Overcoming the resistance to resistance testing: Collecting the data

The hallmark of a direct antiviral could be described as its ability to select for resistant variants. It was thus no surprise that resistance was a concern in the development of direct-acting antivirals (DAAs) for the treatment of hepatitis C virus (HCV) infection. What is perhaps a surprise is the reluctance of the HCV community to embrace resistance testing as part of the management of HCV. The progress in terms of virological cure rates, tolerability and simplicity of treatment with the new agents has been nothing short of remarkable; however, even with the highly effective therapies now available, some patients are not cured. With the enormous numbers of patients treated, even a very low treatment failure rate will generate significant absolute numbers of people who need retreatment. Understanding the patterns of resistance after treatment failure may inform the choice of first-line therapy and will almost certainly influence the optimal strategies for retreatment.

Like all viruses, HCV circulates as a swarm of related but distinct quasispecies that are generated by the error-prone polymerase that replicates the viral genome. Nucleotide substitutions occur randomly throughout the genome and most will have a detrimental effect on viral replication or fitness and thus be selected against. However, substitutions at specific sites in the coding region of viral proteins alter interactions between specific DAAs and their targets, reducing susceptibility of virions carrying these resistance-associated substitutions (RASs) to specific antivirals, a fitness advantage in the presence of therapy. These RASs may occur at baseline, but more frequently are promoted by DAA treatment, especially with agents with a low barrier to resistance due either to the number of substitutions necessary to generate resistance or the fitness level of the resulting variant. Both factors differ by DAA class and by HCV genotype/subtype with major impact on the frequency of RAS development and its consequences.

Part of the confusion surrounding RAS testing comes from reporting of clinical trial results. Studies differ in their methods (population vs deep sequencing), thresholds (1% vs 15%) and degree (two-fold vs 100-fold × EC₅₀) of assessing resistance and whether they report drug-specific or DAA-class-specific RASs by genotype or subtype. Lack of standardization has created major difficulty in interpreting data, with many studies selectively reporting the approach that presents results in the most favourable light, usually downplaying the significance of RASs. As a result, there is ongoing debate about the optimal role of RAS testing in clinical practice, which is further compounded by variable access to high quality RAS testing and reporting and even disagreement between guidelines on when RAS testing is recommended.1, 2

Just as real-world data have been very useful in confirming the efficacy and safety of DAA therapy, real-world RAS data may help clarify their significance. Unfortunately, to date, there are limited data on the profile of RASs that develop in real-world DAA failures. In this issue of Liver International, Di Maio and colleagues present data from 200 virological failures in 197 patients, for whom population sequencing for RASs was performed at treatment failure. A subset of 70 patients also had RAS testing of pre-treatment samples.

This real-world sample of DAA failures was mostly cirrhotic (66%) and treatment-experienced (57%). Peginterferon (IFN) was used in 45% (with a first-generation protease inhibitor (PI) [42%] or sofosbuvir (SOF) [3%]). Only 53 patients (30%) received regimens currently recommended by EASL guidelines, of whom the majority (62%) received SOF plus simeprevir (SIM) +/− ribavirin (RBV). Although relapse was the most common reason for virological failure (57%), breakthrough or non-response was frequently reported, including in over 25% of patients treated with IFN-free DAA regimens.

Resistance-associated substitutions were identified in 55 of 111 (49.5%) of those treated with DAAs alone compared to 65 of 89 (73%) who received a DAA plus IFN and differed depending on the DAA-treatment received. In those who received an NS5A inhibitor, 96% had NS5A RASs. In contrast, in those treated with PIs,70%-76% had NS3 RASs depending on whether they received IFN, and NS5B RASs were found in two of seven (29%) who received dasabuvir. Notably, four of six (67%) patients who received SOF plus IFN and 19 of 93 (20%) who received SOF without IFN had NS5B RASs reported at failure. Of these, three were the classical S282T RAS known to confirm SOF resistance in vitro, which results in a major impairment in replicative fitness, while the others were substitutions at position L159, C316 or S556 alone or in combination, which have been reported to emerge more frequently after SOF failure but are not associated with a significant shift in in vitro EC₅₀ or with SOF failure when present at baseline.3

Multi-class resistance was seen in almost half (47%) of those who received more than one DAA class. Notably, all of those who received a first-generation PI and NS5A inhibitor +/− RBV had both NS3 and NS5A RASs at failure. In seven patients who received three DAA classes, two had RASs in all three genes, two in two genes and two in one gene at failure. In contrast, in those who received an NS5A inhibitor or a PI with SOF, all had NS5A or PI RASs at failure but only two also developed the S282T SOF RAS. These data highlight the higher barrier to resistance of SOF and also suggest that as more robust regimens are used, complex resistance patterns may be less common at the time of failure.

Another interesting observation included the finding that, of 28 patients with known IL28B/IFNL4 genotype who received an IFN-free DAA regimen, zero of six with the favourable CC genotype had RASs at failure compared to 11 of 22 (50%) of those with a CT/TT genotype. This may reflect better innate immune function, and the ability to clear virus that might escape DAA therapy through RAS formation and is consistent with previous data showing that IL-28B CC is associated with improved viral kinetics and higher SVR rates with shortened therapy.4-6

As with most studies, this work raises as many questions as it answers. The authors should first be congratulated for comprehensively reporting their data. It is very helpful to have tables with specific regimens used and RASs identified in each patient. Unfortunately, the data were collected through an anonymized registry with limited clinical information, as it would be interesting to have more clinical data on many patients. The rate of breakthrough and non-response is much higher than reported in clinical trials with the regimens used and certainly raises questions about adherence but also could relate to the high proportion with cirrhosis. The authors partially attribute non-response to genotype misclassification, which they reported in 3% of patients and four of seven non-responders to IFN-free regimens based on the fact that sequencing at treatment failure identified a different genotype than reported at baseline. Although incorrect genotyping is a recognized limitation of commercial assays, other explanations include reinfection or mixed genotype infection with emergence of the non-dominant strain. The authors argue for genotype determination by sequencing in all patients to avoid misclassification. While this is debatable given access and cost issues, as well as the move towards pan-genotypic regimens, the data highlight the importance of re-genotyping after failure before retreatment.

The prevalence of SOF-induced NS5B RASs, especially S282T, is higher than in previous reports. The S282T variant is very unfit with replication efficiency <2% of wild-type virus. In a study of 282 patients who did not achieve SVR after SOF-based therapy, only one (0.3%) had detectable S282T at 4 weeks post-treatment and by 12 weeks, the variant was no longer detectable.3 In another report, of 51 patients with virological failure after ledipasvir (LDV)-SOF therapy, one patient (2%) who had a baseline NS5A RAS developed the S282T at failure.7 In this study, of the three with S282T, one received SOF monotherapy, while the other two received SIM/SOF and LDV/SOF and also developed NS3 and NS5A RASs respectively. Unfortunately, baseline samples were not available to know if the NS3 and NS5A RASs were present before therapy. Compensatory mutations may occur allowing the S282T variant to improve its fitness over time. This is more likely to occur with suboptimal suppression from reduced adherence (“real-world”) or if baseline RASs to other DAAs used are present, which is particularly relevant for retreatment if the same regimen is used, as was shown in the first SOF/LDV retreatment trial in which 25% of those with NS5A RAS at the start of retreatment developed SOF RASs at failure.8 The timing of RAS testing in this study was not specifically stated but it would be interesting to know if the SOF RASs persist long-term off therapy. Although studies have documented successful retreatment with SOF-based regimens after initial SOF failure,9, 10 it will be interesting to see how the three patients with S282T and those with the less well understood SOF “treatment-emergent” RASs, respond to retreatment.

Timing of testing is also important for other RASs. The fitness of treatment-emergent RASs varies considerably. Unlike SOF RASs, NS5A RASs, are very fit, and may persist for years.11 NS3 RASs and non-nucleotide NS5B RASs are less fit than NS5A RASs, but can remain detectable for several months to years post treatment depending on the context (eg genotype 1a vs 1b).12 Whether these “lost” RASs have any impact on subsequent retreatment is not known. The authors support the AASLD recommendations of RAS testing in all patients before retreatment after DAA failure. They emphasize the importance of sequencing all three genes (NS3, NS5A and NS5B) because baseline RASs were found in 20% with samples available but were not all specific for the treatment used. Most patients with baseline RASs developed new/additional RASs in NS3/NS5A at failure but all baseline RASs also persisted. With fit baseline RASs and new treatment-emergent RASs, complex multi-resistant variants may emerge after treatment failure.

The purpose of resistance testing at the time of failure is to help guide retreatment. Because retreatment data are currently limited, particularly in the “real-world”, strategies for retreatment are based on extrapolation from the effect of baseline RASs and in vitro EC₅₀ results, which clearly show that not all RASs are created equal. RASs should be considered for specific DAAs rather than DAA classes. The impact of an individual RAS can vary by orders of magnitude between DAAs of the same class, may vary depending on the presence of other RASs in the same gene, and is generally genotype and subtype-specific.13 These differences can have major clinical impact. In patients treated with elbasvir (NS5A)/grazoprevir (PI), only two of nine (22%) with genotype 1a infection who harboured baseline NS5A RASs with >5× EC₅₀ shift to elbasvir achieved SVR compared to 16 of 17 (94%) with genotype 1b infection with similar baseline NS5A RASs.14 Based on these data, many of the NS5A RASs identified in patients with genotype 1b in this study may not have as much clinical impact as one might first assume. On the other hand, as the authors note, the presence of a second NS5A RAS can greatly enhance the impact of a given RAS in a 1b background.15 For most clinicians, appreciating the subtleties of specific RASs and even worse, RAS combinations, is very difficult. With this in mind, RAS reporting should include specific EC₅₀ shifts by DAA per genotype/subtype but should also recommend preferred regimens based on guidelines rather than reporting whether resistance is “probable” or even worse “possible”, for a specific DAA.

Fortunately, recent data have shown that salvage regimens are likely to be available for most patients who fail DAA-based regimens. Results from the 2016 EASL and AASLD meetings showed that well over 90% of patients who had failed different DAA regimens were successfully retreated with combinations of next-generation DAAs.16-18 As with current regimens, we will need to see how these agents perform in tougher-to-treat real-world populations but hopefully they will make the need for RAS testing less and less important with time. That being said, until these regimens are available, studies like this one are critical for us to better understand what we are up against and whether resistance will be the menace we once feared.

Conflict of Interest

JJF reports receiving consulting fees and/or research support from Abbvie, Gilead, Janssen and Merck. AD has nothing to disclose.