Changes in anti-viral effectiveness of interferon after dose reduction in chronic hepatitis c patients: a case control study

Background High dose interferon induction treatment of hepatitis C viral infection blocks viral production over 95%. Since dose reduction is often performed due to clinical considerations, the effect of dose reduction on hepatitis C virus kinetics was studied. Methods A new model that allowed longitudinal changes in the parameters of viral dynamics was used in a group of genotype-1 patients (N = 15) with dose reduction from 10 to 3 million units of interferon daily in combination with ribavirin, in comparison to a control group (N = 9) with no dose reduction. Results Dose reduction gave rise to a complex viral kinetic pattern, which could be only explained by a decrease in interferon effectiveness in blocking virion production. The benefit of the rapid initial viral decline following the high induction dose is lost after dose reduction. In addition, in some patients also the second phase viral decline slope, which is highly predictive of success of treatment, was impaired by the dose reduction resulting in smaller percentage of viral clearance in the dose reduction group. Conclusions These findings, while explaining the failure of many induction schedules, suggest that for genotype-1 patients induction therapy should be continued till HCVRNA negativity in serum in order to increase the sustained response rate for chronic hepatitis C.


Background
The hepatitis C virus (HCV) causes a slowly progressive liver disease, which may lead to cirrhosis, liver failure and liver cancer. Currently, about 10,000 patients die in the US from HCV related disease yearly and this number is expected to triple in the next 2-3 decades [1] Anti-viral therapy is successful in arresting the progression of the disease in those patients who reach a sustained clearance of the virus, currently only 40% of treated patients [2]. Response to therapy with alpha Interferon injections thrice a week with or without additional Ribavirin is thought to occur gradually over time, and research has focused on improving efficacy by prolonging treatment up to 1 to 2 years [2][3][4]. However, reports on viral dynamics analysis show that response to interferon is very fast and that a 10 to 1000 fold decrease in viral load can be reached within 24 hours of treatment. [5][6][7] The pattern of viral decline seems to be biphasic, with a rapid viral decline within the first 24-48 hours followed by a much slower second phase of viral decline. This biphasic decline is hypothetically caused by a direct anti-viral effect of interferon in blocking virion production from infected cells [5].
A strong dependence of the viral decline in the first phase on the dose of interferon used has been described [5,8]. Nevertheless, it has been shown that it is the second slope which is the best predictor for response to treatment [5,9]. This slower second phase slope of viral decline has large variability between patients, and therefore cross sectional analysis of its dose dependence is hindered. Instead, here we investigated the longitudinal changes in viral dynamics in patients going through a dose reduction in order to asses the effect of dose on the second slope. The current model for HCV dynamics, in which the dynamical parameters are fixed during treatment, fits the observed biphasic viral decline in patients treated with fixed Interferon dosages [5]. However, in this study we adapted the model such that the dynamical parameters can change over time due to a change in dose. We now report that early dose reduction is followed by a rise in viral load, that can only be explained by a decrease in interferon effectiveness; so the potential benefit of a rapid viral decline following high dose induction is often lost.

Methods
A case-control study was performed in an university-based tertiary referral center. Informed consent was obtained from all patients, and the human experimentation guidelines of the University Hospital Rotterdam were followed in the conduct of clinical research.

Study population
All 24 HCV genotype-1 patients enrolled in our high-dose induction studies were evaluated. All patients met the inclusion and exclusion criteria previously described [10]. Note that all patients in these studies were considered "difficult-to-treat", either because they were non-responders to previous treatment or had cirrhosis and/or high baseline viral load. Group 1 patients (N = 9) received 10 million units (MU) of Interferon-α-2b (Intron-A, Schering-Plough) daily for 4 weeks. Group 2 patients (N = 15) received 10 MU of interferon daily for the first 3 days only, followed by 3 MU interferon daily from day 3 until day 28. In both groups, ribavirin was given orally in divided doses of 1,000-1,200 mg daily (according to weight >75 kg). Subsequently, all patients received a maintenance treatment of minimally 3 MU interferon daily for 52 weeks. The patients' baseline characteristics (Table 1) were well balanced, except for a trend for larger number of cirrhotic patients in group 1.

Detection of Serum HCV RNA
Plasma samples were collected frequently during the first 4 weeks of treatment for HCV RNA detection. Blood samples were collected in Vacutainer PPT tubes (Becton-Dickinson) which were spun directly after collection in order to avoid RNA breakdown. The spun PPT tubes [11] were then transported to the virology department where plasma was aliquotted in 5 separate tubes that were stored at -80°C. Plasma samples were obtained at day 0 (0, 4, 8, 12, 16 hours), day 1 (24, 32 and 40 hours), day 2 (48 and 56 hours), day 3 (72 and 80 hours) day 4, 5, 6 7, 10, 14, 17, 21 and 28 after treatment initiation. Viral load was quantified using the Cobas Amplicor Monitor™ version 2 (Roche Molecular Systems). Since the linearity of quantitative assays for high numbers of viral copies has been questionable [12] we routinely diluted samples and re-tested, if the early quantification of that sample was higher than 10 6 copies/ml.

Mathematical Modeling
Viral kinetics were analyzed using a modification of a previously described mathematical model for viral dynamics [5], for which the analytical solution is, This formula contains several dynamical parameters (c, δ, H and ε) which may vary per patient according to the best fit of the actual data, but are constant over time. c de- scribes the clearance rate of free virus, with the corresponding virus half-life of ln(2)/c. δ describes the loss rate of productively infected cells, with the corresponding cellular half-life of ln(2)/δ. The effect of interferon can be modeled here either by a block of de-novo cell infection with effectiveness H (0 ≤ H ≤ 1), or block of virion production with effectiveness ε (0 ≤ ε ≤ 1). The logarithmic drop in viral decline during the first phase (24-48 hours) of treatment can be approximated by log(1-ε). The 2 nd phase slope can be approximated by ε times δ when H << 1, or by δ alone when H ≅ 1.
To investigate the effect of reducing treatment dose, all the above dynamical parameters were allowed to change over time in the solution, e.g. for interferon effectiveness in blocking virion production, ε, we use the function ε (t) : where t 1 is the time of dose change and k is a exponential rate representing how rapid does the change in interferon dose effect the change in the parameter. Thus, the blocking effectiveness starts at ε 1 (for t ≤ t 1 ), and changes with an exponential transition to ε 2 (after (t-t 1 ) >>1/k). The same functional form was used to investigate changes in H (from H 1 to H 2 ), δ (from δ 1 to δ 2 ), and c (from c 1 to c 2 ).
It is important to note that we do not explicitly model in Eq. 1 the dynamics of viral replication after the dose reduction, but rather replace the fixed parameters by time dependent parameters in the original analytical solution obtained with fixed parameters. Nevertheless, we have tested this approximation by simulating a modification of the original differential equation model [5] where changes in the dynamical parameters were allowed to change at the time of dose reduction. We found no significant difference between the simulation of the full modified differential equation model and the modified analytical solution.
Since we only have 2-3 viral measurements immediately after dose reduction, we can not estimate the appropriate replication parameters and thus chose to use the simple approximation given in Eq 1.
To estimate HCV viral kinetic parameters for each patient, the logarithm of V(t) in Eq. 1 (using ε (t) from Eq. 4) was fit to the logarithm of the viral load data by a non-linear least squares method using the Madonna software (R.I. Macey and G.F. Oster, Berkeley, CA, USA). Two patients in group 2 were missing viral load data during the first week of treatment and one patient had a null response (less than 3 fold change in viral load during treatment) and therefore their viral kinetics could not be fitted.

Statistical analysis
The Fisher-exact test (2 × 2 tables) and the Chi-square test (N × N tables) were used to determine the statistical significance of the distribution of categorical variables between groups. The non-parametric independent (or related) Mann-Whitney rank sum test was used to determine the statistical significance of differences in continuous variables between the two groups (or of changes in the parameters within the same patients). Correlation among parameters, or between parameters and baseline values, was evaluated using the Spearman non-parametric test. Significance was established at P < 0.05.

Results
The biphasic decline previously reported [5][6][7]13] describes the viral kinetics during the first month for all patients in group 1 (Fig 1, Fig 2a), for whom interferon dose was kept constant. In contrast, we observed a complex dynamic pattern for 11 out of 12 evaluable patients in group 2 (Fig 1, Fig 2b,c,d). In these patients, a rapid decline occurred during the first day, followed by a slower decline on the second and third days, at which point a rapid increase in viral load (mean 0.8 [range 0-1.3] log copies/ ml) is observed within 24-48 hours after the reduction in interferon dose. Thereafter, viral load again declined with a mean exponential slope comparable to the second phase slope in group 1.
We have tried to fit the viral kinetics of the patients in group 2 with several models, in each one of them allowing to change one parameter (ε, H, c or δ) at the time of dose reduction. The only model that was able to qualitatively reproduce the observed kinetics was the one which allowed a longitudinal change in the interferon effectiveness in blocking virion production (ε) as function of the interferon dose (Eq. 4). By only allowing to change the interferon effectiveness in blocking de-novo infection (H), death rate of infected cells (δ) or the clearance rate of free virions (c), it was not possible to fit the observed data. When assuming the major effect of interferon is to block virion production in a dose dependent way (1 > ε 1 > ε 2 > 0), it was possible to fit the data both with H = 0 or H = 1.
Thus it was not possible to determine if interferon also blocks de-novo infection, in addition to blocking virion production, or not. Varying H between 0 and 1 only gives rise to minimal changes in the estimate of ε and c, while somewhat affecting the estimate of δ when ε is smaller than 0.98 (minimal estimate of δ obtained for H = 1, and maximal estimate for H = 0). Moreover, when allowing ε to change at the time of dose reduction, it was not possible to rule out that the other parameters also change at the same time.
For simplicity, and since our data only implies minute effects due to changes in the other parameters, we have as-sumed a change occurs only in ε when estimating the dynamical parameters. In addition, we needed to estimate the transition rate k (Eq. 4) from ε 1 to ε 2 . It was not possible to get a unique estimate of k for each patient individually, with only 2-3 measurements during the rebound. Since using k = 1 up to 20 did not significantly affect the estimate of the other parameters, we have assumed k = 2 for all patients such that the effect of ε 1 vanishes within 24-48 hours after the dose reduction in accordance to the observed data.
The estimates obtained by non-linear fitting of each patient's viral kinetics individually are given in Table 2. As expected the baseline viral load, the half-life of free virions and the initial effectiveness in blocking virion production, all related to the first phase decline, were similar between the 2 groups. Mean half-life of free virions for all patients was 3.0 hours ± standard deviation of 1.6 hours, similar to that derived in previous studies for interferon-α monotherapy (2.7 hours) [5]. The mean effectiveness in blocking virion production for all patients was 95.6% ± 6.5% for the 10 MU interferon dose. For group 1 the non-linear fit (with Eq. 4) did not give rise to a significant change in ε over time, in accordance to the constant daily interferon dose used in these patients (insert Fig 2a). For group 2, however, a significant (p < 0.005) decrease in effectiveness of blocking production was observed (inserts Fig 2b,c,d,   3a) from an average of ε 1 = 94% ± 8% to ε 2 = 69% ± 27%. The effectiveness of interferon with 10 MU (ε 1 ) and with 3 MU (ε 2 ) were pair wise correlated (R = 0.8, P < 0.001).
Cross-sectionally there was no difference observed between the 2 groups in the death rate of infected cells (Table 2) or in the second phase slope (Fig 3b). However, longitudinal analysis of group 2 patients revealed a significant decrease from the predicted second phase slope, which would have occurred without a dose reduction, to the actual slope after dose reduction in 4 out of 12 pa-tients (Fig 4a). These are the patients with the largest decrease (range 20% to 68%) from ε 1 to ε 2 , since the second slope is a linear function of the interferon effectiveness in blocking production (ε).
The clinical consequences of the viral dynamics data relate to the predicted time of HCV RNA negativity, which is strongly dependent on the second phase slope. The predicted time to HCV-negativity with the biphasic model (median 3.5 weeks) was confirmed by the observed individual data of group 1 patients (median 4 weeks), and the dose-reduction model (with the reduced 3 MU dose) pre-    dicted the time to HCV-negativity in group 2 patients (medians 12 and 14.5 weeks, Table 3). Reducing the interferon dose to 3 MU daily decreased the predicted number of patients that would become HCV-negative within 12 weeks from 12/15 with the 10 MU dose to only 7/15 with the reduced 3 MU dose, as indeed confirmed by the actual individual data (8/15).

Discussion
Our results indicate that the effect of interferon dose reduction on viral dynamics can be completely attributed to decrease in the effectiveness of interferon in blocking vir-ion production (ε). Changes in other parameters, such as blocking de-novo infection (H) loss rate of infected cells (δ) and clearance rate of free virions (c), without a change in blocking production, can not reproduce the observed kinetics. The longitudinal dose dependence of the interferon anti-viral effectiveness observed here corroborates the dose dependence of interferon effectiveness previously described only cross-sectionally [5]. The strength of the current results is that the dose dependence of the effectiveness cannot be attributed to baseline differences between the patients. Since longitudinal changes in the other parameters of this model (such as H, δ and c) do not give rise  to significant differences in the kinetics, we can not rule out combined effects of changes in the other parameters concomitantly with the change in effectiveness of blocking production (ε).
In turn, the dose dependency of the interferon effectiveness determines the result of both the 1 st phase and the 2 nd phase kinetics. The effectiveness in blocking virion production with 10 MU before dose reduction (mean 95.6% and a mean viral decline of 1.8 log cp/ml) was similar to that of a previous study with 10 MU interferon (96%) [5]. However, we show that the benefit of the initial viral decline due to 3 days of high induction dose was lost after the dose reduction in almost all patients (Fig 1,  3). The effectiveness in blocking virion production with 3 MU after the dose reduction (mean 69.1% and a mean viral decline of 0.7 log cp/ml) was similar to that estimated in a previous study with 3 MU interferon initially (70%) [8]. As a consequence, the viral kinetics after the dose reduction in group 2 patients is similar to the kinetics that would have been obtained if the patients had started with the reduced dose to begin with (green line Fig 2).
In contrast to the 1 st phase viral decline, which exponentially depends on the interferon effectiveness, the 2 nd phase slope is a linear function of the effectiveness in blocking virion production. This slope is the most predictive parameter for treatment outcome, with a threshold of 0.13 days -1 below which no sustained response was observed [5] While the increase observed in viral load imme-

Figure 4
The effect of the change in interferon efficacy after the dose reduction (x-axis) for the patients in group 2 is related to the change in slope of the second phase viral decline (open circles: second phase slope before dose reduction; horizontal lines: second phase slope after dose reduction). In 4 patients the slope of the second phase reduces drastically, note that 3/4 of these patients have the largest reduction in interferon efficacy. The dashed line reflects the threshold of 0.13 days -1 below which no sustained response was observed. diately after dose reduction can delay the time to negativity by several weeks at the most (patients 2-J and 2- C Fig 2), the decrease in the second phase slope can radically reduce the chance for HCV-RNA negativity (patient 2- H Fig 2). Therefore the decrease in the second slope could be crucial for their success of treatment. Indeed, the number of patients predicted to become HCV-negative within 12 weeks with the high induction dose was drastically reduced due to the dose reduction (compare the 10 MU column versus 3 MU column in Table 3).
Interestingly, the results obtained here are for interferon and ribavirin combination treatment, while the results from previous studies [5,8] are for interferon monotherapy. On the other hand, in this study most patients are non-responders or cirrhotic rather than normal naïve patients as in the previous studies [5,8]. Thus, we can not conclude if ribavirin has an additive effect on initial viral decline or not.
Is induction treatment beneficial at all considering that following dose reduction the virus rebounds back to the level it would reach anyway with the reduced dose? Previous studies with a longer period of induction treatment (14 days) do not show a consistent viral rebound as observed here [14]. Moreover, studies of prolonged induction treatment (> 28 days) do not show any re-emergence of virus production [15]. Therefore, it could be suggested that an induction period of 3 days is too short, but longer induction periods, which continue until HCVRNA nega- Detection level for HCV RNA negativity is at < 500 copies/ml. 2) Patient 1-D had a viral breakthrough after 12 weeks and was HCV RNA positive at 24 weeks of treatment. 3) Patients with positive HCV RNA at 24 weeks stopped treatment according to protocol. 4) Patient 1-I was non-compliant from 8 weeks on according to self-report tivity in serum, might give rise to continuous suppression of viral replication. In all likelihood this concept may partly explain the increased efficacy of pegylated interferon [16,17] since the rather constant interferon levels with this type of interferon result in a constant block of virus production.

Conclusions
Dose reduction to 3 MU daily after 3 days of 10 MU of interferon daily in HCV genotype-1 patients negated the extra virus suppressive effectiveness of the induction dose. These observations, while explaining the failure of many induction schedules suggests that induction therapy should be continued till HCV RNA negativity in serum in order to increase the sustained response rate in chronic hepatitis C.