Modeling viral kinetics and treatment outcome during alisporivir interferon-free treatment in hepatitis C virus genotype 2 and 3 patients


  • Potential conflict of interest: J.G. has consulted for Gilead; A.S.P. has received research grants from Novartis and Roche and has consulted for Merck, Gilead, Achillion, Bristol-Myers Squibb, Progenics, and Santaris Pharmaceuticals; S.K. was employed by Novartis when this work was done (currently at the Bill and Melinda Gates Foundation); and J.Y., M.L., B.L., and N.V.N. are Novartis employees.


Alisporivir (ALV) is a cyclophilin inhibitor with pan-genotypic activity against hepatitis C virus (HCV). Here, we characterize the viral kinetics observed in 249 patients infected with HCV genotypes 2 or 3 and treated for 6 weeks with different doses of ALV with or without ribavirin (RBV). We use this model to predict the effects of treatment duration and different doses of ALV plus RBV on sustained virologic response (SVR). Continuous viral decline was observed in 214 (86%) patients that could be well described by the model. All doses led to a high level of antiviral effectiveness equal to 0.98, 0.96, and 0.90 in patients treated with 1,000, 800, and 600 mg of ALV once-daily, respectively. Patients that received RBV had a significantly faster rate of viral decline, which was attributed to an enhanced loss rate of infected cells, δ (mean δ = 0.35 d−1 vs. 0.21 d−1 in patients ± RBV, respectively; P = 0.0001). The remaining 35 patients (14%) had a suboptimal response with flat or increasing levels of HCV RNA after 1 week of treatment, which was associated with ALV monotherapy, high body weight, and low RBV levels in patients that received ALV plus RBV. Assuming full compliance and the same proportion of suboptimal responders, the model predicted 71% and 79% SVR after ALV 400 mg with RBV 400 mg twice-daily for 24 and 36 weeks, respectively. The model predicted that response-guided treatment could allow a reduction in mean treatment duration to 25.3 weeks and attain a 78.6% SVR rate. Conclusion: ALV plus RBV may represent an effective IFN-free treatment that is predicted to achieve high SVR rates in patients with HCV genotype 2 or 3 infection. (Hepatology 2014;59:1706–1714)








direct-acting antivirals


half maximal effective concentration




hepatitis C virus


limit of detection






pegylated interferon








response-guided therapy


rapid virologic response


sustained virologic response




viral kinetic


treatment week


wild type

Chronic infection with hepatitis C virus (HCV) affects approximately 160 million people worldwide[1] and is the leading cause of cirrhosis, liver cancer, and liver transplantation.[2] The approval in 2011 of two protease inhibitors, telaprevir and boceprevir, in combination with pegylated interferon-alpha and ribavirin (peg-IFN-α/RBV), has marked an important milestone in improving anti-HCV therapy for patients with HCV genotype 1 (G1).[3, 4] However, these protease inhibitors designed to treat HCV G1 have modest or no effectiveness against other genotypes, particularly HCV genotypes 2 and 3 (G2 and G3),[5] and Peg-IFN/RBV remains the only licensed treatment for these patients. Global prevalence of HCV G2 and G3 varies from 19% to 40% of all HCV-infected patients, thus representing the most common HCV genotypes after G1.1

Alisporivir (ALV) is an oral cyclophilin inhibitor (i.e., a host-targeting antiviral) that has shown potent anti-HCV activity in vitro and in clinical studies.[6] Cyclophilin A has peptidyl-prolyl isomerase enzymatic activity, interacts with the HCV nonstructural (NS)5A protein, and this interaction is essential for HCV replication.[7] ALV blocks this interaction in a dose-dependent manner for all genotypes.[7] In vitro selection of HCV strains with reduced susceptibility to ALV has been observed only after prolonged culture (>3 weeks). Furthermore, individual mutations reduce ALV susceptibility by only 2- to 3-fold, and multiple mutations are required for inducing significant resistance to ALV.[8, 9] Thus, the key attributes of ALV include pan-genotypic antiviral efficacy, a high barrier to resistance, synergistic or additive antiviral effects with direct-acting antivirals (DAAs), and efficacy against the most common resistance-associated variants from the various DAA classes.[10] In the phase IIb VITAL-1 study, approximately 40% of treatment-naïve patients with HCV G2 or G3 that received ALV alone or with RBV achieved a rapid virologic response (RVR). For these patients, IFN-free treatment was maintained for another 20 weeks and 89% of them achieved a sustained viral response to treatment regimens 24 weeks after therapy in per-protocol analysis, demonstrating that ALV/RBV represents a potential alternative to Peg-IFN/RBV in HCV G2 and G3 patients.[11]

Viral kinetic (VK) modeling has provided insights into antiviral effectiveness of different classes of anti-HCV agents.[12] Here, we analyze VK profiles observed during the first 6 weeks of the VITAL-1 study in patients who received ALV, with or without RBV, to estimate its antiviral effectiveness. By simulations, we further evaluated the effect of different ALV IFN-free dosing regimens on virologic response and SVR rate.

Patients and Methods

Study Design

VITAL-1 (DEB025A2211) was an international, multicenter, randomized, open-label phase II study, which included 340 treatment-naïve HCV G2 and G3 patients (ratio, 3:7) randomized to five arms: (A) ALV 1,000 mg once-daily (QD) monotherapy (n = 83); (B) ALV 600 mg QD plus RBV (n = 84); (C) ALV 800 mg QD plus RBV (n = 94); (D) ALV 600 mg QD plus Peg-IFN (n = 39); or (E) Peg-IFN/RBV (n = 40). Patients in ALV-containing arms that achieved RVR, defined as HCV RNA below limit of quantification (<25 IU/mL) at treatment week 4 (W4), continued on the initial treatment for 24 weeks, whereas those with detectable HCV RNA continued with ALV plus Peg-IFN/RBV triple therapy from W6 to W24.

The RBV dose was 800 mg twice-daily (BID). The purpose of the VITAL-1 study was to determine rates of SVR in HCV G2 and G3 patients with IFN-free ALV-based QD regimens and delayed add on of Peg-IFN to ALV/RBV.

Data Included in the VK Analysis

The aim of the analysis was to investigate the VKs of IFN-free regimens. Therefore, VK profiles from the first 6 weeks of the IFN-free arms (arms A, B, and C) were included in this analysis (Supporting Table 1). Patients that did not achieve RVR at week 4 continued with ALV plus Peg-IFN/RBV from W6 to W24, making this later data not relevant for investigating the effect of ALV IFN-free therapy.

Table 1. Mean and Standard Errors (SE) of the Estimated Population Parameters and Between-Patient Variability (ω)
E-max modelEC50cδnkinAICBIC
  1. Parameter values obtained using the E-max model, Equation 2a, and the effect compartment model (Equation 2b). In both cases, the parameter δ did not differ between groups B and C.

Effect compartmentmodelEC50cδnkinAICBIC

Viral Load and Pharmacokinetic Schedule Sampling

ALV plasma concentrations were assessed on day 0 at 0.5, 1, 2, and 4 hours and then predose at days 7, 14, and 28. HCV RNA was assessed by real-time polymerase chain reaction using the COBAS Taqman HCV Test (v2.0; Roche Diagnostics, Indianapolis, IN) at days 0, 7, 14, 21, 28, and 42. Pharmacokinetic (PK) profiles were obtained in a subset of patients with . Twelve patients that did not have at least one viral load and one PK measurement between days 0 and 42 were excluded in our analysis, leaving 249 patients (N = 79; 80 and 90 in arms A, B, and C, respectively), including 53 with additional VK sampling at days 2 and 3 for final analysis.

PK/Dynamic Model

ALV PK profiles were adequately described using first-order absorption and a three-compartment disposition with a parallel linear and nonlinear elimination from the central compartment (see Supporting Fig. 1).

Figure 1.

Viral load data (circles) and their predicted individual fits using the effective compartment model in patients from arms A (upper row), B (middle), and C (lower). Patients were chosen because they have representative patterns of viral decline. Left column: monophasic rapid responders; second column: slow biphasic declines; third column: rapid biphasic declines; right column: rapid initial response followed by a slight increase (at the end of the lead-in phase) and a final decrease. The horizontal line is the LOD (15 IU/mL).

We used the standard model of HCV kinetics to fit the changes in HCV RNA levels.[13] In this model, ε represents drug effectiveness in blocking viral production, c is viral clearance rate, and δ is infected cell loss rate (see Supporting Equation 1).

We examined two different models for the relationship between ALV plasma concentration, as predicted by the PK model, C(t), and drug effect, ε(t): (i) a standard E-max-type model where ε(t) is directly related to the drug concentration (as shown by Equation 1a):

display math(1a)

Half maximal effective concentration (EC50) is the drug concentration needed to achieve an antiviral effectiveness of 50% and n is the Hill coefficient, and (ii) a model with an effect compartment. The latter model assumes that the drug concentration at the site of effect is not always equal to the concentration observed in plasma and accounts for a possible delay between the plasma drug concentration and subsequent antiviral activity, resulting, for instance, to the time needed to metabolize or transport the drug intracellularly. In this model, the concentration at the effect site, Ce(t), is determined by Equation 1b[14]:

display math

and the drug effectiveness is given by:

display math(1b)

Suboptimal Response

The PK/PD (pharmacodynamic) model can capture some aspects of suboptimal response, such as a flat second phase of viral decline or a viral rebound, as long as they can be causally related to low plasma drug levels that lead to loss of treatment effectiveness. This is the case, for instance, when viral load increases because drug doses are missed. However, this simple model cannot capture other mechanisms of suboptimal response, such as viral resistance, or effects of suboptimal exposure to RBV. Thus, using this model in patients having non-ALV PK-related suboptimal response is not appropriate and may lead to a bias in the estimation of VK parameters. Given these limitations, patients having non-ALV PK-related suboptimal response were identified a priori and left out of the final PK/VK analysis (see Supporting Information). The number of patients excluded was 35, which comprised 14% of all patients, and the final analysis involved the remaining 214.

Parameter Estimation Using Mixed-Effects Models

Patient viral load and PK data were fit using nonlinear mixed-effect models,[15] and model fits were compared using the Bayesian Information Criterion (BIC; the lower the better).[16]

Clinical Trial Simulation to Predict Virologic Response With BID Regimens

Because ALV given BID provides higher drug Ctrough exposure and greater antiviral efficacy, compared with QD dosing, we simulated the virologic response that can be expected in patients treated with different BID regimens of ALV plus RBV. A simulated cohort of 10,000 patients whose PK/VK parameter values were drawn from the estimated parameter distributions of PK/VK parameters found in arms B and C were created with an assumed 100% adherence to treatment. In VITAL-1, adherence was >99% during the first 6 weeks in arms A, B, and C.

SVR Prediction

For each simulated patient, viral eradication was considered achieved at time τi where the predicted total HCV RNA, inline imagewas first less than one copy in the entire extracellular fluid volume (i.e., 6.7 × 10−5 HCV RNA/mL).[17] If simulated treatment was stopped before τi, we scored this patient as not eradicating HCV and not attaining SVR. If treatment was stopped after τi, the patient was counted as achieving SVR. Using less than one infected cell as the cure boundary would extend the time for eradication by 1-2 weeks.[18]


VK Modeling

The PK/VK model was able to describe the individual viral load profiles of the 214 patients analyzed. Using the E-max model (Equation. 1a), we found a mean value for the viral clearance rate, c, equal to 2.5 d−1 (Table 1). This value is about 2 times lower than what was repeatedly found in HCV G1 patients[12] and about 3 times lower than what was estimated in HCV G2 and G3 from a small subset of patients.[19] As discussed elsewhere, the viral clearance rate is a physiological quantity that should not depend on treatment and a small value of c is likely to be the result of an inappropriate model structure that overestimates treatment effectiveness in the first days of treatment.[20] In the E-max model, plasma concentration is an instantaneous marker of antiviral effectiveness, which, as a consequence of the rapid accumulation in plasma (Supporting Fig. 3), leads one to conclude that ALV rapidly achieves high levels of antiviral effectiveness. Because this model may overestimate the initial antiviral effectiveness of ALV and thus may lead to inaccurate estimates of the virus clearance rate, we introduced a more sophisticated effect-compartment model that accounts for a possible delay between plasma drug concentration and the intracellular antiviral effectiveness of the drug (Equation 1b). This model has one additional parameter, kin, which determines the time needed to achieve equilibrium between the intracellular (effect) compartment and the plasma compartment. Because both kin and the clearance rate, c, cannot be simultaneously estimated with the sparse sampling design of this study, and given that c is unlikely to be drug-related, we fixed c for all patients to the mean value observed during IFN treatment, that is, c = 6 d−1 to allow estimation of kin.[12]

Both the interpretation and values of the parameters obtained with this model (Table 1) differ from those obtained with the E-max model, but the individual fits obtained were similar (not shown). The model fit to the data from each patient was good (Fig. 1) and the model predictions were very accurate, both in terms of the change in HCV RNA (Table 2) and in the proportion of patients achieving undetectable levels of HCV RNA (Table 3). Because the effect compartment model is more biologically realistic and gave the same quality of fit as the standard E-max model (Table 1), we only discuss the results found with this model.

Table 2. Mean Observed and Estimated log10 HCV RNA Levels Among the 214 Patients Studied in Arms A, B and C After Suboptimal Responders Were Excluded
 Observed log10 HCV RNA/mLEstimated log10 HCV RNA/mLPredicted log10 HCV RNA/mLWith ALV 400 mg BID + 400 mg BID RBV(± SD based on N = 100 Patients)
Treatment WeekABCABC200 mg300 mg400 mg
  1. Estimated HCV RNA/mL values were obtained from the best-fit theoretical curve generated using the estimated best-fit parameters for each of the 214 patients and then averaging. Data under the LOD were treated as equal to the LOD. Predicted log10 HCV RNA/mL was obtained by simulating the effect compartment model with parameters drawn from the distribution of population PK and PD parameters estimated in arms B and C. Although 10,000 patients were simulated for each ALV dose, they were sampled in 100 groups of N = 100 patients to obtain the variability that might be observed in a small clinical trial.

  2. Abbreviation: SD, standard deviation.

W06.126.075.916.086.085.956.00 (0.10)6.00 (0.11)6.00 (0.10)
W13.583.413.463.663.403.505.05 (0.10)4.45 (0.15)3.97 (0.15)
W23.353.123.153.313.103.084.46 (0.12)3.67 (0.17)3.10 (0.16)
W32.852.802.602.842.802.664.02 (0.14)3.15 (0.18)2.58 (0.15)
W42.532.422.372.582.422.393.7 (0.15)2.80 (0.17)2.25 (0.14)
W62.271.911.972.181.861.873.3 (0.16)2.40 (0.17)1.92 (0.12)
Table 3. Percent of Patients Under the LOD As Observed and Estimated From Their Best-Fitting Theoretical Curve Among the 214 Patients Studied in Arms A, B, and C After Suboptimal Responders Were Excluded
 Observed % Under LODEstimated % Under LODPredicted % Under LODWith ALV BID + 400 mg RBV BID (± SD Based on N = 100 Patients Per Arm)
Treatment WeekABCABC200 mg300 mg400 mg
  1. The predicted percentage of patients under the LOD was obtained by simulation as described in Table 2 caption.

  2. Abbreviation: SD, standard deviation.

W00000000 (0)0 (0)0 (0)
W11411131411131.5 (0.9)3.6 (1.9)6.4 (2.5)
W22024271823277.3 (1.8)15.5 (3.1)23.4 (4.18)
W334263830243815.2 (2.5)28.2 (3.6)39.0 (5.23)
W434404536374022.7 (3.0)38.8 (4.5)51.6 (5.5)
W643615642595433.3 (3.3)52.7 (5.4)66.3 (4.71)
W8      41.3 (3.9)62.0 (5.0)75.4 (4.0)
W12      51.5 (4.0)72.4 (4.8)84.3 (3.6)

Estimation of ALV Antiviral Effectiveness

We found that mean effectiveness of ALV in blocking viral production, ε(t), varied substantially over time, as a result of both the time needed to reach steady state in the effect compartment and the change in the dose regimen after the first week, from BID to QD dosing (Supporting Fig. 3). Antiviral effectiveness of ALV was found to be high and consistent across patients during the loading period, with a median effectiveness of 0.98 at the end of the 1-week loading period (Table 4). However, this effectiveness dropped in arms B and C after shifting to QD dosing, with a median ε of 0.90 and 0.96, respectively, at the end of the second week of treatment (Table 4). In arm A, where a higher dose of 1,000 mg QD was given after week 1, a higher drug effectiveness (median ε = 0.98) was maintained beyond the first week of treatment (Table 4). Treatment effectiveness reached steady state and did not change after the second week. Interestingly, no significant differences in EC50 was found between HCV genotypes 2 and 3 (P = 0.22).

Table 4. First and Third Interquartile (Q1, Q3), Median, and Mean Predicted Antiviral Effectiveness for Different Time Periods Using the Effect Compartment Model
 ε Day 2ε Day 6ε Day 14

Effect of RBV on the Second Phase of Viral Decline

The high antiviral effectiveness in arm A does not mean that the viral decline was more profound in arm A than in arms B and C. Indeed, the rate of viral decline in the long run is proportional to εδ,[12] and therefore the loss rate of infected cells, δ, is also instrumental for achieving rapid long-term viral decline. Interestingly, patients who received RBV had a significantly larger estimated value of δ than patients who did not (mean δ = 0.35 vs. 0.21 d−1 in arms B and C vs. A; P = 0.0001). This larger δ explains, in this model, the fact that patients in arms B and C had a better virologic response than patients in arm A (Tables 2 and 3). Importantly, similar to EC50, no differences in δ were found across genotypes (P = 0.24).

We tried to better characterize the relationship between δ and the drug regimen. However, as hinted by the fact that δ was not different in groups B and C, we did not find any relationship with ALV exposure in this population of responders. Likewise, although suboptimal response may be related to suboptimal concentrations of RBV in overweight patients (see below), we did not find any relationship between δ and patient weight in the responder group (P = 0.30).

Characteristics of Patients With Suboptimal Response

Thirty-five patients (14%) had a suboptimal response (i.e., flat or increasing levels of HCV RNA after W1), and their VK profile could not be described using the model. Suboptimal response was significantly more frequent with ALV monotherapy versus ALV/RBV (21.5% vs. 10.6% in arm A vs. B and C; P = 0.02). In patients that received RBV, no difference in proportion of suboptimal responders was found according to the ALV dosing received (12.5% vs. 8.9% in arms B and C, respectively; P = 0.4). Interestingly, weight, but not body mass index, was significantly associated with the occurrence of suboptimal response (with a mean weight of 83.5 vs. 72.1 kg in suboptimal responders vs. responders; P = 0.027).

No parameters describing ALV's PKs were significantly associated with occurrence of suboptimal response. However, a trend was observed toward lower ALV predose concentration between W1 and W6 in suboptimal responders versus responders (P = 0.08). Also, predose RBV concentrations at steady state (trough concentrations at and after W4) were significantly lower in suboptimal responders (median, 1,295 vs. 1,395 ng/mL; P = 0.017).

Sequencing data revealed three mutations in domain II of NS5A enriched in these patients that conferred a low level of reduced susceptibility to ALV in vitro: D320E (∼5.2-fold increase in EC50); R347W (∼3.7-fold increase); and A349V (∼3.3-fold increase). Combination of multiple mutations decreased susceptibility further, but generally at a cost of replication fitness (D320E/R347W: ∼20.7-fold increase with replicative capacity at 2.6%, relative to wild type [WT]; D320E/A349V: ∼11.0-fold increase with replicative capacity at 12.1%, relative to WT). Of 35 suboptimal responders, 23 patients were found to carry one or more of these mutations and 12 had no genotypic changes detected at these positions (Supporting Table 4). A significantly lower ALV exposure predose at W4 (mean Cmin = 151 ng/mL) was detected in patients harboring one or more of these mutations, compared to patients without genotypic changes at these positions (mean Cmin = 636 ng/mL; P value for patients with vs. without mutations at these positions: P = 0.017). Furthermore, all but 1 patient (22 of 23) carrying one or more of the aforementioned NS5A domain II mutations had ALV predose concentrations lower than 400 ng/mL at W4.

Clinical Trial Simulation of Virologic Response With BID Doses of ALV Plus RBV

Simulation Scheme

Next, we aimed to predict the early- and long-term virologic response that could be expected with different doses of ALV given BID (200, 300, or 400 mg) in combination with 400 mg of RBV BID (see Patients and Methods).

PK Concentration

Figure 2 shows predicted plasma ALV concentrations and HCV RNA levels obtained according to the dose of ALV using mean parameter values. Interestingly, the 300- and 400-mg BID doses were predicted to provide higher predose concentrations than the 600- and 800-mg QD doses, respectively. However, maximal exposure would remain lower than what was observed with 1,000 mg QD (Supporting Table 3).

Figure 2.

Predicted trajectories of HCV RNA and plasma ALV concentration using the mean parameters in the three dosing groups (red, 200 mg BID; green, 300 mg BID; blue, 400 mg BID).

Rates of RVR

Both 200- and 300-mg BID dosing of ALV were predicted to achieve lower RVR rates than observed in arms B and C, with rates of 22.7% and 38.8% for 200 mg and 300 mg BID, respectively, compared to the 40% and 45% observed in arms B and C, respectively. Conversely, the model-predicted RVR rate is 51.6% following an ALV 400 mg BID (Table 3), which is higher than the observed response rate following QD doses.

Effects of Dosing on SVR

The predicted lower-than-observed on-treatment efficacy of 200 and 300 mg BID was confirmed by analyzing predicted SVR rates. After 36 weeks of treatment with 200 and 300 mg BID, the predicted SVR rates were 57% and 78%, respectively, compared to a predicted 88% SVR rate in patients receiving 400 mg BID (Table 5). If we take into consideration that 10.6% of patients in arms B and C had suboptimal responses and if we make the conservative assumption that this fraction will remain unchanged with BID dosing, then the predicted SVR rates after 36 weeks of treatment for these three regimens becomes 51%, 70%, and 79%, respectively. Interestingly, with 400 mg BID, the model predicted a SVR rate of 80% (71% correcting for possible suboptimal response) with a treatment duration of 24 weeks.

Table 5. Percentage of Patients Predicted to achieve SVR for Different ALV Doses and Different Treatment Durations, Assuming No or a Fixed Proportion of 10.6% of Suboptimal Response
 % of Viral Eradication Rate
Treatment duration (Weeks)200 mg BID300 mg BID400 mg BID
  1. Numbers in parenthesis are the standard deviation obtained assuming a sample size of N = 100 patients.

1227.5 (3.2) │ 24.6 (2.9)44 (4.0) │ 39.3 (3.6)56 (4.9) │ 50.1 (4.4)
2447.4 (4.1) │ 42.4 (3.7)68 (4.0) │ 60.8 (3.6)80 (3.8) │ 71.5 (3.4)
3657.5 (4.4) │ 51.4 (3.9)78 (3.6) │ 69.7 (3.2)88 (3.0) │ 78.7 (2.7)

Response-Guided Therapy

Our simulation results also suggest that VK analysis might be a tool for designing response-guided therapy (RGT). To explore this possibility, we calculated for the 10,000 simulated patients in the 400-mg BID dosing group the proportion of patients achieving SVR after 12, 24, or 36 weeks of treatment when undetectable HCV RNA (i.e., <15 IU/mL) was achieved after 2, 4, 8, or 12 weeks of treatment. We defined the optimal treatment duration as the shortest duration giving an SVR rate of 95% or more. With this criterion, we predict that 12 weeks of treatment with 400 mg of ALV BID would be sufficient in patients with undetectable HCV RNA by W2, and 24 weeks would be appropriate in patients that go undetectable between W4 and W8 (Table 6). Furthermore, 36 weeks of treatment is predicted to generate a 98.2% SVR rate in patients with their first HCV RNA undetectable at W12 (Table 6). In contrast, patients with detectable HCV RNA at W12 would only achieve 28.8% SVR, even with 36 weeks of treatment.

Table 6. Percentage of Patients Predicted to Achieve SVR With 400 mg ALV Plus 400 mg RBV BID
 SVR Rate in Patients With HCV RNA Below LOD
Treatment Duration (Weeks)W2W4 not W2W8 not W4W12 not W8Not W12
  1. Numbers in parenthesis are the standard deviation obtained assuming a sample size of N = 100 patients.

684.7 (7.5)23.7 (8.9)   
1298.9 (2.3)91.7 (5.3)32.2 (9.6)0 (0)0 (0)
2499.9 (0.6)99.7 (0.9)97.6 (3.0)61.8 (20.0)2.2 (4.0)
3699.9 (0.4)99.9 (0.3)99.9 (0.6)98.2 (4.7)28.8 (12.9)

Overall, our model predicts that using this RGT would allow for achieving an SVR rate of 87.9% with a mean treatment duration of 24 weeks. Of note, if treatment is stopped at W12 in patients with detectable HCV RNA, then the mean treatment duration becomes equal to 20 weeks with an SVR rate of 83.4%.

If we take into consideration that 10.6% of patients in arms B and C had suboptimal response and were excluded from this analysis and assume the same fraction of patients treated with BID dosing have a suboptimal response, then the mean treatment duration becomes 19.1 weeks with an SVR rate of 74.5%, if treatment is stopped at W12 in patients with detectable HCV RNA, and 25.3 weeks if treatment is not stopped with an SVR rate of 78.6%.


ALV is a host-targeting antiviral agent with similar anti-HCV activity against HCV G2 and G3. Analyses of the pooled clinical database, involving over 2,000 patients, confirmed that ALV IFN-free treatment was well tolerated with a markedly better safety profile than IFN-containing treatment, with low rates of anemia, neutropenia, thrombocytopenia, headache, fatigue, nausea, and pyrexia, among others.[21] Hyperbilirubinemia and hypertension were more frequent with ALV treatment (with or without IFN), compared with Peg-IFN/RBV. The most frequent serious adverse events in the ALV/Peg-IFN/RBV triple-therapy group were hypertension, pneumonia, anemia, and pyrexia, none of which were reported with IFN-free ALV treatment.[21] In the present study, we analyzed the VK and PK profiles in three arms from the VITAL-1 trial, where patients received varying doses of ALV, given QD with or without RBV, during the first 6 weeks. During this period, these three groups did not receive IFN and thus provide a basis to evaluate the responses that might be observed in future IFN-free trials.

Although our modeling approach could account for rebounds as a consequence of reduced ALV exposure, we found that 35 of 249 patients (14%) in this study exhibited rebounds or a flat VK profile that could not be solely explained by low ALV concentrations. The origin of this suboptimal response may be diverse and the possibilities encompass reduced sensitivity to ALV, low exposure to RBV, as well as other yet unknown factors. Interestingly, a trend was observed toward lower ALV predose concentration between W1 and W6 in suboptimal responders versus responders (P = 0.08), suggesting that high ALV exposure may reduce the possibility of suboptimal response. Furthermore, 4 patients did not achieve noticeable (<1 log10) viral reduction after 6 weeks of treatment (Supporting Table 4), and in all these patients, ALV exposure was low (ALV Cmin <200 ng/mL at W4), suggesting that the lack of response in this subset of patients was likely a consequence of inadequate therapeutic exposure and, possibly, unknown host factors.

Sequencing and phenotypic analyses of HCV isolates were carried out in these suboptimal responders, and 23 of the HCV isolates had one or two mutations in the domain II of NS5A region (D320E, R347W, and A349V) that confer low-level resistance to ALV (3.3- to 5.2-fold increase in EC50) in vitro. Importantly, a significantly lower ALV exposure was observed in patients harboring one or more mutations (P = 0.017), consistent with the hypothesis that low ALV exposure could lead to a suboptimal response.

Interestingly, RBV was critically associated with prevention of suboptimal responses, whose proportion was significantly higher with ALV monotherapy than in the RBV-containing arms (21.5% vs. 10.6%; P = 0.02). In patients receiving RBV, higher body weight was associated with suboptimal response, suggesting that RBV exposure may be important in preventing rebounds and flat second-phase responses. This observation was confirmed by an analysis of RBV exposure because patients with suboptimal response had lower levels of plasma RBV at steady state (P = 0.017). The possibility that higher doses of RBV may reduce the occurrence of suboptimal response warrants further investigation and suggests that, in future studies, weight-based RBV dosing should be considered.

Because of the difficulty to precisely identify the origin of suboptimal response, we restricted our kinetic analysis to the 214 patients (86%) that showed continuous viral decline. For this population of patients, we found that a combined PK/VK model fit the patient viral load data well (Fig. 1). By day 14, the median antiviral effectiveness of ALV in blocking viral production, ε, reached steady state and was equal to 0.98, 0.96, and 0.90 in patients receiving 1,000, 800, and 600 mg of ALV QD, respectively. Rate of second phase viral decay was approximately εδ and was significantly enhanced in patients that received RBV (mean δ = 0.35 vs. 0.21 d−1 in patients ± RBV, respectively; P = 0.0001). The fact that RBV could increase the second-phase decay rate was predicted by a previous viral kinetic model[17] and has been observed in other studies.[6, 22]

In order to inform the design of future studies using ALV plus RBV as IFN-free therapy, we used the distributions of PK and VK parameter values estimated by analyzing arms B and C to create synthetic populations of patients from which we could predict response rates to various doses of ALV. We focused on analyzing the effects of giving ALV at 200, 300, or 400 mg of ALV BID plus 400 mg of RBV BID, assuming full compliance. However, to validate our simulation method, we also predicted the percentage of patients in our synthetic population that would go undetectable at weeks 1 through 6 and compared these predictions to the observed values (Table 3). Here, because we excluded the suboptimal responders in our parameter estimation procedure, we also excluded the patients showing suboptimal response from this comparison. The predicted and observed fraction of patients under the limit of detection (LOD) agreed well (Table 3). Using the same method, we predicted that, in a new trial using 200, 300, or 400 mg of ALV BID plus BID RBV, that 22.7%, 38.8 %, and 51.6% would exhibit RVR.

With a BID regimen, the chances of achieving SVR are high. We predicted that 68% and 80% would achieve SVR with 300 and 400 mg BID, respectively, after 24 weeks of treatment (Table 5). These percentages were increased with 36 weeks of treatment to SVR rates of 78% and 88%, respectively (Table 5). These predicted SVR rates did not take into account patients showing suboptimal response. Using the conservative assumption that the proportion of non-PK suboptimal responders would be 10.6%, as observed in arms B and C (i.e., in patients that received RBV), these predicted SVR rates would be reduced to 70% and 79% with 300 and 400 mg BID, respectively, after 36 weeks of treatment (Table 5). Noncompliance could further lower these predicted SVR rates.

Our simulations suggest that SVR rates could be further optimized by using RGT. We found that in patients having their first undetectable HCV RNA between W4 and W8, 24 weeks is the optimal treatment duration. However, this duration can be shortened to 12 weeks in patients with undetectable HCV RNA by W2 and it should be extended to 36 weeks in patients that become undetectable by W12 or later. Our simulations predict that this approach would allow one to achieve an overall SVR rate of 79%, assuming the same rate of rebound and flat second-phase response as in the dual-therapy arms in the study, with a mean treatment duration of 25 weeks in patients treated with 400 mg of ALV BID plus RBV (i.e., a 8% increase), compared to a fixed course of 24 weeks treatment. However, with this regime, only 28.8% of patients that did not achieve an early virologic response (defined by undetectable HCV RNA by week 12) are predicted to achieve SVR after 36 weeks of treatment (Table 6).

In summary, we have shown that for patients with HCV G2 or G3 infection, a regimen consisting of ALV plus RBV may be an effective IFN-free treatment option that could achieve high rates of SVR.