Polymorphism in the Human Major Histocompatibility Complex and Early Viral Decline during Treatment of Chronic Hepatitis C

The dynamics of the viral decline immediately after the startof therapy for chronic hepatitis C virus (HCV) infection mayhave prognostic potential for ultimate sustained virologic response.Considerable interindividual variability in the decline hasbeen reported, including differences by race. The human majorhistocompatability complex (MHC) genes encode the human leukocyteantigens, which are important in the immune response to viralinfections. We examined whether carriage of specific human MHCalleles are associated with the rate of the early viral decline.Longitudinal viral level data (baseline and days 1, 2, 7, 14,and 28 of treatment), medium resolution MHC genotyping, andrandom coefficients models were used to examine associationsbetween MHC class I and class II allele carriage and the dynamicsof the viral decline in 180 African-Americans (AAs) and 194Caucasian Americans (CAs) with genotype-1 HCV infection overthe first 28 days of treatment with peginterferon ${alpha}$ 2a plus ribavirin.Baseline viral levels were similar by race, irrespective ofallele carriage. However, the rate of change in the viral declinewas associated with both allele and race. Among the four subgroupsdefined by race and specific allele, the fastest rates of declinewere observed (in terms of estimated mean viral declines log₁₀IU/ml during the first four weeks) in CA noncarriers for A*03(2.75; P = 0.018), in CA carriers for Cw*03 (2.99; P = 0.046),and in CA noncarriers for DQA1*04 (2.66; P = 0.018) or DQB1*0402(2.65; P = 0.018). MHC alleles are associated with the viraldecline during the first 28 days of peginterferon therapy.

INTRODUCTION

Top

INTRODUCTION

Hepatitis C virus (HCV) infection affects an estimated 170 millionindividuals worldwide and 4 to 5 million in the United Statesand may result in cirrhosis of the liver and primary hepatocellularcarcinoma (2). Currently, therapy for chronic HCV infectionconsists of a combination pegylated alpha interferon in combinationwith the nucleoside analogue, ribavirin, and is effective ineliminating virus from ca. 50 to 80% of those treated (8, 20).Significant racial disparities also exist, with African-Americans(AAs) having significantly lower response rates than those ofCaucasian Americans (CAs) (5, 12, 13).

Early disappearance of viremia during the first 4 weeks afterthe start of treatment is believed to be highly predictive ofthe ultimate sustained response to antiviral therapy. In a recentstudy, 91% of individuals who were PCR negative for HCV-RNAby week 4 achieved sustained virologic response (SVR), whereasonly 60% of patients who had a <2-log₁₀ drop in HCV levelsby week 4 but were PCR negative for HCV by week 12 achievedSVR (7). The importance of the early viral response is furtherunderscored in studies of the kinetics of viral decline, wheremultiple and very frequent measurements of HCV levels are obtainedduring the initial hours after the start of treatment (9, 19).Rapid rates of viral decline are associated with SVR (undetectablelevels of HCV 6 months after discontinuation of therapy). Accordingly,viral dynamics may serve as a potential predictor of SVR (18,19). Racial differences have also been reported with respectto viral dynamics, with AAs exhibiting slower viral dynamicdeclines than CAs (16, 17). The mechanisms behind the differentdynamics of the viral decline are not known, but host geneticvariation may contribute.

Spontaneous clearance of HCV is associated with vigorous androbust CD4⁺ and CD8⁺ T cells (3), and several studies have indicatedan important role for T cells in antiviral therapy-induced clearance(6, 26, 28), although other studies have failed to find an association(14). The genes of the human major histocompatibility complex(MHC) encode the highly polmorphic human leukocyte antigens(HLA), which are crucial in antigen presentation and mediationof the T-lymphocyte response. Studies of the potential roleof host genetic diversity and the dynamics of the early viraldecline have been hampered by the difficulty in obtaining sequentialHCV viral levels in a large cohort of individuals. We took advantageof the availability of repeated viral level data from a clinicalstudy of AA and CA patients undergoing anti-HCV therapy andexamined whether carriage of specific MHC alleles is associatedwith the dynamics of the HCV viral decline during the first28 days of peginterferon therapy for chronic HCV and whetherhost genetic factors are associated with differences in viraldynamics by race.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Study population. This study utilized participants from the Study of Viral Resistanceto Antiviral Therapy of Chronic Hepatitis C (Virahep-C), a multicenterstudy sponsored by the National Institutes of Health aimed atunderstanding the mechanisms of resistance to antiviral therapyfor chronic HCV infection among interferon treatment-naive individualsinfected with genotype 1 HCV (1a and 1b), as well as the differencesin outcome by race among AAs and CAs (5). All subjects wereborn in the United States, and race was determined by a self-administeredquestionnaire.

MHC genotyping. Genotyping of class I MHC A, B, and C loci and the class IIloci, DRB1, DQA1, DQB1, DPB1, and DPA1, were conducted by usingmedium resolution HLA typing methods with biotin-labeled PCRprimers and hybridization of the labeled amplicons to sequence-specificoligonucleotide probes immobilized on nylon membranes developedby Roche Molecular Systems. All typing of MHC genes were conductedby Roche Molecular Systems (Alameda, CA).

HCV viral assessment. Viral levels were assessed for all participants in the studyat baseline, as well as day 1, 2, 7, 14, and 28 after the initiationof treatment. Although more sensitive assays have been subsequentlydeveloped, the Roche HCV Amplicor version 2 assay, which hasa lower detection limit of 600 copies/ml, was used in the presentstudy.

Evaluation of population structure. Using data from 161 ancestry-informative single nucleotide polymorphisms,we derived estimates of individual admixture for participantsin the genetics study and utilized the structured associationmethod developed by Pritchard et al. (23, 24) to evaluate thepopulation structure. We observed two distinct ancestral groupsthat had a strong correlation with self-reported race (30).Consequently, we used self-reported race in these analyses.

Statistical analysis. Carriage of MHC alleles that occurred with a frequency of $≥$ 5%in either racial group was assessed for the potential effecton the change in viral decline during at baseline and days 1,2, 7, 14, and 28 after the start of treatment (4, 27).

All viral level measurements were log₁₀ transformed to approximatea normal distribution. Due to the fact that the Roche Amplicorassay had a lower limit of detection of 600 IU/ml, we imputedthe viral level values for those below detectable using a uniformdistribution in the log scale. The time variable was expressedas "weeks" (rather than days or hours) in our analyses. A square-roottransformation was applied to the time variable to attain linearity.

Longitudinal analysis was conducted using a multilevel modelfor change (also known as a linear random coefficients modelor linear growth model) (4, 27). Detailed mathematical representationsof the model are available online in Table S1 in the supplementalmaterial, along with the variance components (see Table S2 inthe supplemental material). Briefly, a two-level modeling strategywas utilized; a level 1 submodel describes how each person changesover time, and a level 2 submodel relates "interindividual differencesin change" to predictors such as allele carriage and race. Amultilevel modeling was chosen because the primary questionof interest was to determine whether allele and/or race hadan effect on the rate of change in viral levels over time (thedynamics of viral decline). We also adjusted for potential confoundingfactors such as patient age, sex, fibrosis score (Ishak >3 [versus $≤$ 3]), and weight, which we have previously reportedto be associated with the interferon response (5). Since thesevariables did not significantly change our final qualitativeconclusions, we omitted them from the final models for the simplicityand focused on the effects of allele and race only. By fittingeach "patient" as a random effect, we accounted for the interindividualvariability of viral levels at the baseline (i.e., to allowintercepts to vary between patients) "patient-by-time interaction"as a random effect to account for the interindividual variabilityin rates of change of viral levels (i.e., to allow slopes ofthe decline to vary between patients) during the course of thefirst 28 days of therapy. The variables "time," "allele," and"race" were included as fixed effects. To evaluate the differentialeffects of time on HCV RNA by race and allele (i.e., statisticalinteractions), we also included "race-by-time," "allele-by-time,""race-by-allele," and "allele-by-race-by-time interaction" asfixed effects.

The SAS v.9.1 system (Proc Mixed; SAS Institute, Cary, NC) wasused for data analyses, and the R v.2.3.1 statistical packagewas used for graphical presentation of data (R Language andEnvironment for Statistical Computing [www.r-project.org]).Statistical significance was set at ${alpha}$ = 0.05.

RESULTS

Top

RESULTS

Cohort characteristics. Of the 401 individuals who were treated in the Virahep-C study(5), 373 consented to participate in host genetics studies (30).Table 1 summarizes the baseline characteristics of the participantsin the present study. The distribution of viral levels at baseline,and days 1, 2, 7, 14, and 28 pre- and posttransformation arepresented in Fig. 1.

View this table:
[in this window]
[in a new window]

TABLE 1. Baseline characteristics of the 373 participants with MHC genotyping data in the Virahep-C genetics study categorized by race

View larger version (11K):
[in this window]
[in a new window]

FIG. 1. Average viral levels (based on imputed and observed viral levels) in the cohort overall before and after transformation. (A) Viral levels before transformation of the time variable (time is expressed as days); (B) viral levels after transformation of time (time is expressed as weeks) utilizing a square-root transformation. Dark circles indicate the mean viral levels (in log₁₀ IU/ml) at each time point. Vertical bars depict the 95% confidence intervals (mean ± 1.96 [standard error of the mean]). Solid lines represent the estimated linear relationship based on general linear models. A slightly better linear relationship can be seen between mean viral level and time in panel B than in panel A.

Twenty-four patients required an alteration in the dose of interferondue to side effects. We conducted our analyses with these individualsincluded, as well as with them excluded, and observed no significantdifferences in the results. Therefore, we report data with these24 patients included.

MHC allele carriage and the dynamics of viral decline. We observed a total of 47 alleles in our cohort with a frequencyof $≥$ 5%. These alleles, along with their corresponding frequencies,are presented in Table 2. Of the alleles tested, four were significantlyassociated with the dynamics of the viral decline (P < 0.05).The models and results presented in Table 3 evaluate the effectsof allele carriage on the initial baseline viral level, as wellas on the rates of change in viral levels, over time, controllingfor the effects of race.

View this table:
[in this window]
[in a new window]

TABLE 2. Distribution of MHC class I and class II alleles in the Virahep-C cohort^a

View this table:
[in this window]
[in a new window]

TABLE 3. Summary of multilevel models^a

Based on the two-level model including DQA1*04 allele, the estimateddifference between initial viral levels for the noncarriersand carriers of this allele was (–0.08 log₁₀ IU/ml; P= 0.574; Table 3) and AAs had slightly lower baseline virallevel than for the CAs (by –0.03 log₁₀ IU/ml; P = 0.756;Table 3). Nonetheless, these differences were not statisticallysignificant. This model also estimated the population initialmean viral level to be 6.27 log₁0 IU/ml with a standard errorof 0.146. The results were similar for all four models in Table3, in that allele carriage did not have a significant impacton baseline viral levels between carriers and noncarriers, eitherindependently or when adjusted for race.

However, we did observe differences in the dynamics of the viraldecline (Table 3). For DQA1*04 allele carriage, the rate ofviral decline was higher among noncarriers than carriers (P< 0.001) and the magnitude of the rate of decline dependedon race (Table 3 and Fig. 2A). The estimated mean viral leveldeclines for the first 4 weeks were 2.66 for CA noncarriers,1.51 for AA noncarriers, 1.19 for AA carriers, and 0.98 forCA carriers (log₁₀ IU/ml P = 0.018, Fig. 2A), suggesting thatCA noncarriers exhibited the fastest decline. This trend wassimilar when we examined the estimated mean viral level declineduring the first week (Fig. 2A).

View larger version (22K):
[in this window]
[in a new window]

FIG. 2. The predicted mean decline of HCV viral levels by allele carriage and race. The vertical axis presents log₁₀-transformed predicted mean viral levels from the random coefficients model, while the horizontal axis presents time. The time variable was expressed as days. In addition, predicted baseline viral levels along with the rates of viral change, are provided below each figure by race and the carriage of each of the alleles over time (with a ${surd}$ week transformation). (A) DQA1*04; (B) DQB1*0402; (C) A*03; (D) Cw*03.

The effect of DQB1*0402 was very similar to that of DQA1*04;CA noncarriers exhibited the fastest decline (Fig. 2.B) andhad higher rates of viral decline compared to carriers. Theestimated mean viral level declines during the first 4 weekswere: 2.65 for CA noncarriers, 1.49 for AA noncarriers, 0.94for CA carriers, and 1.21 for AA carriers (log₁₀ IU/ml, P =0.018).

CA noncarriers of A*03 had the highest rates of viral declinecompared to CA carriers; the estimated mean viral level declinesfor the first 4 weeks were 2.10 for CA carriers, 2.75 for CAnoncarriers, 1.65 for AA carriers, and 1.39 for AA noncarriers(log₁₀ IU/ml, P = 0.018) (Fig. 2C).

CA carriers of Cw*03 had the largest rates of viral declinecompared to CA noncarriers, while AA noncarriers had higherrates of viral decline than did the AA carriers. The estimatedmean viral level declines for the first 4 weeks were 2.99 forCA carriers, 2.49 for CA noncarriers, 1.51 for AA noncarriers,and 1.19 for AA carriers (log₁₀ IU/ml, Fig. 2D). For all fourof these alleles, the viral subgenotype (1a or 1b) did not havean impact on outcome (P > 0.05).

DISCUSSION

Top

DISCUSSION

This study examines the effects of genetic diversity in thehuman MHC on the dynamics of viral decline in subjects treatedwith peginterferon for chronic HCV. We utilized data from alarge, well-characterized cohort of AAs and CAs with genotype1 HCV infections who underwent a standard treatment protocol.We genotyped the A, B, and C class I loci, as well as the DRB1,DQA1, DQB1, DPA1, and DPB1 class II loci, and observed significantassociations of two class I alleles and two class II alleleswith the dynamics of the viral decline.

We also observed that the carriage of DQA1*04 or DQB1*0402 alleleswas associated with slower rates of viral decline. Interestingly,the frequencies of these alleles are greater among AAs thanamong CAs (Table 2). In addition, carriage of each of thesealleles is associated with a greater difference in the levelof viral decline among CAs than among AAs. The frequency ofDQA1*04 is 10.1% among AAs and 2.8% among CAs; the frequencyof DQB1*0402 is 8.1% among AAs and 2.8% among CAs. Differentialdistribution of the frequencies of these alleles by race maycontribute to the observed population differences in responseto peginterferon by race on the population level.

The DQA1*04 and DQB1*04 alleles have been previously implicatedas a risk for Crohn's disease (31) and cirrhosis (10) in chronicHCV infections. The DQB1*0401/*0402 alleles are associated withviral persistence in several Japanese studies (1, 10, 15, 23).Previous studies that have examined the structure of DQB1*04have shown that a leucine substitution exists at position 56(21), which is located in the antigen-binding groove of theDQ molecule and probably contributes to the ability to presentHCV peptides. The leucine substitution is present only in theDQB1*0401 and DQB1*0402 alleles. All other DQB1 alleles havea proline at this position. Future studies are needed to specificallyexamine HCV antigen presentation in the context of these twoclass II MHC alleles.

In addition, the DQB1*0402 and DQA1*04 alleles exist as a haplotypeand, therefore, our findings of similar magnitude and directionof effects for the individual alleles are not surprising. Becausethe DQA1*04 and DQB1*0402 alleles often exist on a haplotypewith DRB1*03 or DRB1*08, we examined whether carriage of theDRB1*08-DQA1*04-DQB1*0402 or DRB1*03-DQA1*04-DQB1*0402 haplotypeswas associated with viral dynamics. However, due to the relativelylow frequency of individuals with DRB1*08-DQA1*04-DQB1*0402(4 AA and 10 CA patients) and DRB1*03-DQA1*04-DQB1*0402 (24AA and 0 CA patients) in the cohort overall and the extremedistributions within each race, we were not able to obtain meaningfulresults with our longitudinal analyses (data not shown) and,therefore, are not able to exclude the possibility that anotherallele in linkage disequilibrium with the DQA1*04 or DQB1*0402may negatively affect viral decline during therapy. Extendedhaplotypes involving DQB1*0401/*0402 have also been associatedwith greater HCV-induced liver injury (15), although the immediatefunctional relevance with respect to viral dynamics is not clear.

CA noncarriers of the class I allele A*03 had greater ratesof viral decline than CA carriers. In contrast, AA noncarriershad slower rates of viral decline than AA carriers (Fig. 2C),unlike the effect within CAs. The effects of allele carriageappear to be greater among CA noncarriers and carriers thanamong AA noncarriers and carriers. Overall, the frequency ofthe A*03 allele is slightly lower among AAs than CAs (11.2%among AAs and 14.2% among CAs). It is possible that the maineffects of this allele are seen among CAs, and that the smallerdifferences in viral dynamics among AA noncarriers and carriersis due to other undetermined genetic and/or nongenetic factors.It is also possible that AA and CA haplotype differences maycontribute to the observed differences. Alternatively, it ispossible that our observations are due to chance. Larger studiesare needed to evaluate the potential effects of extended haplotypeswithin each race.

Carriers of Cw*03 had more pronounced viral dynamic declinesthan CA noncarriers (Fig. 2D). Viral declines among AA carriersand noncarriers were similar. Class I alleles are importantin CD8⁺ T cells responses, which have been shown to play a rolein the clearance of HCV during therapy (28). Future studiesare needed to understand the role of Cw*03 in HCV antigen presentation.Overall, the frequency of the Cw*03 allele is similar amongAAs and CAs (10.9% among AAs and 10.3% among CAs).

In summary, our observations suggest that carriage of DQA1*04and DQB1*0402, or another allele in linkage disequilibrium withthese alleles, may represent a risk for slower rates of viraldecline during peginterferon therapy. With both DQA1*04 andDQB1*0402, CA noncarriers had better rates of viral declineduring the course of therapy compared to CA carriers, AA noncarriers,and AA carriers, respectively (Fig. 2C and D).

The 28-day viral response represents a critical biological pointfor the study of the response to therapy due to its prognosticpotential (16, 17, 19). Unlike SVR, early viral responses areless likely to be affected by dose adjustments and patient complianceand may therefore represent a more purely biologic picture ofthe role of host genetics in the interferon response. It islikely that in a complex process such as the response to peginterferontreatment for chronic HCV infection, multiple genes are involved,with different genes possibly operating at different time pointsduring the course of therapy. Therefore, it is possible thatthe alleles that we describe here play an important role, eitherindependently, as a haplotype, or in combination with otherunidentified genetic markers during the early period after thestart of therapy, while other markers may exert a more influentialrole at subsequent time points during the course of therapy.In the future, larger studies are needed to refine the potentialinfluence of extended MHC haplotypes on HCV viral dynamics duringtreatment.

Previous mathematical models using differential equations havesuggested that the shape of the viral decline with standardalpha interferon or peginterferon is nonlinear. For example,biphasic and triphasic declines have been reported (9, 22).However, these models are based on very frequent sampling ofviral levels spaced only hours apart, as opposed to days apartas with our study, meaning that the frequency and spacing ofviral level measurements are critical to the modeling of viralkinetics. In the absence of intensive sampling of viral levels,we used a square root transformation of the time scale to linearizethe trajectory of the viral decline for our statistical modeling.This approach allowed us to focus our analysis on the linearrelationship of MHC allele carriage on the overall dynamicsof the viral decline, rather than the smaller fluctuations inthe viral decline after each administration of interferon. Consequently,these linear individual change trajectories were modeled, andthe effects of allele and race on the rate of change were examined.This analysis allowed a more simplified summary measure of therelationship between MHC allele carriage and the trajectoriesof viral levels over time.

A key factor in genetic association studies is the potentialfor confounding from population stratification. We have previouslyutilized recently developed techniques in mathematical geneticsto infer individual admixture and found self-reported race tobe highly correlated with individual admixture. Consequently,we have utilized self-reported race in our study (30). In addition,due to the hypothesis-generating nature of the present study,we have deliberately elected to not adjust for the number ofcomparisons made rather than reject an association solely basedon statistical grounds (29). Although it is possible that theassociations that we have observed are false positives, theconsistent patterns within each race provide additional evidenceof potential biological effects (Fig. 2).

Our study offers an important first picture of the role of diversityin the human MHC on the dynamics of viral decline during thefirst 28 days of therapy. Future studies are needed to confirmour observations, as well as expand our understanding of thebiological mechanisms of these alleles in the context of HCVtherapy. We have previously reported that the A*02, B*58, andDPB1*1701 alleles are associated with the SVR to alpha peginterferon-plus-ribavirintherapy (25). The response to therapy is likely a complex processthat occurs over the course of time. Most likely, differentcombinations of genes are involved at different time pointsto clear viremia. For example, studies of Caucasian populationshave consistently shown that the class II alleles DRB1*11 andDQB1*0301 are associated with natural clearance of HCV viremia(11, 29). However, a clear understanding of the biological mechanismsof these alleles is still lacking. Additional research, particularlywith understanding of the differences in the presentation ofHCV peptides by different HLA alleles, is needed to better understandthe differences that we have observed between early viral dynamicsand SVR. Additional nongenetic factors, such as decreased immunefunction through coinfections, may also play a role. Furtherresearch is needed to help evaluate these factors, as well asthe functional characteristics of these alleles on HCV elimination.

ACKNOWLEDGMENTS

This clinical study was a cooperative agreement funded by theNIDDK and cofunded by the National Center on Minority Healthand Health Disparities, with a Cooperative Research and DevelopmentAgreement with Roche Laboratories, Inc., under grants U01 DK60329,U01 DK60340, U01 DK60324, U01 DK60344, U01 DK60327, U01 DK60335,U01 DK60352, U01 DK60342, U01 DK60345, U01 DK60309, U01 DK60346,U01 DK60349, and U01 DK60341. Other support was provided bythe National Center for Research Resources, the Intramural ResearchProgram of the National Institutes of Health, the National CancerInstitute, the Center for Cancer Research, and General ClinicalResearch Centers program grants M01 RR00645 (New York Presbyterian),M02 RR000079 (University of California, San Francisco), M01RR16500 (University of Maryland), M01 RR000042 (University ofMichigan), and M01 RR00046 (University of North Carolina). Additionalsupport for L.J.Y. was provided by a National Institutes ofHealth Clinical Research Career Development Award Grant (1KL2RR024154-02).

The members of Virahep-C included Nezam Afdhal and Tiffany Geahiganfrom the Beth Israel Deaconess Medical Center, Boston, MA; RobertS. Brown, Jr., Lorna Dove, and Shana Stovel from the New York-PresbyterianMedical Center, New York, NY; Norah Terrault, Stephanie Straley,Eliana Agudelo, and Melissa Hinds from the University of California,San Francisco, CA; Thelma E. Wiley and Monique Williams fromRush University, Chicago, IL; Charles D. Howell and Karen Callisonfrom the University of Maryland, Baltimore, MD; Lennox J. Jeffers,Shvawn McPherson Baker, Maria DeMedina, and Carol Hermitt fromthe University of Miami, Miami; Hari S. Conjeevaram, RobertJ. Fontana, and Donna Harsh from the University of Michigan,Ann Arbor, MI; Michael W. Fried, Scott R. Smith, Dickens Theodore,Steven Zacks, Roshan Shrestha, Karen Dougherty, Paris Davis,and Shirley Brown from the University of North Carolina, ChapelHill, NC; John E. Tavis, Adrian Di Bisceglie, Ermei Yao, MaureenDonlin, Nathan Cannon, and Ping Wang from St. Louis University,St. Louis, MO; Huiying Yang, George Tang, and Dai Wang fromCedars-Sinai Medical Center, Los Angeles, CA; Hugo R. Rosen,James R. Burton, and Jared Klarquist from the University ofColorado Health Sciences Center, Denver; Scott Weston from theVeteran's Administration, Portland, OR; Milton W. Taylor, CorneliuSanda, Joel Schaley, and Mary Ferris from Indiana University,BloomingtonN; Steven H. Belle, Richard A. Bilonick, GeoffreyBlock, Jennifer Cline, Marika Haritos, KyungAh Im, StephanieKelley, Sherry Kelsey, Laurie Koozer, Sharon Lawlor, StephenB. Thomas, Abdus Wahed, Yuling Wei, and Leland J. Yee from theData Coordinating Center, Graduate School of Public Health atthe University of Pittsburgh, Pittsburgh; Patricia Robuck, JamesEverhart, Jay H. Hoofnagle, Edward Doo, T. Jake Liang, and LeonardB. Seeff from the National Institute of Diabetes and Digestiveand Kidney Diseases; and David E. Kleiner from the NationalCancer Institute, Bethesda, Maryland.

FOOTNOTES

* Corresponding author. Mailing address: Department of Epidemiology, University of Pittsburgh, A511 Crabtree Hall, 130 DeSoto Street, Pittsburgh, PA 15261. Phone: (412) 624-5326. Fax: (412) 624-7397. E-mail: YeeL{at}edc.pitt.edu

Published ahead of print on 13 October 2008.

Supplemental material for this article may be found at http://aac.asm.org/.

Present address: Department of Discovery, Medicine, and Epidemiology,AstraZeneca Pharmaceuticals LP, Wilmington, Delaware.

Contributing members of the Virahep-C Study are listed in theAcknowledgments.

REFERENCES

Top