We thank O'Sullivan et al. for their comments on our article and for sharing the results of their analysis of samples with low concentrations of hepatitis C virus (HCV) RNA. Their analysis nicely illustrates one of the technical drawbacks of using a target detected/target not detected cutoff for treatment decision making, as this cutoff is inherently not reproducible over a certain range of HCV RNA levels (in this case, ∼1 to 10 IU/mL). Their analysis also demonstrates that lower concentrations of HCV RNA are often not detected by the assay, even if the samples are analyzed repeatedly. Based on this analysis the authors raise the important consideration that transiently detected but not quantifiable (Detected/<LLOQ [lower limit of quantitation]) HCV RNA during follow-up for patients who have achieved a sustained virologic response (SVR) may indicate incomplete clearance of HCV, as opposed to “false-positive” detection of HCV RNA.
As discussed in our article,1 we interpreted many of the transiently Detected/<LLOQ follow-up HCV RNA results from SVR-achieving subjects as likely reflecting false-positive detection of HCV RNA. Our interpretation was based on the totality of the evidence we had from our reviews of the Phase 3 boceprevir and telaprevir trials. Much of that evidence came from observations between Vendor A and Vendor B. We believe the on-treatment results reported by Vendor A are more consistent with the fact that, on average, HCV RNA levels reported as Not Detected (i.e., “Target Not Detected”) are qualitatively lower than HCV RNA levels reported as Detected/<LLOQ, which led us to question the higher rate of transiently Detected/<LLOQ follow-up HCV RNA results reported by Vendor B. We also determined that the predictability of virologic relapse based on Detected/<LLOQ HCV RNA at the end of treatment (EOT) varied by assay vendor. In a pooled analysis of all arms in clinical trials P05216 and C216 (both used Vendor A), only 14 subjects had an HCV RNA result of Detected/<LLOQ at EOT (or last available on-treatment sample), of whom five (36%) achieved SVR and nine (64%) experienced virologic relapse. For all five SVR subjects, the Detected/<LLOQ results were not reproduced with repeat testing or subsequent follow-up analyses. In contrast, 87 subjects in clinical trial 108 had EOT results reported as Detected/<LLOQ by Vendor B, and 74 subjects (85%) achieved SVR, whereas 13 subjects (15%) experienced virologic relapse. Again, the poor predictability of virologic relapse for EOT results by Vendor B raise concerns of a higher false-positive detection rate. Importantly, Detected/<LLOQ results from Vendor B were more likely to be “Not Detected” when reanalyzed by Vendor A using the same assay, particularly for samples collected later during treatment or during follow-up, times when HCV RNA would be expected to be absent or extremely low for patients who ultimately achieved SVR.
An alternative interpretation is that the assay in the hands of Vendor B had superior sensitivity in detecting very low-level HCV RNA, which would imply that SVR reflects the reduction and maintenance of HCV RNA just below a threshold somewhere between the sensitivity of the assay in the hands of Vendors A and B. In our opinion this explanation seems inconsistent with our general understanding of SVR and the biology of HCV.
Nevertheless, as we acknowledged in our article, our analyses do not rule out the possibility that transiently Detected/<LLOQ HCV RNA measurements during follow-up reflect accurate detection of HCV RNA. We agree with O'Sullivan et al. that we should remain cautious when interpreting results from an HCV nucleic acid-based test. Lack of HCV RNA detection by such a test does not distinguish between complete viral clearance and the presence of extremely low levels of virus or viral RNA. Furthermore, positive detection of HCV RNA does not distinguish between fully replication-competent virus and noninfectious viral RNA, or even a rare nonspecific amplification signal. Finally, it is necessary to reiterate the most important point of our article: Detected/<LLOQ measurements on-treatment reflected reduced virologic responses and thus should be considered differently from Target Not Detected measurements when making response-guided clinical decisions, whereas measurements at EOT and off-treatment follow-up will not affect these clinical decisions.