Strategic treatment optimization for HCV (STOPHCV1): a randomised controlled trial of ultrashort duration therapy for chronic hepatitis C [version 1; peer review: awaiting peer review]

Background: The world health organization (WHO) has identified the need for a better understanding of which patients with hepatitis C virus (HCV) can be cured with ultrashort course HCV therapy.


Introduction
The recent and rapid development of treatment for hepatitis C virus (HCV) has enabled an ambitious strategy for the elimination of viral hepatitis as a global public health threat by 2030, with the target of treating 80% of those chronically infected with HCV 1 . Although licensed durations of [8][9][10][11][12] weeks' therapy with directly acting antivirals (DAAs) are significantly shorter, more tolerable and more effective than previous interferon-based therapies 2 , there are patients who still find it challenging to complete a full treatment course. Such patients will become an increasingly important part of clinical practice as treatment coverage expands to reach marginalised groups, and world health organization (WHO) treatment guidelines highlight the need to understand the factors that could be used to select patients for successful short course treatment 3 .
Shorter treatment courses of licensed therapies are likely to improve adherence, including in those with active illicit drug use 4,5 . In acute or recent HCV infection, shortened courses of licensed therapy may have sufficiently high efficacy to be recommended routinely 6 . However, in chronic infection there is limited data to select patients able to achieve high cure rates with short duration therapy. Unselected short duration treatment has been able to achieve cure rates of 20-40% with 4 weeks and 57-95% with 6 weeks treatment in small Phase II studies [7][8][9] , but few of these combinations or durations were subsequently licensed for use. For licensed therapies, baseline viral load (<6,000,000 IU/mL) 10 and subgenotype 11 have been recommended to shorten therapy from 12 to 8 weeks, but there are no validated criteria to recommend less than 8 weeks therapy in chronic infection.
For a clinician deciding whether to start treatment in a patient considered at high risk of not completing therapy, there is a potential concern that emerging resistance with virological failure may compromise future treatment options. However, it is also possible that with shorter courses of treatment, less resistance may emerge. In vitro evidence suggests the addition of ribavirin, a generically available guanosine analogue, may improve rates of virological cure with shorter treatment courses 12 and could reduce the emergence of resistance in those failing treatment when added to short-course therapy 13 . However, these hypotheses have not been tested in a randomised trial.
We performed a strategic post-licensing randomised controlled trial in HCV-infected participants with mild liver disease to evaluate strategies for short-course treatment, the impact of treatment failure on retreatment and the role of adjunctive ribavirin in short-course therapy.

Design
We conducted a multi-centre, randomised, open-label, factorial, parallel group non-inferiority trial of adults with mild chronic HCV in 14 UK centres ( Figure 1). The trial was approved by the Cambridgeshire South Research Ethics Committee (15/EE/0435). The trial was registered at ISRCTN (37915093, 11 th April 2016), and EudraCT (2015-005004-28, 31 st December 2015). This study is reported in line with the Consolidated Standards of Reporting Trials (CONSORT) 14 .

Participants
Eligible participants were adults (≥18 years) infected with HCV genotype 1a/1b/4 for ≥6 months, with consistently detectable viremia 6 months before randomisation, no evidence of significant liver fibrosis (Fibroscan score ≤7.1kPa equivalent to F0-F1 15 ), body mass index (BMI) ≥18kg/m 2 , HCV viral load (VL) detectable but <10 million IU/ml at screening, no previous DAA exposure for current infection (previous pegylated-interferon/ribavirin allowed) and with laboratory results meeting minimum thresholds (platelets ≥60×10 9 /L, haemoglobin >12g/dL (male) or >11g/dL (female), creatinine clearance (estimated using Cockcroft-Gault) ≥60ml/min, international normalised ratio (INR) <1.5). Individuals co-infected with HIV were eligible if HIV VL had been <50 copies/ml for >24 weeks on anti-HIV drugs. Participants were excluded if they had a history of malignancy within 5 years, any history of pre-existing cardiac disease or haemoglobinopathies, or a current disorder which may cause ongoing liver disease, may negatively impact the participant's ability to adhere to the study or might limit the participant's life expectancy. Participants were also excluded if they were hypersensitive to any of the study drugs, currently taking any medication known to interact with study medication or had used other investigational products within 60 days of screening. Female participants were excluded if lactating, pregnant, planning to become pregnant or not willing to use effective contraception (excluding products containing ethinyl-oestradiol) including up to four months after the study. Male participants were excluded if they were planning pregnancy with a partner or not willing to use effective contraception for up to seven months after the study. Participants gave written informed consent after explanation of the aims, methods, benefits and potential hazards of the trial and before any trial-specific procedures were performed/ any blood taken for the trial. The main criteria for selecting participating hospitals was that they had the potential for recruiting the required number of chronic (>6 months) HCV genotype 1a/1b/4 infected Figure 1. Trial schematic. Note: DAA: direct-acting antivirals. The ribavirin randomisation was a partial factorial in those randomised to a shorter course than the full licensed duration of therapy (the vast majority of participants recruited to the trial).
participants within the agreed recruitment period. This was established by the use of a trial specific site survey. Site also needed to meet the following criteria: no competing studies that would impact on the ability to enrol quickly to the trial; turnaround of no more than 7 days for HCV viral load test results; ability to provide 24 hour cover for trial patients; local governance approval likely to take <3 months.

Intervention and procedures
For participants allocated to VUS, the duration of first-line therapy varied between 28 and 42 (mean 32; before 1 April 2017) or 49 (mean 39; after 1 April 2017) days determined by the baseline screening VL using a continuous scale (Table 1,  Table 2, Figure 2). Blinding was not used because the primary end-point was an objective measure of viraemia blinded to clinical data measured in routine laboratories without knowledge of randomisation. All patients failing treatment were retreated as soon as practicable with 12 weeks of oral sofosbuvir 400mg/ledipasvir 90mg once-daily and weight-based oral ribavirin twice-daily (dosing as above). Treatment failure was defined as (i) two consecutive measurements of HCV VL above the lower limit of quantification (LLOQ) (taken at least one week apart) after two consecutive visits with HCV VL <LLOQ at any time, with the latter confirmatory measurement also being >2000 IU/mL, or (ii) two consecutive measurements of HCV VL (taken at least one week apart) that were >1log 10 increase above the nadir on treatment and >2000 IU/mL at any time.
All participants were followed by the site teams for 24 weeks after the end of first-line treatment or re-treatment (where applicable) for evaluation of virological response and toxicity. Participants on first-line therapy had clinical assessments on days 3, 7, 10, 14, 28 and end of treatment (EOT, where EOT was not day-28) followed by weeks 2, 4, 8, 12 and 24 after EOT. All outcome measures were assessed at these clinic visits.

Randomisation
Randomisation was performed via a computer-generated programme at the STOP-HCV-1 Co-ordinating Centre (MRC CTU at UCL). Patients were allocated 1:1 using a factorial design to each of: biomarker-stratified variable ultrashort vs fixed duration; adjunctive ribavirin or not (a partial factorial in those randomised to a shorter course than the full licensed duration of therapy). Randomisation was stratified by study centre, HCV genotype and study drug regimen using a minimisation algorithm incorporating a probabilistic element incorporated securely into the online trial database. Randomisation determined the duration of first-line therapy rather than the choice of DAAs which was pre-specified by the investigator before randomisation based on local availability.
Each allocation was generated within the trial database only at the point of randomisation after it was confirmed the participant was eligible and was to be randomised. Allocations were generated using minimisation with a probabilistic element, so there was no pre-determined allocation sequence to conceal. To further conceal the potential allocation, study centres were not informed of the randomisation strata.  On the day of randomisation, participant eligibility was checked at sites and the data confirming eligibility was entered onto a case record form and sent to MRC CTU. The data was entered onto the database at MRC CTU and checked for eligibility again. Once confirmed that the participant was eligible, the database would perform randomisation using the computer-generated programme. Sites were then informed of the allocation and length of DAA treatment required for the participant.

Sample size
A priori power calculation assumed 88% of participants would achieve SVR12 on first-line fixed-duration, and that SVR12 would be 85% on retreatment (significantly lower than the actual retreatment success rate, below), leading to an overall cure rate of 98% (first-line plus retreatment) in the control group. Assuming 98% cure rate in the fixed-duration group, 80% power, one-sided alpha 0.025, and a 5% loss to follow-up, 408 participants were needed to demonstrate noninferiority with a 4% margin. The choice of the 4% margin was based on clinical judgement and to ensure that overall cure rates in the variable-duration group would be well over 90% if non-inferiority was demonstrated. Interim data were reviewed by an independent Data Monitoring Committee (DMC) (4 biannual meetings). The protocol stipulated that the DMC could alter the first-line treatment strategy if there was strong evidence the SVR12 rate for VUS1 was less than 65%. After the DMC meeting in April 2017, the DAA duration strategy was changed from 4-6 weeks (VUS1) to 4-7 weeks (VUS2) ( Table 1). All participants randomised from the 1st April 2017 were treated under VUS2. The trial closed in August 2018 when no further recruitment was possible. By this time, the great majority of patients with viraemia were unable to engage with treatment per se, and not suitable for inclusion in this study.

Outcomes
The primary outcome was sustained virological response 12 (SVR12, plasma HCV VL <LLOQ without prior failure 12 weeks after the end of the combined first and any re-treatment phases). For ribavirin comparison the primary outcome was SVR12 after first-line treatment only. Secondary outcomes were SVR12 after first-line treatment (where not the primary outcome), SVR12 after the end of the combined first and any re-treatment phases (where not the primary outcome), SVR24 (24 weeks) after the end of the combined first and any re-treatment phases, SVR24 after first-line treatment only, lack of initial virological response, viral load rebound after becoming undetectable, serious adverse events, grade 3/4 adverse events, grade 3/4 adverse events judged definitely/probably related to interventions, treatment-modifying adverse events (any grade), grade 3/4 anaemia and emergence of resistanceassociated HCV variants. Adverse events were graded following the Division of AIDS grading tables. Full genome sequence was obtained by RNA Sequencing on Illumina Miseq platform following target enrichment with a library of genotype specific HCV capture probes as described previously 16 .

Statistical analysis
Randomised groups were compared following the principle of intention-to-treat (including all follow-up regardless of changes to treatment) using binomial regression (risk difference scale) to estimate risk differences for binary outcomes, Kaplan-Meier and Cox regression for time to first-line failure (with competing risks methods for its components, primary failure and VL rebound), and generalised estimating equations with independent working correlation for global tests of repeated measures (adjusted for baseline for continuous measures). A per-protocol analysis included patients receiving >90% and <100% of the prescribed duration of first-line treatment and where the difference between screening and enrolment HCV RNA values would have led to a difference of ≤2 days in allocated duration of DAAs had they been allocated to the variable-duration group. Primary analyses of outcomes restricted to first-line therapy were stratified by first-line DAA strategy in place (VUS1 (before 1 April 2017) or VUS2 (after 1 April 2017)) as a main effect, and as an interaction with randomised group (fixed-duration vs variable-duration), where the p-value for the interaction term was <0.05. Primary analyses of outcomes including retreatment were unstratified, reflecting the overall strategy comparison and because no patients failed after receiving retreatment. Analyses used Stata v15.1. No adjustment was made for multiple testing. All subgroups within the subgroup analyses were pre-specified in the protocol. Further information on statistical methods and results can be found as extended data 17 .
Baseline characteristics were broadly balanced between randomised groups (Table 3) and across strategies (before vs after 1 April 2017). Median screening and enrolment HCV VL were 711423 and 741946 IU/mL in samples taken a median (IQR) 19 (13,33) days apart. Whilst the median (IQR) difference was 0.01 (-0.19,+0.21) log 10 IU/mL, absolute differences were greater (Figure 4), leading to 57 (28%) participants being excluded from the per-protocol analysis because they would have received a difference of ≥3 days of DAAs had this been determined by the enrolment rather than the screening VL (31 (23%) VUS1 vs 26 (39%) VUS2, because the second strategy received more drug overall).

Follow-up and treatment received
Each pre-or post-first-line EOT visit was missed by no more than eight (4%) participants. One participant was lostto-follow-up at day-28 on first-line and three (all randomised to fixed-duration, two with ribavirin) stopped first-line treatment >1 week early. All other participants completed first-line treatment ( Figure 5). Adverse events caused one participant to stop both first-line DAAs and ribavirin three days early    Excludes one patient lost-to-follow-up during first-line whose outcome on first-line is unknown.
The difference in first-line SVR12 between fixed-duration vs variable-duration was significantly smaller in 2 of 16 subgroups pre-specified in the protocol, suppression at day-7 or day-14 (heterogeneity p=0.02, 0.03 respectively) (Figure 7(b) and Figure 8). Considering the time when individuals first became undetectable , all 10 individuals who became undetectable at day-3 of treatment achieved first-line SVR12 regardless of treatment duration (as did 31/38 (82%) of those first undetectable at day-7) (Figure 7(c)). In the ribavirin randomisation, no subgroup had a difference that was significantly larger or smaller (Figure 9).
Corresponding to the timing of failures, the percentages with undetectable VL decreased more rapidly post-EOT in VUS1 vs VUS2 (p<0.0001) or VUS1 vs fixed-duration (p<0.0001) (Figure 12(a)). There was no evidence of differences in the proportions with detectable VL at EOT in VUS1 compared to VUS2 (p=0.33) or fixed-duration (p=1.00). There was no evidence that ribavirin was associated with higher percentages detectable post-EOT overall (p=0.48) or, within the variable-group, between VUS1 vs VUS2 (interaction p=0.17, Figure 12(b)).

Emergence of resistance
In total, 14 participants who failed on first-line treatment developed a new resistance mutation (not present at baseline) to at least one of their prescribed drugs ( Table 4, Table 5). Within paired samples available at baseline and failure, there was no evidence of a difference in emergent resistance between fixed-duration and variable-duration (3/10 (30%) vs 11/46 (24%, respectively, p=0.77). However, ribavirin was associated with lower emergence of resistance to first-line drugs (3/27 (11%) with vs 11/29 (38%) without ribavirin, p=0.01; Figure 13) with significantly lower rates of resistance to any DAA and to NS5a inhibitors (both p=0.01).

Discussion
This large strategic post licensing trial demonstrated overall non-inferiority of strategies using first-line ultrashort treatment durations, with both variable duration and fixed duration groups achieving 100% SVR12 rate after retreatment. The initial shortening strategy (VUS1) was able to cure only 36% of participants first-line but, strikingly, a relatively small increase in ultrashort treatment duration (from a mean of 32 days to 39 days) resulted in a doubling of SVR12 rates (from 36% to 72%). The 8-week fixed-duration strategy, with an SVR12 of 91%, did not have higher efficacy than previous phase II trials of shorter treatment courses 19 , despite limiting the maximal baseline viral load in those enrolled, which might have been expected to reduce the risk of failure. The trial's findings suggest a high proportion of patients can be cured with, on average, approximately 60% of the licensed duration of first-line therapy with the agents used in the trial (VUS2 strategy). However, first-line cure rates were not sufficient to be routinely recommended for stable patients able to adhere to 8-12 weeks' therapy.
Previous work has found adherence to DAA therapy declines as treatment progresses, with patients citing "feeling as if the treatment is working" as a reason for decreasing adherence 5 . Despite trial participants generally being considered to have better adherence, we found similarly decreasing adherence with time on first-line, with 28% reporting missed first-line doses, and poorer adherence to retreatment (despite its 100% SVR12) ( Figure 5). That we observed 72% SVR12 with VUS2 (mean duration 39 days) despite 28% reporting missing doses, suggests that intermittent non-adherence may be less important than overall adherence during weeks 4 to 8 of first-line treatment, and emphasise the importance of supporting adherence after week 4 of therapy to ensure good cure rates in hard-toreach populations 10,20 .
The risk of virological failure, in a patient unlikely to complete the recommended treatment course, is a clinical concern. Emergent resistance could compromise retreatment, particularly where retreatment does not include a protease inhibitor (as in this trial, in contrast to licensed retreatment options). This is also an ethical consideration in short-course therapy trials. We found the first-line treatment strategy did not compromise participants' ability to ultimately achieve SVR12. Whether this would be the case with pan-genotypic first-line treatment regimens remains to be tested, although it seems plausible. The 100% SV12 rates for retreatment is reassuring from an ethical perspective and suggests that, in certain circumstances, the combination may represent a viable retreatment option for patients failing therapy where access to licensed retreatment options (such as sofosbuvir/velpatasvir/voxilaprevir) remains limited.
Although ribavirin side-effects, particularly anaemia and fatigue, increasingly limit its use 2,12 it still has a role for some patients 13 with limited evidence that it can increase efficacy 21 and in vitro evidence that it may reduce the emergence of resistance with short course therapy 22 . Here, additional ribavirin was well tolerated, with only 2-4% participants experiencing adverse events. Across randomised groups, there was no evidence of improvement in SVR12. However, the emergence of resistance was significantly lower in those failing therapy (12% with ribavirin v 38% without), the first time this has been demonstrated in a randomised trial. This suggests that adjunctive ribavirin may have a role for some patients considered at high risk of not completing therapy in order to reduce the risk of compromising retreatment and further work is required to understand better the mechanism of action in this setting.
Response-guided therapy, shortening treatment based on initial virological response, was commonplace for interferon-based therapy 23 but this approach is not currently recommended         for DAA therapy 24,25 . Our findings suggest, for the first time, that very early responses to treatment (undetectable at day-3, or even day-7) may be helpful in predicting success of shortened treatment courses. Whilst not widely applicable, in specific supervised clinical settings (including in-patients, prisoners or directly observed daily therapy in the community) such an approach may help guide management and deserves further confirmation in prospective studies [25][26][27] .
This trial was designed to test treatment strategies, rather than specific regimens. Almost all recruitment happened when ombitasvir, paritaprevir, dasabuvir and ritonavir (Viekirax, Abbvie) was the preferred first-line treatment in the UK National Health Service (NHS). This combination remains a recommended NHS option, part of the WHO Essential Medicines List and is used in a number of countries. However, its use has been superceded by pangenotypic options in many settings and the extent to which these findings can be generalised to other combinations with broader genotype coverage is unknown. Give the similar declines in HCV VL between this and other DAA combinations (Figure 12), it seems plausible that the relationship between treatment duration and SVR12 is similar for other DAAs combinations approved for 12 weeks for patients with mild disease.
The trial required a population able to adhere to a schedule with significantly more visits than standard of care. Nonattendance was low and self-reported adherence reasonably high ( Figure 5), despite one-third participants actively using recreational drugs. Following expanded access to DAA therapy, trial recruitment completed short of its original target when there were very few patients in need of treatment able to adhere to the follow-up schedule. However, higher than anticipated success of retreatment (predicted to be 85%, actually 100%) and lower than expected loss to follow-up meant that the trial was able to demonstrate non-inferiority according to its pre-specified margin, providing confidence that either strategy would result in an overall SVR12 rate of at least 96% (higher than originally specified).
The treatment of individuals who are unlikely to complete recommended treatment courses is crucial for elimination strategies. Our findings suggest that ultrashort-courses of treatment can cure a significant proportion of patients with mild liver disease, without compromising retreatment in those not cured. Additional ribavirin in those unlikely to complete a course of treatment may be helpful to prevent the emergence of resistant virus. Statistical analysis was performed by LMc. Sequencing and bioinformatic analyses were performed by DN, DS, AA. The first draft was written by GC and ASW with review and comment from all authors.