TITLE: Cost-effectiveness analysis of therapeutic options for chronic hepatitis C genotype 3 infected patients.

BACKGROUND
This study provides a cost-effectiveness analysis of therapeutic strategies for chronic hepatitis C genotype 3 infected patients in Spain.


METHODS
A Markov model was designed to simulate the progression in a cohort of patients aged 50 years over a lifetime horizon.


RESULTS
Sofosbuvir (SOF) plus peginterferon and ribavirin for 12 weeks was a cost-effective option when compared to standard of care (SoC) in the treatment of both 'moderate fibrosis' and 'cirrhotic' patients. Incremental cost-effectiveness ratios were €35,276/QALY and €18,374/QALY respectively. ICERs for SOF plus daclatasvir (DCV) regimens versus SoC were over the threshold limit considered, at €56,178/QALY and €77,378/QALY for 'moderate fibrosis' and 'cirrhotic' patients respectively.


CONCLUSION
Addition of SOF to IFN-based regimens for genotype 3 was cost-effective for both 'moderate fibrosis' and 'cirrhotic' patients. IFN-free options including SOF and DCV association required price reductions lower than the list prices to be considered cost-effective.

The BOSON CT provided the evidence for SOF/pIFN/RBV (dosed as in non-cirrhotic) where the achieved SVR was 88%.
In the absence of CTs evaluating the SOF/DCV/RBV 24 weeks option in cirrhotic patients, efficacy parameters were obtained from the European Multicenter Compassionate Use Program in real-life clinical setting, where the achieved SVR was 92% [5].
The SVR rate for pIFN/RBV group was obtained from EPIC CT, with figures of 48% for this subpopulation [6].
Different variables introduced clinical and demographic heterogeneity in CTs populations (fibrosis, age, viral load, race…). However, patients included in CTs where SVR was obtained matched well (Table I).

2.-Cost estimations
Drug costs were estimated on the basis of the dosing and therapeutic schemes included in the CTs and SmPC. A perfect adherence was considered on estimations. Cost estimations for pIFN/RBV option took over a 13% of treatment discontinuation at week number 12 due to inadequate response [7]. The costs of the different therapeutic options are shown in the manuscript Table 1. Drug adverse reactions are often associated with an additional cost related to drugs consumption, medical visits, hospital admissions and other resources generated by their treatment. Oral DAA combinations are generally associated with a good tolerance and no additional cost was considered. IFN containing regimens are not as well tolerated, and flu-like symptoms and anemia are the most relevant side effects [4,6]. The first one could not have relevant direct associated costs, and the second one is often managed with RBV dose reduction. The impact of these side effects on HRQL has been included in associated treatment disutility.
A poorer tolerance to IFN containing regimes is regarded for cirrhotic patients, so an associated side effects cost was considered for this subpopulation with an incidence of 12% neutropenia and 15% anemia [8,9] (Table II).
Health care resource costs were obtained from hospitals in the Basque Health Service in 2014. The estimated cost for each disease state is summarized in Table III. The transition costs were differentiated from state costs. The former corresponds to the in-hospital care of patients owing to different complications related to chronic liver disease. The latter includes the cost of resources used in the follow-up [10]. Inhospital care costs for decompensated cirrhosis, hepatocellular carcinoma and liver transplant were obtained from the Basque Country Public Hospitals Statistics Costs System in 2014, based on the patient classification system DRG. Categories corresponding to each liver chronic condition were used for all public hospitals in the Basque Country.
Unit costs were obtained from an ad hoc micro-costing study carried out in all Basque Public Health-Care System hospitals. The costs of in-hospital events were based on the DRG system. Accounting Departments deliver annually resources unit costs and DRGs categories costs for the whole Basque Health Care System.
The National Health Care System in Spain share the same funding system for every region in function of the population what means that costs from all regions are similar. These unit costs were checked with those in the Spanish Network of hospital costs (RECH) (http://www.rechosp.org/rech/faces/es/jsf/index.jsp) and the results matched well.

3.-Health Related Quality of life -Utilities
Health-state utilities are summarized in Table IV. Side effects associated with treatment affect patient's HRQL. Although there were no available data comparing disutility associated with the options evaluated, estimations were performed based on HRQL reduction for dual therapy (pIFN+RBV) published by Grieve et al. [11] Good tolerance of IFN-free regimens was considered to have a 50% decrease disutility respect to the SoC. Length of treatment was considered for final estimations in all cases.

4.-Markov model
The model starts at "moderate fibrosis" hepatitis C or "compensated cirrhosis" state where the patient may progress to more advanced disease. Patients in those initial states are susceptible to being treated or not with different antiviral therapeutic alternatives. A SVR from moderate hepatitis, considered the patient to be cured, progressing as the general population. Cirrhotic patients with SVR may evolution to decompensated cirrhosis or HCC at lower transition rates [12,13]. Mild fibrosis population was not considered in the analysis.
When no SVR is achieved, the disease follows a process with different steps: moderate hepatitis, cirrhosis, decompensated cirrhosis, hepatocellular carcinoma, liver transplant and death.
Transition probabilities were obtained from the literature and the revision by Townsend et al. [12][13][14][15][16][17][18][19]. Those probabilities are presented as annual probabilities in the manuscript (to facilitate comparison between studies) and as quarterly probabilities in Table V. Given the fast changes that occur during the advanced stages of chronic liver disease, the Markov cycle length was one trimester (3 months), as an annual cycle does not adequately reflect the changes that a patient undergoes. Quarterly probabilities were obtained first by calculating the annual rates and by dividing them by 4 to apply the formula p = 1-e -r (where p is the probability and r the quarterly rate) [20]. Deterministic probabilities are transformed to quarterly rates by applying the formula r=-Ln(1-p)/4 and then converted to quarterly probabilities with the previous formula.
The representation of decompensated cirrhosis in cost-effectiveness models varies widely in the literature because of its complex nature. The patient can begin with one or more types of decompensated cirrhosis and become subsequently hospitalized for an alternative form of decompensation. It was followed an approach based on a retrospective study that classified patients into four groups as a function of the first decompensation: ascites, hepatic encephalopathy, gastrointestinal hemorrhage due to portal hypertension and severe bacterial infection [16]. Based on the specific information from each group as a function of the initial decompensation, the re-hospitalization rates for each type of decompensation and death, hepatocellular carcinoma and transplantation rates were calculated. Thus, the structure of the model ensured that each type of decompensation could occur in each group. As a result, the costs as a function of the numbers and types of re-hospitalizations as well as the probability of death can be calculated separately for each group.
The patient could reach hepatocellular carcinoma status from decompensated cirrhosis or cirrhosis. The probability of transition from compensated cirrhosis to hepatocellular carcinoma is 1.4% per annum, according to Fattovich et al. [15] Transition probability from decompensated cirrhosis to hepatocellular carcinoma was obtained from Planas et al. [16] The annual mortality rate was 42.7% per annum in the hepatocellular carcinoma state [15]. Liver transplantation may be the appropriate treatment for two possible states of the disease, decompensated cirrhosis and hepatocellular carcinoma. Other models do not include this second case, but as 23.9% of all liver transplants performed in Spain in 2010 were for this reason, this transition is justified [18]. Post-transplantation mortality is higher during the first year after transplantation (15% annually) than in subsequent years (3%) [18]. Table VI presents the basic figures of events and follow-up by type of initial decompensation obtained from Planas et al, [16] which served to calculate both deterministic probabilities and beta distributions. Table VII shows the quarterly probabilities mean values applied in the deterministic models and the alpha and beta parameters used for each transition in the probabilistic models to calculate the quarterly rates. Those rates were converted to probabilities (p) with the equation p = 1-e -r . The alpha parameter was the number of events for each transition and the beta was the follow-up time in quarterly minus the number of events because the mean rate is alpha/(alpha+beta).

5.-Base case results
Two figures have been included in the Technical Annex to represent the results of the base case in moderate fibrosis and cirrhotic patients (Figure I and II).

6.-Probabilistic sensitivity analyses
This approach is based on randomly varying all the parameters within a distribution range at the same time in Beta distributions parameters were calculated based on the confidence intervals obtained from the literature.
Beta distribution is the appropriate function for probabilities given its range from zero to one. Its parameters are alpha and beta and the mean is calculated as alpha/(alpha+beta). Minimum and maximum figures described by Townsend et al. were used as an approximation of 95% confidence intervals when they were not available. Therefore, the beta functions were estimated with the adequate quarterly mean and 95% confidence intervals. The availability of patient-level data for "transition costs" allows applying a log-normal distribution and the inclusion of the information in the probabilistic sensitivity analysis ( Table VIII). The cost of followup was based on the expert advice regarding resource consumption and the unit resource cost from Basque hospitals. Estimations were based on previous analysis [8]. Resources consumption and unit costs are shown in Table III.
"Utilities" were introduced as beta distributions ( Table IV). The parameters for different beta distributions were calculated based on the confidence intervals from the literature [11].
Probabilistic sensitivity analysis results are shown in manuscript figures 2 and 3 for "base-case" costeffective options (SOF/pIFN/RBV for 12 weeks either in "moderate fibrosis" or "cirrhotic" patients).

Figures III and IV
in the Technical Annex, show the results for "base-case" non-cost-effective options at official prices (IFN free regimes for both "moderate fibrosis" and "cirrhotic" patients).

7.-Model validation
Estimations of life expectancy in different health states of hepatitis C were made to validate the model.
Both "life expectancies" from different states and the percentage of patients who progress to cirrhosis (21% at 20 years and 37.4% at 30 years) were calculated (Table IX). Those were similar to other models' results [17]. Hepatic mortality (advanced liver disease, liver transplant, and liver related mortality) was separately calculated from other-cause mortality in a 49-years-old chronic hepatitis C patients' cohort. Liver-related mortality accounted for 24% of total mortality. SVR patients' survival was similar to that of the general population. As expected, early stage chronic hepatitis C patients die from other causes and when decompensated cirrhosis stage is reached, death occurs due to liver disease complications.
SOF_DCV: SOF+DCV for 12 weeks SOF_DCV_DISCOUNT: SOF+DCV for 12 weeks with a drug acquisition cost reduction of 40% Figure IV. Base-case "Cost-effectiveness" plane and "acceptability curve" for "SOF/DCV/RBVx24w" vs "pIFN/RBVx48w" in cirrhotic patients (a 40% discount in drug acquisition costs capped to 12 weeks was also drawn in the acceptability curve).
SOF_DCV_RBV: SOF+DCV+RBV for 24 weeks SOF_DCV_DISCOUNT: SOF+DCV+RBV for 24 weeks with a drug acquisition cost reduction of 40% and capped to 12 weeks