|Home | About | Journals | Submit | Contact Us | Français|
Patients with chronic hepatitis C virus (HCV) infections are treated with pegylated interferon and ribavirin (PEG-IFN/RBV), which is effective in less than 50% of those infected with HCV genotype 1. Genome-wide association studies have linked response to PEG-IFN/RBV with common single nucleotide polymorphisms in the vicinity of IFN-λ genes on chromosome 19. We investigated the association between the polymorphism rs12979860 and treatment response in a diverse cohort of chronic HCV patients.
A cross-sectional study was performed using data from 1021 consecutive patients enrolled in the Duke Hepatology Clinic Research Database and Biorepository. We analyzed DNA, clinical, and demographic data, along with validated data of the response of 231 subjects to PEG-IFN/RBV. The study included Caucasians (n=178), African Americans (n=53), and HCV genotypes 1 (n=186) and 2/3 (n=45). The rs12979860 genotype was tested for an association with sustained virologic response, defined as undetectable levels of HCV RNA 24 weeks after treatment ended.
The rs12979860 CC genotype (found in ~40% of Caucasians) predicted a sustained virologic response to therapy among Caucasians (odds ratio 5.79; 95% confidence interval 2.67–12.57; p=9.0 × 10-6), independent of HCV genotype and other covariates. Rs12979860 CC predicted a sustained response with 78% specificity and 65% sensitivity in patients infected with HCV genotype 1—better than HCV genotype (currently used to predict treatment response).
rs12979860 genotype is a significant independent predictor of response to PEG-IFN/RBV in patients with chronic HCV infection; tests for this genotype might be used to determine the best course of treatment for patients considering antiviral therapy.
Infection with the Hepatitis C Virus (HCV) is a global health problem, with worldwide estimates of 120 to 130 million carriers1. In the U.S. alone, 3-4 million people are thought to be infected, representing 1.6% of the population2, 3. Chronic HCV infection can lead to progressive liver disease, resulting in cirrhosis and complications including decompensated liver disease and hepatocellular carcinoma leading to the need for transplantation 2, 3. The current standard of care treatment for suitable patients with chronic HCV infection consists of pegylated interferon alpha 2a or 2b (PEG-IFN) given by injection in combination with oral ribavirin (RBV), for 24 or 48 weeks, dependent on HCV genotype. This treatment is not only costly, but associated with significant side-effects resulting in reduced compliance and fewer patients completing treatment. Only about one-half of treated subjects achieve a sustained virologic response (SVR)4-6. For all these reasons, decisions regarding treatment with current standard of care IFN-based therapy are complex and based on the balanced assessment of host and viral determinants that aid in predicting virologic response. HCV genotype, in particular, is used in making treatment decisions: patients with HCV genotype 2/3 have a relatively high rate of SVR (70-80%) with 24 weeks of treatment, while those infected with HCV genotype 1 (representing about 70-75% of infected persons in the U.S.) have a much lower rate of SVR (40%) despite 48 weeks of treatment6.
Recently, several highly correlated common single nucleotide polymorphisms (SNPs) on a linkage disequilibrium block in the vicinity of three interferon lambda genes on chromosome 19, encoding interferon λ1 (IL29), λ2 (IL28A) and λ3 (IL28B), have been implicated in response to PEG-IFN/RBV among patients infected with HCV genotype 1 from three genome-wide association studies7-9. The risk genotype associated with non-response to therapy was more common in African Americans and thought to account for half of the observed ethnic variation in treatment response. Paradoxically, the ‘responder’ genotype was associated with higher viral load.
We sought to confirm and extend our understanding of this recently reported genetic association in a cohort of patients with chronic HCV infection from a large tertiary care setting, including subjects infected with HCV genotypes 2 and 3.
The Duke Hepatology Clinical Research Database and Repository is an ongoing registry of HCV-infected subjects initiated in 1992, representing a large, well-phenotyped collection of North American chronic HCV patients10. All subjects referred to the Duke Liver Clinic with a diagnosis of HCV infection are eligible to be included in this database. Patients are enrolled at the time of their initial clinic visit and informed consent is obtained for the collection and storage of serum, liver tissue and peripheral blood for DNA extraction. HCV status is confirmed by presence of detectable serum HCV RNA. The database includes clinical and demographic data extracted from the medical record, laboratory reports and case-report forms for those patients enrolled in clinical trials.
The degree of liver fibrosis was determined in liver biopsies scored by a panel of expert histopathologists using the METAVIR scoring system (F0-F4) and coded as a three-level variable (Mild – F0, F1; Moderate – F2; Severe – F3, F4) for analysis. HCV genotype was determined by the INNOLIPA HCV assay (Innogenetics, Zwijnaarde, Belgium)11. Treatment history was defined as naïve (no previous treatment for HCV infection) versus prior (one or more previous courses of treatment for HCV infection). For the analysis of response to standard of care PEG-IFN/RBV, subjects with sustained virologic response (SVR) were defined as having undetectable HCV RNA levels 24 weeks after cessation of treatment. Those who did not achieve SVR consisted of patients whose HCV RNA levels remained detectable at the end of treatment (end-of-treatment non-responders), as well as relapsers, who had undetectable levels of HCV RNA at the end of treatment, but detectable HCV RNA levels at 24 weeks after cessation of treatment. Race, an important potential confounder of genetic association studies, was self-defined by participants.
For our study, we identified all subjects from the database who had high quality DNA available for genetic studies (n=1040). Among these, 1021 were successfully genotyped for the rs12979860 polymorphism (98.2% success rate). For analysis of treatment response, we excluded those who were not treated at all as well as those having only received treatment other than PEG-IFN/RBV (n= 611). Among patients treated with PEG-IFN/RBV, we included all who attained a SVR, but to restrict our analysis to biological non-responders, we excluded non-responder subjects who had not received the full course of the per protocol planned treatment, or in whom treatment response information was not available (n=97). We also excluded 48 patients that may have also been included in the IDEAL genetic study that first identified rs12979860 associated with treatment response7. Finally, we excluded 34 subjects who were either co-infected with hepatitis B virus or human immunodeficiency virus-1; who had undergone liver transplantation; were of a race that was mixed, unknown or other than Caucasian or African American; had HCV genotypes that were not 1, 2 or 3; or were missing data on key covariates (see Figure 1).
We limited our analysis of viral load to available measurements taken either on treatment naïve subjects or in treated patients prior to treatment or after completion of treatment for non-responders. These subjects were drawn from the larger dataset and not restricted to those included in the analysis of treatment response; However, overlapping subjects within the IDEAL study were excluded. All serum HCV RNA quantitations were measured using either the National Genetics Institute SuperQuant assay (Culver City, Los Angeles, CA) or the Cobas Taqman HCV Test (Roche Molecular Systems, Pleasanton, California) and classified as low (<600,000 IU/ml) or high (≥600,000 IU/ml) for analysis. Odds ratios were calculated comparing high and low viral load.
This study, the database and repository were approved by the Duke University Institutional Review Board. All patients provided written informed consent and the study was conducted in accordance with provisions of the Declaration of Helsinki and Good Clinical Practice guidelines.
The associated interval reported in the three genome-wide association studies contains several highly correlated SNPs around the IL28B gene. We selected the most strongly associated SNP from the Ge et al. study, rs12979860, located upstream of the gene for genotyping in our cohort using the 5’ nuclease assay with allele specific TaqMan probes12. This SNP was associated with treatment response in both Caucasian and African Americans and is in strong linkage disequilibrium (correlated) with the SNPs reported in the other two studies7. Genotyping was performed in the Duke Institute for Genome Sciences and Policy Genotyping Core and was conducted in a blinded fashion relative to HCV treatment status and other characteristics. Genotyping calls were manually inspected and verified prior to release. Hardy-Weinberg Equilibrium was assessed in Caucasians and African Americans separately.
SAS statistical software, Version 9.1 (SAS Institutes, Cary, NC) was used to carry out logistic regression analysis of rs12979860 genotypes associated with treatment response, and subsequently viral load. Rs12979860 genotypes were coded to test both an additive model, defined as 0, 1 or 2 copies of the C allele, and a recessive model, comparing subjects homozygous (CC) to those with one or no copies (CT/TT) of the C allele. Homogeneity of genotype effects across strata was tested by introducing an interaction term. Analysis of response to PEG-IFN/RBV was carried out in subjects with valid treatment response data, while analysis of rs12979860 associations with HCV genotype and with viral load was carried out in a larger number of patients from the extended cohort. Primary analysis of the association between rs12979860 and treatment response controlled for age, sex, HCV genotype, fibrosis score and treatment history and was stratified by race. To assess the relative contribution of key established predictors of treatment response, a multivariable model was run in the combined population of African Americans and Caucasians controlling for race and the same covariates above, as well as viral load (available on a subset of the population). Likelihood ratios were calculated as: sensitivity/(1-specificity) for LR+ and (1-sensitivity)/specificity for LR-. Overall model fit was assessed using the Hosmer-Lemeshow-Cressie goodness of fit statistic13, the C statistic and Nagelkerke R2 (both bias corrected using the Design library of the R statistical language)14-16. To assess the value of rs12979860 in reclassifying treatment response we computed the Net Reclassification Improvement17, which quantifies reclassification of the outcome when adding a predictor to the model. Multinomial regression was used to model the association of rs12979860 with treatment response as a three-level variable: SVR, relapser and end-of-treatment non-response. Association between rs12979860 genotype and viral load was carried out in a logistic regression model controlling for age, sex, HCV genotype and fibrosis.
A description of the cohort used in our analysis of response to PEG-IFN/RBV is shown in Table 1. Patients were predominantly Caucasian, infected with HCV genotype 1, and broadly representative of a tertiary care chronic HCV cohort. Among patients treated with PEG-IFN/RBV, those included in this analysis differed only slightly from those excluded, having more HCV genotype 2/3-infected subjects (20% vs. 11%; p=0.02). As expected, responders to standard of care PEG-IFN/RBV treatment were characterized as having significantly lower viral load, being younger, of Caucasian race, infected with HCV genotype 2 or 3 and being treatment-naive. Rs12979860 allele frequencies differed by race and were consistent with those previously reported (Caucasian C=.63, T=.37; African American C=.40, T=.60). In the overall cohort, a nominal deviation in Hardy-Weinberg Equilibrium was found for Caucasians (p=0.02), which could not be attributed to genotyping error, but possibly reflects an underlying association between genotype and chronic HCV infection.
Figure 2 illustrates how among Caucasians, rs12979860 genotype was significantly associated with response to PEG-IFN/RBV, with the greatest effect seen when comparing subjects with two copies of the C allele (CC genotype) to those with one or no copies; Therefore, all subsequent analyses considered the CC genotype versus CT/TT. This association remained significant in multivariable analysis controlling for age, sex, HCV genotype, treatment history and fibrosis, where presence of the CC genotype conferred a nearly six-fold increased odds of SVR relative to the CT/TT genotype (odds ratio 5.79; 95% CI 2.67, 12.57; p=9.0 × 10-6). Among Caucasians, there was no difference in the effect of rs12979860 on SVR comparing subjects with HCV genotype 1 and 2/3, or between those with a prior treatment history and those treatment naïve at baseline (rs12979860 genotype by trait interactions p>0.05).
African Americans comprised only 23% of our cohort and were predominantly infected with genotype 1 HCV (98%) and 57% were treatment naïve. As illustrated in Figure 2, no association was found between rs12979860 genotype and treatment response in African Americans in our study, nor did the difference in the rs12979860-treatment response association between African Americans and Caucasians reach significance (rs12979860 by race interaction p=0.34). However, it should be noted that the estimated proportion of SVR among patients with the CC genotype, 0.125 (1/8), is unstable due to small sample size, resulting in a wide confidence interval (0.003, 0.53) and thus making it difficult to draw inference.
To assess the relative contribution of important covariates to treatment response, we examined the effect of each covariate in a multivariable model. Table 2 shows the relative effect of all covariates on treatment response in univariable and multivariable models. HCV genotype was the strongest predictor of treatment response in univariable and multivariable analysis that did not include the rs12979860 genotype. Inclusion of rs12979860 in the model attenuated the effects of HCV genotype and enhanced the association with viral load. Rs12979860 genotype emerged as the strongest factor associated with treatment response; other independently associated factors included HCV genotype, viral load, age and treatment history. Inclusion of rs12979860 genotype in the model improved the C statistic (from 0.75 to 0.82) as well as the Nagelkerke R2 (from 0.25 to 0.38). In the current study, inclusion of IL28 genotype in the model resulted in reclassification of responders (30%) and non-responders (56%), for an overall Net Reclassification Improvement (0.86) that was highly significant (p= 2.3 × 10-7).
To assess the relative value of rs12979860 compared to HCV genotype, we calculated sensitivity, specificity and likelihood ratios in our cohort of Caucasian subjects (Table 3). Comparing rs12979860 CC genotype to HCV genotype 2/3 revealed HCV genotype as having somewhat better specificity, but rs12979860 genotype has better sensitivity and is more common, found in 38% of chronic HCV patients (compared to ~25% prevalence of HCV genotypes 2/3). Performance of rs12979860 among the subset of HCV genotype 1 subjects, or among the subset of treatment-naïve subjects, mirrors that in all subjects with HCV genotype 1, 2 and 3 combined.
Among 111 (62%) Caucasian subjects who failed to achieve SVR to PEG-IFN/RBV, 29 (26%) were classified as relapsers. The frequency of the rs12979860 CC genotype in this group was in between that observed for the SVR and end of treatment non-responder groups. However, only the difference found between relapsers and end of treatment non-responders was significant, suggesting that rs12979860 CC genotype influences end-of-treatment response and does not accurately distinguish between patients with a relapse response and those with a sustained virologic response (Figure 3).
HCV genotype data were available on a larger group of patients than were used for analysis of treatment response (n=649 Caucasians). We compared the rs12979860 genotype frequencies between Caucasians infected with HCV genotypes 1, 2 and 3 (Figure 4). Rs12979860 CC was most common in genotype 3 patients (55%), followed by genotype 2 (46%) and genotype 1 (33.5%; p=0.0007). To correct for possible referral bias, we examined this association in the subset of naïve patients undergoing treatment with PEG-IFN/RBV and observed the same trend (data not show).
Off-treatment viral load measures were available on 301 Caucasians with chronic HCV infection from our cohort, 277 of whom had complete data on covariates as well and could be used in our analysis to determine the association between rs12979860 genotype and viral load. The odds of having a higher quantitative viral load in patients with rs12979860 CC genotype were twice that of patients with genotype CT/TT in analysis controlling for age, sex, HCV genotype, and fibrosis (OR=2.13; 95% CI 1.24, 3.66; p=0.0061).
Treatment decisions in patients with chronic hepatitis C infection are currently based on clinical, demographic and virologic characteristics of infected patients. While these may be helpful from a population point of view, for any individual patient and provider, these baseline pretreatment characteristics are inaccurate in predicting treatment response in hepatitis C patients infected with genotype 1, the most common genotype found in the United States. We have confirmed in this study that the recently identified common genetic variant in the interferon lambda gene region is strongly associated with response to standard of care PEG-IFN/RBV in a tertiary care setting. The association among Caucasians was thus validated outside of the clinical trial setting. Because of the limited sample size of African Americans and other racial groups in this current study, we had inadequate power to detect statistical association in these groups, as was observed by Ge et al.7. Nonetheless, inclusion of rs12979860 genotype and race in a multivariable model caused an attenuation of the effect of race on treatment response, indicating that the effect of race may be mediated in part through this polymorphism.
Our results extend the observations of Ge et al. and indicate that, among Caucasians, rs12979860 has a high specificity, similar to HCV genotype, the current best baseline indicator used for identifying patients who will achieve a SVR to anti-viral therapy. Used in conjunction with HCV genotype, rs12979860 may provide additional discriminatory power to identify likely responders to treatment.
Another interesting and previously unreported observation was the significantly different frequency of the rs12979860 genotype according to HCV genotype in Caucasian patients. HCV genotype 1 infected patients were less likely to have the rs12979860 ‘response’ genotype than were patients infected with HCV genotypes 2 or 3. While interesting, this relationship does not fully explain why patients infected with HCV genotypes 2 or 3 have significantly higher sustained response rates compared to those infected with HCV genotype 1; both the gene variant and HCV genotype remain independently associated with treatment response in the multivariable models we ran. Nonetheless, this observation could theoretically have some biological bearing on development of persistent or chronic (versus spontaneously cleared) HCV infection. The data could be interpreted in the context of a case-only study design, where an association between two factors (rs12979860 and HCV genotype) among cases of chronic HCV may reflect interaction between rs12979860 and HCV genotype in the development of chronic HCV 18, 19. Based on these results, we would hypothesize that subjects with both HCV genotype 2/3 and rs12979860 CC would be more likely to develop chronic HCV and hence, less likely to spontaneously clear the virus than those with HCV genotype 1 and rs12979860 CT/TT. While this relationship may seem counterintuitive on account of HCV genotype 2/3 subjects having a significantly higher rate of treatment-induced viral clearance, it warrants further exploration in appropriate prospective studies of spontaneous clearance. We also considered that the association could have been due to referral bias, whereby chronic HCV patients who are difficult to treat may be over-represented among the Duke cohort, and these patients would be more likely to be HCV genotype 1 and rs12979860 CT/TT genotype. This bias could lead to an overestimate of the association between rs12979860 genotype and HCV genotype. However, when we analyzed a subset of treatment eligible patients with no prior treatment history we found a similar trend of association.
Although the rs12979860 SNP showed a remarkably strong association with treatment response in both the study by Ge et al.7 and our replication study, the exact causal variant underlying the observed genetic association has not been identified. This SNP lies closest to two genes encoding for interferon lambda proteins, λ2 and λ3. Interferon lambdas have previously been studied in the context of HCV infection and shown to suppress HCV replication in vitro20-22. It is possible that the causal variant affects expression or function of one of these anti-viral genes, which may in turn affect viral control.
Interestingly, Ge et al.7 reported a paradoxical relationship to viral load in that the ‘responder’ allele of rs12979860 was statistically associated with a higher baseline viral load, inconsistent with the clinical observation that higher baseline viral load is typically associated with a poorer treatment response4-6. We also confirmed this paradoxical association with off-treatment viral load in our expanded cohort of Caucasian patients from a tertiary care setting. The exact mechanism underlying this genetic association with high viral load and yet increased likelihood of treatment response remains to be determined.
There are several important clinical implications of this genetic association, originally reported by Ge et al 7 in a clinical trial population and confirmed in the present study in a tertiary care population. First, the rs12979860 SNP may be useful in determining which chronic HCV genotype 1 patients would be most likely to respond to treatment with PEG-IFN/RBV. Given the high prevalence of the rs12979860 ‘responder genotype’ in Caucasians (~40%), this could have a substantial impact on treatment decision-making. Our data support the potential utility of rs12979860 in predicting SVR in Caucasians, regardless of treatment history. However, additional prospective studies are needed to determine the true predictive value of this marker among all treatment-eligible patients including those from other racial/ethnic and HCV genotype groups, and taking into account non-compliant and relapsing patients as well. Ultimately, chronic HCV patients with the rs12979860 responder genotype may be more motivated to comply with treatment, or to undergo treatment in the presence of mild underlying liver disease, knowing that they have a higher likelihood of SVR.
These important genetic findings related to treatment response in patients with chronic hepatitis C infection suggest the possibility of personalized medicine for the treatment of this disease. Clinical trials are now necessary to determine whether HCV genotype 1 infected patients with the favorable rs12979860 responder genotype would benefit equally from shorter treatment duration with our current therapies or future therapies, thus reducing the cost and side effects associated with longer term treatment. How the rs12979860 responder genotype influences the outcomes to future anti-viral strategies, including those based upon protease and polymerase inhibition requires immediate attention and investigation. Understanding the clinical implications of this genetic, biologically plausible finding will be a major research agenda and priority.
We would like to thank all of the study participants who contributed their biospecimens and data to the Duke Hepatology Clinic Database and Biorepository and acknowledge Diane Uzarski, Crystal Cates, Chris Delionbach and Melissa Austin for continued maintenance of this valuable resource.
Grant Support : This study was funded in part by a generous grant from the David H Murdock Institute for Business and Culture via the M.U.R.D.O.C.K. Study and award 1 UL1 RR024128-01 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH) and NIH Roadmap for Medical Research, and its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH. Dr. Thompson received funding support from the Duke Clinical Research Institute, a generous research gift from the Richard B. Boebel Family Fund, the National Health and Medical Research Council of Australia and the Gastroenterology Society of Australia.
Author contribution: J.H.L., S.S., X.Q.L. and J.J.M. performed statistical analysis; A.T., J.J.M. and J.G.M. designed the study and wrote and edited the manuscript; H.T., A.M. and K.P. consulted on manuscript preparation. J.J.M. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors reviewed the final manuscript prior to submission.
Disclosures: Drs. Thompson and McHutchison are co-inventors with Schering Plough on a patent application on the original finding of rs12979860 association with PEGIFN/RBV treatment response in HCV infection. Drs. McHutchison and Muir have received research funding and acted in an advisory capacity for Schering Plough.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.