Introduction

Hepatitis C virus (HCV) is a ribonucleic acid (RNA) virus causing acute and chronic hepatitis [1]. Worldwide, the HCV prevalence is about 3% [2]. In Europe and the US, HCV infection through injection drug use has become the major transmission route [3]. Although most patients infected with HCV are symptomless, chronic hepatitis C (CHC) poses a significant risk of developing cirrhosis and hepatocellular carcinoma, if left untreated [4]. Thus, CHC is a cause of major health burden causing substantial morbidity and mortality [5]. In the US, costs of about 6.5 billion per year are estimated [6] despite the availability of antiviral therapy.

The aim of therapy in chronic hepatitis C is to achieve a sustained virological response (SVR). SVR is defined as undetectable serum HCV RNA after the end of treatment, signalling eradication of HCV infection [7]. SVR at 12 weeks has shown high concordance with SVR at 24 weeks [7] and has been accepted by regulators in the US and Europe as an appropriate endpoint indicating treatment success [8]. Response to HCV treatment differs according to HCV genotype, disease stage, and HCV treatment history [9]. Pegylated interferon alpha and ribavirin have long been considered standard of care [8] with SVR rates of 40–50% in genotypes 1 and 4 [10] and SVR rates up to 80% in patients with genotypes 2 and 3 [11]. Due to significant side effects and contraindications associated with pegylated interferon alpha and ribavirin therapy, direct-acting antivirals have been developed [9]. Protease inhibitors such as telaprevir and boceprevir have been licensed since 2011 for HCV genotype 1 and have increased SVR rates, but major safety and efficacy issues persist [9].

Sofosbuvir, a newly developed uridine nucleotide analogue HCV NS5B polymerase inhibitor, has shown high antiviral activity across genotypes and few severe side-effects in a range of clinical trials including various patient populations [12–16]. In treatment-naïve genotype 1 patients, triple therapy with sofosbuvir, pegylated interferon alpha and ribavirin for 12 weeks reached a SVR of 89% in a phase III trial [13]. SVR was 96% and 83% in a phase II trial enrolling treatment-experienced patients with genotypes 2 and 3 receiving 12 weeks of triple therapy with sofosbuvir, pegylated interferon alpha and ribavirin [16]. Treatment with sofosbuvir in combination with pegylated interferon alpha and/or ribavirin recently received approval for reimbursement by the Swiss statutory health insurance in patients with CHC and fibrosis stage 3 or 4, or symptomatic patients with extra hepatic manifestations [17].

The aim of this cost-effectiveness analysis was to estimate clinical effectiveness in terms of quality-adjusted life years (QALYs) gained, the direct medical cost, and the cost-effectiveness in terms of cost per QALY gained, of sofosbuvir-based treatment strategies compared with the current standard treatment of mono-infected patients with CHC genotypes 1–4. The article follows the CHEERS statement for reporting health economic evaluations [18].

Materials and Methods

Health economic model characteristics

We used a Markov state-transition model with a life-long time horizon implemented in MS Excel; the structure is shown in Fig 1. The model structure of two health economic models developed by the Southampton Health Technology Assessment Centre [21,22] that were submitted to the UK National Institute for Health and Care Excellence (NICE) served as basis and was adapted to achieve reconciliation with available data from clinical trials. Specifically, mild (METAVIR [23] score F0-F1, absent or portal fibrosis) and moderate HCV (METAVIR score F2 or F3, portal fibrosis with few or several septa) states included in the original models were combined to a non-cirrhotic state to reflect available data from clinical trials. In the resulting structure, virtual mixed cohorts of 10,000 patients with no cirrhosis or compensated cirrhosis entered the model at the beginning of CHC treatment. Possible health states included non-cirrhotic with SVR, cirrhosis with SVR, decompensated cirrhosis, hepatocellular carcinoma, liver transplant, post-liver transplant and death (Fig 1). Disease-specific mortality (excess mortality) and general mortality were included separately; patients were exposed to age-specific probabilities of death in the general population [24] in each health state. The possibility of recurrences and relapses were tested in sensitivity analysis (dotted arrow in Fig 1). The disease was assumed not to progress while patients were on treatment and 12 weeks after the end of treatment. Patients could not die during treatment. A half-cycle correction was used for costs as well as effects. Costs (Swiss Francs, CHF 2014) and effects were discounted at 3% (range in sensitivity analysis: 0–5%) as recommended in current literature [25]. We calculated the cost-effectiveness of sofosbuvir in terms of cost per quality-adjusted life year (QALY) from the perspective of the Swiss health care system (indirect costs not included). There is no published or generally accepted cost-effectiveness threshold for Switzerland but other Swiss cost-effectiveness studies have used CHF 100,000 per QALY gained to distinguish potentially favourable from unfavourable incremental cost-effectiveness ratios (ICERs) [26–28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Structure of the Markov state transition model.

Patients can transition from each health state to death from any cause.

https://doi.org/10.1371/journal.pone.0126984.g001

Model input parameters and assumptions

Transition probabilities, quality of life weights, treatment efficacy and safety data, resource use data, and treatment-related adverse event (AE) costs were obtained from published clinical trial reports, other peer-reviewed literature and treatment guidelines. Patient characteristics, all-cause mortality, treatment-related costs and health state costs were extracted from Swiss-specific sources.

Annual transition probabilities between health states were extracted from published literature as referenced in Table 1. Probabilities of death in the general population were obtained from the Swiss Federal Statistical Office [24].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Base-case values for model input parameters and ranges used in the sensitivity analyses.

https://doi.org/10.1371/journal.pone.0126984.t001

Treatment duration, efficacy and safety were obtained from randomised and non-randomised phase II and phase III clinical trials as presented and referenced in Table 2. The percentages of patients and the time point at which these patients have discontinued treatment due to adverse events (AEs) or other reasons before the end of the planned treatment duration are provided in Table 2. Incidences of adverse events (AEs) such as anaemia, nausea, vomiting and rash are provided in S1 Table. Efficacy data were obtained from phase II trials if the corresponding phase III trials had missing data, e.g. did not report SVR rates according to cirrhosis status, or the treatment strategy did not match, e.g. SVR rates were reported for 12 instead of 24 weeks of treatment. There was one head-to-head trial available in genotypes 2 and 3 patients which compared a combination of sofosbuvir and ribavirin with a regimen consisting of pegylated interferon alpha and ribavirin [13]. Other efficacy and safety data were obtained from available randomised trials and single-arm studies with comparable patient characteristics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Genotype-specific parameters including treatment efficacy, duration and safety.

https://doi.org/10.1371/journal.pone.0126984.t002

Quality of life weights (utilities) for the health states represented in the model, based on the European quality of life questionnaire (EQ-5D), were obtained from published literature (Table 1). For patients on treatment with sofosbuvir or a comparator, a utility decrement obtained in a sofosbuvir trial based on the short-form health survey (SF-6D) was applied [13]. When non-cirrhotic or cirrhotic patients reached SVR after treatment, utilities obtained in a UK randomised controlled trial of interferon and ribavirin based on the EQ-5D [29] were added to the utility of the corresponding health state (Table 1).

Resource use parameters such as tests and procedures related to CHC diagnosis and treatment were obtained from EASL Clinical Practice Guidelines [8,20] (S2 Table). Swiss clinical experts checked whether they reflected Swiss clinical practice. Experts were selected to represent all Swiss regions and university hospitals. In Swiss regions without a university hospital, clinical experts from hospitals experienced in the treatment of HCV patients were included.

For every resource use parameter, we obtained unit costs, if possible, using published Swiss sources or adjusting foreign sources to reflect Swiss practice. Drug unit costs for HCV treatment and treatment-related AEs were derived from the list of pharmaceutical specialties of the Swiss Federal Office of Public Health [40]. AE management costs were obtained by multiplying incidences of AEs such as anaemia, nausea, vomiting and rash with drug unit costs for their treatment. Unit costs for monitoring were obtained from Swiss tariff lists [41] or the analyses list of the Swiss Federal Office of Public Health [42]. If not otherwise indicated, unit costs were for the year 2014. Health state costs for the year 2012 were collected at the Gastroenterology and Hepatology clinic, University Hospital of Zurich. They thus represent a single centre experience. All cost parameters are shown in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Cost parameters and their ranges used in the deterministic and probabilistic sensitivity analyses.

https://doi.org/10.1371/journal.pone.0126984.t003

Results

Base-case analysis

In the base-case analysis including a mixed population of about 25% cirrhotic and 75% non-cirrhotic CHC patients as seen in the SCCS, the incremental cost-effectiveness of therapy with sofosbuvir depended on genotype, treatment history, and interferon tolerability (Table 4). Sofosbuvir-containing treatment regimens in patients with CHC genotype 1–4 led to base-case ICERs between CHF 10,337 and CHF 91,570 per QALY gained. ICERs are below CHF 50,000 per QALY in treatment-naïve interferon eligible genotype 1 patients and in treatment-naïve interferon ineligible genotype 2 and 3 patients.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Summary of base-case, sensitivity and subgroup analysis results.

https://doi.org/10.1371/journal.pone.0126984.t004

ICERs for treatment-naïve genotype 1 and 4 patients receiving sofosbuvir, pegylated interferon alpha and ribavirin for 12 weeks ranged from CHF 10,337 to CHF 36,108 per QALY. The lowest ICERs for genotype 1 were obtained against telaprevir-containing regimens (CHF 10,337 per QALY) and boceprevir-containing regimens (CHF 13,276 per QALY). Underlying drug cost differences were smaller in these cases than in comparisons of sofosbuvir, pegylated interferon alpha and ribavirin with pegylated interferon alpha and ribavirin alone (Table 5). In treatment-naïve, interferon-ineligible genotype 1 patients, the treatment-related and total costs for a sofosbuvir-based treatment compared to no treatment were higher, as reflected in a higher ICER (CHF 86,648 per QALY).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Cost (CHF) per patient and cost-effectiveness results in genotype 1.

https://doi.org/10.1371/journal.pone.0126984.t005

In genotype 2, treatment-naïve patients we observed a substantial difference in the ICERs (Table 4) for interferon-eligible and interferon-ineligible patients (CHF 76,526 vs. CHF 10,471 per QALY). With a sofosbuvir-based treatment, interferon-ineligible patients gained 2.6 QALYs compared to no treatment, whereas interferon-eligible patients gained 0.6 QALYs compared to treatment with pegylated interferon alpha and ribavirin (S3 Table).

In genotype 3, ICERs of interferon-ineligible patients were in the same range for treatment-naïve (CHF 34,826 per QALY) and treatment-experienced (CHF 45,935 per QALY) patients (Table 4). The limited difference in ICERs was due to moderately lower QALY gains in treatment-experienced patients (2.1 vs. 2.5 QALY). We also noted a large difference in the cost per QALY gained of two different sofosbuvir-containing regimens, versus the same comparator, for the treatment of genotype 3, interferon-eligible patients (S4 Table). A treatment strategy of sofosbuvir in combination with pegylated interferon alpha and ribavirin for 12 weeks was clearly more cost effective than a 24 week-alternative without interferon in treatment-experienced (ICERs CHF 16,235 per QALY vs. CHF 74,805 per QALY, respectively) and treatment-naïve patients (ICERs CHF 38,512 per QALY vs. CHF 91,570 per QALY) (Table 4).

Total and disaggregated costs and QALYs gained per patient are presented in Table 5 for genotype 1. Cost and QALY results for the other HCV genotypes are presented in S3–S5 Tables. In genotype 1 treatment-naïve patients, drug costs for sofosbuvir-containing regimens were substantially higher than drug costs for comparator strategies (e.g. CHF 62,862 for sofosbuvir+pegylated interferon alpha+ribavirin vs. CHF 20,454 for pegylated interferon alpha+ribavirin). Monitoring costs were generally lower for sofosbuvir-containing regimens than comparator strategies (CHF 3,279 for sofosbuvir+ pegylated interferon alpha+ribavirin vs. CHF 4,650 for pegylated interferon alpha+ribavirin, respectively), except when the comparator strategy was no treatment. Health state costs were usually higher in comparator strategies than sofosbuvir-containing regimens (e.g. CHF 24,403 for pegylated interferon alpha+ribavirin vs. CHF 8,646 for sofosbuvir+ pegylated interferon alpha+ribavirin). QALYs per patient were always higher with treatment regimens including sofosbuvir, by 0.4 to 2.6 QALYs.

Discussion

Although the initial costs of sofosbuvir-containing regimens were high, AE costs, monitoring costs and health state costs were generally lower than in active comparator strategies. Additionally, QALYs gained per patient were consistently higher in sofosbuvir-containing regimens. In the base-case representing a mixed cirrhotic and non-cirrhotic cohort, ICERs for sofosbuvir-containing treatment regimens were better than CHF 100,000 per QALY, for all comparisons. In the majority of sensitivity analyses, results were robust. ICERs of sofosbuvir-based regimens were also lower than CHF 100,000 per QALY in cirrhotic patients, but were above CHF 100,000 per QALY for treatment-naïve, interferon eligible genotype 2 and 3 non-cirrhotic patients.

SVR, the goal of antiviral therapy, can be achieved with a shorter duration of sofosbuvir-containing treatment than with other treatment regimens. For patients who are ineligible, intolerant or unwilling to take interferon-based regimens, the combination of sofosbuvir and ribavirin addresses an unmet medical need. There has been a shift in European and American guidelines in the way that EASL [20] and the American Association for the Study of Liver Diseases (AASLD/IDSA/IAS-USA) [47] no longer recommend treatment with telaprevir or boceprevir in genotype 1 patients. Given the recent marketing authorisation of a combination pill of sofosbuvir and ledipasvir for the treatment of HCV genotype 1 patients in October 2014 by the Food and Drug Administration [48] and more recently by the European Medicines Agency [49], further guideline adaptations may follow.

In our modelling study, sofosbuvir-containing treatment regimens compared to standard regimens led to gains in QALYs. This is in contrast to the German Institute for Quality and Efficiency in Health Care's (IQWiG) conclusion [50] that “there is no data documenting an additional benefit of sofosbuvir-containing regimens compared to other treatment regimens because no appropriate analyses were provided by the pharmaceutical company”. IQWiG’s position was partially due to a lack of head-to-head trials and the use of mixed treatment comparisons. Heterogeneity between patient characteristics may limit the validity of mixed treatment comparisons [51]. As patient characteristics such as age and genotype were comparable between studies, we regard mixed treatment comparisons as an acceptable approach to integrate available evidence, in the present case.

Leleu et al. published a cost-effectiveness study of sofosbuvir in mono-infected and co-infected CHC patients in France [52]. The resulting ICERs were better than in our study. Leleu et al. used a lower utility decrement for patients during treatment with sofosbuvir than for patients during treatment with comparators [52]. We used the same utility decrement for all treatments. In the French study, the percentages of cirrhotic patients were up to twice as high as in the base-case of our study. We showed that ICERs for cirrhotic patients were better than ICERs for non-cirrhotic patients. Hence, these differences in model input parameters may explain the variation in the magnitude of the ICERs. Another cost-effectiveness study in France by Deuffic-Burban et al. [53] compared sofosbuvir-containing treatment regimens and standard treatments in genotype 1 treatment-naïve CHC patients. The authors concluded that treatment with sofosbuvir is cost-effective in CHC patients with fibrosis stage 2 or higher [53]. An Italian cost-effectiveness study by Petta et al. [54] compared triple therapies with sofosbuvir, boceprevir, and telaprevir. Sofosbuvir-based regimens were cost-effective compared to boceprevir, except in cirrhotic and IL28B CC patients, and mostly cost-effective compared to telaprevir. A third cost-effectiveness study in the US concluded that sofosbuvir-based treatment regimens generally dominated telaprevir or boceprevir-based regimens [55]. The authors of a Spanish cost-effectiveness study reported that treatment regimens with sofosbuvir, pegylated interferon alpha and ribavirin for 12 weeks were below the considered cost-effectiveness threshold [56]. Overall, the results of the international cost-effectiveness studies were similar to the present analysis. However, transferability of cost-effectiveness results between different countries is limited due to e.g. differences in epidemiology of the disease, clinical practice, consumer preferences and price levels [57].

In Switzerland, there is no formally accepted cost-effectiveness threshold. In 2010, the Swiss Federal Court of Justice decided that treating a very rare orphan disease at CHF 500,000 per year was too expensive for the compulsory health insurance to cover, given its small health benefits [58]. The Court further stated that beyond a threshold of CHF 100,000 per QALY, health insurers cannot be obliged to pay for a treatment [58]. This was the first time that a formal cost-effectiveness threshold has been suggested for Switzerland to distinguish favourable from unfavourable ICERs. On this basis, sofosbuvir-containing regimens would be cost-effective compared to standard treatment. However, this remains arguable and there will be further discussion, also given widely varying approaches in other countries. NICE's acceptable cost-effectiveness thresholds for the UK are in the range of GBP 20,000 to 30,000 per QALY gained and are intended to represent the opportunity cost to a fixed National Health Service budget [59]. The cost-effectiveness threshold commonly adopted in studies for the US varies between USD 50,000 to 100,000 per QALY [60]. The World Health Organization (WHO) suggests cost-effectiveness thresholds between 2–3 times the gross domestic product (GDP) per person [61]. The Swiss GDP per person was CHF 74,010 in 2012 [62]. According to the threshold suggested by WHO, all results including subgroup analyses would be cost-effective.

An additional point of discussion relates to the fact that thresholds representing the societal willingness to pay for any intervention leading to health gains may vary depending on the context. This was pointed out in a recent article that emphasized how the cost-effectiveness of sofosbuvir may dramatically clash with its affordability if the treatment is to be implemented on a large scale [63]. Not surprisingly, economic studies consider ICERs in distinct subgroups of patients. This appears palatable in those patients in whom the benefit is clear cut, i.e. patients with advanced liver fibrosis at high risk of complications, while it becomes increasingly difficult to accept by payers for patients in whom the life-long risk of complications is relatively small, i.e. patients with no or minimal fibrosis. Those patients are significantly less likely to die of HCV-related consequences and treatment-induced viral clearance may lead to less clear health gains.

Our study has several strengths and limitations relating to the decision-analytic model’s structural assumptions and available input data. Patient characteristics were obtained from the SCCS [19]. Swiss patient characteristics are thus reflected and the local applicability of the results is improved. The validity of the model was assessed by scrutinizing outputs for internal consistency. Model elements and formulae were thoroughly checked for correctness. We also compared the structure with that of other models in the field, which supported the approaches taken.

Clinical input parameters such as transition probabilities and utilities were derived from international literature. Transition probabilities reflect the biological course of CHC and mainly depend on disease characteristics and treatment. European countries with similar treatment options available should have comparable transition probabilities. The utilities used in our model were mainly based on UK studies [21,22,29]. Although the EQ-5D is a generic instrument to measure quality of life, the value sets applied to obtain utilities differ between countries [64]. The values available to us were based on the UK value set. There is no Swiss-specific value set, but a recent study reported that the UK value set generates lower utilities than the German or French value set [65]. Because of the small difference, we have no reason to assume that this had a substantial impact on our results, at least not in the sense of favouring sofosbuvir. For sofosbuvir and alternative treatment regimens we used the same utilities according to health state because treatment-specific utilities were not available. Given the lifelong time horizon of the model, this should not substantially impact the findings. Treatment characteristics and medical resource use based on clinical trial data, international literature and European guidelines were checked by Swiss clinical experts and generally found to be applicable to the Swiss situation. A small number of assumptions were, however, modified. For example, all HCV patients in Switzerland are tested for hepatitis B virus infection while there is no cryoglobulin test performed.

To the extent feasible, the model was populated with Swiss input parameters for patient characteristics, health state costs and other unit costs. This implied some uncertainties. For the patient characteristics implemented in the model, we used data from the SCCS, which reflects local patient characteristics and medical practice. However, patients in the clinical trials of sofosbuvir were younger than patients in the SCCS. If relevant SVR effects, adverse event rates, etc. were age-dependent, this age difference could distort the study results. In addition, we received health state costs from one big university hospital in Switzerland, which were based on a data collection of one year. These costs could differ from those accrued in other settings in Switzerland, which might lead to different results. Despite these aspects and because the model structure, underlying assumptions and model parameters were reviewed for appropriateness by Swiss clinical experts, we believe to have achieved reasonable estimates of the cost effectiveness of CHC treatment with sofosbuvir in Switzerland.

In the base-case analysis, we only included drug costs for the treatment of AEs due to a lack of Swiss specific data. In this respect, this modelling study may provide a conservative estimate of the cost-effectiveness of sofosbuvir-based HCV treatment regimens. In a scenario analysis, total costs of severe AE management were obtained from a US study [43] and adjusted to Swiss costs to the extent possible. ICERs in this scenario analysis changed by less than 5% in comparison with the base case, except for telaprevir-containing and boceprevir-containing regimens. Given differences in regimen-specific AE profiles, data allowing for the inclusion of total AE management costs (i.e. not only drug costs to treat AEs) might improve ICERs of sofosbuvir treatment.

Given the model structure and underlying model input parameters (mild and moderate HCV states (fibrosis stage F0-F3) were combined to a non-cirrhotic state to reflect available data from clinical trials), it was not possible to calculate ICERs for sofosbuvir-based regimens compared to standard treatment according to fibrosis stage. Therefore, the model is not able to reflect the current Swiss treatment limitations. In some cases, SVR rates were obtained from phase II trials due to missing or inapplicable data from phase III trials. Nevertheless, SVR rates from phase II and III trials were consistent. The cost-effectiveness for genotype 1 treatment-experienced could not be assessed due to lack of clinical data. As there are no data available regarding the reversibility of complications after treatment and any assumption made would add uncertainty to the model, we did not consider this and assumed a similar post-treatment clinical course independent from the type of treatment used. Clinical efficacy data were derived from registration trials and may not fully reflect effectiveness in routine clinical practice due to compliance and adherence issues, which limits the generalisability of the results. Finally, the analysis was conducted from the health care system perspective. A broader societal perspective including indirect costs (e.g. due to loss of productivity, absence from work) would provide additional insights into the implications of sofosbuvir-based therapy.

Several authors have criticised the high price of sofosbuvir [66–68]. Given a relevant pool of prevalent, previously untreated HCV cases, budget impact remains an issue among payers and policy makers, especially due to the high up-front costs. Swiss regulators restricted the reimbursement of sofosbuvir-containing treatment to patients with fibrosis stage 3 or 4, or symptomatic patients with extra-hepatic manifestations [17], which reduces initial budget impact. At the same time, our findings for Switzerland generally indicate acceptable ICERs for sofosbuvir-containing regimens, which may initiate further discussion on reimbursement issues. Our results further provide a benchmark for future treatment regimens in patients with CHC that are expected to become available in the near future.

Supporting Information

Acknowledgments

We thank Dr. David Semela, Dr. Philip Bruggmann and Dr. Marcel Stöckle for reviewing the model input parameters.