictors of failure to achieve SVR (P < 0.0001, odds ratio [OR] favoring SVR = 0.45, 95% confidence interval [CI] = 0.34-0.60). These results were comparable when serum ferritin was analyzed as a continuous variable (P < 0.0001), or when selecting a different cutoff for high serum ferritin (P < 0.0001 for ferritin cutoff >200 and >250 μg/L). The association between elevated ferritin and response to treatment remained highly significant in multivariate analyses (P = 0.0002, OR = 0.35, 95% CI = 0.20-0.61) after adjustment for known predictors of SVR, including the IL28B genotype, baseline viral load, HCV genotype, presence of diabetes, and liver fibrosis stage (Table 2). When free serum iron and transferrin saturation were entered into the model, the association between ferritin and treatment failure remained significant (OR = 0.42 [0.23-0.78], P = 0.006).

Table 2. Uni- and Multivariate Analyses of Variables Associated With Treatment Response in Patients With CHC
 Univariate AnalysisMultivariate Analysis
VariablesOR (95% CI)P ValueOR (95% CI)P Value
  • *

    Iron was excluded from the above-shown multivariate analysis, but the association with ferritin remained significant when iron was entered into the model (OR = 0.42 [0.23-0.78], P = 0.006).

  • A sex-specific median was applied (80 μg/L in women, 200 μg/L in men). Results of this multivariate analysis were comparable when ferritin serum concentrations were included as continuous variables or with fixed cutoffs ≥200/250 μg/L. Overall, 376 patients with complete datasets were included in multivariate analysis.

Age (years, continuous)0.96 (0.95-0.98)<0.00010.98 (0.95-1.01)0.13
Male sex0.81 (0.60-1.07)0.140.68 (0.39-1.18)0.17
BMI (kg/m2, continuous)0.95 (0.92-0.99)0.02  
Diabetes, presence0.49 (0.30-0.80)0.0050.40 (0.14-1.14)0.09
Alcohol (40 g/day ≥5 years)0.76 (0.51-1.14)0.2  
ALT (U/L, continuous)1.00 (1.00-1.00)0.6  
Activity, none to mild versus moderate to high0.93 (0.63-1.37)0.7  
Fibrosis, F3-F4 versus F0-F20.52 (0.36-0.74)<0.00010.47 (0.26-0.85)0.01
Steatosis, presence0.62 (0.42-0.93)0.02  
IL28B rs8099917, TT versus TG/GG0.35 (0.25-0.48)<0.00010.20 (0.11-0.34)<0.0001
Iron (ug/dL, continuous)*0.98 (0.97-1.00)0.01  
Transferrin saturation (%, continuous)1.00 (0.99-1.00)0.5  
Ferritin ≥ median0.45 (0.34-0.60)<0.00010.35 (0.20-0.61)0.0002
HCV genotype 2/3 versus 1/45.12 (3.78-6.93)<0.00013.34 (1.95-5.74)<0.0001
HCV RNA (log10 RNA, continuous)0.38 (0.31-0.46)<0.00010.30 (0.21-0.44)<0.0001
HIV coinfection0.59 (0.24-1.47)0.3  

The number of HCV genotype 1– and genotype 3–infected patients was sufficient for stratified analyses. In both genotypes, serum ferritin was a strong, independent predictor of failure to achieve SVR (Supporting Table 1). In HCV genotype 3–infected patients, ferritin was an even stronger predictor than IL28B rs8099917 genotype (P = 0.0005, OR = 0.38, 95% CI = 0.22-0.66 for ferritin versus P = 0.10, OR = 0.58, 95% CI = 0.31-1.10 for rs8099917). In the multivariate model, ferritin remained a strong, independent predictor of SVR in HCV genotype 3 patients (OR = 25 [0.08-0.78], P = 0.02), though its effect was slightly smaller than that of IL28B rs8099917 (OR = 20 [0.06-0.65], P = 0.007). Detailed test characteristics of serum ferritin versus IL28B rs8099917 genotype as predictors of SVR are shown in Supporting Table 2.

To analyze the contribution of both ferritin levels and IL28B polymorphisms to response to treatment, patients were stratified according the IL28B rs8099917 genotype (i.e., unfavorable TG or GG versus favorable TT) and the ferritin level (i.e., high versus Low, if above or below the sex-specific median; Table 3). In both HCV genotype 1– and genotype 3–infected patients with unfavorable IL28B genotype, SVR rates differed by >30% after stratification for low versus high serum ferritin (P = 0.0001 for HCV genotype 1, P = 0.0002 for HCV genotype 3), whereas the association between serum ferritin and SVR was less significant in patients with favorable IL28B genotype (P = 0.033 for HCV genotype 1, P = 0.2 for HCV genotype 3; Table 3). Therefore, serum ferritin appears to add to the value of IL28B genotype in predicting treatment outcome, especially in patients with an unfavorable IL28B genotype.

Table 3. SVR Rates of HCV Genotype 1 and 3 Patients Stratified According to IL28B Genotype and Serum Ferritin
HCV Genotypers8099917 GenotypesFerritinSVR YesSVR NoSVR Rate (%)P Value
  1. Low and high ferritin indicate a value below or above the sex-specific median. P values were calculated using the Fisher's exact test.

  2. Abbreviation: NS, not significant.

1 (N = 283)TTLow542766.70.033
High363848.7
TG/GGLow253144.60.0001
High106213.9
3 (N = 218)TTLow691384.2NS
High411474.6
TG/GGLow40687.00.0002
High171848.6

Serum Ferritin Levels and Liver Fibrosis Stage As Well As Necroinflammatory Activity in Patients With CHC.

Elevated ferritin (>sex-specific median) was associated with severe liver fibrosis (F3-4, P < 0.0001, OR = 3.27, 95% CI = 2.44-4.38; Table 4). This association was still significant in a multivariate model (P < 0.0001, OR = 2.67, 95% CI = 1.68-4.25) after adjustment for other known predictors of fibrosis, including male sex, diabetes, steatosis, high necroinflammatory activity, and heavy alcohol intake. Elevated serum ferritin was also associated with moderate to high necroinflammatory activity (univariate P < 0.0001, OR = 2.08, 95% CI = 1.50-2.87), but this association lost significance after adjustment for relevant covariates (P = 0.3; Table 4).

Table 4. Uni- and Multivariate Analyses of Variables Associated With Liver Fibrosis Stage and Necroinflammatory Activity
 Univariate AnalysisMultivariate Analysis
VariablesOR (95% CI)P ValueOR (95% CI)P Value
  • *

    A sex-specific median was applied (80 μg/L in women, 200 μg/L in men). Results of this multivariate analysis were comparable when ferritin serum concentrations were included as continuous variables or with fixed cutoffs ≥200/250 μg/L and when a cutoff for fibrosis stage of F0-F1 versus F2-F4 was applied. Overall, 312 and 519 patients with complete datasets were included in multivariate analyses of fibrosis and activity, respectively.

Fibrosis stage, F0-2 versus F3-4    
 Age at infection (years, continuous)1.02 (1.01-1.04)0.0021.01 (0.99-1.04)0.3
 Male sex1.35 (1.00-1.83)0.051.57 (0.93-2.65)0.09
 BMI (kg/m2, continuous)1.05 (1.02-1.09)0.003  
 Diabetes4.81 (2.86-8.09)<0.00016.68 (2.63-17.0)0.0001
 Alcohol (40 g/day ≥5 years)2.76 (1.98-3.86)<0.00012.62 (1.57-4.39)0.0002
 ALT (U/L, continuous)1.00 (1.00-1.00)0.03  
 Activity (none-mild versus moderate to high)5.95 (4.24-8.35)<0.00014.46 (2.48-8.02)<0.0001
 Steatosis3.43 (2.37-4.97)<0.00012.02 (1.17-3.49)0.01
 IL28B rs8099917 (TT versus TG/GG)0.97 (0.68-1.39)0.9  
 Ferritin, ≥median*3.27 (2.44-4.38)<0.00012.67 (1.68-4.25)<0.0001
 HCV RNA (log10 RNA, continuous)0.95 (0.84-1.09)0.5  
 HIV coinfection1.50 (0.86-2.63)0.16  
Necroinflammatory activity (none to mild versus Moderate to high)    
 Age at infection (years, continuous)1.04 (1.02-1.06)<0.00011.03 (1.00-1.05)0.04
 Male sex1.12 (0.80-1.57)0.51.36 (0.73-2.52)0.3
 BMI (kg/m2, continuous)1.04 (1.01-1.08)0.02  
 Diabetes2.54 (1.49-4.31)0.0006  
 Alcohol (40 g/day ≥5 years)1.38 (0.94-2.03)0.1  
 ALT (U/L, continuous)1.00 (1.00-1.00)0.007  
 Steatosis3.15 (2.05-4.86)<0.00013.68 (1.56-8.67)0.003
 Fibrosis (F0-F2 versus ≥F3-F4)5.95 (4.24-8.35)<0.00014.71 (2.67-8.29)<0.0001
  IL28B rs8099917 (TT versus TG/GG)0.74 (0.51-1.07)0.11  
 Ferritin, ≥median*2.08 (1.50-2.87)<0.00011.38 (0.77-2.48)0.3
 HCV RNA (log10 RNA, continuous)0.82 (0.71-0.94)0.005  
 HIV coinfection0.58 (0.26-1.31)0.2  

Serum Ferritin Levels and Liver Steatosis in Patients With CHC.

Serum ferritin ≥ the sex-specific median was strongly associated with the presence of steatosis in >5% of hepatocytes (univariate P < 0.0001, OR = 2.65, 95% CI = 1.97-3.56). Other variables associated with steatosis in >5% of hepatocytes were high ALT serum levels, high BMI, IL28B genotype, male sex, infection with HCV genotype 3, presence of diabetes, high necroinflammatory activity, advanced liver fibrosis, and heavy alcohol intake (Table 5). In a multivariate analysis, high ferritin levels remained a strong, independent predictor of steatosis (P = 0.002, OR = 2.29, 95% CI = 1.35-3.91), together with BMI, infection with HCV genotype 3, high necroinflammatory activity, and severe liver fibrosis (Table 5). These findings were comparable when using another steatosis cutoff (i.e., present or absent).

Table 5. Uni- and Multivariate Analyses of Variables Associated With Liver Steatosis (≤5% Versus >5% of Hepatocytes)
 Univariate AnalysisMultivariate Analysis
VariablesOR (95% CI)P ValuesOR (95% CI)P Values
  • *

    A sex-specific median was applied (80 μg/L in women, 200 μg/L in men). Results of this multivariate analysis were comparable when steatosis was analyzed as present versus absent. Overall, 392 patients with complete datasets were included in multivariate analysis.

Age at infection (years, continuous)1.01 (1.00-1.03)0.141.01 (0.98-1.04)0.4
Male sex1.69 (1.26-2.27)0.00051.65 (0.98-2.79)0.1
BMI (kg/m2, continuous)1.13 (1.09-1.18)<0.00011.15 (1.07-1.22)<0.0001
Diabetes5.02 (2.34-10.75)<0.0001  
Alcohol (40 g/day ≥5 years)1.56 (1.08-2.26)0.02  
ALT (U/L, continuous)1.00 (1.00-1.01)0.0004  
Fibrosis, F0-F2 versus ≥F3-F43.13 (2.19-4.48)<0.00011.79 (0.96-3.35)0.07
Activity, none to mild versus moderate to high2.60 (1.73-3.92)<0.00012.56 (1.14-5.77)0.02
IL28B rs8099917, TT versus TG/GG1.44 (1.00-2.05)0.05  
Ferritin, ≥median*2.65 (1.97-3.56)<0.00012.29 (1.35-3.91)0.002
HCV genotype 3 versus non-32.40 (1.71-3.36)<0.00014.08 (2.33-7.14)<0.0001
HCV RNA (log10 RNA, continuous)1.07 (0.90-1.28)0.4  
HIV coinfection0.96 (0.54-1.70)0.9  

Genetic Study of Determinants of Serum Ferritin Levels in CHC Patients.

First, we analyzed the associations of nine SNPs from four genes that were previously associated with elevated serum ferritin at the genome level in healthy subjects (Table 6).32-36 These SNPs included two major genetic variations in the HFE gene associated with hemochromatosis (rs1800562, resulting in C282Y, and rs1799945, resulting in H63D).37 In our cohort, only candidate SNPs in the HFE gene (rs1800562; P = 0.03) and in LRRC16A (rs2274089; P = 0.02), a gene involved in the organization of the cytoskeleton, were weakly associated with serum ferritin levels, although these associations would be no more significant after correction for multiple testing (Table 6). Importantly, patients with proven or suspected hemochromatosis were excluded from the present study. None of these SNPs were significantly associated with SVR, though some P values reached close to the level of statistical trends, suggesting that a larger sample size may reveal significant associations (Supporting Table 3).

Table 6. Associations Between Candidate SNPs and Serum Ferritin Levels in Patients With CHC
GeneSNP rs#ChrNoneffect AlleleEffect Alleler2_hatBetaSEP Value
  1. Candidate SNPs included in this analysis were identified by previous GWAS to be associated with serum ferritin concentrations in healthy populations.

  2. Abbreviations: Chr, chromosome; r2_hat, imputation quality score; SE, standard error.

TMPRSS6rs85579122AG0.990.0720.0510.16
HFErs18005626AG0.99−0.2550.120.033
 rs17999456CG1.000.0650.0680.34
 rs173427176CT0.990.0990.0900.27
 rs14082726GT0.99−0.2650.130.032
Transferrinrs38116473AG1.00−0.0420.0530.435
 rs17998523CT0.99−0.0950.0790.26
 rs22806733AC1.000.0110.0540.85
LRRC16Ars22740896CT1.000.2370.1030.022

We next performed a GWAS to test whether genetic variations other than those observed in the general population influence serum ferritin levels in CHC patients. However, no SNP reached genome-wide significance for an association with serum ferritin levels as a continuous variable (Table 7) or as a categorical variable (< versus ≥ the sex-specific median or ≤ versus ≥ the 90% percentile; not shown). To our knowledge, no clear biological evidence for a role in iron metabolism is known for each of the top SNPs of this GWAS (Table 7).

Table 7. Genome-Wide Association Study of Determinants of Serum Ferritin Levels in Patients With CHC (Top 10 SNPs Are Shown)
GeneSNP rs#ChrPositionDist (kb)AaAbr2_hatBeta (Ab)SEP Value
  1. Aa, allele a; Ab, allele b; chr, chromosome; dist, distance from gene; kb, kilobase; SE, standard error.

KIAA0494rs107894911469518970AG0.990.3120.0649.57 10−07
ARHGEF10Lrs225413511770732931,587CT0.970.2690.0583.41 10−06
DNASE2Brs121447151846378180AT0.91−0.6360.146.55 10−06
SCG2rs1686496822241679271,978AG0.99−0.8260.1773.28 10−06
CWH43rs117259574487576490AG1.000.7270.1637.74 10−06
SLC22A3rs250491661607440180AT0.990.3060.0652.01 10−06
RP11-157J24.1rs112427046148099720,474AG0.98−0.2570.0565.47 10−06
EPHA7rs1252781869396784040,023CT0.980.2740.0617.29 10−06
LAMA4rs1078217261126876755,133CT0.960.330.0759.73 10−06
OLFM1rs10776934913716952216,671GT0.95−0.2810.0581.05 10−06

Discussion

In the present study, we show that serum ferritin is a strong predictor of failure to achieve SVR after treatment of CHC with PEG-IFN-α and ribavirin independently of other predictors of virologic response, including the IL28B genotype. In addition, serum ferritin was independently associated with advanced liver fibrosis and the presence of steatosis, but not with necroinflammatory activity. In a candidate gene approach and in a GWAS, we did not identify significant genetic determinants of elevated serum ferritin levels in patients with CHC.

Importantly, serum ferritin was one of the strongest predictors of treatment response in our cohort, with an OR for treatment failure comparable to the IL28B genotype. In this regard, the effect of serum ferritin on treatment response, in comparison to the IL28B genotype, appeared to be somewhat less pronounced in HCV genotype 1 patients than in patients infected with HCV genotype 3. However, in both HCV genotype 1 and 3 patients, serum ferritin appeared to provide significant additional information to IL28B genotyping in patients with a poor-response IL28B genotype, in whom SVR rates differed by >30% after stratification by serum ferritin levels. An association between serum ferritin levels and outcome of IFN-α-based therapy has been reported previously in smaller studies, although it remained controversial as to whether serum ferritin is an independent predictor of treatment response or whether it is merely a surrogate marker for advanced stages of liver fibrosis.18, 24-27 The present, to our knowledge, largest, most comprehensive study, which was restricted to adherent patients who had received the recommended dose of PEG-IFN-α and ribavirin during the first 12 weeks of therapy, had the power to clearly demonstrate that serum ferritin is an independent predictor of treatment failure, even though serum ferritin was also independently associated with advanced liver fibrosis and steatosis. This observation indicates that serum ferritin is a promising candidate to be included in the panel of predictors of response to IFN-α-based therapy.12, 26, 38-42 This may be especially relevant in patients infected with HCV genotype 3, in whom the utility of IL28B genotyping is limited, even though a substantial number of patients cannot be cured with PEG-IFN-α and ribavirin alone.43, 44 In addition, the development of directly acting antivirals for HCV genotype 3 patients is less advanced, compared to HCV genotype 1.3 In this context, it may also be worthwhile to assess the relationship of serum ferritin and response to triple therapy, including telaprevir, in treatment-experienced patients with previous null response—a group of patients with limited chances to achieve cure, even by triple therapy.6

For our statistical analyses, we have chosen a sex-specific median as the cutoff between high and low serum ferritin levels, but such cohort-specific values are not suitable for clinical decision making. Nevertheless, the sex-specific median for both women (85 ug/L) and men (203 ug/L) in our cohort of HCV-infected patients was relatively close to the upper limit of normal of the general population (e.g., 150 ug/L in women and 300 ug/L in men, with some variations according to the assay used). Therefore, a threshold of serum ferritin at the upper limit of normal, or even slightly below, might be used to predict treatment failure, though additional studies are required to validate these suggestions.

In our study, serum ferritin levels were also strongly associated with advanced liver fibrosis, as well with the presence of steatosis. These associations remained highly significant in multivariate models, and serum ferritin did not appear to be a simple surrogate of either steatosis or advanced fibrosis. In our study, we could confirm that infection with HCV genotype 3 is an independent risk factor of liver steatosis (Table 5), as described previously.45-47 However, according to subanalyses, the association between serum ferritin and steatosis remained highly significant in uni- and multivariate models in patients infected with HCV genotype 1, but only in the univariate analysis in HCV genotype 3 patients (Supporting Table 4). Altogether, serum ferritin appears to provide additional information in describing a distinct phenotype of CHC characterized by advanced liver fibrosis and steatosis as well as a poor response to IFN-α-based therapy.

We also aimed to investigate potential genetic determinants of serum ferritin levels in patients with CHC. No significant associations between SNPs and serum ferritin levels were identified in a GWAS including more than 700 patients with CHC, suggesting that at least common genetic variants do not substantially influence ferritin levels in this context. Among the top hits of this GWAS was rs16864968 very close to SCG2, encoding secretogranin II, a protein with immune-modulating properties.48 However, for none of the top hits was an obvious link to iron metabolism described thus far. The known association between serum ferritin and rs1800562 and rs1408272 in HFE and rs2274089 in LRRC16A could be confirmed by a candidate gene approach, although the association appeared to be weak, compared to individuals not infected with HCV.

Our study had some limitations, because stainable hepatic iron deposition was not quantitatively assessed in the SCCS, and it therefore was not possible to assess the relationship between serum ferritin and intrahepatic iron load. Thus, our study does not provide any evidence on a causal role of serum ferritin in treatment response. In addition, we were unable to take potential diurnal and nonfasting variability of serum ferritin and serum iron levels into account, and because of the lack of on-treatment viral load data, we could not assess associations between serum ferritin and rapid or early virologic response. Moreover, it remains unclear whether our findings could be extrapolated to patients of other ethnicities, because almost all patients included in our study were of Caucasian origin.

In principle, increased serum ferritin levels in the context of HCV infection could reflect, at least partially, an increased hepatic iron load, which might be explained by dysregulated hepcidin expression.17 However, serum ferritin is also frequently increased as a consequence of inflammatory responses during acute and chronic infectious diseases, autoimmune diseases, or cancer, a finding that may substantially contribute to increased serum ferritin levels in patients with CHC. Both increased hepatic iron and an ongoing systemic inflammatory response can contribute to an unfavorable course of CHC, of which serum ferritin appears to be a surrogate. In this regard, it was shown recently that a strong host response against translocated gut microbial products correlates with an unfavorable outcome of chronic hepatitis B and C as well as with elevated serum ferritin levels.49 Our finding, that free serum iron, but not transferrin saturation, is associated with treatment failure may point to such a role of serum ferritin levels as a surrogate for a systemic inflammatory response in CHC. Nevertheless, it has been shown that iron reduction by phlebotomy results in a significant reduction of serum ferritin levels in patients with CHC, and iron reduction may improve liver histology and treatment response in CHC patients.15, 21, 22, 50 This may be especially relevant in patients who are unable to receive or who did not respond to triple therapy.

In conclusion, we show that serum ferritin is strongly and independently associated with failure to achieve HCV eradication by IFN-α-based therapy as well as with advanced liver fibrosis and steatosis. Serum ferritin may be included in clinical decision making in patients with CHC, especially in patients with a poor IL28B genotype.

*The members of the Swiss Hepatitis C Cohort Study Group are Francesco Negro (Geneva, Chairman), Antoine Hadengue (Geneva, Chairman of Scientific Committee), Laurent Kaiser, and Laura Rubbia-Brandt (Geneva); Darius Moradpour, Cristina Cellerai (Lausanne), and Martin Rickenbach (Lausanne Data Center); Andreas Cerny and Gladys Martinetti (Lugano); Jean-François Dufour, Meri Gorgievski, and Virginie Masserey Spicher (Berne); Markus Heim and Hans Hirsch (Basel); Beat Müllhaupt, Beat Helbling, and Stephan Regenass (Zurich); Raffaele Malinverni (Neuchatel); David Semela and Guenter Dollenmaier (St. Gallen); and Gieri Cathomas (Liestal).

Ancillary