Determining the Lower Limit of Detection Required for HCV Viral Load Assay for Test of Cure Following Direct-acting Antiviral-based Treatment Regimens

Evidence From a Global Data Set

Jake R. Morgan; Elizabeth Marsh; Alexandra Savinkina; Sonjelle Shilton; Shaun Shadaker; Tengiz Tsertsvadze; George Kamkamidze; Maia Alkhazashvili; Timothy Morgan; Pam Belperio; Lisa Backus; Waheed Doss; Gamal Esmat; Mohamed Hassany; Aisha Elsharkawy; Wafaa Elakel; Mai Mehrez; Graham R. Foster; Constance Wose Kinge; Kara W. Chew; Charles S. Chasela; Ian M. Sanne; Yin M. Thanung; Anne Loarec; Khawar Aslam; Suna Balkan; Philippa J. Easterbrook; Benjamin P. Linas


J Viral Hepat. 2022;29(6):474-486. 

In This Article


Data Sources: Observational Cohorts and Clinical Trials Registries

We assembled a data set of patients with detectable HCV viral load at week 12 after completion of DAA treatment from clinical observational cohorts in nine countries, in addition to international clinical trial registry databases from two pharmaceutical companies. We identified potential cohorts for inclusion from four sources: (1) cohorts included in a previously published analysis of 12 countries for LoD at diagnosis;[11] (2) cohorts that had previous collaborative projects with WHO or were known by our research team; (3) a PubMed literature search using the search terms 'HCV SVR' and 'cohort study'; and (4) conference abstracts from 2018 to 2020 at the American Association for the Study of Liver Diseases and the European Association for the Study of the Liver. We sent our study protocol (Appendix S1) and an invitation to join a research collaborative to the principal investigators of each identified cohort. We approached 19 cohorts and registries with HCV-infected patients who had received treatment.

Eleven observational cohorts representing nine countries and two clinical trial registries agreed to collaborate and share data. To be included in the global data set, cohorts and trial databases were required to have the following patient-level data: Detectable HCV RNA test at 12 weeks post-treatment and linked demographic data per protocol (see Appendix S1). Observational cohorts were characterized by one of the following: (1) registries from country-wide national HCV ministries of health (Georgian and Egypt national programs); (2) large healthcare systems (United States Department of Veteran Affairs or the UK); (3) non-governmental organizations with programmatic data including across multiple countries (Médecins Sans Frontières (MSF) sites in Mozambique, Cambodia and Pakistan); or (4) grant-funded research projects (Myanmar and Ukraine).

We also included clinical trials databases from two pharmaceutical companies who were responsible for originator DAAs (Gilead and AbbVie). We pooled data from relevant DAA trials into a single repository. Patient eligibility criteria varied by study trial, and the majority required a pre-treatment viral load over 1000 IU/mL for enrolment and censored individuals who experienced on-treatment virological failure. Both clinical trial databases were used individually to determine LLoD and the distribution of HCV RNA at treatment failure based on summary data at SVR12 assessment. However, only one database was able to provide patient-level data to contribute to the multivariable regression analyses of factors associated with low-level viraemia (LLV).

Characteristics of Study Cohorts

Most of the included cohorts have been well described in the literature.[19–26] Table 1 summarizes key characteristics of these cohorts, including number of HCV-treated patients, gender, age, and genotype distribution, DAA regimens and the proportion who achieved SVR following treatment. There was a high degree of heterogeneity in patient characteristics across cohorts, reflecting varying HCV epidemic profiles in different countries.

Data Concatenation

We modified a protocol for data concatenation from the previous study of viral load at diagnosis,[27] to create comparable variables across a global data set.[27] First, we requested a core set of demographic and clinical variables for each included individual: age (dichotomized as under 60 or 60 years or older based on the format of data received), sex, HCV genotype, fibrosis stage, type and duration of DAA regimen, and presence of human immunodeficiency virus (HIV) or hepatitis B virus (HBV) co-infection. Second, we harmonized data for three key variables—fibrosis stage; HCV genotype; and HCV DAA treatment regimens. We standardized fibrosis stage by calculating the Fibrosis-4 (FIB4) score, which correlates well with staging based on transient elastography and liver biopsy.[28–30] We assigned the corresponding Metavir state to the FIB4 scores: FIB4 score <1.45, (Metavir stage F0-F1); FIB4 1.45–3.25 (Metavir stage F2—F3); and FIB4 scores above 3.25 (Metavir stage F4). We imputed genotype data for one country, Egypt, which has predominantly genotype 4 infection, and where genotyping is no longer performed routinely.[31] We used the distribution of genotypes from the literature as probabilities of having each given genotype and employed a Markov chain Monte Carlo technique to stochastically assign a genotype to each individual in the Egyptian cohort. All other cohorts had patient-level data on genotype. Finally, since more than 20 different DAA treatment regimens were in use during our data collection period, we assigned these regimens to three categories based on the time period of introduction: early era DAAs (sofosbuvir/ribavirin and sofosbuvir/simeprevir with or without ribavirin), mid-era (sofosbuvir/daclatasvir, sofosbuvir/ledipasvir and elbasvir/grazoprevir- or ombitasvir/paritaprevir-containing regimens) and recent-era (glecaprevir/pibrentasvir and sofosbuvir/velpatasvir regimens).

Statistical Analyses

We evaluated the distribution of HCV viral load at 12 weeks post-treatment assessment among all patients in the combined data set using both standard IU/mL measures and normalized using a log10 transformation, to allow for better visualization of the viral load distribution. From these measures, we identified the lower 95th, 97th and 99th percentiles of the HCV RNA distribution. To estimate the 95% confidence interval for each viral load threshold, we use the method described by Hahn and Meeker, 1991, which corrects for a non-normal distribution of values.[32]

We next identified the subgroup of individuals who had HCV RNA that is detectable, but <1000 IU/mL at the time of SVR assessment. Such individuals were defined as having 'low-level viremia' (LLV) at the time of treatment failure. We chose the threshold of <1000 IU/mL because that is the LLoD for HCV core antigen assay and a reasonable proxy for newer platforms being developed that have potential to be near point-of-care in use and so promote access to diagnosis and treatment.[33,34] We described the characteristics of this subgroup and employed logistic regression to assess factors associated with having LLV compared to no LLV among those with detectable viral load at SVR12. Analyses were conducted using SAS (version 9.4; SAS Institute; Cary, NC). The Boston University Institutional Review Board ruled this study not human subjects research.

Sensitivity Analyses

We conducted two additional subgroup and sensitivity analyses. First, we examined whether results differed between clinical trial and observational cohorts to determine whether conclusions potentially differ in investigational and real-world settings. Second, we examined the HCV RNA viral load at SVR24 after treatment completion to assess whether checking for cure at 24 weeks rather than 12 weeks would provide similar conclusions. This was undertaken in the cohort from Georgia, the US Department of Veterans Affairs, and one of the clinical trial cohorts that reported patient HCV RNA test results at both 12- and 24 weeks post-treatment.