Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Relevance of Baseline Viral Genetic Heterogeneity and Host Factors for Treatment Outcome Prediction in Hepatitis C Virus 1b-Infected Patients

  • Verónica Saludes,

    Current address: Department of Experimental and Health Sciences, Universitat Pompeu Fabra, Barcelona, Spain

    Affiliations Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain, CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain

  • Elisabet Bascuñana,

    Affiliation Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain

  • Elena Jordana-Lluch,

    Affiliation Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain

  • Sònia Casanovas,

    Affiliation Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain

  • Mercè Ardèvol,

    Affiliation Hospital Pharmacy, Hospital Universitari Germans Trias i Pujol, Badalona, Spain

  • Esther Soler,

    Affiliations Liver Unit, Hospital Universitari Germans Trias i Pujol, Badalona, Spain, CIBER Enfermedades Hepáticas y Digestivas (CIBEREHD), Barcelona, Spain

  • Ramón Planas,

    Affiliations Liver Unit, Hospital Universitari Germans Trias i Pujol, Badalona, Spain, CIBER Enfermedades Hepáticas y Digestivas (CIBEREHD), Barcelona, Spain

  • Vicente Ausina,

    Affiliations Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain, CIBER Enfermedades Respiratorias (CIBERES), Bunyola, Spain

  • Elisa Martró

    emartro.igtp.germanstrias@gencat.cat

    Affiliations Microbiology Service, Fundació Institut d’Investigació en Ciències de la Salut Germans Trias i Pujol, Hospital Universitari Germans Trias i Pujol, Universitat Autònoma de Barcelona, Badalona, Spain, CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain

Relevance of Baseline Viral Genetic Heterogeneity and Host Factors for Treatment Outcome Prediction in Hepatitis C Virus 1b-Infected Patients

  • Verónica Saludes, 
  • Elisabet Bascuñana, 
  • Elena Jordana-Lluch, 
  • Sònia Casanovas, 
  • Mercè Ardèvol, 
  • Esther Soler, 
  • Ramón Planas, 
  • Vicente Ausina, 
  • Elisa Martró
PLOS
x

Abstract

Background

Only about 50% of patients chronically infected with HCV genotype 1 (HCV-1) respond to treatment with pegylated interferon-alfa and ribavirin (dual therapy), and protease inhibitors have to be administered together with these drugs increasing costs and side-effects. We aimed to develop a predictive model of treatment response based on a combination of baseline clinical and viral parameters.

Methodology

Seventy-four patients chronically infected with HCV-1b and treated with dual therapy were studied (53 retrospectively −training group−, and 21 prospectively −validation group−). Host and viral-related factors (viral load, and genetic variability in the E1–E2, core and Interferon Sensitivity Determining Region) were assessed. Multivariate discriminant analysis and decision tree analysis were used to develop predictive models on the training group, which were then validated in the validation group.

Principal Findings

A multivariate discriminant predictive model was generated including the following variables in decreasing order of significance: the number of viral variants in the E1–E2 region, an amino acid substitution pattern in the viral core region, the IL28B polymorphism, serum GGT and ALT levels, and viral load. Using this model treatment outcome was accurately predicted in the training group (AUROC = 0.9444; 96.3% specificity, 94.7% PPV, 75% sensitivity, 81% NPV), and the accuracy remained high in the validation group (AUROC = 0.8148, 88.9% specificity, 90.0% PPV, 75.0% sensitivity, 72.7% NPV). A second model was obtained by a decision tree analysis and showed a similarly high accuracy in the training group but a worse reproducibility in the validation group (AUROC = 0.9072 vs. 0.7361, respectively).

Conclusions and Significance

The baseline predictive models obtained including both host and viral variables had a high positive predictive value in our population of Spanish HCV-1b treatment naïve patients. Accurately identifying those patients that would respond to the dual therapy could help reducing implementation costs and additional side effects of new treatment regimens.

Introduction

Hepatitis C virus (HCV), with an estimated 150 million people chronically infected worldwide, is the major causative agent of chronic liver disease, cirrhosis and hepatocellular carcinoma [1]. HCV has a positive single-stranded RNA genome that exhibits significant genetic variability, leading to the circulation within an infected individual of a dynamic mosaic of closely related viral variants usually referred to as quasispecies. This phenomenon has been associated with chronic infection establishment, pathogenicity and resistance to antiviral drugs [2].

Pegylated-interferon alpha (PegIFN-α) and ribavirin (RBV) combination therapy constitutes the current standard of care for the treatment of chronic hepatitis C by non-1 genotypes [3]. However, triple therapy adding an HCV-specific protease inhibitor (PI) [4], [5] has recently been approved for chronic infection by HCV genotype 1 (HCV-1) in several countries in America, Europe, the Middle East, Asia and Australia. Treatment failure rates for naïve HCV-1-infected patients decrease from 40–50% with PegIFN-α and RBV [6], [7], to about 25–33% with the triple therapy [4], [5], [8]. However, the addition of a PI increases the costs, side effects and drug-drug interactions of the dual therapy, and the efficacy of triple therapy depends largely on susceptibility to PegIFN-α and RBV. Therefore, there remains a need to identify those patients most likely to respond to this dual therapy before starting treatment in order to decrease the implementation costs of novel triple therapies, as well as the additional side effects. This will constitute a step forward towards personalized treatment regimens of chronic hepatitis C.

A number of host-related factors have been associated with IFN-α-based treatment failure in HCV-1-infected patients, such as African-American ancestry, advanced liver fibrosis or cirrhosis, older age, male gender, metabolic disorders, transaminase levels and, more recently, several host genetic polymorphisms and the level of certain chemokines in serum (reviewed in [9] and [10]). Baseline virological factors include high viral loads, high levels of genetic variability within the E1–E2 and NS5A regions, as well as mutations in the so-called Interferon Sensitivity Determining Region (ISDR) and the core region [11], [12]. Nevertheless, such associations have not always been found and remain controversial.

Although most studies focussed on the prediction of treatment outcome have been based on the predictive value of single host or viral factors, more recently predictive models combining several variables have been proposed for chronic infection by HCV-1. However, most of these models have not been validated and/or have a variable accuracy (reviewed in [13]). We previously developed a predictive model based on baseline host and viral variables [14]. In the current study, we considered additional variables including the IL28B polymorphism, and increased the sample size used to generate and validate new predictive models of PegIFN-α and RBV therapy response in HCV-1b infected patients.

Materials and Methods

Ethics Statement

This study was approved by the Clinical Research Ethics Committee at our institution (“Comité Ético de Investigación Clínica”, CEIC). Written informed consent was obtained for all patients.

Patients and Specimens

A total of 74 patients with chronic hepatitis C by HCV-1b treated with combination therapy at “Hospital Universitari Germans Trias i Pujol” were included. All of them were Caucasian and of Spanish origin. Exclusion criteria were: previous IFN-α-based treatment, HIV or HBV coinfection, alcohol abuse or having other causes of chronic liver disease. The patients had started antiviral therapy with PegIFN-α2a (180 µg/week) plus weight-based doses of RBV (1000–1200 mg/day) for 48 weeks between 2003 and 2011. The patients were considered either as responders (SVR, defined as undetectable HCV-RNA in serum 24 weeks after treatment cessation) or non-responders (continued presence of HCV-RNA during therapy −null response−, rebound of HCV-RNA while on therapy −breakthrough−, or 24 weeks after the end of treatment −relapse−). Included patients were classified into two groups: the training group consisted of 53 patients (retrospective study) and the validation group included 21 patients (prospective study). All virological analyses were performed using serum specimens obtained before treatment initiation and conserved at −80°C until testing.

Baseline Clinical and Epidemiological Host Parameters

The following variables were obtained by clinical record review: gender, age, weight, body mass index (BMI), stage of fibrosis according to the Forns index [15], serum levels of cholesterol, platelets, ALT, AST, and GGT. Enzyme levels were transformed into a quotient expressing the factor times upper limit of normal (ULN) according to gender. A good treatment adherence was considered when having received ≥80% of total maximum dose prescribed of both drugs for ≥80% of the expected duration of therapy [16].

Besides, the rs12979860 polymorphism near the human IL28B gene was determined by real-time PCR using the LightMix® Kit IL28B (Roche Diagnostics GmbH, Mannheim, Germany) according to manufacturer’s instructions starting from whole blood specimens collected in tubes containing EDTA.

Finally, serum levels of human Interferon-γ Inducible Protein 10 (IP-10) were quantitatively determined with the Quantikine ELISA Human CXCL10/IP-10 Immunoassay (R&D Systems, Abingdon, UK), following manufacturer’s instructions. Patients were classified as having low or high IP-10 values using a 600 pg/mL cut-off [17].

Baseline viral Parameters

Serum viral load.

HCV-RNA had been quantified by RT-PCR (Cobas® Amplicor HCV Monitor test, Roche Molecular Systems, Pleasanton, CA, USA) or by real-time RT-PCR (Abbott RealTime HCV assay, Abbott Molecular Inc., Des Plaines, IL, USA), according to manufacturer’s instructions and was recorded as Log10IU/mL.

RNA extraction and reverse transcription.

Total RNA was extracted from 220 µl of serum, using the QIAamp® viral RNA kit (QIAGEN® GmbH, Hilden, Germany) according to the manufacturer’s protocol. Reverse transcription was performed using random hexamers in order to prevent any bias during the reaction, as previously described [18].

PCR-cloning and sequencing of the E1–E2 region.

A 532-bp sequence encompassing the E1 C-terminal and the E2 N-terminal regions (including the HVR-1, HVR-2 and HVR-3) was amplified, cloned and sequenced as previously described [14] and referred to as E1–E2 region (nucleotides 1322–1853 in the H77 reference sequence, GenBank accession number AF009606). For the prospective patients, E1–E2 PCR products were cloned using the Zero Blunt TOPO PCR cloning kit for sequencing (Invitrogen). Between 24 and 35 colonies were selected and subjected to PCR followed by purification and sequencing of both DNA strands. Sequence readings were assembled and edited with the STADEN package v1.6. [19].

PCR and direct sequencing of the core region.

The whole core region (573 bp, H77 positions 342–914) was amplified and sequenced as previously described [14]. Sequences were assessed for the presence of amino acid polymorphisms associated with treatment outcome.

PCR and direct sequencing of the NS5A region.

A fragment of the NS5A region containing the ISDR was amplified and directly sequenced as described by Torres-Puente et al. [20]. The number of amino acid substitutions with respect to the HCV-J strain was determined.

Phylogenetic analysis of the core and E1–E2 regions.

Phylogenetic analysis of the HCV core region derived from the patients included in the study together with reference sequences was used to confirm that the HCV genotype and subtype was 1b. E1–E2 sequences were also subjected to phylogenetic analysis in order to rule out potential contamination between specimens. Sequences were aligned by ClustalW implemented in MEGA 4 [21]. Maximum-likelihood phylogenetic trees were obtained with PHYML [22].

Genetic variability analysis of the E1–E2 region.

Multiple alignments were generated with all clones generated from each patient for the complete E1–E2 region, and the HVR-1, HVR-2 and HVR-3 subregions (H77 nucleotide positions 1491–1571, 1761–1787, and 1632–1739, respectively). The following genetic variability estimates were obtained for each multiple alignment with DnaSP v4.50 [23]: total number of polymorphic sites (S), total number of mutations (η), nucleotide diversity corrected by Jukes-Cantor method (π), and number of viral variants (number of haplotypes, nHap). The number of substitutions per synonymous site (Ks) and number of nonsynonymous substitutions per nonsynonymous site (Ka) were obtained using the Nei-Gojobori method.

Statistical analysis.

Clinical and virological values were compared between responders and non-responders in bivariate analysis. Student’s t test (Normal distribution) or Mann-Whitney U test (non-Normal distribution) were used for quantitative variables, and Chi-square, Fisher’s exact test or Likelihood ratio test were used for categorical variables. Data was expressed as mean ± standard deviation, median and range, or relative frequency.

Statistical models were developed to predict treatment outcome. Firstly, a multivariate discriminant analysis [24] was carried out to develop a predictive model on the training group, which was then validated in the validation group. Covariates initially included in the discriminant model were explanatory variables that achieved a p-value ≤0.15 in the bivariate analyses (training group). Some variables were transformed (square root) in order to achieve normality. The discriminant function was obtained using a backward stepwise variable selection procedure. Secondly, a regression tree analysis [25] was performed using the same initial variables as in the discriminant analysis. The JMP10 software (SAS Institute Inc., Cary, NC, USA) was used to choose the variable and its optimal cut-off that was able to generate the most significant division of the training group of patients into two prognostic subgroups that were as homogeneous as possible for the probability of SVR. Then, this process was repeated on each subgroup of patients in a step-wise manner until no additional significant variable was identified. A ROC curve was obtained for each model and the effectiveness of prediction was measured by: area under the ROC curve (AUROC), sensitivity (proportion of responders which are correctly identified), specificity (proportion of non-responders which are correctly identified), negative predictive value (NPV) and positive predictive value (PPV). Cut-off values that yielded highest PPV and specificity were selected from the ROC curve. The reproducibility of the models was tested with the data from the validation group of patients. Statistical analyses were performed using SPSS v15.0 and JMP10. P-values<0.05 were considered significant.

Accession numbers.

All sequences obtained in this study were deposited in the EMBL Nucleotide Sequence Database (http://www.ebi.ac.uk/embl/) under the following accession numbers: FN675941–FN675983 for core, FN675984–FN676976 and HF572064–HF572784 for E1–E2 and NS5A regions.

Results

Patient Groups and Treatment Adherence

The training group consisted of 53 patients with a 47% SVR rate, and the validation group included 21 patients with a 57% SVR rate, adding up to a total of 74 patients. Both patient groups were comparable in terms of descriptive clinical-epidemiological characteristics (Table 1). All patients were on treatment for the complete expected time and adherence to both drugs was overall >80%.

In order to develop predictive multivariate models, firstly, the association between each baseline variable and treatment outcome was studied in the training group of patients.

Baseline Clinical variables Associated with Treatment Outcome

Baseline clinical characteristics of patients associated with treatment outcome in the training group are shown in Table 2. The IL28B polymorphism was the variable most strongly associated with treatment outcome (p = 1.53×10−4) with only 1/13 patients with the favourable C/C genotype not responding to therapy. The AST/ALT ratio (p = 0.022) and the GGT quotient (p = 0.055) were higher in non-responders, while the ALT quotient was higher in responders (p = 0.028). Non-responder patients tended to have a higher Forns fibrosis index score. Both groups of patients were comparable for the rest of variables. Regarding the IP-10 levels, although two patients had insufficient serum volume left to perform the assay, non-responders tended to have high levels of this chemokine more frequently than responders (7/20, 38.1% vs. 8/21, 35.0%), but this difference did not reach statistical significance (Figure S1).

thumbnail
Table 2. Baseline variables used for model development in the training group (p-value <0.15).

https://doi.org/10.1371/journal.pone.0072600.t002

Baseline viral Variables Associated with Treatment Outcome

All patients were confirmed to be infected with HCV-1b by phylogenetic analysis of the core region (Figure S2). Amino acid composition analysis of this genetic region also showed that the absence of amino acid arginine (R) at position 70 and leucine (L) at position 91 was more frequent in non-responders (p = 0.015). Regarding the E1–E2 genetic variability estimates, while non-responders tended to have higher values than responders for most of the parameters, those most strongly related to treatment outcome were the nHap in the whole E1–E2 studied region (p = 4.23×10−4) and in the HVR-1 subregion (p = 0.027). The phylogenetic analysis of E1–E2 sequences confirmed the absence of contamination events (Figure S3). Regarding the ISDR region, all patients showing four or more mutations belonged to the responder group (p = 0.034). Finally, the viral load tended to be higher in non-responders (Table 2).

Statistical Models for the Prediction of Treatment Outcome using Baseline Host and viral Variables

Discriminant analysis.

The variables that persisted in the multivariate discriminant predictive model in decreasing order of significance were: nHap_E1–E2 (F ratio = 14.441), the core amino acid substitution pattern (F ratio = 12.219), the IL28B polymorphism (F ratio = 5.189), GGT ratio (F ratio = 4.623ALT ratio (F ratio = 1.696and viral load (F ratio = 0.774)This model was able to accurately predict the achievement of a sustained virological response in the training group (AUROC = 0.9444; 96.3% specificity, 94.7% PPV, 75% sensitivity and 81% NPV) when a 0.86 cut-off was used to maximize the PPV (Table 3). These values remained high when the model was applied to the validation group (AUROC = 0.8148, 88.9% specificity, 90.0% PPV, 75.0% sensitivity and 72.7% NPV). On the other hand, a 0.4 cut-off could be used to better predict non-response to treatment, maximizing the NPV (92% sensitivity and NPV, 85.2% specificity, and 84.6% PPV in the training group; 83.3% sensitivity, 80.0% NPV, 88.9% specificity, and 90.9% PPV in the validation group).

thumbnail
Table 3. Sensitivity, specificity, and predictive values for the predictive models obtained.

https://doi.org/10.1371/journal.pone.0072600.t003

Decision tree analysis.

The generated decision tree is shown in Figure 1. The variables that persisted in this predictive model in decreasing order of significance were: the IL28B polymorphism (G2 = 14.1257), the ALT ratio (G2 = 12.8909), the nHap_E1–E2 (G2 = 12.1293), and the Forns index (G2 = 6.6038). This model was able to predict treatment outcome accurately in the training group (AUROC = 0.9072, 84.4% specificity, 80.0% PPV, 95.2% sensitivity and 96.4% NPV) (Table 3). In the validation group these values decreased to 70% specificity, 75.0% PPV, 81.8% sensitivity and 77.8% NPV (AUROC = 0.7361).

thumbnail
Figure 1. Decision tree model generated in the training group.

The factors used for splitting and their cut-offs are indicated. Pie charts represent the rate of sustained virological response in white (the percentage is indicated) for each group of patients after each split. nHap_E1E2, number of haplotypes in the E1–E2 studied region; ALT quotient, square root of the alanine transaminase levels expressed as factor times upper limit of normal used in our center for males and females (41 and 31 U/L, respectively).

https://doi.org/10.1371/journal.pone.0072600.g001

Discussion

The new standard of care for chronic HCV-1 infection based on the administration of an HCV-specific PI, PegIFN-α and RBV has increased the treatment success rate [3]. However, this triple drug combination is associated with additional side effects and markedly higher health care costs than for PegIFN-α and RBV. It is important to consider that about 50% of HCV-1 patients successfully respond to the dual therapy [6], [7], which still is the current standard of care for HCV-1 chronic infection in many countries where PI are still not available or remain unaffordable. Moreover, in those countries where PI are already being administered, the triple therapy may not be appropriate for all patients; naïve patients with the IL28B-C/C genotype and F0–F2 fibrosis stage may still be treated with PegIFN-α plus RBV [26]. Therefore, a reliable prediction of response to dual therapy at baseline would be highly beneficial for the development of more effective and personalized treatment selection algorithms in order to optimize both patient wellbeing and health care expense.

Predictive Models of Response to PegIFN-α and RBV Therapy

In this study, we developed two predictive models including host and viral variables that could help to improve treatment selection algorithms and assist clinicians in decision making.

The predictive model obtained by discriminant analysis generated an aggregate probability of response to treatment based on the IL28B polymorphism, and serum GGT and ALT levels as host variables, as well as the E1–E2 number of haplotypes, the core amino acid substitution pattern, and the viral load asviral variables. This model, which could be easily implemented in a computer-based application, showed an AUROC of 0.9444 and a high PPV both in the training and the validation groups (94.7 and 90.0%, respectively), thus offering a reliable prediction of SVR. As predictive models obtained by decision tree analysis might be easier to implement and interpret in the clinical setting, a second predictive model was generated. However, this model showed a lower PPV (80% and 75% in the training and validation groups, respectively) and a worse reproducibility than the discriminant one.

Other predictive models have been generated but only a few have been validated. To the best of our knowledge, those that have been developed for HCV-1b-infected patients showed a lower predictive accuracy than the ones described in this study. E. Martínez-Bauer et al. [27] developed a score based on multiple regression analysis including the AST/ALT ratio, cholesterol levels, the Forns index and the HCV viral load, and predicted SVR in a subgroup of patients with a high PPV (96% in the training group and 90% in the validation group); however, response could not be predicted in the group of patients with intermediate score values (50% of the total number of patients). M. Kurosaki et al. [28] developed a predictive model based on decision-tree analysis using the IL28B polymorphism, platelet levels, the viral load and the number of ISDR mutations, and predicted SVR with 78% sensitivity and 70% specificity. T. Takayama et al. [29] found that artificial neural networks analysis predicted SVR with more accuracy than regression analysis, and obtained a 59% sensitivity and 71% specificity based on a number of host variables and the HCV viral load. A. Tsubota et al. [30] developed a multiple regression model using the variables gender, age, platelet count, the IL28B and SLC9A1 (a major ribavirin transporter gene) polymorphisms, and viral load, achieving a 73.3% PPV (71.4% in the confirmatory group). D. Miki et al. [31], using a prediction score based on multiple regression analysis including the variables BMI, IL28B polymorphism, and plasminogen activator inhibitor-1 (PAI-1) levels were able to predict SVR with 63% PPV (46% in the validation cohort).

Relevance of Host Factors for Treatment Outcome Prediction

Among host-related factors associated with IFN-α-based treatment response, the polymorphisms upstream the IL28B gene constitute the strongest predictive factor of SVR identified so far [32][36]; however, European-American patients not having the most favourable rs12979860 genotype (C/C) still have approximately 40% chance of responding to therapy [34]. Similarly, this variable showed some limitations as a predictor in our study; among our population of Spanish patients only 33.3% were C/C, and while 87.5% of them responded to therapy, 31.3% of those who did not have this genotype also did. Consensus guidelines state that IL28B testing may be considered, but recommendations in favour of the use of this marker are not strong as its individual predictive value is low [3], [37].

While it is well established that patients with an advanced fibrosis stage respond worse than those with null or mild fibrosis, liver biopsy had only been performed for 35.1% of the patients and we had to rely on the non-invasive Forns index. Whereas this fibrosis indicator is able to reliably differentiate between patients with and without advanced fibrosis, intermediate stages are not classifiable, which may explain why this variable was not as strongly associated with treatment outcome as expected.

ALT levels were significantly higher in responder patients as previously reported [11], [38], while GGT levels were higher in non-responders. High GGT levels have been reported as an important independent predictor of treatment failure [39][41]. Higher GGT levels have been related to advanced fibrosis, steatosis and insulin resistance, which are more common among non-responders [42]. Furthermore, J. Everhart et al. suggested that GGT reflects a state of oxidative stress and that it should be regarded as a marker of disease activity, as GGT levels were found to predict both treatment response and liver disease outcomes [39].

We also took into consideration other variables that had been previously associated with SVR such as age, gender, BMI, AST/ALT ratio, and cholesterol, platelets and IP-10 levels. However, none of them persisted in the final predictive models. IP-10 seems to be associated with a stronger first-phase decline in the HCV viral load, and low levels of this chemokine have been associated with SVR [17], [43], [44]. However, other authors have found an association with rapid virological response but not SVR [45].

Relevance of viral Factors for Treatment Outcome Prediction

The nHap_E1–E2 is an indicator of viral genetic heterogeneity, and a high value at baseline has been previously associated with dual therapy failure [14], [46], [47], either through the pre-existence or the generation of drug-resistant viral variants. This variable showed a greater significance in the discriminant predictive model than the rest of the variables. To the best of our knowledge, this is the first validated predictive model that includes a marker for baseline HCV quasispecies heterogeneity. While studying this variable by cloning and sequencing is labour intensive, alternative methodologies can be used; among them, next generation sequencing techniques have the capacity to simultaneously analyse several samples, which can decrease associated costs [48].

Baseline core amino acid substitutions at positions 70 (R by Q) and/or 91 (L by M) have been described as useful independent predictors of treatment failure in HCV-1b-infected patients [49][52]. However, this association has not been found in other studies [53] and it has been excluded from other predictive models [28], [54]. Several mechanisms have been proposed for the role of the Core protein in IFN resistance [55][57], and it has been suggested that this predictor could maintain its value in the era of triple therapy including Telaprevir [58].

A low HCV load has been suggested as a predictor of SVR [42], but the threshold to distinguish between low and high viral loads in not well established [37]. In our study, the viral load was treated as a continuous variable and, despite being marginally significant in the bivariate analysis, it was considered to be relevant in the discriminant model.

Finally, the association between the presence of ≥4 mutations in the ISDR and treatment response was initially described in Japanese patients [59] but it is less pronounced in European patients [60]. Only four patients in our study showed ≥4 mutations and all of them were responders, but this variable did not persist in any of the two generated models.

Study Limitations

Our study has several limitations: (i) recent studies have suggested that other single nucleotide polymorphisms in several human genes are associated to treatment outcome in HCV-1-infected patients, including the human leukocyte antigen C (HLA-C) and the killer cell immunoglobulin-like receptor (KIR) genes [61], the programmed cell death-1 (PD-1) gene [62], and the inosine triphosphatase (ITPA) gene [63]. These polymorphisms were not considered in our study and could have increased the accuracy of the predictive models; and (ii) the sample size was limited to 74 patients given the laboriousness of the assessment of HCV genetic heterogeneity in the E1–E2 region. However, a similar number of patients were included in each group, accounting for the fact that about 50% of patients infected by HCV-1b achieve an SVR. In addition, we performed a validation of the obtained models in a comparable group of patients in terms of ethnicity, clinical background and HCV subtype.

Conclusions

Achieving a rapid virological response at treatment week 4 has a high PPV (91%) for obtaining an SVR to PegIFN-α and RBV therapy, however, only 15–20% of persons with HCV-1 achieve this type of response [64], [65]. A sustained virological response to dual therapy could be predicted with a similarly high PPV (90.0% in the validation group) in our population of Spanish naïve HCV-1b-infected patients using the generated discriminant model, which was based on pretreatment host and viral variables. Those patients identified as responders could be treated with dual therapy with high chances of achieving an SVR; such a strategy could decrease the additional costs and side-effects associated with the triple therapy. Furthermore, most non-responders (88.9% specificity in the validation group) would also be identified as possible candidates for novel treatment regimens. Further studies should be performed to assess the applicability of the generated models to other populations.

Supporting Information

Figure S1.

IL-10 levels in responder and non-responder patients in the training group.

https://doi.org/10.1371/journal.pone.0072600.s001

(PDF)

Figure S2.

Genotype 1 phylogenetic subtree of the core region. Genotyped reference sequences available in the Los Alamos National Library HCV sequence database (http://hcv.lanl.gov/content/index) are shown in bold with the accession number and the HCV-1 subtype. The patients included in this study are identified with the patient identification number, accession number, and the treatment response group (R, responders; NR, non-responders). Substitution model: GTR+I+G (proportion of invariable sites: 0.369, gamma shape parameter: 0.449). Nodes supported with bootstrap values >70% (1000 replicates) are indicated. The scale bar represents substitutions per nucleotide position.

https://doi.org/10.1371/journal.pone.0072600.s002

(PDF)

Figure S3.

Unrooted phylogenetic tree of the E1–E2 region. All viral sequences obtained for each patient are identified with a vertical line, the patient identification number and the response group (R, responders; NR, non-responders). Substitution model: GTR+I+G (proportion of invariable sites: 0.311, gamma shape parameter: 1.094). All nodes corresponding to each individual patient were supported with bootstrap values >70%. The scale bar represents substitutions per nucleotide position. This tree shows the sequences derived from 31 patients; the phylogenetic tree for the rest of patients included in this study can be found at doi:10.1371/journal.pone.0014132.s001.

https://doi.org/10.1371/journal.pone.0072600.s003

(PDF)

Acknowledgments

The authors would like to thank Rosa Maria Morillas and Margarita Sala for patient recruitment under informed consent, Oliver Valero (Statistics Service, Universitat Autònoma de Barcelona) for his help in statistical analysis, Aída Ramírez and Lurdes Matas for clinical record review, and Oriol Pavón for sequence editing.

Author Contributions

Conceived and designed the experiments: VS VA EM. Performed the experiments: VS EB EJL SC MA ES EM. Analyzed the data: VS RP EM. Contributed reagents/materials/analysis tools: RP VA EM. Wrote the paper: VS EM. Critically revised the manuscript: EB EJL SC MA ES RP VA.

References

  1. 1. World Health Organization (WHO) (2012) Hepatitis C: WHO Fact Sheet N° 164. Available: http://www.who.int/mediacentre/factsheets/fs164/en/. Accessed 22 July 2013.
  2. 2. Farci P, Purcell RH (2000) Clinical significance of hepatitis C virus genotypes and quasispecies. Semin Liver Dis 20: 103–126.
  3. 3. Ghany MG, Nelson DR, Strader DB, Thomas DL, Seeff LB (2011) An update on treatment of genotype 1 chronic hepatitis C virus infection: 2011 practice guideline by the American Association for the Study of Liver Diseases. Hepatology 54: 1433–1444.
  4. 4. Poordad F, McCone J Jr, Bacon BR, Bruno S, Manns MP, et al. (2011) Boceprevir for untreated chronic HCV genotype 1 infection. N Engl J Med 364: 1195–1206.
  5. 5. Jacobson IM, McHutchison JG, Dusheiko G, Di Bisceglie AM, Reddy KR, et al. (2011) Telaprevir for previously untreated chronic hepatitis C virus infection. N Engl J Med 364: 2405–2416.
  6. 6. Manns MP, McHutchison JG, Gordon SC, Rustgi VK, Shiffman M, et al. (2001) Peginterferon alfa-2b plus ribavirin compared with interferon alfa-2b plus ribavirin for initial treatment of chronic hepatitis C: a randomised trial. Lancet 358: 958–965.
  7. 7. Fried MW, Shiffman ML, Reddy KR, Smith C, Marinos G, et al. (2002) Peginterferon alfa-2a plus ribavirin for chronic hepatitis C virus infection. N Engl J Med 347: 975–982.
  8. 8. Jacobson IM, Pawlotsky JM, Afdhal NH, Dusheiko GM, Forns X, et al. (2012) A practical guide for the use of boceprevir and telaprevir for the treatment of hepatitis C. J Viral Hepat. 19 Suppl 21–26.
  9. 9. Soriano V, Poveda E, Vispo E, Labarga P, Rallon N, et al. (2012) Pharmacogenetics of hepatitis C. J Antimicrob Chemother. 67: 523–529.
  10. 10. Tai AW, Chung RT (2009) Treatment failure in hepatitis C: mechanisms of non-response. J Hepatol 50: 412–420.
  11. 11. McHutchison JG, Lawitz EJ, Shiffman ML, Muir AJ, Galler GW, et al. (2009) Peginterferon alfa-2b or alfa-2a with ribavirin for treatment of hepatitis C infection. N Engl J Med 361: 580–593.
  12. 12. Wohnsland A, Hofmann WP, Sarrazin C (2007) Viral determinants of resistance to treatment in patients with hepatitis C. Clin Microbiol Rev. 20: 23–38.
  13. 13. Saludes Montoro V, Ausina Ruiz V, Martró Català E (2011) [Current options for predicting therapeutic response in chronically infected patients with hepatitis C virus genotype 1]. Enferm Infecc Microbiol Clin 29 Suppl 551–58.
  14. 14. Saludes V, Bracho MA, Valero O, Ardèvol M, Planas R, et al. (2010) Baseline prediction of combination therapy outcome in hepatitis C virus 1b infected patients by discriminant analysis using viral and host factors. PLoS One 5: e14132.
  15. 15. Forns X, Ampurdanes S, Llovet JM, Aponte J, Quinto L, et al. (2002) Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. Hepatology 36: 986–992.
  16. 16. McHutchison JG, Manns M, Patel K, Poynard T, Lindsay KL, et al. (2002) Adherence to combination therapy enhances sustained response in genotype-1-infected patients with chronic hepatitis C. Gastroenterology. 123: 1061–1069.
  17. 17. Darling JM, Aerssens J, Fanning G, McHutchison JG, Goldstein DB, et al. (2011) Quantitation of pretreatment serum interferon-gamma-inducible protein-10 improves the predictive value of an IL28B gene polymorphism for hepatitis C treatment response. Hepatology 53: 14–22.
  18. 18. Jiménez-Hernández N, Torres-Puente M, Bracho MA, García-Robles I, Ortega E, et al. (2007) Epidemic dynamics of two coexisting hepatitis C virus subtypes. J Gen Virol 88: 123–133.
  19. 19. Staden R, Beal KF, Bonfield JK (2000) The Staden package, 1998. Methods Mol Biol 132: 115–130.
  20. 20. Torres-Puente M, Cuevas JM, Jiménez-Hernández N, Bracho MA, García-Robles I, et al. (2008) Genetic variability in hepatitis C virus and its role in antiviral treatment response. J Viral Hepat 15: 188–199.
  21. 21. Tamura K, Dudley J, Nei M, Kumar S (2007) MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Mol Biol Evol 24: 1596–1599.
  22. 22. Guindon S, Gascuel O (2003) A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol 52: 696–704.
  23. 23. Rozas J, Sanchez-DelBarrio JC, Messeguer X, Rozas R (2003) DnaSP, DNA polymorphism analyses by the coalescent and other methods. Bioinformatics 19: 2496–2497.
  24. 24. McLachan GJ (1992) Discriminant analysis and statistical pattern recognition. New York: Wiley. 526 p.
  25. 25. Witten IH, Frank E, Hall MA (2011) Data mining: practical Machine Learning Tools and Techniques. Burlington, MA: Morgan Kaufmann Publishers/Elsevier. 585 p.
  26. 26. Bruguera M, Esteban R, Forns X, Planas R, Quer J, et al. (2012) [Position paper of the Catalan Society of Gastroenterology: treatment of Genotype 1 Chronic Hepatitis C Virus with Triple Therapy]. Gastroenterol Hepatol 35: 667–674.
  27. 27. Martínez-Bauer E, Crespo J, Romero-Gómez M, Moreno-Otero R, Solà R, et al. (2006) Development and validation of two models for early prediction of response to therapy in genotype 1 chronic hepatitis C. Hepatology. 43: 72–80.
  28. 28. Kurosaki M, Tanaka Y, Nishida N, Sakamoto N, Enomoto N, et al. (2011) Pre-treatment prediction of response to pegylated-interferon plus ribavirin for chronic hepatitis C using genetic polymorphism in IL28B and viral factors. J Hepatol 54: 439–448.
  29. 29. Takayama T, Ebinuma H, Tada S, Yamagishi Y, Wakabayashi K, et al. (2011) Prediction of effect of pegylated interferon alpha-2b plus ribavirin combination therapy in patients with chronic hepatitis C infection. PLoS One 6: e27223.
  30. 30. Tsubota A, Shimada N, Yoshizawa K, Furihata T, Agata R, et al. (2012) Contribution of ribavirin transporter gene polymorphism to treatment response in peginterferon plus ribavirin therapy for HCV genotype 1b patients. Liver Int 32: 826–836.
  31. 31. Miki D, Ohishi W, Ochi H, Hayes CN, Abe H, et al. (2012) Serum PAI-1 is a novel predictor for response to pegylated interferon-alpha-2b plus ribavirin therapy in chronic hepatitis C virus infection. J Viral Hepat 19: e126–e133.
  32. 32. Su X, Yee LJ, Im K, Rhodes SL, Tang Y, et al. (2008) Association of single nucleotide polymorphisms in interferon signaling pathway genes and interferon-stimulated genes with the response to interferon therapy for chronic hepatitis C. J Hepatol. 49: 184–191.
  33. 33. Tsukada H, Ochi H, Maekawa T, Abe H, Fujimoto Y, et al. (2009) A polymorphism in MAPKAPK3 affects response to interferon therapy for chronic hepatitis C. Gastroenterology. 136: 1796–1805.
  34. 34. Ge D, Fellay J, Thompson AJ, Simon JS, Shianna KV, et al. (2009) Genetic variation in IL28B predicts hepatitis C treatment-induced viral clearance. Nature 461: 399–401.
  35. 35. Suppiah V, Moldovan M, Ahlenstiel G, Berg T, Weltman M, et al. (2009) IL28B is associated with response to chronic hepatitis C interferon-alpha and ribavirin therapy. Nat Genet 41: 1100–1104.
  36. 36. Tanaka Y, Nishida N, Sugiyama M, Kurosaki M, Matsuura K, et al. (2009) Genome-wide association of IL28B with response to pegylated interferon-alpha and ribavirin therapy for chronic hepatitis C. Nat Genet. 41: 1105–1109.
  37. 37. EASL Clinical Practice Guidelines: management of hepatitis C virus infection (2011) J Hepatol. 55: 245–264.
  38. 38. Berg T, Sarrazin C, Herrmann E, Hinrichsen H, Gerlach T, et al. (2003) Prediction of treatment outcome in patients with chronic hepatitis C: significance of baseline parameters and viral dynamics during therapy. Hepatology 37: 600–609.
  39. 39. Everhart JE, Wright EC (2013) Association of gamma-glutamyltransferase (GGT) activity with treatment and clinical outcomes in chronic hepatitis C (HCV). Hepatology 57: 1725–1733.
  40. 40. Innes HA, Hutchinson SJ, Allen S, Bhattacharyya D, Bramley P, et al. (2012) Ranking predictors of a sustained viral response for patients with chronic hepatitis C treated with pegylated interferon and ribavirin in Scotland. Eur J Gastroenterol Hepatol 24: 646–655.
  41. 41. Weich V, Herrmann E, Chung TL, Sarrazin C, Hinrichsen H, et al. (2011) The determination of GGT is the most reliable predictor of nonresponsiveness to interferon-alpha based therapy in HCV type-1 infection. J Gastroenterol 46: 1427–1436.
  42. 42. Kau A, Vermehren J, Sarrazin C (2008) Treatment predictors of a sustained virologic response in hepatitis B and C. J Hepatol. 49: 634–651.
  43. 43. Diago M, Castellano G, Garcia-Samaniego J, Perez C, Fernandez I, et al. (2006) Association of pretreatment serum interferon gamma inducible protein 10 levels with sustained virological response to peginterferon plus ribavirin therapy in genotype 1 infected patients with chronic hepatitis C. Gut. 55: 374–379.
  44. 44. Askarieh G, Alsio A, Pugnale P, Negro F, Ferrari C, et al. (2010) Systemic and intrahepatic interferon-gamma-inducible protein 10 kDa predicts the first-phase decline in hepatitis C virus RNA and overall viral response to therapy in chronic hepatitis C. Hepatology. 51: 1523–1530.
  45. 45. Fattovich G, Covolo L, Bibert S, Askarieh G, Lagging M, et al. (2011) IL28B polymorphisms, IP-10 and viral load predict virological response to therapy in chronic hepatitis C. Aliment Pharmacol Ther. 33: 1162–1172.
  46. 46. Yeh BI, Han KH, Lee HW, Sohn JH, Ryu WS, et al. (2002) Factors predictive of response to interferon-alpha therapy in hepatitis C virus type 1b infection. J Med Virol 66: 481–487.
  47. 47. Cuevas JM, Torres-Puente M, Jiménez-Hernández N, Bracho MA, García-Robles I, et al. (2008) Refined analysis of genetic variability parameters in hepatitis C virus and the ability to predict antiviral treatment response. J Viral Hepat 15: 578–590.
  48. 48. Farci P (2011) New insights into the HCV quasispecies and compartmentalization. Semin Liver Dis 31: 356–374.
  49. 49. Akuta N, Suzuki F, Sezaki H, Suzuki Y, Hosaka T, et al. (2005) Association of amino acid substitution pattern in core protein of hepatitis C virus genotype 1b high viral load and non-virological response to interferon-ribavirin combination therapy. Intervirology 48: 372–380.
  50. 50. Alestig E, Arnholm B, Eilard A, Lagging M, Nilsson S, et al. (2011) Core mutations, IL28B polymorphisms and response to peginterferon/ribavirin treatment in Swedish patients with hepatitis C virus genotype 1 infection. BMC Infect Dis 11: 124.
  51. 51. Kurosaki M, Sakamoto N, Iwasaki M, Sakamoto M, Suzuki Y, et al. (2011) Sequences in the interferon sensitivity-determining region and core region of hepatitis C virus impact pretreatment prediction of response to PEG-interferon plus ribavirin: data mining analysis. J Med Virol 83: 445–452.
  52. 52. Akuta N, Suzuki F, Hirakawa M, Kawamura Y, Sezaki H, et al. (2012) Amino acid substitution in HCV core/NS5A region and genetic variation near IL28B gene affect treatment efficacy to interferon plus ribavirin combination therapy. Intervirology 55: 231–241.
  53. 53. Umemura T, Joshita S, Yoneda S, Katsuyama Y, Ichijo T, et al. (2011) Serum interleukin (IL)-10 and IL-12 levels and IL28B gene polymorphisms: pretreatment prediction of treatment failure in chronic hepatitis C. Antivir Ther. 16: 1073–1080.
  54. 54. Hashimoto Y, Ochi H, Abe H, Hayashida Y, Tsuge M, et al. (2011) Prediction of response to peginterferon-alfa-2b plus ribavirin therapy in Japanese patients infected with hepatitis C virus genotype 1b. J Med Virol 83: 981–988.
  55. 55. de Lucas S, Bartolome J, Carreno V (2005) Hepatitis C virus core protein down-regulates transcription of interferon-induced antiviral genes. J Infect Dis 191: 93–99.
  56. 56. Blindenbacher A, Duong FH, Hunziker L, Stutvoet ST, Wang X, et al. (2003) Expression of hepatitis c virus proteins inhibits interferon alpha signaling in the liver of transgenic mice. Gastroenterology 124: 1465–1475.
  57. 57. Funaoka Y, Sakamoto N, Suda G, Itsui Y, Nakagawa M, et al. (2011) Analysis of interferon signaling by infectious hepatitis C virus clones with substitutions of core amino acids 70 and 91. J Virol 85: 5986–5994.
  58. 58. Akuta N, Suzuki F, Hirakawa M, Kawamura Y, Yatsuji H, et al. (2010) Amino acid substitution in hepatitis C virus core region and genetic variation near the interleukin 28B gene predict viral response to telaprevir with peginterferon and ribavirin. Hepatology 52: 421–429.
  59. 59. Enomoto N, Sakuma I, Asahina Y, Kurosaki M, Murakami T, et al. (1995) Comparison of full-length sequences of interferon-sensitive and resistant hepatitis C virus 1b. Sensitivity to interferon is conferred by amino acid substitutions in the NS5A region. J Clin Invest 96: 224–230.
  60. 60. Pascu M, Martus P, Höhne M, Wiedenmann B, Hopf U, et al. (2004) Sustained virological response in hepatitis C virus type 1b infected patients is predicted by the number of mutations within the NS5A-ISDR: a meta-analysis focused on geographical differences. Gut 53: 1345–1351.
  61. 61. Suppiah V, Gaudieri S, Armstrong NJ, O’Connor KS, Berg T, et al. (2011) IL28B, HLA-C, and KIR variants additively predict response to therapy in chronic hepatitis C virus infection in a European Cohort: a cross-sectional study. PLoS Med 8: e1001092.
  62. 62. Vidal-Castiñeira JR, López-Vázquez A, Alonso-Arias R, Moro-García MA, Martínez-Camblor P, et al. (2012) A predictive model of treatment outcome in patients with chronic HCV infection using IL28B and PD-1 genotyping. J Hepatol 56: 1230–1238.
  63. 63. Fellay J, Thompson AJ, Ge D, Gumbs CE, Urban TJ, et al. (2010) ITPA gene variants protect against anaemia in patients treated for chronic hepatitis C. Nature. 464: 405–408.
  64. 64. Shiffman ML, Suter F, Bacon BR, Nelson D, Harley H, et al. (2007) Peginterferon alfa-2a and ribavirin for 16 or 24 weeks in HCV genotype 2 or 3. N Engl J Med 357: 124–134.
  65. 65. Ferenci P, Fried MW, Shiffman ML, Smith CI, Marinos G, et al. (2005) Predicting sustained virological responses in chronic hepatitis C patients treated with peginterferon alfa-2a (40 KD)/ribavirin. J Hepatol 43: 425–433.