Naturally Occurring Resistance-Associated Variants of Hepatitis C Virus Protease Inhibitors in Poor Responders to Pegylated Interferon-Ribavirin

Abstract

The pretherapeutic presence of protease inhibitor (PI) resistance-associated variants (RAVs) has not been shown to be predictive of triple-therapy outcomes in treatment-naive patients. However, they may influence the outcome in patients with less effective pegylated interferon (pegIFN)-ribavirin (RBV) backbones. Using hepatitis C virus (HCV) population sequence analysis, we retrospectively investigated the prevalence of baseline nonstructural 3 (NS3) RAVs in a multicenter cohort of poor IFN-RBV responders (i.e., prior null responders or patients with a viral load decrease of <1 log IU/ml during the pegIFN-RBV lead-in phase). The impact of the presence of these RAVs on the outcome of triple therapy was studied. Among 282 patients, the prevalances (95% confidence intervals) of baseline RAVs ranged from 5.7% (3.3% to 9.0%) to 22.0% (17.3% to 27.3%), depending to the algorithm used. Among mutations conferring a >3-fold shift in 50% inhibitory concentration (IC₅₀) for telaprevir or boceprevir, T54S was the most frequently detected mutation (3.9%), followed by A156T, R155K (0.7%), V36M, and V55A (0.35%). Mutations were more frequently found in patients infected with genotype 1a (7.5 to 23.6%) than 1b (3.3 to 19.8%) (P = 0.03). No other sociodemographic or viroclinical characteristic was significantly associated with a higher prevalence of RAVs. No obvious effect of baseline RAVs on viral load was observed. In this cohort of poor responders to IFN-RBV, no link was found with a sustained virological response to triple therapy, regardless of the algorithm used for the detection of mutations. Based on a cross-study comparison, baseline RAVs are not more frequent in poor IFN-RBV responders than in treatment-naive patients and, even in these difficult-to-treat patients, this study demonstrates no impact on treatment outcome, arguing against resistance analysis prior to treatment.

INTRODUCTION

Direct-acting antiviral agents (DAAs) (1) targeting the nonstructural 3 (NS3)/4A protease, the NS5A protein, or the NS5B RNA-dependent RNA polymerase of hepatitis C virus (HCV) are increasingly used in the treatment of chronic hepatitis C, either as part of triple combination therapies (triple therapy) with pegylated interferon (pegIFN) and ribavirin, or in combination with several other DAAs in an IFN-free regimen (2, 3). Due to the high rate of viral turnover and the error-prone activity of the HCV polymerase, HCV replication results in the constant production of numerous variants that are selected to constitute the viral quasispecies. Among them, resistance-associated variants (RAVs) that confer resistance to DAAs are likely to be naturally present before treatment and, when present in high and detectable amounts, might alter the result of DAA-containing therapies (4).

Using population sequence analysis (i.e., direct sequencing), baseline RAVs against NS3/4A protease inhibitors (PIs) telaprevir and boceprevir have been detected in 2 to 28% of treatment-naive patients in previous studies (1, 5–11). During triple therapies combining pegIFN and ribavirin with telaprevir or boceprevir, the presence of preexisting RAVs at baseline did not decrease the sustained virological response (SVR) rates (rates of infection cure) in patients who naturally responded to pegIFN-ribavirin; however, lower SVR rates have been observed in patients with baseline RAVs who were also poor pegIFN-ribavirin responders. In pooled phase II and III boceprevir studies, a lower SVR rate was observed in poor IFN responders with baseline RAVs than in those without baseline RAVs (23% versus 34%, respectively; P = 0.002). In this population, the presence of mutations conferring a >3-fold shift in the concentration needed to inhibit HCV replication by 50% in vitro (IC₅₀) for telaprevir or boceprevir (V36M, T54S, V55A, or R155K) at baseline was associated with non-SVR in boceprevir-treated patients (12). Moreover, in the REALIZE study with telaprevir, no prior null responders with the preexisting variants T54S or R155K achieved an SVR (13).

This study was performed in a real-life multicenter cohort, including a large number of patients receiving pegIFN-ribavirin plus telaprevir or boceprevir triple therapy who were either null responders to a prior course of pegIFN-RBV or poor responders (<1 log IU/ml viral load decrease) during a 4-week dual-therapy lead-in phase. Our goal was to describe the prevalence of protease inhibitor RAVs prior to therapy in this patient population and to investigate the impact of these mutations on the SVR to triple therapy.

MATERIALS AND METHODS

Patients.

Analyses were performed on pretreatment prospectively collected and retrospectively analyzed plasma samples from a multicenter cohort of 282 patients with chronic hepatitis C treated with pegIFN-ribavirin and either boceprevir or telaprevir triple therapy in 22 French university hospitals.

Sixty-four patients started treatment in early 2011 within the framework of French temporary authorizations for the French Early Access Programme (ANRS CO20-CUPIC) observational cohort (14). The other treatments were started between July 2011 and April 2013 after full marketing authorizations were obtained for the use of these two anti-HCV protease inhibitors, according to the French clinical practice guidelines (15).

The main inclusion criteria were a poor response to IFN-RBV (i.e., a null response to a prior course of pegIFN-α-RBV dual therapy or a viral load decrease of <1 log IU/ml during the dual-therapy lead-in phase of 4 weeks) and the availability of a frozen plasma sample taken at triple-therapy baseline (<6 months before the start of triple therapy). Exclusion criteria were any prior treatment, including an HCV protease inhibitor, HIV or hepatitis B virus (HBV) coinfections, and withdrawal from triple therapy due to adverse effects.

The study was performed according to the French and European biomedical ethics recommendations, including ethics committee approval (CECIC Rhône-Alpes-Auvergne, Clermont-Ferrand, institutional review board [IRB] no. 5891), and written patient consent was obtained for the use of samples for research purposes.

Amplification and sequence analysis of the HCV NS3/4A region.

Each of the 22 participating laboratories performed its own PCR amplification and sequence analysis of HCV NS3/4A using either the ANRS protocol for genotype 1 (n = 13 laboratories) (11), the pangenotypic NS3 amplification described by Besse et al. (n = 5 laboratories) (7), or their own laboratory method (n = 4 laboratories). All centers but one participated in the ANRS NS3 quality control, which was recently reported (16).

Data collection.

The data set for this study was collected using an adapted version of the ANRS Greg+ software. This software is freely available for routine analysis of HIV, HCV, and HBV sequences and was available for download by the participating laboratories for local use. Each participating center selected patients matching the inclusion criteria of the study from their local database and sent the anonymized epidemiological, clinical, and virological data to a centralized server working as a shared platform that was accessible to all participants.

RAV analysis.

The NS3 sequences were aligned with the HCV-1a strain H77 reference sequence (AF009606). As no algorithm for the interpretation of HCV resistance has been recommended in European, American, or Asian guidelines, we selected three regularly used algorithms for NS3 RAV detection. Algorithm 1 has been described by the HCV Drug Development Advisory Group of the Forum for Collaborative HIV Research, University of California Berkeley, Berkeley, CA, USA (http://www.hivforum.org/). Algorithm 2 integrates more mutations and was extracted from the website Geno2Pheno (http://hcv.geno2pheno.org/), developed by the Max-Planck-Institut für Informatik, Saarbrücken, Germany. Algorithm 3 was restricted to mutations that confer a >3-fold shift in HCV replicon activity (V36M plus T54S, V55A, R155K, and A156T/V) and has been used in clinical trials with boceprevir (1, 17–19). Table 1 summarizes the considered mutations for these three algorithms.

TABLE 1.

Variants that were considered with each RAV counting algorithm

Algorithm	Variant(s) at aa position:
	36	41	43	54	55	80	87	109	117	132	138	155	156	158	168	170	174	175
1 (HIV forum)	A/G/I/L/M			A/S/G/C	A					V		G/K/M/T	F/N/S/T/V	I	N	A/T		L
2 (Geno2Pheno)	A/G/I/L/M	R	C/I/V/S		A	G/H/K/L/R	T	K	H	V	T	G/I/K/M/Q/T	F/G/N/S/T/V		A/E/G/H/I/N/T/V/Y	A/T	F
3 (BOC clinical trials)	M			S	A							K	T/V

Phylogenetic analysis.

Phylogenetic analysis was performed after alignment of the 282 NS3 sequences (550 nucleotides [nt], positions 3420 to 3970, according to H77 numbering) downloaded by the 22 centers in the Greg+ NRMUT software with 10 genotype 1 HCV NS3 reference sequences using FFT-NS-i (20, 21). The phylogenetic tree was constructed by means of the neighbor-joining method, including all gap-free sites and a Jukes-Cantor substitution model using the MAFFT version 7 online software (http://mafft.cbrc.jp/alignment/software/) (22). The reliability of the various inferred clades was estimated by bootstrapping (1,000 replicates). Visualization of the tree and branch and node coloring were performed using Archaeopteryx (23).

Statistical analysis.

The study population is described using frequencies for categorical variables with a 95% confidence interval, and means ± standard deviations were used for continuous variables or medians and the interquartile range were used for non-Gaussian continuous-level variables.

The prevalence of mutations (primary endpoint) with its 95% confidence interval is shown for each algorithm (Table 2). To assess the secondary endpoints, univariate analyses were performed to describe: (i) the risk of no SVR to triple therapy associated with each mutation, and (ii) prognostic factors of no SVR to triple therapy. Multivariate analysis (logistic regression) was performed to take significant confounding factors into account. Continuous data were compared using a t test if the variable was normally distributed or the Mann-Whitney test for nonparametric variables. The chi-square test (Fisher's exact test if necessary) was used for categorical variables.

TABLE 2.

Prevalence of NS3 RAVs by means of the three decisional algorithms, and SVR rates according to the viral subtype

HCV genotype 1 subtype by algorithm used	Prevalence of RAV (n = 282) (% [95% CI])a	SVR rate (n = 282)b
Algorithm 1	8.5 (5.5–12.4)	45.8
1a	11.8 (7.3–17.8)	9/19 (47.3)
1b	4.1 (1.4–9.4)	2/5 (40)
Algorithm 2	22.0 (17.3–27.3)	38.3
1a	23.6 (17.3–30.9)	14/37 (37.8)
1b	19.8 (13.1–28.1)	9/23 (39.1)
Algorithm 3	5.7 (3.3–9.1)	50
1a	7.5 (3.9–12.7)	6/12 (50)
1b	3.3 (0.9–8.2)	2/4 (50)

^aCI, confidence interval.

^bThe SVR data are presented as the no. of patients with SVR/total no. of patients (%), or simply with the percentage.

Statistical significance was set at a P value of ≤0.05. All statistical analyses were performed using Stata SE version 11.0 software (StataCorp LP, TX, USA).

RESULTS

Prevalence of NS3 RAVs at baseline in poor IFN-RBV responders.

Two hundred eighty-two patients met the inclusion criteria; 161 were infected with genotype 1a, and 121 were infected with genotype 1b. Among the 282 patients, 227 were null responders to a prior course of pegIFN-α-RBV, whereas the remaining 55 patients had experienced an HCV RNA level decrease of <1 log IU/ml during a 4-week dual-therapy lead-in phase. Fifty-eight percent of patients (159/274) were cirrhotic. When retreated with triple therapy, 152 patients received telaprevir, and 130 patients received boceprevir. About one-third of the subjects (38%) achieved an SVR.

In this difficult-to-treat population, RAVs were found at baseline in 5.7% (3.3% to 9.1%) to 22.0% (17.3% to 27.3%) of patients, depending on the algorithm used, and frequencies varied depending on the viral subtype (Table 2).

Using the most restrictive algorithm (algorithm 3), the most commonly found mutations were T54S (11/16) in 7 viral genotype 1a and 4 viral genotype 1b patients. Two R155K mutations were observed in the genotype 1a group in one patient with an SVR and one patient without an SVR treated with telaprevir and boceprevir, respectively. Two strains also carried the A156T mutation, either alone in a genotype 1a patient who did not respond to triple therapy with telaprevir or in association with the T54S mutation in a genotype 1b patient who responded to triple therapy with boceprevir. The V36M and V55A single mutations were also observed in two nonresponding patients (Table 3). By combining all algorithms, double mutants were detected in eight patients: the association of A156T with T54S cited above, a double R155K/V158I substitution in a genotype 1a telaprevir-treated SVR patient, a V36L/T54S mutant in a boceprevir-treated non-SVR patient, and five genotype 1a patients with the double T54S/V55I substitution (4 SVR and 1 non-SVR). The Q80K mutation was found in 19 (6.7%) patients (all with genotype 1a), and 6 more patients had another amino acid substitution at this position (4 Q80L and 1 Q80G in genotype 1b patients and 1 Q80H in a genotype 1a patient).

TABLE 3.

Characteristics of the patients with detectable NS3 RAVs at baseline with algorithm 3

Patient no.	Mutation(s)	HCV subtype	PI useda	Change in IC50 (fold change)b	Fibrosis stage	IL28B	Baseline viral load (log IU/ml)
SVR+
P1	T54S	1b	TPV	4.2*	F3	TT	6.49
P1P2	T54S	1a	TPV	4.2*	F4	CT	6.6
P1P3	R155K	1a	TPV	10**	F4	—c	6.44
P1P4	T54S	1a	TPV	4.2*	F4	CC	6.95
P1P5	T54S	1a	BOC	8.5*	F4	—	5.72
P1P6	T54S + A156T	1b	BOC	8.5 and 65**	F4	—	7.1
P1P7	T54S	1a	BOC	8.5**	F4	—	3.71
P1P8	T54S	1a	TPV	4.2*	—	—	5.68
SVR−
P1P9	T54S	1b	BOC	8.5**	F4	—	3.8
P1P10	V36M	1a	BOC	1.8**	F4	—	4.66
P1P11	R155K	1a	BOC	4.7**	F4	—	5.79
P1P12	T54S	1a	BOC	8.5**	F4	—	5.74
P1P13	T54S	1a	BOC	8.5**	F3	CT	6.45
P1P14	V55A	1a	TPV	2.7***	F4	—	6.1
P1P15	T54S	1b	TPV	4.2*	F2	TT	6.05
P1P16	A156T	1a	TPV	>62*	F4	TT	7.93

^aPI, protease inhibitor; TPV, telaprevir; BOC, boceprevir.

^b*, reference 12; **, reference 15; ***, reference 21.

^c—, missing data.

Factors linked with the presence of RAVs.

According to algorithm 1, subtype 1a was associated with a higher prevalence of RAVs (P = 0.030). When using algorithm 2, the clonal complex (CC) interleukin 28B (IL28B) genetic polymorphism was associated with a lower percentage of RAV. Similar nonsignificant trends were observed with the other algorithms. No other sociodemographic or viroclinical characteristics were significantly associated with a higher prevalence of RAVs (Table 4).

TABLE 4.

Sociodemographic and viroclinical characteristics of the study population according to the presence or not of detectable NS3 RAVs with the three decisional algorithms

Characteristic	Data using algorithm:						Total population
	1		2		3
	Mutation (n = 24)	No mutation (n = 258)	Mutation (n = 62)	No mutation (n = 220)	Mutation (n = 16)	No mutation (n = 267)
Age (mean ± SD) (yr)	53 ± 7	55 ± 10	54 ± 9	55 ± 10	54 ± 7	55 ± 10	55 ± 9
Male (%)	62.5	70.5	67.7	70.5	56.3	70.7	69.9
Viral type (%)
1a	79.2a	55.0a	61.3	55.9	75.0	56.0	57.1
1b	20.8	45.0	38.7	44.1	25.0	44.0	42.9
Fibrosis stage (%) (n = 274)
Stage 0–3	34.8	42.6	45.9	40.9	26.7	42.9	42.0
Stage 4	65.2	57.4	54.1	59.2	73.3	57.1	58.0
Protease inhibitor used (%)
BOC	37.5	46.9	50.0	45.0	50.0	45.9	46.1
TVP	62.5	53.1	50.0	55.0	50.0	54.1	53.9
IL28B (n = 133)
CC	10.0	7.3	21.4b	3.8b	16.7	7.1	7.5
Non-CC	90.0	92.7	78.6	96.2	83.3	92.9	92.5
ALT (IU/liter) (mean ± SD)	102 ± 74	104 ± 91	114 ± 139	101 ± 72	108 ± 80	103 ± 91	104 ± 90
Duration of infection (mean ± SD) (yr) (n = 109)	25.5 ± 12	25.5 ± 12	24.3 ± 14	25.8 ± 11	24.5 ± 10	25.6 ± 12	25.5 ± 12
Viral load at wk 12
Detectable	50.0	48.5	50.0	48.2	50.0	48.5	51.4
Undetectable	50.0	51.5	50.0	51.8	50.0	51.5	48.6
Viral load at wk 4 (n = 231)
Detectable	72.2	78.4	82.2	76.9	80.0	77.8	77.9
Undetectable	27.8	21.6	17.8	23.1	20.0	22.2	22.1

^aP = 0.030.

^bP = 0.002.

Phylogenetic tree analysis of the NS3 region showed no clustering of sequences carrying NS3 RAVs, suggesting that the presence of naturally occurring RAVs at detectable levels was not influenced by the transmission of resistant viral strains (Fig. 1). Subtypes 1a and 1b were clearly separated on the tree. The NS3 sequences from SVR patients did not cluster distinctly from those from patients who did not achieve an SVR, indicating that no specific genetic pattern was associated with therapeutic failure.

FIG 1.

Phylogenetic tree comparing the 282 NS3 sequences from our study population with 10 reference strains. A 550-nt long fragment was analyzed (nt 3420 to 3970, according to H77 numbering). Phylogenetic analysis was conducted with MAFFT (http://mafft.cbrc.jp/alignment/server/) (22) using the neighbor-joining method, the substitution model of Jukes-Cantor, and a bootstrap resampling of 1,000. Branches are colored purple for subtype 1b and blue for 1a. Nodes are shown in green when at least one mutation was detected (according to algorithm 3) and the patient achieved an SVR, in yellow when a mutation was detected and the patient did not achieve an SVR, and in red when no mutation was detected and the patient did not achieve an SVR.

Impact of baseline RAVs on failure to achieve an SVR.

Regardless of the algorithm used, including the most restrictive one, the presence of RAVs at baseline in these poor responders to IFN-RBV did not impact the SVR (proportion of patients with or without an SVR, respectively, with detectable substitutions at baseline are 10.2% versus 7.5%, P = 0.4 with algorithm 1; 21.3% versus 22.4%, P = 0.8 with algorithm 2; 7.4% versus 4.6%, P = 0.3 with algorithm 3) (Fig. 2). Among all the tested parameters, only subtype 1b (54.6% versus 35.6%, respectively), an undetectable HCV RNA at week 12 (87.0% versus 29.3%, respectively), and the use of telaprevir (63.9% versus 47.7%, respectively) were significantly associated with a higher rate of SVR in univariate analysis.

FIG 2.

Prevalence of mutations at baseline according to the algorithm used for analysis, and the virological outcome (SVR or no SVR) of triple therapy using a protease inhibitor. For each algorithm, the open circles correspond to the prevalence of mutation(s) with the 95% confidence interval for each algorithm for patients achieving a sustained virological response to triple therapy. The closed circles correspond to the prevalence of mutation(s) at baseline for patients who responded to triple therapy.

In multivariate analysis, only subtype 1b and the undetectability of HCV RNA at week 12 were significantly associated with an SVR to triple therapy (Table 5).

TABLE 5.

Multivariate analysis of the factors associated with a lack of SVR to triple therapy with the mutation rates provided by the three algorithms

Factor	Algorithm 1a		Algorithm 2		Algorithm 3
	ORadj	95% CI	ORadj	95% CI	ORadj	95% CI
Mutation	0.52	0.17–1.60	1.12	0.53–2.36	0.47	0.12–1.77
Viral type 1b	0.38	0.19–0.76	0.40	0.20–0.79	0.38	0.19–0.77
Undetectable viral load at wk 12	0.05	0.02–0.10	0.05	0.02–0.10	0.05	0.02–0.10
Male sex	1.14	0.55–2.37	1.17	0.57–2.44	1.13	0.54–2.35
Age	1.00	0.96–1.04	1.00	0.97–1.05	1.00	0.97–1.04
Cirrhosis stage F4	1.42	0.76–2.67	1.38	0.74–2.58	1.43	0.76–2.69
TVP used	1.22	0.63–2.38	1.20	0.62–2.33	1.17	0.60–2.29

^aOR_adj, adjusted odds ratio; CI, confidence interval.

Effect of RAVs on baseline viral load.

On examining viral replication levels in patients carrying dominant RAVs according to the broadest algorithm (i.e., algorithm 2), 208 patients (73.8%) displayed viral loads in the range of >500,000 to 8.5 × 10⁷ IU/ml (Fig. 3), including 1 patient with an R155K substitution and 7 out of the 8 patients carrying a double mutant. This suggests that drug-resistant strains were not necessarily impaired in their ability to replicate in vivo (P = 0.912).

FIG 3.

Baseline viral loads in patients with and without NS3 RAVs, according to algorithm 2. The no-mutation group included 220 patients, and the mean ± standard deviation (SD) viral load was 6.03 ± 0.05 log IU/ml. The mutation group contains 62 patients, and the mean ± SD viral load was 5.96 ± 0.13 log IU/ml. The estimated P value (Mann-Whitney U test) shows no significant difference between the two groups (P = 0.9121).

DISCUSSION

In this large real-life multicenter cohort of IFN-RBV null-responder patients exhibiting advanced stages of fibrosis and a long history of infection (median, 28 years), the prevalence of NS3 RAVs at baseline ranged from 5.7% to 22.0%, depending on the interpretation algorithm used. These results are in the same range as those reported in a treatment-naive population, although the lack of a reference algorithm hinders comparisons between studies. Using the broad Geno2Pheno algorithm (algorithm 2 in our study), another French team reported 19% of RAVs in 63 naive patients, whereas we found 22% in our poor responders to IFN-RBV (7). In clinical studies with telaprevir, the prevalence of telaprevir-resistant variants at baseline was reported to be 2 to 3.5% in treatment-naive patients, including <1% of patients carrying the V36M or R155K mutation (5, 6). Analysis of phase 3 telaprevir trials in patients with prior treatment failure also showed 3% of RAVs at baseline (n = 652) and 2.7% (n = 185) in the subgroup of null responders. Using our most restrictive algorithm, we found a slightly higher rate of RAVs (5.3% [n = 282]), but they were not as rare as V36M (0.4%) and R155K (0.7%) mutants were in previous studies. T54S was the most prevalent variant in all studies, occurring in 2.6% of treatment-naive patients, 1.6% of null responders in telaprevir studies (5), and in 3.9% of our population.

Another large European multicenter study has reported an RAV prevalence of 8.6% in genotype 1a and 1.6% in genotype 1b patients (1). Using a similar algorithm (algorithm 1), we also found a significant difference between subtypes 1a and 1b but with higher prevalence rates (11.8% for genotype 1a and 4.1% for genotype 1b). The genetic barriers to resistance to the first-wave, first-generation protease inhibitors (telaprevir or boceprevir) have been reported to be lower in genotype 1a than those in genotype 1b viruses, explaining the low SVR rate in genotype 1a. This can be explained by the fact that only one transition is needed for substitutions V36M and R155K in subtype 1a versus two nucleotide changes (one transition plus one transversion) for these substitutions in the consensus 1b sequence (24). Although more frequently found in subtype 1a (8/11), the most common mutation in our study was the T54S mutation, which occurs with one transversion whatever the codon of origin (mostly ACT in 1a and ACC in 1b). This suggests that other parameters may influence the differential mutation rate between the genotype 1 subtypes.

A second-wave first-generation protease inhibitor, simeprevir, is currently commercialized. Detection of the Q80K polymorphism at baseline is associated with a poorer response to regimens including this drug, and the manufacturer recommends checking its absence before starting triple therapy (25). In the European population of our study, 19 patients (6.7%) infected with HCV subtype 1a would not have been eligible for the combination of simeprevir, pegIFN, and ribavirin.

Previous telaprevir and boceprevir clinical trials suggested that the presence of baseline RAVs does not influence the response to triple therapy in treatment-naive patients. However, a relationship was reported in boceprevir studies when the analysis was restricted to poor IFN responders, particularly when only mutations conferring a >3-fold shift in the replicon assay were considered in the analysis (5, 12). In our HCV-infected cohort restricted to poor IFN-RBV responders, we observed no relationship between the presence of baseline NS3 RAVs and protease inhibitor-based triple-therapy outcomes. Unlike clinical trials driven by the manufacturers, in which the determination of RAV prevalence and its impact on SVR were not the primary goals, our study was specifically designed with the aim of assessing the influence of baseline RAVs on the virological response. It included real-life patients treated with either telaprevir (54%) or boceprevir (46%). The significant disequilibrium in SVR according to the protease inhibitor used (36% with boceprevir and 64% with telaprevir; P = 0.008) might explain the discrepancy with the boceprevir clinical trial results described by Barnard et al. (12). Indeed, of 16 patients in our study presenting with baseline RAVs according to the same algorithm as that of Barnard et al. (12), 5/8 treated with telaprevir achieved an SVR, whereas only 3/8 treated with boceprevir did.

Some authors suggested a more subtle implication of certain types of mutations, such as V36A/M or R155K/T/Q, in the failure of telaprevir-based treatments (26). Despite the selection of poor responders to IFN-RBV, this association was not found in our study, suggesting that the presence of these mutations at baseline was not the unique reason for the previously described failures. Moreover, in our study, two patients were infected with strains carrying the A156T mutation, which confers a very high level of resistance in the replicon system (65- to 75-fold and 105- to 112-fold increase in IC₅₀s for boceprevir and telaprevir, respectively) (19). One patient did not respond to triple therapy with telaprevir, and the other achieved an SVR with boceprevir. Their respective baseline viral loads were 7.9 and 7.1 log IU/ml, suggesting that despite a high IC₅₀ fold change, this mutation did not impair viral fitness. Other mutations detected using algorithm 2 did not affect the plasma HCV viral load, as already described by Kuntzen et al. (1). As a result, our study demonstrates that the presence of RAVs at baseline, even when detected using population sequence analysis or when they are associated with a previously reported large increase in IC₅₀, is not sufficient in itself to induce treatment failure. The observation of the ineffectiveness of IFN-RBV dual therapy in the patients in our study might be helpful when making treatment strategy decisions for other DAAs used in IFN-free combinations. Nevertheless, differences in the mechanism of viral inhibition between the different classes of drugs necessitate the realization of other studies.

This study has some limitations: (i) the number of patients found exhibiting baseline RAVs is quite low, depending on the algorithm used, and does not allow us to draw strong conclusions concerning the effect of a given mutation (e.g., R155K or A156T); (ii) the IC₅₀ phenotypic data were not assessed in the patients in this study but were extrapolated from other studies; (iii) the use of population sequencing in this study was less sensitive for the detection of RAVs than deep sequencing. Nevertheless, it can be assumed that if RAVs detected using population sequencing had no impact on the SVR rate, it is likely that minority RAVs potentially detected by deep sequencing also have no effect on SVR, even when associated with an already known large increase in IC₅₀.

In conclusion, our study in real-life patients treated with telaprevir or boceprevir shows that baseline NS3 RAVs are detected by population sequence analysis in 5.7% to 22.0% of IFN-RBV null responders and that their presence does not impact the virological outcome of triple therapy. Thus, resistance testing prior to therapy is not needed.