To the Editor:

We read with interest the recent article by Akuta and colleagues,1 who reported the association between interleukin-28B (IL-28B) polymorphism and the treatment response in a cohort of Japanese patients treated with telaprevir and pegylated interferon-alfa (peg-IFN)/ribavirin (RBV) combination therapy. The good-response genotype (C/C rs12979860 or T/T rs8099917) was strongly associated with an increased rate of sustained virological response despite the addition of a directly acting antiviral agent. This suggests that patient IL-28B genotype will remain relevant in the dawning era of specifically targeted antiviral therapy for hepatitis C virus because a combination with peg-IFN and RBV is required to restrict the development of antiviral resistance. It will, therefore, be important to consider IL-28B genotype in clinical development programs; because of the population frequency of the good-response IL-28B variant and its association with rapid viral decline during peg-IFN therapy,2 it is possible for small early-phase efficacy trials to be confounded by an imbalance in the IL-28B genotype across treatment arms.

We statistically modeled the probability of an imbalance in the good-response IL-28B variant (C/C rs12979860) between treatment arms for three hypothetical situations: a phase 1 trial (n = 60), a phase 2a trial (n = 120), and a phase 2b trial (n = 240). Each involved three randomized arms (Fig. 1). The probability of an imbalance in one treatment arm of ±10% (<23% or >43% when the C/C genotype frequency was assumed to be 33%2) was 31%, 18%, and 6% for the phase 1, 2a, and 2b trials, respectively, and the probability of an imbalance in one treatment arm of ±20% was 10%, 0.4%, and <0.01% for the phase 1, 2a, and 2b trials, respectively. We assumed a Caucasian population for this analysis; the inclusion of other ethnic groups would be expected to increase the risk of sampling error.2

thumbnail image

Figure 1. Probability of an imbalance in the IL-28B genotype frequency in one arm of an early-phase clinical trial involving three randomized treatment arms. The columns represent the probability of an imbalance at the specified levels; the curves represent the continuous probability distribution. The phase 1 study was assumed to involve 60 patients (20 per treatment arm), the phase 2a study was assumed to involve 120 patients, and the phase 2b study was assumed to involve 240 patients.

Download figure to PowerPoint

We then modeled the implications of such an imbalance for the primary outcome of viral load reduction at week 4 in studies combining a direct antiviral agent with peg-IFN and RBV (Table 1). An imbalance in the IL-28B genotype of 10% to 20% could lead to differences in HCVRNA reduction of 0.2- to 0.5-log10 IU/mL between treatment arms due to peg-IFN alone. This has great relevance for dose-finding studies in which the dose-related antiviral potency must be weighed against toxicity. In the setting of more extreme mismatching (e.g., in a mixed-ethnicity cohort), confounding by IL-28B genotype might even affect the decision to advance a compound from proof of concept to the next stage of clinical development. Indeed, Anadys Pharmaceuticals recently reported an imbalance in the frequency of the C/C genotype that confounded the week 12 results of a phase 2 trial (the C/C genotype frequency was 21% in the active treatment arms and 56% in the control arm).3

Table 1. Apparent Differences in the Antiviral Efficacy Resulting from an Imbalance in the IL-28B Genotype Frequency Between Treatment Arms in Early-Phase Trials
Imbalance Between Treatment ArmsArm A: Placebo and Peg-IFN/RBV (n = 20)Arm B: Antiviral (Dose 1) and Peg-IFN/RBV (n = 20)Arm C: Antiviral (Dose 2) and Peg-IFN/RBV (n = 20)
  1. We assumed that (1) the candidate drug had no antiviral activity, (2) the frequency of the good-response (C/C) IL-28B type was 33%, (3) the viral load reduction with the good-response (C/C) IL-28B type at week 4 was 3.8-log10 IU/mL, and (4) the viral load reduction with the poor-response (non-C/C) IL-28B type at week 4 was 1.4-log10 IU/mL.2

 C/C IL-28B type (%)233343
 Non-C/C IL-28B type (%)776757
 Mean viral load reduction  at 4 weeks (log10 IU/mL)
 C/C IL-28B type (%)133353
 Non-C/C IL-28B type (%)876747
 Mean viral load reduction  at 4 weeks (log10 IU/mL)

IL-28B genotype distribution is therefore an important consideration when early-phase hepatitis C virus clinical trials are being undertaken and their results are being interpreted. In the research setting, the typing of the IL-28B polymorphism is a straightforward, inexpensive test. We believe that treatment arms should be stratified by the IL-28B genotype. This is particularly relevant to early studies of direct antivirals in combination with peg-IFN and RBV for which antiviral efficacy is the primary outcome.


  1. Top of page
  • 1
    Akuta N, Suzuki F, Hirakawa M, Kawamura Y, Yatsuji H, Sezaki H, et al. Amino acid substitution in HCV core region and genetic variation near IL28B gene predict viral response to telaprevir with peginterferon and ribavirin. Hepatology 2010; 52: 421429.
  • 2
    Thompson AJ, Muir AJ, Sulkowski MS, Ge D, Fellay J, Shianna KV, et al. IL28B polymorphism improves viral kinetics and is the strongest pre-treatment predictor of SVR in HCV-1 patients. Gastroenterology; doi:10.1053/j.gastro.2010.04.013.
  • 3
    ANADYS announces completed 12-week results from phase II combination study of ANA598 with interferon and ribavirin. Accessed June 2010.

Alexander J. Thompson MD, PhD*, Andrew J. Muir MD*, Mark S. Sulkowski MD†, Keyur Patel MD*, Hans L. Tillmann MD*, Paul J. Clark MD*, Susanna Naggie MD*, Jacques Fellay MD, PhD‡, Dongliang Ge PhD‡, Jeanette J. McCarthy PhD§, David B. Goldstein PhD‡, John G. McHutchison MD*, * Duke Clinical Research Institute, Duke University Medical Center, Durham, NC, † Division of Infectious Diseases, Johns Hopkins University School of Medicine, Baltimore, MD, ‡ Center for Human Genome Variation, Duke University, Durham, NC, § Institute for Genome Sciences and Policy, Duke University, Durham, NC.