The analysis of a data set of 5973 cases of detectable viraemia following treatment with a wide range of different DAA treatment regimens from nine different country observational cohorts and two international clinical trials databases shows that 95% of HCV treatment failures identified 12 weeks after the end of DAA therapy have a VL greater than 227 IU/mL (2.36 log IU/mL) and 97% have greater than 70 IU/mL (95% CI 48–86). There were important differences in the distribution of viral load at treatment failure in those who were participants in clinical trials compared to observational studies. The median viral load at treatment failure was nearly 10-fold higher (2,344,229 IU/mL in clinical trial registries vs. 264,809 IU/mL in observational cohorts) and 95% had a viral load greater than 4030 IU/mL (95% CI 24–4100) compared to 214 IU/mL (95% CI 166–266). Just 3% of clinical trial participants had viral load under 1000 IU/mL, compared to 10% in those from observational databases. This is broadly consistent with results from an analysis of 34 phase 2/3 clinical trials which showed less than 1% had had viral load under 1000 IU/mL. The reasons for this higher viral load in treatment failures among trial participants may relate to the more stringent selection criteria for clinical trials and exclusion of those with LLV (only those with VL >1000 IU/ml were enrolled, and on-treatment failures were excluded from analysis). This highlights importance of reporting analyses separately for clinical trial and observational databases.
We also found that several independent demographic, clinical and treatment characteristics were associated with LLV (<1000 IU/mL) including female sex (relative to male), fibrosis stage F0-F1 (compared to F4, and medications from early DAA treatment era (vs. mid), although the biological reasons for these associations are not clear. In a sensitivity analysis, using data from SVR week 24 yielded similar results on viral load distribution or predictors of low-level viraemia.
There are several practical implications for these findings. Currently, there are five assays that have WHO prequalification for HCV confirmatory viral load: three laboratory-based assays (Abbott Real time HCV PCR, Alinity m HCV RT-PCR and Abbott Architect HCV Ag) and two point-of-care assays (Xpert HCV viral load and GeneDrive HCV). The majority of existing lab-based assays with a LloD of between 5 and 15 IU/mL as well as approved PoC assays (HCV RNA PoC GeneXpert assays has a LLOD of 10 IU/mL for venous blood,[12,13] or 100 IU/mL using fingerstick capillary blood) would detect more than 97% of treatment failures and is also therefore appropriate for testing for HCV cure. The recently WHO prequalified portable PoC Genedrive instrument has a reported LLoD of 2362 IU/mL as well as existing HCV core antigen (ref) with a LLOD of 1000 IU/mL. A clinical trial is currently underway evaluating the Truenat HCV RNA assay from Molbio Diagnostics that uses capillary blood and a battery-powered mobile platform (LLoD is not yet available).[36,37] Currently, the European Association for the Study of the Liver Diseases, IDSA-AASLD HCV guidance panel, recommend a minimum LLoD of 1000 IU/mL for HCV diagnosis, with no specification of minimal test characteristics for test of cure. End users should be aware that some low-level virological failures may therefore be missed may therefore be missed with an assay with a LLoD greater than 1000 IU/mL and that there will need to be a trade-off with the convenience of lower cost and more available viral load assays that may have a higher limit of detection (LLoD) and thus lower analytic sensitivity than standard laboratory-based assays. The results of our analysis provide an additional valuable evidence base for guidance panels and regulatory authorities to assess use of platforms for monitoring of SVR12 as well as diagnosis. Given the differences in viral load distribution between clinical trials and observational studies, more work is needed to better understand which data sources should be used to inform WHO LOD standards.
The primary strength of this study is that the analysis was based on the largest global data set to date of nearly 6000 of HCV treatment failures. The high cure rate associated with pan-genotypic DAAs, routinely exceeding 90%, has previously made it difficult to assemble a large enough cohort, representing different geographic regions with different genotypes, range of stage of disease and use of different DAA regimens with adequate rates of follow-up SVR measurement—to reflect real-life distribution of viral loads at treatment failure. We had data from both clinical trials with high level of follow-up, as well as from observational cohorts reflecting real-world treatment experience.
There are several limitations to the data and analysis. First, we are not able to measure all potential factors contributing to HCV RNA level at treatment failure, such as risk characteristics (injection drug behaviours, sex work, etc.), but the initial analysis did not show that drug regimen, genotype or stage of disease were important determinants of low-level treatment failure. It is important to study and understand those unmeasured confounders, because if the underlying causal relationship is between a measurable or identifiable trait and the likelihood of having LLV at the time of treatment failure, then it may be possible to tailor guidance to identify venues or subgroups of people in whom it is still appropriate to employ available, close to patient assays to test for HCV cure. Second, our data set only included those who initiated treatment and returned for follow-up HCV RNA testing 12 weeks post-treatment. It is likely that those who fail to return for SVR may be at higher risk of treatment failure, and it is unclear whether they will be at lower or higher risk for LLV. However, this primarily affects the observational cohort and not clinical trial registry data. Third, more than 70% of our global data set cohort of treatment failures came from either Egypt (predominantly genotype 4) or the United States (predominantly genotype 1). We were not able to assemble a cohort of individuals with HCV treatment failure that represented all HCV genotypes, and all stages of disease. Finally, our global data set included data from both national and health system-wide observational databases reflecting real-world treatment experience, as well as from clinical trials with strict inclusion and exclusion criteria.
This study assessed the distribution of detectable viral loads 12 weeks following the end of treatment for HCV infection in an international cohort to inform the lower limit of detection of viral load assays for test of cure to identify treatment failure as well as for diagnosis of chronic hepatitis C infection. Based on a combined data set of clinical trials and observational data, a LLoD of 227 IU/mL (4030 IU/mL in the clinical trials subsample) would identify 95% of patients with a detectable viral load 12 weeks after treatment. While more than 10 times higher than the analytical sensitivity of laboratory-based NAATs, it is more than 10 times lower than the LLoD for HCV diagnosis. These findings demonstrate it might be prudent and necessary to consider different LLoD standards for HCV diagnosis and for test of cure. Development of a point-of-care HCV test for cure with a low enough limit of detection to identify 95% of patients and is affordable, is an important aspect of expanding access to HCV treatment and a vital component of the WHO's HCV elimination targets.
DAA, direct-acting antiviral; FIB4, Fibrosis-4; HBV, hepatitis B virus; HCV, hepatitis C virus; HIV, human immunodeficiency virus; LLoD, limit of detection; LLV, low-level viraemia; MSF, Médecins Sans Frontières; NAAT, Nucleic Acid Amplification Testing; WHO, World Health Organization.
National Institute on Drug Abuse; UNITAID
The authors thank the Georgia National Hepatitis C Elimination Program and Hepatology and Gastroenterology Department of the Medical Center Mrcheveli, Tbilisi Georgia for their contribution. The authors thank Liyun Ni, Anand Chokkalingam and Betty Chiang of Gilead Sciences for the provision of data and thoughtful comments on the draft manuscript. Further, the authors wish to acknowledge the role of the HCV Research UK (Funded by the Medical Research Foundation [award number C0365]) in collecting and making available the data used in the generation of this publication and the United States Department of Veteran Affairs and the Government of Egypt for the provision of data for this project. Part of the data used for this study was provided by Medecins sans Frontieres and Epicentre. This project was supported through a grant from UNITAID, and the FIND contribution was supported by UNITAID as part of HEAD-Start (Hepatitis Elimination through Access to Diagnostics). The National Institute on Drug Abuse (grant P30DA040500, P30AI042853) funded this project. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the United States Centers for Disease Control and Prevention.
J Viral Hepat. 2022;29(6):474-486. © 2022 Blackwell Publishing