Ribavirin exposure after the first dose is predictive of sustained virological response in chronic hepatitis C


  • Potential conflict of interest: Nothing to report.


The impact of ribavirin exposure on sustained virological response (SVR) in patients with chronic hepatitis C is unknown. Preliminary studies showed marked inter-individual variability of ribavirin concentrations despite dose adjustment for body weight (BW) and suggested there was a correlation between single time point concentrations and SVR. None of them evaluated the global exposure to ribavirin. This study was conducted to determine whether early ribavirin global exposure is related with SVR. An exploratory pharmacokinetic-pharmacodynamic (PK-PD) study was conducted in genotype 1 hepatitis C patients treated with peginterferon alfa-2a and ribavirin (dose-adjusted for BW) for 12 weeks, to which amantadine was added for the following 36 weeks. Full and abbreviated ribavirin area under the concentration time curves (AUC0–12h, AUC0–4h) were derived from plasma concentration profiles at day 0 (D0), week 12 (W12), W12 + 1 day, and W24. Virological follow-up was performed at D0 (0, 12, and 24 hours), W2, W4, W6, and monthly until W72 (TaqMan polymerase chain reaction, cut-off 15 international units/mL). Twenty-eight patients were enrolled in the study and 24 completed it. Patients with a SVR had a significantly higher D0 AUC0–12h (3695 [1571–6916] versus 2937 [1266–4913] μg/hour/L, P = 0.03) and D0 AUC0–4h (2010 [615–3175] versus 1340 [622–2246] μg/hour/L, P = 0.03). Patients with D0 AUCs above the cut-off values defined by receiver operating characteristic curves (3014 μg/hour/L and 1755 μg/hour/L for AUC0–12h and AUC0–4h, respectively) had a significantly better chance of achieving an SVR than patients with AUCs under the thresholds (odds ratio = 16.0, 95% confidence interval 1.54–166.6, P = 0.02 and odds ratio = 8.9, 95% confidence interval, 1.4–56.6; P = 0.02). Conclusion: Ribavirin exposure at D0 is significantly related to SVR. To our knowledge, this is the first study to give an early pharmacokinetic predictor of SVR. We propose a minimum AUC0–4h threshold of 1755 μg/hour/L at D0 as a target for ribavirin dose adjustment. (HEPATOLOGY 2008;47:1453–1461.)

The association of peginterferon and ribavirin is the standard approved treatment for patients with chronic hepatitis C virus (HCV).1, 2 Unfortunately, 50% of patients relapse or remain virological nonresponders. The main predictive factors of a poor sustained virological response (SVR) are an HCV genotype other than 2 or 3, a high baseline viral load, and a dose of ribavirin <10.6 mg/kg of body weight (BW).1 However, the impact of the pharmacological properties of ribavirin has not been fully investigated and requires complementary studies. Two clinical trials1, 2 have suggested that ribavirin dose adjustment for BW may be beneficial. Manns et al.1 showed that the cumulative dose of ribavirin per kilogram of BW is a significant predictor of SVR. Ribavirin has a large distribution volume and large interindividual variability in concentrations.3 Preliminary pharmacological studies have been published.3–8 Two studies in patients with chronic HCV who were receiving combination therapy4, 6 found no correlation between ribavirin dose adjusted for BW and a single ribavirin time-point serum concentration at week 4 (W4) or at steady state (W12), that is, a poor dose-concentration relationship. However, a low ribavirin serum concentration was associated with a low SVR rate, suggesting a better concentration-effect relationship.5, 6

The aim of this study was to further investigate the pharmacokinetic-pharmacodynamic (PK-PD) behavior of ribavirin, in particular the relationship between early ribavirin exposure and SVR in patients with chronic hepatitis C who were cotreated with peginterferon or peginterferon and amantadine.


AUC, area under the curve; BMI, body mass index; BW, body weight; Clcreat, creatinine clearance; D, day of treatment; EVR, early virological response; HCV, hepatitis C virus; HR, hazard ratio; NPV, negative predictive value; PCR, polymerase chain reaction; PK-PD, pharmacokinetic-pharmacodynamic; PPV, positive predictive value; ROC, receiver operating curve; RVR, rapid virological response; SVR, sustained virological response; T, time; W, week of treatment.

Patients and Methods

An observatory PK-PD bicenter clinical trial was performed in treatment-naïve patients with hepatitis C genotype 1 given peginterferon alfa 2a (40 kDa) 180 μg weekly and ribavirin dose-adjusted for BW (<75 Kg BW at 1000 mg/day, ≥75 Kg BW at 1200 mg/day) for 12 weeks; to this treatment protocol 200 mg amantadine was added daily at W12 and for 36 weeks, in order to evaluate the eventual impact of amantadine on ribavirin PK, where the patients were their own controls.9 Twenty-eight consecutive patients were included in this exploratory study. The inclusion and exclusion criteria were those usually required in hepatitis C therapeutic clinical trials.

Ethical Considerations.

The protocol was approved by the regional human research and ethics committees and all patients gave their written informed consent.

Clinical virological follow-up was performed by HCV viremia detection using the Cobas Roche Amplicor polymerase chain reaction (PCR) assay (cut-off 50 IU/mL; Roche Diagnostics) at weeks 12, 24, 48, and 72. Additional serum samples were collected at day 0 (D0) (0, 12, and 24 hours) and at weeks 2, 4, 6, 8, 12, 16, 20, 24, 36, 48, 60, and 72; samples were stored at −80°C for viral load follow-up. The quantitative real-time PCR assay HCV Ampliprep TaqMan (cut-off 15 IU/mL; Roche Diagnostics) was used after January 2006 in the routine patient follow-up. All these additional samples were retrospectively tested using this assay.

Ribavirin plasma concentration profiles were planned for each patient after the first dose (D0), at treatment W12, W12+1day (that is, first day of amantadine dose), and W24. Ribavirin PK was assessed in normal life conditions: hour 0 (h0) was taken after 12 hours fasting; the patients then had standard meals during the 12-hour dosing interval.

No comedications taken by the patients have been described as influencing ribavirin PK. Particularly, none of them took nucleoside analogs. Patients were hospitalized for each profile and blood samples were collected immediately before and at 0.5, 1, 1.5, 2, 3, 4, 6, 8, 10, and 12 hours after the morning dose.

Peripheral blood samples were collected on ethyene diamine tetraacetic acid tubes. After 15 minutes of centrifugation at 1733 g (GR422 centrifuge; Jouan SA, Saint-Herblain, France) at + 4°C, plasma was divided into aliquots and stored at −80°C in brown 1.5 mL propylene tubes (Eppendorf) until analysis.

All ribavirin levels were measured in a central laboratory with a validated, specific, selective reverse-phase high performance liquid chromatography–tandem mass spectrometry method in the positive-ion mode, following two ionic transitions per compound. Sample preparation involved solid-phase extraction of a plasma spiked with 6-methylaminopurine 9-ribofuranoside (internal standard) using Bond Elut Phenyl Boronic Acid cartridges 100mg (gel phase quantity in the cartridge), 1 mL (volume capacity per cartridge) (Varian, France) previously equilibrated with an ammonium acetate solution. A methanol-formic acid mixture (97:3 by volume) was used as elution solvent. The calibration curves were linear from 5 μg/L up to 5000 μg/L. The within-day and between-day coefficients of variation and bias were less than 15% over this range. A set of calibration standards at 0, 5, 10, 25, 50, 100, 250, 500, 1000, 2000, and 5000 μg/L of ribavirin were prepared, extracted, and analyzed with each series, together with internal quality controls at three levels.

Pharmacokinetic Analysis.

Ribavirin area under the concentration-time curves (AUC) in plasma were calculated using the linear trapezoid method. The interdose AUC (here AUC0–12h) is a measure of global exposure and is the gold standard for the dose adjustment of many drugs. However, single concentration values and abbreviated profiles with samples collected over a shorter period of time were also investigated in this case, because they would be more practical and acceptable for monitoring outpatients.

Virological Response.

The Cobas PCR was used to evaluate HCV viremia as a surrogate marker of treatment response. Patients with a negative PCR at the end of the follow-up (W72) were regarded as sustained virological responders. Those with detectable viremia at W24 of treatment were considered nonresponders and treatment was stopped. However, the TaqMan quantitative assay was retrospectively used for the purpose of statistical analysis.

Patients exhibiting a virological response at W48 but with a virological relapse before W72 were considered as nonresponders. Rapid virological response (RVR) was defined by either a nondetectable viral load or a viral load drop >2 log IU/mL at W4, and early virological response (EVR) was defined by a nondetectable viral load at W12.

Statistical Analysis.

Analyses were performed in the per-protocol population. Quantitative variables were expressed as median (range). To compare the groups of responders and nonresponders, the Mann-Whitney nonparametric test was used for continuous variables whereas the chi-squared or exact Fisher test were used for categorical variables. The Wilcoxon signed-rank test was used to compare values from the same patients at different periods.

The continuous exposure indices (that is, AUCs), dose, and dose/BW were dichotomized if univariate analysis showed they were significantly different (P < 0.1) between responders and nonresponders. Threshold values were obtained using receiver operating curve (ROC) analysis, and these thresholds were used only in case the area under the ROC curve was significantly different from 0.5. Exposure was then considered as “high” above the threshold and “low” below it. Finally, the threshold was tested for sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV).

Statistical analysis was performed using the S-PLUS (Insightful Corp., Seattle, WA) software package; ROC curves were obtained with MEDCALC 9.0 (Medcalc Software, Mariakerke, Belgium).

Stepwise logistic regression was used to investigate the multivariate association of sustained response with ribavirin exposure (that is, AUCs, residual concentration, peak concentration, and concentration values collected around the peak), dose, dose/BW as well as if significant after univariate analysis, age, sex, initial viral load, BW, lean BW, body mass index (BMI), and hepatic and renal function. Results are presented as odds ratio (OR) with 95% confidence interval (CI).


This study was based on a homogeneous population of 28 patients with chronic hepatitis C infected with genotype 1. One patient was excluded at D0 (that is, before collection of the first PK profile) because he developed lung tuberculosis. Nineteen male patients and eight female patients, median age 43 (20–62), median viral load 2,350,000 (5500–56,400,000) IU/mL, BW = 73 (50–125) kg, BMI = 25 (19–38), median fibrosis score (distribution: 14 patients were F1, eight patients were F2, three were F3, and two were F4) according to the METAVIR scoring system, median prothrombin time 91 (69–125)% and albumin 43 (34–51) g/L, median creatinine 65 (44–91) μmol/L, and Cockroft Clcreat 117 (58–241) mL/minute/1.70 m2 were analyzed.

Three patients developed severe side effects (at W8, giant urticaria [n = 1] and severe anemia [n = 1] and at W28, severe neutropenia [n = 1]), so that the treatment had to be stopped before completion of the study. These patients were withdrawn from the statistical analysis of the exposure-efficacy (AUC-SVR) relationship. Twenty-four patients completed the study.

Pharmacodynamic data were available at W12 for 25 patients and at W72 for 24 patients. Fifteen of 27 and 14 of 26 patients exhibited an RVR and a EVR, respectively. Eleven of 24 patients had an SVR.

D0 pharmacokinetic data were available for 27 patients. One patient was excluded from pharmacokinetic analysis at W12 and W12+1day because he took a wrong ribavirin dose. Overall, 103 concentration-time curves were obtained, representing 1111 blood samples.

There was a wide distribution of ribavirin AUC0–12h at D0 and W12 among patients, with a median of 3407 (1266–6916) μg/hour/L at D0 and 28526 (15937–52950) μg/hour/L at W12 (Fig. 1). The AUC accumulation ratio (AUC0–12h W12/AUC0–12h D0) was 7.9 (2.8–15.0). There was no significant correlation between W12 AUC0–12h and D0 AUC0–12h (r2 = 0.08, P = 0.15).

Figure 1.

Concentration-time profiles of ribavirin at D0 (n = 27) and W12 (n = 25) in patients with hepatitis C.

No pharmacokinetic interaction of amantadine on ribavirin PK was observed. Indeed, no difference was found neither between median ribavirin AUC0–12h values at W12 and W12+1day (28155 [15937–52950] μg/hour/L versus 27685 [14938–48278] μg/hour/L, P = 0.84) that is, before and after amantadine, nor between ribavirin AUC0–12h values at W12 and W24 (27882 [16410–46147], P = 0.52). No significant correlation was observed between BW and ribavirin AUCs at D0 and W12 (r2 = 0.07 P = 0.11, r2 = 0.004 P = 0.31, respectively). Table 1 summarizes the baseline characteristics and Table 2 the ribavirin dose and exposure indices of responders and nonresponders. Median D0 AUC0–12h values in sustained responders were significantly higher than in nonresponders [3695 (1571–6916) μg/hour/L versus 2937 (1266–4913) μg/hour/L, P = 0.03] but the W12 AUC values were not significantly different [28800 (15937–52950) μg/hour/L versus 26372 (17385–33787) μg/hour/L, P = 0.19] (Fig. 2). Median D0 AUC0–12h values were also significantly higher in patients with a RVR than in nonresponders [3686 (2390–6916) μg/hour/L versus 2975 (1266–4914) μg/hour/L P = 0.028]. Finally, there was a trend toward a higher median D0 AUC0–12h value in patients exhibiting EVR P = 0.057.

Table 1. Patients with or without SVR: Baseline Characteristics
CharacteristicPatient Group with SVRPatient Group Without SVRMann-Whitney
Number of patients1113 
Gender ratio (M/F)1.752.25 
 Median (range)Median (range)P
Mean viral load Taqman (UI/mL)1970000 (5500–18700000)4380000 (598000–56400000)0.09
Age (Years)43 (38–59)43 (20–62)0.48
Body weight (kg)73 (56.4–109)75.8 (50.8–125)0.56
Body mass index (BMI)25 (20–33)25 (19–38)0.93
Lean body weight (kg)51.1 (40.5–72.9)50.9 (40.0–80.3)0.83
Fibrosis score (METAVIR)1 (1–3)2 (1–4)0.50
Creatinin (μmol/L)77 (52–91)62 (44–81)0.03
Prothrombin time (%)90 (62–119)94 (82–125)0.27
Albumin (g/L)45 (37–51)42 (34–44)0.04
Table 2. Comparison Between Doses and Exposure Indices of Ribavirin in Patients with or Without SVR
CharacteristicPatient Group with SVRPatient Group Without SVRMann-Whitney
(Median [Range])(Median [Range])P
Number of patients1113 
AUC0–12 D0 (μg/hour/L)3695 (1571–6916)2937 (1266–4913)0.03
AUC0–4D0 (μg/hour/L)2010 (615–3175)1340 (622–2246)0.03
Residual concentration D0 (μg/L)435 (237–768)332 (218–738)0.26
Peak concentration D0 (μg/L)850 (225–1630)558 (210–901)0.03
Dose (mg)1000 (800–1200)1000 (1000–1200)0.49
Dose/WT (mg/kg) (D0)13.62 (11–17.85)13.86 (9.6–19.6)0.89
Dose/WT/month (mg/kg) (D0-W24)14.89 (11.32–18.43)14.27 (1.30–20.92)0.37
AUC0–12W12 (μg/hour/L)28800 (15937.75–52950)26372.5 (17385.5–33787.5)0.19
AUC0–12 W24 (μg/hour/L)27905 (21255–46147)24687 (16410–42967)0.23
Residual concentration W12 (μg/L) (without Patient 11)2230 (1130–4450)2170 (1090–2940)0.78
Residual concentration W12+1 (μg/L)2440 (1130–4430)1982 (1060–2990)0.11
Residual concentration W24 (μg/L)2700 (2040–3210)2010 (1260–2830)0.007
Decrease in Hb level D0-W12−2.8 (−5.9/−1.6)−2.8 (−4.8/−0.8)0.97
Figure 2.

Box plots showing the distribution of exposure variables at D0: [(A) D0 AUC0–12h, (B) D0 AUC0–4h, (C) D0 Cmax, (D) W12 AUC0–12h] in sustained virological responders (n = 11) and in nonsustained virological responders (n = 13) (middle lines represent the median, the upper and lower extremities of the boxes represent the 25th and 75th percentiles, the whiskers represent the 10th and 90th percentiles, and the dots the outliers).

The AUC ROC curve surface-area was 0.755 (P = 0.0128). The optimal cut-off value for D0 AUC0–12h was 3014 μg/hour/L with a sensitivity of 90.9% (95% CI 58.7–98.5), a specificity of 61.5% (95% CI 31.6–86.0), a PPV of 66.7% and a NPV of 88.9% for SVR. The observed cumulative percentages of SVR were 67% and 21% in the high and low AUC0–12h groups, respectively. In patients with SVR, the time to reach undetectable viral load ranged from 2 to 12 weeks and from 8 to 16 weeks in the high and low AUC0–12h groups, respectively.

At D0, the abbreviated AUC0–4h was highly correlated with AUC0–12h (r2 = 0.913, P < 0.0001) and AUC0–4h was also significantly higher in the SVR group than in the nonresponder group (Table 2). The relationship between D0 AUC0–4h and SVR was also tested using ROC curve analysis. An optimal cut-off value of 1755 μg/hour/L was obtained with a sensitivity of 72.7% (95% CI 39.1–93.7), a specificity of 84.6% (95% CI 54.5–97.6), a PPV of 80%, an NPV of 78.6%, and a ROC AUC of 0.755 (P = 0.0128).

Among the variables available at D0, the only ones found to be associated with an increased risk of SVR by logistic regression analysis were, in decreasing order, high exposure indices at D0 (AUC0–12h, AUC0–4h), creatinine level, peak concentration value (Cmax), and C1.5h. Table 3 summarizes the results obtained with univariate analysis and after adjustment for the initial viral load, which is usually a well-known predictive factor of SVR (even if not statistically significant in this study). No multivariate model was better than the univariate models including either D0 AUC0–12h or D0 AUC0–4h. Patients with D0 AUC0–12h > 3014 μg/hour/L achieved significantly more frequent SVR than patients with an AUC0–12h below this threshold (OR = 16.0 [95% CI, 1.54–166.6]; P = 0.02). The same stood for patients with D0 AUC0–4h ≥ 1755μg/hour/L (OR = 8.9 [95% CI, 1.4–56.6]; P = 0.02). RVR was also predictive of SVR (OR = 6.0 [95% CI, 1.02–35.4]; P = 0.048). The initial viral load, BMI, fibrosis score, dose, dose/BW, Clcreat, BW, lean BW, and ribavirin trough concentration (C0) at D0 were not significantly associated with the clinical response.

Table 3. Logistic Regression Analysis: Odds Ratio Between Sustained Virogical Responders and Nonresponders, Obtained Using Univariate Analysis and After Adjustment for the Initial Viral Load
ParameterOdds Ratio (95% CI)
Unadjusted on Initial Viral LoadP ValueAdjusted on Initial Viral LoadP Value
D0 AUC0–12h(high AUC group)16.0 (1.5–166.6)0.0212.5 (1.1–138.0)0.038
D0 AUC0–4h(high AUC group)8.9 (1.4–56.6)0.027.0 (1.0–47.8)0.045
D0 Cmax (μg/L)1.004 (1.000–1.007)0.0421.003 (1.000–1.007)0.08
D0 C1.5h(μg/L)1.004 (1.000–1.009)0.0451.004 (1.000–1.008)0.08
D0 Creatinine (μmol/L)1.1 (1.0–1.2)0.0361.1 (1.0–1.2)0.046
D0 AUC0–12h (μg/hour/L)1.001 (1.000–1.002)0.0561.000 (1.000–1.000)0.087
D0 AUC0–4h (μg/hour/L)1.002 (1.000–1.003)0.0521.001 (1.000–1.003)0.09
RVR (yes vs. no)6.0 (1.0–35.3)0.0476.7 (1.0–43.8)0.045

Interestingly, all patients (8 of 8) who had an RVR associated with a D0 AUC0–4h ≥ 1755 μg/hour/L at D0 reached SVR whereas none of the four patients with an RVR associated with an AUC0–4h < 1755μg/hour/L at D0 reached an SVR (Table 4).

Table 4. Sustained Virological Responses as a Function of RVR, EVR, and Ribavirin AUC0–4h at D0
AUC0–4hViral LoadNumber of PatientsETR*SVR†
  • *

    ETR, end of treatment response (W48); †SVR, sustained virological response (W72).

> cut off (1755 μg/hour/L)NegNegNeg222
 ≥2 log ↓NegNeg666
 <2 log ↓PosNeg110
≤ cut off (1755 μg/hour/L)≥2 log ↓NegNeg110
 <2 log ↓NegNeg222

The median hemoglobin levels were 15.3 (12.3–16.9) g/dL at D0 and 11.6 (8.4–15.3) g/dL at W12. At W12, 18 of 26 (69%) patients had hemoglobin levels below normal values. Hemoglobin levels at W12 were correlated with ribavirin exposure at D0 (r2 = 0.28, P = 0.005), whereas no significant correlation was observed between hemoglobin levels at W12 and ribavirin trough levels at W12 (P = 0.11).


In this study, interdose (AUC0–12h) or abbreviated (AUC0–4h) ribavirin plasma exposure after the first dose were significantly linked with SVR, as well as with RVR and, to a lesser extent EVR. To our knowledge, this is the first study to propose such early PK predictors of SVR.

In patients with hepatitis C, the well-known predictors of SVR are RVR and EVR. However, McHutchison et al. recently reported that the total dose of ribavirin received during the first 12 weeks was predictive of SVR.10 Very early dose adjustment of ribavirin might therefore be effective. Data from the literature show that after a single oral dose, the plasma concentration of ribavirin exhibits a three-phase profile: a rapid absorption phase with a mean time to the maximum concentration (tmax) of about 1.5 hours; a rapid distribution phase (half life of approximately 3.7 hours); and a long terminal elimination phase, the last measurable concentration time point being at approximately 100 hours after dose.3, 11 As a consequence, global exposure after a single dose should be measured between at least h0 and h100. However, this study was purposely conducted in a real-life situation where patients are dosed with ribavirin twice daily. Therefore, at D0, AUC0–12h mainly describes the absorption and distribution phases of ribavirin PK and is only slightly affected by the elimination phase, whereas AUC0–4h is clearly limited to the first two phases. In vitro studies showed that the intracellular concentration of ribavirin is maximal at 6 hours after dosing,12 suggesting that AUC0–12h and to a lesser extent AUC0–4h could reflect the maximum ribavirin effect.

In our study and according to the literature,3 patients' global exposure to ribavirin (that is, AUC) was widely distributed, despite dose-adjustment for BW; the AUC/dose ratio was highly variable (range 1.3 × 10−3 to 8.7 × 10−3 ng/hour/L), supporting individual dose-adjustment of the drug based on individual factors linked with the dose-concentration relationship or on measured exposure.

Individual factors probably influence ribavirin PK. In our study, the correlation between ribavirin D0 AUC and total or weight-standardized dose was not significant. This reflects the large interindividual variability of the dose-concentration relationship, that is, of ribavirin PK. Moreover, Wade et al. recently showed that lean BW was the only covariate with a clinically significant influence on ribavirin clearance and distribution volume, contrary to total BW.13 This study seems to partly confirm this result. Indeed, a statistically significant (though loose) correlation was observed between lean BW and AUC/dose at D0 (r2 = 0.16, P = 0.046). Bruchfeld et al. recommended that ribavirin dosage be mainly adjusted to Clcreat and not, as done in practice, to BW. However, in their study, the effect of renal function on ribavirin clearance was only apparent when Clcreat was less than 34 mL/minute.7 In all our patients renal function was close to normal. Other studies showed that ribavirin clearance was affected by BW, gender, age, and serum creatinine, but that these four covariates only explained 27%–40% of its interindividual variability.4, 7 Metabolism accounts for most of ribavirin elimination, but the sites involved are unknown. It is possible that the liver plays a role, although Glue et al.14 showed that hepatic dysfunction had no substantial influence on the apparent clearance of ribavirin. Only the mean Cmax and concentration-time profiles between 0 and 6 hours (that is, intermediate between our AUC0–12h and AUC0–4h), were different between the groups with normal liver function or moderate or severe liver disease, which is consistent with the present results. Anyway, all patients in this study had normal liver function. The gastrointestinal (GI) tract has recently been suggested to be the major site of ribavirin first-pass elimination.11 None of our patients had a known GI tract disease which may have induced abnormal absorption or altered GI metabolism of ribavirin. In particular, none had portal hypertension. Finally, high-fat meals can increase ribavirin bioavailability by 46% related to the fasting state,13 but in our study, all patients were given standard meals.

Consequently, it is unlikely that hepatic or intestinal dysfunction or comorbidity or food intake may have influenced these results.

The first attempts of ribavirin PK-PD studies in the literature were based on single-point sampling strategies. In their response model, Jen et al. concluded that serum ribavirin concentrations at “steady state” (W4) played a role in the SVR but was less influential than genotype or viral load.5 However, the interval between the last dose of ribavirin and blood sampling was not predetermined in this protocol and was not reported either. Larrat et al. also showed a relationship between SVR and a single serum concentration taken 2-4 hours after the morning dose of ribavirin at W12 (n = 24; P = 0.03).6 One must be careful when using exposure estimates based on poorly timed, single concentrations, even at the steady state. In this study, the ribavirin plasma concentration profiles at “steady-state”, that is at W12, still showed large differences between the concentrations measured at T0 and T12h (Fig. 1). The full interdose AUC, which is a measure of global drug exposure, was found to be more pertinent in terms of exposure-effect relationship than any single time point.

The choice of plasma instead of serum for ribavirin concentration measurements is also important. Plasma ribavirin concentrations have been shown to be more reproducible than serum in intensive-sampling pharmacokinetics studies,11 and the wide variability of the results obtained in PK-PD studies in the literature,4, 6 mostly retrospective, may be partly explained by the use of serum instead of plasma.

An assumption recently made by Dahari et al.15 is that measuring intraerythrocyte ribavirin concentrations could be a more relevant approach than measuring plasma levels. Their hypothesis is that intraerythrocyte ribavirin accumulation early on during treatment would be higher in patients more likely to achieve a SVR. But the irreversible intraerythrocyte accumulation of ribavirin-triphosphate is probably not representative of the drug turn-over in other target cells such as hepatocytes where ribavirin activation into triphosphate metabolites is reversible. The intraerythrocyte dosing of ribavirin should certainly be further investigated, but so far there is no scientific proof that it is a better approach than plasma concentrations.

Ribavirin AUC0–12h after the first dose was significantly linked with all types of responses (RVR, EVR, and SVR) in the per-protocol population. Indeed, the overexposed patients excluded for toxic symptoms, who are likely to also have a high probability of response, would have been considered as nonresponders in an “intent-to-treat” analysis, which would be contrary to the purpose of the present analysis aimed at defining a threshold exposure for achieving SVR.

The logistic regression analysis that took into account AUCs as continuous variables, and was conducted to explore the possible mechanisms of the impact of early exposure on SVR, led to a P value very close to 0.05 (Table 3). This observation supports the hypothesis that the probability of response increases with ribavirin exposure. However, achieving a minimum threshold AUC value might optimize the probability of response.

RVR was also predictive of SVR whereas the initial viral load was not. However, responders' initial viral load was 45% that of nonresponders'; although the difference was not statistically significant, this may have contributed to the better outcome in the former. The important point however, is that in logistic regression analysis the initial viral load was less predictive than AUCs and RVR, which remained as significant predictive factors of SVR even when initial viral load was taken into account.

For a more practical approach in routine clinical use, we propose the use of abbreviated AUC0–4h, which was highly correlated with the AUC0–12h as well as with all types of virological responses and required samples over a shorter time period (4 hours instead of 12).

In this study, we found no difference between sustained virological responders and non–sustained virological responders for full or abbreviated AUC at W12. A first explanation could be found in the data provided by Dahari et al.,15 suggesting that differences in ribavirin kinetics between sustained virological responders and nonresponders appear mainly during the first week. It seems therefore pertinent to focus attention on the very early phase of ribavirin tissue distribution. Moreover, after multiple dosing, ribavirin gradually accumulates in plasma and reaches a maximum asymptotic mean concentration in about 4 weeks. In our study, as well as in the literature, the concentration variability resulting from absorption and distribution is less apparent on the concentration-time curves at steady state; in other words, the absorption and distribution phases are drowned by the steady-state baseline at W12.16 Finally, Lindhal et al.8 suggested in their pilot study that the early use of very high doses of ribavirin led to a high SVR rate in patients with genotype 1 as well as to a high incidence of side effects.

Several hypotheses can be raised to explain the PD relevance of very early ribavirin PK. The hypothesis of an influence on the first-phase decline of HCV RNA is not likely. In the literature, Pawlotsky et al. found a small initial viral load decline (about 0.5 log) in a fraction of patients treated with ribavirin alone, but ribavirin did not enhance the first-phase decline in combination with interferon.17 In this study, there was no correlation between the initial viral load decline (0-24 hours) and ribavirin D0 AUCs (r2 = 0.07, P = 0.25).

Dixit et al. suggested that ribavirin might influence the second-phase decline of HCV RNA in a dose-dependent manner in patients with low interferon efficacy.18, 19 They raised the hypothesis that ribavirin would not increase the death rate of infected cells but would lower the viral infectivity by virus mutagenesis. The inherent frequency of viral mutagenesis might be low in ribavirin monotherapy. In the presence of interferon, however, viral production might be diminished and the relative ribavirin concentration per replicating genome in infected cells is elevated, which could increase the mutation frequency. In the HCV replicon system, ribavirin was shown to inhibit the ability of progeny subgenomic replicons to transfect new cells, although the replication rate remained unaffected,20, 21 However, a very recent in vivo study undermined this hypothesis.22

So far, the antiviral mechanisms of ribavirin remain unknown. However, increasing hepatocyte exposure to this drug at the time of the onset of interferon (first injection), might be crucial for virus eradication. A recent article by Feld et al.23 suggests that ribavirin enhances the response of interferon-stimulated genes to peginterferon and that pretreatment with ribavirin may heighten this induction, making cells more responsive to interferon and increasing the production of endogenous interferon. Other mechanisms such as the down-regulation of the genes involved in interferon-inhibitory pathways and hepatic stellate cell activation, or having a positive effect on the apoptotic pathways, have also been suggested in this study. Similar observations had been made in vitro with respiratory syncytial virus.24 In accordance with these findings or hypotheses, our results suggest that a high enough early ribavirin exposure (absorption and distribution phases) after the first dose of both ribavirin and interferon, reflecting the exposure of all types of cells to ribavirin, might favorably influence the second-phase decline of HCV RNA.

In addition, in this study, ribavirin AUC values at D0 but not at W12 were highly correlated with the hemoglobin level at W12, confirming the relevance of the early exposure indices. Jen et al. had similar results with a single ribavirin concentration at W4 and the W4 hemoglobin nadir, but they underlined the high variability of their data.4 The same authors in their retrospective toxicity model showed that steady-state ribavirin concentration and the baseline hemoglobin level were the most important factors in the occurrence of toxic events.5

The strong relationship observed in our study between D0 AUC and SVR suggests that individual dose adjustment in the first days of treatment may improve all types of virological responses. We also found that the AUC0–4h was as predictive as AUC0–12h and was a simpler tool because it required only a few samples to be collected within the first 4 hours after dosing.

We therefore propose a target ribavirin AUC0–4h ≥ 1755 μg/hour/L after the first dose for early individual dose-adjustment of ribavirin. Though the strategies previously proposed for ribavirin in the literature were based on later exposure, studies dealing with other drugs found significant relationships between exposure indices after the first dose and response of a cumulative nature.25

In this study, one patient with the highest AUC at D0 (6120 μg/hour/L) presented a high hematological toxicity of ribavirin at W8 and was excluded from the protocol despite a negative PCR at W4. This observation suggests that a maximum target exposure should be also defined.

If the exposure level required for treatment efficacy (AUC above the cut-off) is also responsible for increased hematological toxicity, use of erythropoietin can be proposed.

In conclusion, in this PK-PD study, we show that patients presenting a genotype 1 chronic hepatitis C disease and given interferon in association with ribavirin had a higher probability of being rapid, early, and sustained virological responders when they exhibited high interdose or abbreviated ribavirin D0 AUC. A large prospective, randomized trial is necessary to validate, in terms of improvement of drug efficacy, the individual ribavirin dose adjustment based on D0 abbreviated AUC, using the target value proposed herein.


We gratefully acknowledge for their collaboration: Dr. Gérard Babany, Dr. Isabelle Lonjon, and Dr. Natalia Kharlov; Roche France. We are indebted to Dr. Marianne De Vinzelle, Lydie Lemonnier, Sandrine Naturel, Annick Rigout the nurse from Limoges and the nurses from Angers, and Fabrice Quet from the department of Clinical Research Biostatistics.