Real world SOF/VEL/VOX retreatment outcomes and viral resistance analysis for HCV patients with prior failure to DAA therapy

Sustained viral response (SVR) rates for direct‐acting antiviral (DAA) therapy for hepatitis C virus (HCV) infection routinely exceed 95%. However, a small number of patients require retreatment. Sofosbuvir, velpatasvir and voxilaprevir (SOF/VEL/VOX) is a potent DAA combination primarily used for the retreatment of patients who failed by DAA therapies. Here we evaluate retreatment outcomes and the effects of resistance‐associated substitutions (RAS) in a real‐world cohort, including a large number of genotype (GT)3 infected patients. 144 patients from the UK were retreated with SOF/VEL/VOX following virologic failure with first‐line DAA treatment regimens. Full‐length HCV genome sequencing was performed prior to retreatment with SOF/VEL/VOX. HCV subtypes were assigned and RAS relevant to each genotype were identified. GT1a and GT3a each made up 38% (GT1a n = 55, GT3a n = 54) of the cohort. 40% (n = 58) of patients had liver cirrhosis of whom 7% (n = 4) were decompensated, 10% (n = 14) had hepatocellular carcinoma (HCC) and 8% (n = 12) had received a liver transplant prior to retreatment. The overall retreatment SVR12 rate was 90% (129/144). On univariate analysis, GT3 infection (50/62; SVR = 81%, p = .009), cirrhosis (47/58; SVR = 81%, p = .01) and prior treatment with SOF/VEL (12/17; SVR = 71%, p = .02) or SOF+DCV (14/19; SVR = 74%, p = .012) were significantly associated with retreatment failure, but existence of pre‐retreatment RAS was not when viral genotype was taken into account. Retreatment with SOF/VEL/VOX is very successful for non‐GT3‐infected patients. However, for GT3‐infected patients, particularly those with cirrhosis and failed by initial SOF/VEL treatment, SVR rates were significantly lower and alternative retreatment regimens should be considered.


| INTRODUC TI ON
Sustained virological response (SVR) rates for patients chronically infected with hepatitis C virus (HCV) and treated with direct-acting antiviral (DAA) therapies in clinical trials and real-world cohorts are often >95%. [1][2][3] Treatment failure is associated with multiple host and viral factors including advanced fibrosis or cirrhosis, HCV subtype [4][5][6] and the presence of resistance-associated substitutions (RAS) in HCV-encoded proteins that are targeted by DAA. 7 This leaves a small percentage of HCV-infected patients who have been failed by first-line therapies and are therefore by definition "difficult to treat".
For patients who have been failed by pan-genotypic regimens such as SOF/VEL and glecaprevir and pibrentasvir, (GLE/PIB) retreatment options are limited. Currently the European Association for the Study of the Liver (EASL) and American Association for the Study of Liver Disease (AASLD), recommend a combination of sofosbuvir, velpatasvir and voxilaprevir (SOF/VEL/VOX) for these patients. 8,9 Retreatment of patients previously failed by predominantly firstgeneration NS5A inhibitor-containing DAA therapy with SOF/VEL/ VOX has been evaluated in both the POLARIS-1 and POLARIS-4 phase-II and III studies. 10 In these studies, SOF/VEL/VOX showed a very high SVR rate in GT1-infected patients (222/228; 97% SVR); however, a slightly lower SVR rate was observed in GT3 patients (126/132; 95% SVR). In addition, a number of studies have evaluated the observed real-world outcomes of SOF/VEL/VOX retreatment.

| Patients and samples
This study included 215 individuals from across England who did not achieve SVR with previous interferon-free DAA treatment.
Blood samples were taken prior to retreatment and samples sent to Oxford and Glasgow for HCV whole-genome sequencing (WGS).
Clinical, demographical and treatment outcome data were collected from patients, who were enrolled in HCV Research UK (n = 37), following informed consent and ethical approval. 14

| Whole genome sequencing of HCV
Whole HCV genome sequencing was conducted at three sites to define HCV subtype and identify RAS: (i) The MRC-University of particularly those with cirrhosis and failed by initial SOF/VEL treatment, SVR rates were significantly lower and alternative retreatment regimens should be considered. cDNA. An Illumina sequencing library was generated and enriched for HCV viral sequences using specific oligonucleotide probes.

K E Y W O R D S
Illumina sequencing reads were then processed and mapped to the closest HCV reference genome. Precise details of how the sequencing was performed at each site are described in the supplementary materials.

| Resistance-associated substitution calling using HCV-GLUE
HCV subtypes were assigned and NS3, NS5A and NS5B RAS relevant to each genotype were identified (15% of reads cut off) using HCV-GLUE and RAS definitions provided by PHE 16 (Supplementary   Table 3). A RAS was considered relevant if there is in-vitro evidence of a >50 Fold change in EC50 and/or in-vivo evidence that the RAS is treatment emergent or associated with treatment failure.
HCV-GLUE is a resource created as part of the STOP-HCV consortium based on the GLUE (Genes Linked by Underlying Evolution) platform. 17 HCV-GLUE uses the ICCT HCV reference sequences and collates publicly available HCV sequences, to construct phylogenies using RAxML to assign sequences to HCV clades. It also contains a database of HCV RAS created by PHE 16 and this is used to identify RAS relevant to the HCV genotype and subtype. 18

| Statistical analysis
All statistical analysis was performed using custom code and R version 3.6.2. Logistic regression was used for the analysis of association between categorical variables.

| Effectiveness of retreatment with SOF/VEL/ VOX
The overall SVR12 rate for patients retreated with SOF/VEL/VOX was 91% (n = 129) ( Figure 2). The majority of patients who did not achieve an SVR (n = 15) experienced a post-treatment relapse (n = 10). One patient experienced an on-treatment breakthrough (patient's HCV RNA was undetectable during treatment and then became detectable again during treatment.) and two were nonresponders (patients who never achieve undetectable HCV RNA).
Relapse/breakthrough/non-response was not reported by the treating clinician for the remaining two patients. Poor adherence to retreatment was not reported for any of the patients who did not achieve an SVR.
However, with the exception of one GT1 patient previously treated with SOF/VEL, all patients with prior SOF/VEL or SOF +DCV treatment had GT3 infection which confounded this result (Table 1).
When GT3-infected patients were analysed separately, cirrhosis and prior treatment regimen were no longer significantly associated with outcome, in both the univariable and multivariable analysis.  (Table 2). When each RAS (Full list in Table 3) was considered independently using logistic regression, A30K and Y93H RAS in NS5A were associated with retreatment failure in the whole cohort (A30K p = .02, Y93H p = .01) ( Table 3). The A30K RAS is unique to GT3a (it is wild-type in GT3b) and the Y93H substitution was much more common in GT3 patients compared to other genotypes (Supplementary Figure 2). When the analysis is limited to GT3 patients, the SVR rates for patients with these RAS were still reduced (for A30K SVR, 71% and Y93H SVR, 78%) but this was no longer statistically significant (A30K p = .2, Y93H p = .3). In addition, when the genotype is added as a cofactor to the logistic regression Y93H is no longer associated with the outcome. Whilst individually the A30K and Y93H RAS are common, the combination of A30K + Y93H is found rarely and is not associated with treatment outcome in this cohort (Supplementary Figure 3). No other RAS was significantly associated with outcome within the whole cohort or within a specific genotype of patients.  In summary, our study shows that retreatment outcomes for patients failed by first-line DAA therapy are very successful for non-GT3-infected patients. However, for GT3-infected patients, particularly those with cirrhosis and failed by initial SOF/VEL treatment, SVR rates were significantly lower and alternative retreatment regimens such as GLE/PIB +SOF should be considered.

ACK N OWLED G EM ENTS
The following clinicians contributed to data collection for this pro-