Cost-effectiveness of Sofosbuvir-based Regimens for HCV
Cost-effectiveness of Sofosbuvir-based Regimens for HCV
A Markov model was developed to estimate costs and clinical outcomes, from the beginning of the therapy with a lifelong time horizon, using different SOF-based therapeutic strategies. We simulated the progression between the different health states presented in hepatitis C monoinfection, including death by disease or by other causes.
A discount rate was applied to reflect the time preference, transforming future costs and outcomes (from first year) to current values, and thus, compare them with the initial cost of therapy. The annual discount rate considered was 3%, as recommended in the literature. This study was conducted from the perspective of the Spanish National Healthcare System and considered a threshold limit of €40 000/quality-adjusted life-years (QALY) for cost-effectiveness.
We evaluated SOF-based regimens for naive and pretreated patients, separately for each genotype on the basis of different CTs, according to SmPC. Efficacy estimations were based on the sustained viral eradication 12 weeks after completion of treatment. If discrepancies between SmPC and CTs publications existed, data from SmPC were recorded.
Naive Patients. Genotype 1: The SmPC covers two therapeutic regimens. On the one hand, the triple therapy based on NEUTRINO CT with an SVR of 90%. On the other hand, SmPC recorded dual therapy with SOF+RBV for 24 weeks based on the phase II CTs QUANTUM and 11-I-0258, where SVR was 65%. Given the lack of control group in the cited CTs, the SVR rate for pIFN+RBV group was obtained from an alternative source. A comparison versus PI triple therapy was also performed as could be considered in most cases the SoC for genotype 1 patients.
Genotype 2: The label information recorded treatment option with SOF+RBV for 12 weeks. Efficacy parameters were based on the FISSION CT, which evidenced an SVR of 95% for SOF+RBV and 77.6% for pIFN+RBV.
Genotype 3: The label information included two options. First, SOF triple therapy based on phase II ELECTRON and PROTON CTs that included 39 patients with an estimated SVR of 97%. Second, dualtherapy with SOF+RBV for 24 weeks based on VALENCE CT with an SVR of 93%.
The main methodological limitation of the CTs VALENCE, ELECTRON and PROTON was the absence of control group with the SoC, so it was necessary to obtain the pIFN+RBV option SVR from the FISSION CT.
Pretreated Patients. Genotype 2: The label information included treatment with SOF plus RBV for 12 weeks with an 82% of SVR based on FUSION CT. Control group efficacy estimations with pIFN+RBV were obtained from the genotype 2 subgroup included in the EPIC CT.
Genotype 3: The SmPC included two treatment options. First, SOF triple therapy based on the LONESTAR-2 phase II CT, which reached an 83% SVR. Second, dual therapy with SOF+RBV for 24 weeks based on the VALENCE CT, which achieved a 77% SVR.
The absence of a control group for the LONESTAR-2 and VALENCE CTs required another source of information for SoC SVR, obtained from the EPIC study genotype 3 subpopulation.
Drugs costs were estimated on the basis of the dosing and therapeutic schemes included in the CTs and SmPC. The absence of stopping rules, in contrast to the boceprevir and telaprevir CT, simplified the average cost estimates of therapeutic options. A perfect adherence was considered on estimations.
Due to its recent approval, SOF does not yet have a registered price so the one for the 'expanded access' programme was applied. The acquisitions of pIFN and RBV in our setting are subjected to considerable discounts with regards to the official price due to the competition among therapeutic alternatives. As this is widely practised, actual prices were used instead of the official prices.
Cost estimations were performed considering a regimen of 400 mg/day SOF, 1000 mg/day of RBV and a weekly dose of 180 μg peginterferon α-2a with a variable duration (12–24 weeks) according to the SmPC (Table 1).
Medical monitoring costs were estimated according to health resource consume related to the length of therapy (Table 2).
Drug adverse reactions are associated with an additional cost due to drug consumption, medical visits, hospital admissions and other resources generated from their treatment. The association of SOF+RBV seems to be better tolerated than the pIFN+RBV association according to FISSION CT. No additional cost was considered for pIFN+RBV derived from its dose reduction managing. The impact of drug toxicity on health-related quality of life (HRQL) was taken into account as mentioned in the next section.
Healthcare resource costs associated to disease progression were obtained from hospitals in the Basque Health Service in 2013. The estimated cost for each disease state is summarised in Table 1. We differentiated transition costs from state costs. The former correspond to the in-hospital care of patients due to different complications related to chronic liver disease. The latter include the cost of resources used in the follow-up. The model included costs associated for SVR states, which derived from the consumption of healthcare resources associated with patients' monitoring.
Additional information about costs estimations is detailed in the Technical Annex.
Quality of life was assessed by applying a specific value of utility to each health state. Because values for Spanish patients were not available, we used the figures elicited by applying EQ-5D and Time Trade-Off tariffs to a sample of UK patients with hepatitis C that have been widely used in literature. Different utility figures were used for SVR states according to the health state they came from. The HRQL weights are recorded in Table 2, and more detailed information is given in the Technical Annex. Estimations were also performed with utility values from a different source in sensitivity analysis.
A Markov-based decision model was built to transform SVR figures on QALY establishing the impact of long-term therapeutic alternatives and analysing the efficiency of the SOF-based regimens. The Markov diagram outlined in figure 1 represents the natural history of hepatitis C and was carried out using TreeAge Pro 2014 software (Tree-Age Software, Inc, Boston, Massachusetts, USA). We used a lifelong time horizon to estimate QALYs expectancy and lifetime direct costs. The initial cohort population candidate for treatment was defined according to the patient's average characteristics of CTs, with 50 years of age, and assuming a distribution of 50% of patients in 'mild hepatitis' estate (F0–F1) and other 50% in 'moderate hepatitis' (F2–F3). The cycle duration was 3 months, and based on the recommendations, a mid-cycle correction was carried out. Transition rates were obtained from the literature and had been widely used in previous models.
(Enlarge Image)
Figure 1.
Chronic hepatitis C Markov model.
Estimations of life expectancy in different states of hepatitis C were made to validate the model. We calculated the life expectancies from different states and the percentage of patients who progress to cirrhosis (21% at 20 years and 37.4% at 30 years). These data are similar to those of other models. Detailed information about Markov model framework and validation is shown in the Technical Annex.
A one-way sensitivity analysis was carried out for each therapeutic option analysed in the base case to show the influence of different variables on incremental cost-effectiveness ratio (ICER).
We analysed the consequences of starting treatment at different ages by establishing a framework from 30 to 70 years. Furthermore, several scenarios were developed using an interval of transition probability values between states of moderate hepatitis to cirrhosis; these were the maximum and minimum values offered by Townsend et al in their review. A different acquisition cost for SOF was also considered including a hypothetical 20% and 40% cost reduction. Different discount rates were also applied. Conducting a sensibility analysis for every susceptible parameter was avoided, as introducing a large number of variables into the estimates would complicate its interpretation. Changes from base-case scenario were evaluated assuming a different distribution of the initial cohort for 100% of patients in mild hepatitis state first, and 100% of patients in moderate hepatitis state later, as well as for a cohort of cirrhotic patients ('supplementary digital content'). The impact of other included variables was also considered in supplementary digital content.
A probabilistic sensitivity analysis was developed to offset the limitations of a traditional sensitivity analysis. All the distribution parameters applied in this analysis are shown in the Technical Annex.
The manuscript content was structured on the basis of the established recommendations by the 'Health Economic Evaluation Publication Guidelines Good Reporting Practices Task Force' (Consolidated Health Economic Evaluation Reporting Standards checklist).
Methods
Design and Analysis Perspective
A Markov model was developed to estimate costs and clinical outcomes, from the beginning of the therapy with a lifelong time horizon, using different SOF-based therapeutic strategies. We simulated the progression between the different health states presented in hepatitis C monoinfection, including death by disease or by other causes.
A discount rate was applied to reflect the time preference, transforming future costs and outcomes (from first year) to current values, and thus, compare them with the initial cost of therapy. The annual discount rate considered was 3%, as recommended in the literature. This study was conducted from the perspective of the Spanish National Healthcare System and considered a threshold limit of €40 000/quality-adjusted life-years (QALY) for cost-effectiveness.
Treatment Strategies and Effectiveness
We evaluated SOF-based regimens for naive and pretreated patients, separately for each genotype on the basis of different CTs, according to SmPC. Efficacy estimations were based on the sustained viral eradication 12 weeks after completion of treatment. If discrepancies between SmPC and CTs publications existed, data from SmPC were recorded.
Naive Patients. Genotype 1: The SmPC covers two therapeutic regimens. On the one hand, the triple therapy based on NEUTRINO CT with an SVR of 90%. On the other hand, SmPC recorded dual therapy with SOF+RBV for 24 weeks based on the phase II CTs QUANTUM and 11-I-0258, where SVR was 65%. Given the lack of control group in the cited CTs, the SVR rate for pIFN+RBV group was obtained from an alternative source. A comparison versus PI triple therapy was also performed as could be considered in most cases the SoC for genotype 1 patients.
Genotype 2: The label information recorded treatment option with SOF+RBV for 12 weeks. Efficacy parameters were based on the FISSION CT, which evidenced an SVR of 95% for SOF+RBV and 77.6% for pIFN+RBV.
Genotype 3: The label information included two options. First, SOF triple therapy based on phase II ELECTRON and PROTON CTs that included 39 patients with an estimated SVR of 97%. Second, dualtherapy with SOF+RBV for 24 weeks based on VALENCE CT with an SVR of 93%.
The main methodological limitation of the CTs VALENCE, ELECTRON and PROTON was the absence of control group with the SoC, so it was necessary to obtain the pIFN+RBV option SVR from the FISSION CT.
Pretreated Patients. Genotype 2: The label information included treatment with SOF plus RBV for 12 weeks with an 82% of SVR based on FUSION CT. Control group efficacy estimations with pIFN+RBV were obtained from the genotype 2 subgroup included in the EPIC CT.
Genotype 3: The SmPC included two treatment options. First, SOF triple therapy based on the LONESTAR-2 phase II CT, which reached an 83% SVR. Second, dual therapy with SOF+RBV for 24 weeks based on the VALENCE CT, which achieved a 77% SVR.
The absence of a control group for the LONESTAR-2 and VALENCE CTs required another source of information for SoC SVR, obtained from the EPIC study genotype 3 subpopulation.
Cost Estimations
Drugs costs were estimated on the basis of the dosing and therapeutic schemes included in the CTs and SmPC. The absence of stopping rules, in contrast to the boceprevir and telaprevir CT, simplified the average cost estimates of therapeutic options. A perfect adherence was considered on estimations.
Due to its recent approval, SOF does not yet have a registered price so the one for the 'expanded access' programme was applied. The acquisitions of pIFN and RBV in our setting are subjected to considerable discounts with regards to the official price due to the competition among therapeutic alternatives. As this is widely practised, actual prices were used instead of the official prices.
Cost estimations were performed considering a regimen of 400 mg/day SOF, 1000 mg/day of RBV and a weekly dose of 180 μg peginterferon α-2a with a variable duration (12–24 weeks) according to the SmPC (Table 1).
Medical monitoring costs were estimated according to health resource consume related to the length of therapy (Table 2).
Drug adverse reactions are associated with an additional cost due to drug consumption, medical visits, hospital admissions and other resources generated from their treatment. The association of SOF+RBV seems to be better tolerated than the pIFN+RBV association according to FISSION CT. No additional cost was considered for pIFN+RBV derived from its dose reduction managing. The impact of drug toxicity on health-related quality of life (HRQL) was taken into account as mentioned in the next section.
Healthcare resource costs associated to disease progression were obtained from hospitals in the Basque Health Service in 2013. The estimated cost for each disease state is summarised in Table 1. We differentiated transition costs from state costs. The former correspond to the in-hospital care of patients due to different complications related to chronic liver disease. The latter include the cost of resources used in the follow-up. The model included costs associated for SVR states, which derived from the consumption of healthcare resources associated with patients' monitoring.
Additional information about costs estimations is detailed in the Technical Annex.
Quality of Life
Quality of life was assessed by applying a specific value of utility to each health state. Because values for Spanish patients were not available, we used the figures elicited by applying EQ-5D and Time Trade-Off tariffs to a sample of UK patients with hepatitis C that have been widely used in literature. Different utility figures were used for SVR states according to the health state they came from. The HRQL weights are recorded in Table 2, and more detailed information is given in the Technical Annex. Estimations were also performed with utility values from a different source in sensitivity analysis.
Markov Model
A Markov-based decision model was built to transform SVR figures on QALY establishing the impact of long-term therapeutic alternatives and analysing the efficiency of the SOF-based regimens. The Markov diagram outlined in figure 1 represents the natural history of hepatitis C and was carried out using TreeAge Pro 2014 software (Tree-Age Software, Inc, Boston, Massachusetts, USA). We used a lifelong time horizon to estimate QALYs expectancy and lifetime direct costs. The initial cohort population candidate for treatment was defined according to the patient's average characteristics of CTs, with 50 years of age, and assuming a distribution of 50% of patients in 'mild hepatitis' estate (F0–F1) and other 50% in 'moderate hepatitis' (F2–F3). The cycle duration was 3 months, and based on the recommendations, a mid-cycle correction was carried out. Transition rates were obtained from the literature and had been widely used in previous models.
(Enlarge Image)
Figure 1.
Chronic hepatitis C Markov model.
Estimations of life expectancy in different states of hepatitis C were made to validate the model. We calculated the life expectancies from different states and the percentage of patients who progress to cirrhosis (21% at 20 years and 37.4% at 30 years). These data are similar to those of other models. Detailed information about Markov model framework and validation is shown in the Technical Annex.
Sensitivity Analyses
A one-way sensitivity analysis was carried out for each therapeutic option analysed in the base case to show the influence of different variables on incremental cost-effectiveness ratio (ICER).
We analysed the consequences of starting treatment at different ages by establishing a framework from 30 to 70 years. Furthermore, several scenarios were developed using an interval of transition probability values between states of moderate hepatitis to cirrhosis; these were the maximum and minimum values offered by Townsend et al in their review. A different acquisition cost for SOF was also considered including a hypothetical 20% and 40% cost reduction. Different discount rates were also applied. Conducting a sensibility analysis for every susceptible parameter was avoided, as introducing a large number of variables into the estimates would complicate its interpretation. Changes from base-case scenario were evaluated assuming a different distribution of the initial cohort for 100% of patients in mild hepatitis state first, and 100% of patients in moderate hepatitis state later, as well as for a cohort of cirrhotic patients ('supplementary digital content'). The impact of other included variables was also considered in supplementary digital content.
A probabilistic sensitivity analysis was developed to offset the limitations of a traditional sensitivity analysis. All the distribution parameters applied in this analysis are shown in the Technical Annex.
The manuscript content was structured on the basis of the established recommendations by the 'Health Economic Evaluation Publication Guidelines Good Reporting Practices Task Force' (Consolidated Health Economic Evaluation Reporting Standards checklist).
Source...