Summary and Introduction

Summary

Both absolute viral load and log decline in viral load from baseline were found clinically useful in predicting sustained virological response and lack of sustained virological response (non-sustained virological response, NSVR) to treatment. We assessed the clinical utility of hepatitis C virus (HCV) RNA quantitation and changes in viral load using the VERSANT® HCV RNA 3.0 Assay (bDNA) in 351 HCV-infected individuals treated with interferon plus ribavirin. We show that viral load decision thresholds provided negative predictive values (NPVs) of >95% at week 4 using a 100 000 IU/mL cut-off and at weeks 8 and 12 using 10 000 IU/mL cut-offs. A 2-log decline from baseline provided NPVs >95% at weeks 8 and 12. Combinations of absolute viral loads and changes in viral load from baseline did not enhance the performance of the decision rules for predicting NSVR. The positive predictive values (PPVs) at weeks 8 and 12 were 59.1 and 67.3%. This study highlights the critical importance of viral quantitation in gauging therapeutic response in patients with chronic HCV infection on antiviral therapy. Early changes in viral load, measured as absolute viral loads or change in viral load from baseline, are highly predictive of NSVR at 8 and 12 weeks. PPVs are modest but these data may provide encouragement to patients who are in the early phases of treatment when side effects are frequent. Additionally, we demonstrated the need for cautious interpretation of stopping rules when the values are at or near the decision thresholds.

Introduction

Quantitative hepatitis C virus (HCV) RNA tests have been used prior to initiating interferon-based anti-HCV treatment to assess the likelihood of sustained virological response (SVR) and early on in treatment to try to identify patients who will be responders or nonresponders to treatment.^[1-8] As interferon-based antiviral treatment is costly and associated with frequent side-effects, early identification of nonresponders is particularly useful to reduce the costs and morbidity of ineffective antiviral therapy.

The search for early predictors of treatment outcome began with the first large clinical trials of interferon monotherapy. The data from these early trials suggested that patients who did not reach undetectable levels of HCV RNA by week 12 would not attain a SVR with continued treatment.^[1,9] Subsequently, quantitative viral tests became more widely used and the prediction of response focused on assessment of viral load changes during the first few months of treatment. In clinical trials using interferon plus ribavirin treatment for 24 weeks in naïve patients, a viral load >400 000 copies/mL at week 4 of treatment had a negative predictive value (NPV) of 100% for a non-sustained virological response (NSVR). In patients treated for 48 weeks with combination interferon plus ribavirin, this decision rule had NPVs of 96.7 and 100% at weeks 4 and 12, respectively.^[10] Most recently, the predictive value of viral load determinations early in treatment was determined with pegylated interferon plus ribavirin treatment. A 2-log decline from baseline or an undetectable level of HCV RNA at week 12 had a NPV of 98%.^[6,11,12] The importance of early prediction of response and nonresponse in patient management is reflected in the National Institutes of Health (NIH) Consensus Development Conference Statement, Management of Hepatitis C: 2002,^[13] in which the use of a 2-log decline (or reaching undetectable levels of HCV RNA) decision rule at week 12 was recommended.

The current paper evaluates the utility of different viral loads (absolute levels and change during treatment) as predictors of SVR and NSVR to interferon plus ribavirin therapy in both treatment-naïve and experienced patients with chronic HCV infection. Various decision rules and combinations of rules were studied to define the decision rules that accurately predicted NSVR in the largest number of patients as early in treatment as possible, i.e. within the first 3 months of treatment initiation. Additionally, the performance of these stopping rules at the decision thresholds was evaluated.