The reporting of three independent genome-wide association studies has heralded a burst of excitement for the use of interleukin-28B (IL-28B) polymorphisms in the prediction of spontaneous or treatment-induced hepatitis C virus (HCV) clearance. In the year following the initial reports,1-3 there were more than 20 publications on the subject. Several studies,4-6 including a recent publication in Hepatology,7 argue that IL-28B genotyping will be of central importance for the management of patients. Although we agree that this discovery holds promise for understanding the variable host responses to interferon-α–based regimens, we believe that there may be some confusion with respect to the difference between odds ratio (OR) and the predictive value of the IL-28B genotype as a standalone biomarker.
OR describes the strength of association between an IL-28 single-nucleotide polymorphism and virological response. Tanaka et al.1 reported ORs of 17.7 and 27.1 for rs12980275 and rs8099917 with astronomical P values of 2.84 × 10−27 and 2.68 × 10−32, respectively. Ge et al.3 determined a combined OR of 3.1 for rs12979860, and Suppiah et al.2 found a combined OR of 1.98 for rs809917. It is challenging to ascertain the predictive value of a particular IL-28B allele in the first two studies cited; however, Suppiah et al. were careful to note that “according to a model of dominant inheritance, the rs8099917 G allele predicts non-response with 57% sensitivity and 63% specificity.” They also reported a negative predictive value (NPV; i.e., a value indicating the correct prediction of treatment failure) of 64%.
We recently reported data from a prospective cohort collected for plasma biomarker discovery.8 Although our cohort was considerably smaller than the populations used for genome-wide association studies, we found that the rs12979860 C/C genotype increased the odds of early virological response (EVR; OR = 2.53) with a positive predictive value (PPV) of 75% (Fig. 1A). Notably, biomarkers are used in the management of chronic patients with the intention of identifying those individuals who will fail to respond to therapy and thus can be directed to alternative treatment options (e.g., the failure to achieve EVR is used in the clinic as a negative predictor of sustained virological response). As such, the important index is the NPV, which was found to be 42% for the patients in our study (Fig. 1A). We thus argue the need for phenotypic markers to complement markers of genetic susceptibility. One plasma biomarker that has received attention is interferon induced protein 10 (IP-10 and also referred to as CXCL10), with higher concentrations predicting treatment failure.9 We confirmed these data and found an NPV of 86% (Fig. 1B). One exciting possibility is the combination of assays; if they are taken together, predictions based on the C/C genotype or a low plasma concentration of CXCL10 yield an NPV of 100% (Fig. 1C).
Our growing knowledge of epigenetics and the impact of environmental factors makes clear that phenotypic markers and/or functional assays (measured with validated assays) will be required to fully exploit the knowledge gained by genetic studies. We conclude that there is a real need to continue the effort to identify predictive biomarkers that will be clinically useful for managing patients with HCV disease.