Cost-effectiveness of sacubitril / valsartan in chronic heart-failure patients with reduced ejection fraction

Ademi Zanfinaae, Pfeil Alena Ma, Hancock Elizabethb, Trueman Davidc, Haroun Rolad, Deschaseaux Célined, Schwenkglenks Matthiasa a Institute of Pharmaceutical Medicine (ECPM), University of Basel, Switzerland b PHMR, London, United Kingdom c Source HEOR, Oxford, United Kingdom d Novartis Pharma AG, Basel, Switzerland e Monash Centre of Cardiovascular Research and Education in Therapeutics, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia


Introduction
Heart failure is a progressive and incurable disease, with high morbidity and mortality in high-income countries including Switzerland.The reported prevalence of heart failure varies from between 1 and 2%, and increases for individuals aged above 65 years [1].Estimates for 2010 expected 15 million people with heart failure in Europe and 6.6 million in the United States [2,3].Chronic heart failure has a prevalence of 1 to 2% and heart failure with reduced ejection fraction (HFrEF) accounts for about 50% of all heart failure cases [4].In general, the condition requires complex management and treatment protocols that require substantial effort from patients, care givers, and healthcare services, and therefore poses a high cost burden on society [5].Morbidity is very prominent in terms of severity of symptoms, reduced quality of life, hospitalisations and continuous need for treatment [6,7].Previous guidelines recommend angiotensin-converting enzyme inhibitors (ACEIs) and beta-blockers as initial treatment, as well as diuretics if there is a fluid overload [8].These treatments appear to reduce the risk of death and improve exercise capacity.Angiotensin receptor blockers (ARBs) are controversial and less well tolerated than ACEIs, but remain a treatment option where ACEIs are not tolerated.Other treatments such as anti-platelets and lipid-lowering agents are added if necessary [9].Advances in chronic heart failure treatment have been quite limited in the last decade.Sacubitril/valsartan, an angiotensin-receptor-neprilysin-inhibitor (ARNI), is a novel oral therapy proposed in the cur-rent guidelines for the treatment of heart failure in patients with reduced left ventricular ejection fraction (LVEF) [9].The phase-III prospective double-blind randomised controlled trial PARADIGM-HF (prospective comparison of ARNI with ACEI to determine impact on global mortality and morbidity in heart failure) compared morbidity and mortality between sacubitril/valsartan and the ACEI enalapril in a population with HFrEF [10].The primary outcome was a composite of death from cardiovascular causes or hospitalisation for heart failure.After a median follow-up of 27 months, sacubitril/valsartan was associated with a significant reduction in time to the primary outcome (hazard ratio [HR] 0.80, 95% confidence interval [95% CI] 0.73-0.87;p <0.001), all-cause mortality (HR 0.84, 95% CI 0.76-0.93;p <0.001) and cardiovascular mortality (HR 0.80, 95% CI 0.71-0.89;p <0.001).In addition, sacubitril/valsartan was also associated with a reduced risk of hospitalisation for heart failure of 21% (p <0.001) and a reduction in the symptoms and physical limitations of heart failure (p = 0.001) [10].The aim of this study was to assess the clinical effectiveness in terms of quality-adjusted life years (QALYs) gained, the direct medical cost, and the cost-effectiveness of sacubitril/valsartan (in addition to standard care) compared to ACEIs (in addition to standard care) from the perspective of the Swiss healthcare system.

Overview of approach and model
A model-based cost-utility analysis was undertaken comparing sacubitril/valsartan and standard care to ACEI and standard care.The incremental cost-effectiveness ratio (ICER) was expressed as cost per QALY gained.The analysis was conducted from the perspective of the Swiss healthcare system.Costs and effects occurring after one year were discounted by 3% in the base-case analysis.A two-state Markov model [11] was implemented for the current analyses.In brief, the model is structured as a two-state Markov model (with health states "alive" and "dead").Regression models were used to predict events and outcomes such as mortality, hospitalisations, adverse events and health-related quality of life over the lifelong time horizon of the model, based on patient characteristics and treatment received (fig.1).This type of model was chosen as the benefits of treatment and costs continue to accrue beyond the observation period of the PARADIGM-HF trial.Cycle length is one month and a half-cycle correction is applied.The model permits both deterministic (DSA) and probabilistic sensitivity analyses (PSA).Death can occur at any point in time.Model outcomes include survival time (i.e., life years), QALYs, medical resource utilisation, accrued lifetime and total and disaggregated costs, and other clinical events such as number of hospitalisations and adverse events.

Patient population
The patient population considered for the economic model was the same as that enrolled on the PARADIGM-HF trial [10] i.e., adult HErEF and a mean age of 64 years.The following eligibility criteria were applied: age of at least 18 years, NYHA class II-IV symptoms, ejection fraction of 40% or less (which was changed to 35% or less) [12], and plasma B-type natriuretic peptide (BNP) level of at least 150 pg/ml or hospitalisation for heart failure within the previous 12 months and a BNP of at least 100 pg/ ml.Patients taking stable doses of ACEIs or ARBs four weeks before screening were considered for participation in the study.Implantable cardioverter-defibrillators (ICDs) and cardiac resynchronisation therapy (CRT) are increasingly used in patients with HFrEF.In the PARADIGM-HF trial [10], 1,857 (22%) of the eligible patients used either ICDs or CRT at baseline.After screening, patients had a run-in phase with enalapril or sacubitril/valsartan, which was followed by the main double-blind randomised treatment phase [13].Of 8,442 patients randomised, 43 patients were excluded for the full analysis set (FAS) due to invalid randomisation (n = 6) and good clinical practice (GCP) violations (n = 37).The analysis population consisted of 4187 patients receiving sacubitril/valsartan and 4212 patients receiving enalapril.The baseline characteristics of the trial population are presented in supplementary table S1 in appendix 2.

Treatment strategies
The average daily dose at the end of the PARADIGM-HF trial [10] for sacubitril/valsartan (in addition to standard care) was 375 mg compared to a treatment strategy with a daily dose of the ACEI enalapril (in addition to standard care) of 18.9 mg.Standard care included the use of diuretics, beta-blockers, aldosterone antagonists, digoxin, anticoagulants, aspirin, adenosine diphosphate antagonists and lipid-lowering medications.The choice of standard care is based on medication classes observed in the PARADIGM-HF trial [10].

Clinical model inputs
Clinical information regarding all-cause mortality, hospitalisation rates, health-related quality of life and adverse events was obtained from the PARADIGM-HF trial [10].

Mortality
The base-case analysis used a multivariable parametric survival model of all-cause mortality, which was based on the treatment arm, baseline characteristics of the patients,

and time since randomisation (supplementary table S2 in appendix 2).
An alternative scenario analysis used multivariable parametric survival for cardiovascular mortality from the PAR-ADIGM-HF trial [10] (supplementary table S3), and noncardiovascular mortality from Swiss national life-tables (table S6).The monthly probability of non-cardiovascular mortality was obtained by subtracting the probability of cardiovascular mortality from the probability of all-cause mortality as calculated with data provided by the Swiss Federal Office of Public Health (SFOPH) [14] and the Swiss Federal Office of Statistics (SFOS) tables [15].A death rate including cardiovascular death for five-year age bands was calculated by dividing the number of deaths obtained from Swiss life tables [15] by the number of persons in the relevant age group sourced from the SFOPH [14].The death rates were converted to yearly probabilities of death using the formula p = 1 − e −central death rate*time (time is 1/12 years in this case, as we derived monthly probabilities).All-cause mortality and cardiovascular mortality were assumed to be constant within the 5-year age bands provided by the SFOS, and constant in the age group of persons aged 85 years or above, as we had no additional data for this age category.Additional information about Swiss population and related mortality is available in tables S4, S5 and S6.

Hospitalisation and adverse events
The model predicted the risk of all-cause hospitalisation beyond the PARADIGM-HF trial using negative binomial regression [11].Briefly, predicted hospitalisation rates were adjusted for baseline characteristics of the subjects included in the PARADIGM-HF trial such as age, race, and region, and were dependent on the treatment arm.The model for all-cause hospitalisation showed that a treatment strategy with a daily dose of sacubitril/valsartan compared to ACEI treatment reduced all-cause hospitalisation (supplementary table S7).More serious adverse events were considered to be covered by all-cause hospitalisations (table S7), whereas less serious adverse events were considered independently.Rates of these adverse events (hypotension, elevated serum creatinine and potassium, cough and non-severe angio-oedema) were estimated from the PARADIGM-HF trial) [10].Occurrence of less serious adverse events can be found in the additional material provided in table S8.

Health-related quality of life (HRQoL)
A mixed-effects regression model derived from the PAR-ADIGM-HF trial based on patient-level EQ-5D data was estimated to allow the prediction of the EQ-5D-based utility values as a function of baseline characteristics (including baseline EQ-5D), hospitalisations, adverse events, treatment arm and time since randomisation.The EQ-5D 3-level questionnaire was administered at baseline and at months 4, 8, 12, 24, 36, and end of study.The UK EQ-5D tariff published by Dolan et al. was applied to EQ-5D patient responses [16].Details are available in table S9.Hospitalisations in the 30 days before EQ-5D measurement (to capture the acute effect of hospitalisation), and hospitalisations 30-90 days before EQ-5D measurement (to capture any long term effect during rehabilitation) were imple-mented.Utility decrements in the model were applied to subjects experiencing hospitalisations or adverse events.

Resource use
Drug dosage (primary and background drug therapy) data from the PARADIGM-HF trial were validated by using the recommendations of the Swiss Heart Failure Working Group of the Swiss Society of Cardiology [17], which are based on 2012 European Society of Cardiology (ESC) guidelines [8].Drug dosages used for the Swiss model can be found in supplementary table S10.The proportional occurrence of hospitalisations for surgical procedures (4.0%), interventional procedures (8.0%), or medical management only (88.0%), was obtained from the Western European population of the PARADIGM-HF trial (table S11).A background medical resource utilisation per unit of time was assumed to be the same in both treatment strategies of the model.We assumed that patients with heart failure would need to have at least 12 primary-care physician (PCP) visits per year.This was based on an article by Muntwyler et al. [18], which measured the quality of the diagnosis and management of heart failure in primary care in 1999 in Switzerland.Over 82 PCPs from all over Switzerland participated in the study.A total of 474 patients were included.Milder adverse events reported in the PARADIGM-HF trial [10] were modelled separately, as mentioned previously.The following assumptions were applied for resource use associated with adverse events; (a) if a patient experiences hypotension, he/she needs 2 additional PCP visits, (b) if a patient experiences cough, he/she needs 2 additional PCP visits and blood tests, (c) in the case of angio-oedema, patients can experience milder or severe angio-oedema.Milder angio-oedema patients require use of antihistamines and 2 additional outpatient visits, while patients experiencing more severe angio-oedema need 2 additional outpatient visits and use of glucocorticoids, (d) if patients show signs of elevated serum creatinine, they need 2 additional PCP visits and a blood test, (e) if patients show signs of elevated serum potassium, they need 2 additional PCP visits and a blood test.

Unit costs
The cost of sacubitril/valsartan per day in the base-case analyses was CHF 5.79 (375 mg per day), and unit costs of background therapies were sourced from SFOPH data (Spezialitätenliste) relevant to 2015 [19].For each reported therapeutic substance used in the PARADIGM-HF trial, we collected and mapped drugs representing the same substance, based on the number of available producers in the Swiss pharmaceutical market.For example, if there were three pharmaceutical producers of enalapril 10 mg on the market, then the average cost per tablet strength was calculated.Monthly costs were calculated by multiplying the daily costs by 365.25/12.Daily costs of primary therapies and background therapies can be found in supplementary table S10.Unit costs of hospitalisations were estimated on the basis of diagnosis-related group (DRG) costs, and by mapping each reported hospitalisation in the PARADIGM-HF trial to relevant Swiss DRG codes [20].For this mapping procedure, we used the proportional occurrence of hospitalisations involving surgery, interventional procedures or med-ical management.Where several suitable Swiss DRG codes were identified, the weighted mean was used based on their activity and cost as reported for 2012, which is when the latest data was published [20].Details about hospitalisations, the proportional occurrence of diagnoses reported in the PARADIGM-HF trial, and Swiss DRGs assigned are provided in table S11.The weighted mean cost per hospitalisation provided information on the unit cost per hospitalisation event rather than cost per day in hospital.For the year 2012, the average cost per hospitalisation was CHF 13 847; these costs were then updated to 2014 values using the Swiss consumer price index [21].The consumer price index values for 2012 and 2014 were 99.9 and 98.1.The resulting cost per hospitalisation in 2014 was CHF 13 598.As described previously, the estimated number of PCP visits per year was informed by the European IMPROVE-MENT-HF study [18].The unit costs of a PCP visit were derived from the santésuisse web page, and amounted to CHF 113 in 2007.Based on the consumer price index [21] the updated value for one PCP visit in 2014 was CHF 110.30.Unit costs for the treatment of each relevant type of adverse event were estimated from Swiss literature and information publicly available from the SFOPH, Tarmed, and santésuisse websites [19] [22].

Subgroup analyses
Subgroup analyses based on the a priori subgroups in PARADIGM-HF were undertaken to understand variation of the main results between subgroups of patients enrolled in the PARADIGM-HF trial.

Sensitivity analysis
To assess the impact of different assumptions on the model results, a series of scenario analyses were performed.Some were of general relevance.Additional analyses were regarded as specifically relevant for the Swiss setting (see appendix 1 for description).A series of deterministic sensitivity analyses (DSA) were performed to assess the impact of uncertainty surrounding key input parameters.Important parameters were varied independently over plausible ranges determined by the 95% confidence intervals (CI) surrounding point estimates.Where 95% CIs were not available, upper and lower values of ±25% surrounding point estimates were used (supplementary table S12).The ICERs resulting from each analysis were recorded for the upper and lower value and are presented in a Tornado diagram.Probabilistic sensitivity analysis was undertaken to explore joint parameter uncertainty (details about their respective distributions in table S12).A total of 10 000 iterations were run and the results are shown as a cost-effectiveness plan.

Base-case analyses
In the base-case analysis, the sacubitril/valsartan strategy compared to enalapril showed a decrease in the number of hospitalisations (6.0%/year absolute reduction) and lifetime hospital costs by 8.0% (discounted).Total QALYs per person over a lifetime horizon were 4.99 and 4.56 in the sacubitril/valsartan and ACEI treatment strategies respectively (table 1).This led to an incremental difference of 0.425 QALYs.The total incremental costs difference was CHF 10 926 (table 1) and the ICER for sacubitril/valsartan treatment versus ACEI was CHF 25 684 per QALY gained.Alternatively, the use of Swiss life-tables for non-cardiovascular mortality and cardiovascular mortality rates from the PARADIGM-HF trial led to an ICER of CHF 24 490.Cost-effectiveness results for subgroups of patients are presented in full in supplementary table S13 in appendix 2. The ICER was quite stable, with ±1-11% variation from the base-case result.In brief, if the baseline eGFR was <60, the ICER decreased by 8.0%.No use of beta-blockers at baseline decreased the ICER by 11.0%, and where ≤1 year since diagnosis of heart failure was recorded, this increased the ICER by 8.0%.

Sensitivity analysis
The most influential parameters in univariate sensitivity analysis were related to all-cause mortality, hospitalisations and HRQoL (fig.2).Table 2 shows the scenario analysis results.An ICER of > CHF 48 000 per QALY occurred if all treatment effects of sacubitril/valsartan were assumed to cease after year 5 (while the treatment costs of sacubitril/valsartan continued for life).An analysis based on two years of follow-up led to an increased ICER of CHF 58 679.An ICER of CHF 30 812 per QALY gained was observed where there was assumed to be no effect of sacu- bitril/valsartan on HRQoL.Using the German and French EQ-5D value sets instead of that for the UK led to slightly more favourable ICER results.Other scenario analyses did not have a major impact on the ICER.PSA results are presented in figure 3 as a cost-effectiveness plane.All simulation fell within the northeast quadrant of the cost-effectiveness plane (meaning in all simulations sacubitril/valsartan was both more effective and costlier than enalapril), with a 95% confidence interval range of CHF 18 798 to CHF 43 974 per QALY gained.The costeffectiveness threshold of CHF 000 per QALY gained  French EQ-5D tariff used Using EQ-5D tariff instead of UK tariffs 23 German EQ-5D tariff used Using EQ-5D German tariffs instead of UK tariff 24

NT-pronBNP test inclusion 26
HF management outpatient visits ( 40  was met in 78.0% of 10 000 runs, and threshold of CHF 50 000 per QALY gained was met in 99.0% of 10 000 runs.

Discussion
Given limited healthcare budgets throughout the world, the health economic aspects of new drug evaluations can be as important as efficacy, safety and the ability to serve important medical needs under routine clinical practice.In most developed countries, heart failure poses a great economic burden.Estimates show that management of heart failure accounts for 2-5% [2, 3] of total healthcare budgets.Longterm drug treatment is a cornerstone of heart failure therapy.
The cost-effectiveness of sacubitril/valsartan plus standard care compared to enalapril plus standard care has been assessed, from the perspective of the Swiss health care system.The base-case analysis indicated an ICER of CHF 25 684 per QALY gained.The findings presented were robust to changes in assumptions, and the ICER results were similar across multiple patient subgroups.When model input parameters were varied on the basis of their 95% confidence intervals (as observed in the PARADIGM-HF trial or estimated in regression analyses based thereupon), the ICER remained below CHF 50 000 per QALY gained in most of the cases.In the scenario analyses performed, the ICER also remained below CHF 50 000 per QALY gained, except in the extreme scenarios of the treatment effect of sacubitril/valsartan that persisted for only two or five years.
It should be noted that there is no formally accepted costeffectiveness threshold in Switzerland.In this study, we tentatively assume a threshold of CHF 30 000 and CHF 50 000 per QALY gained to distinguish between favourable and unfavourable ICER results [23,24].This threshold level is similar to the upper limit of the threshold range of £20 000-£30 000 accepted in the United Kingdom (UK) [25].A few years ago, a court in Switzerland hinted at a CHF 100 000 per QALY threshold [26].Another recent study from the Netherlands, using a Markov model and using the effectiveness data from the PARADIGM-HF trial over a lifetime horizon, showed that sacubitril/valsartan was considered cost effective at an ICER of 19 113 per QALY gained [30]. Differences observed with our study were mainly in terms of model structure, but also in terms of input parameters, such as quality of life.The Dutch study was not able to utilise patient level PARADIGM-HF trial data.The reported incremental QALY gained was 0.29 for the Dutch population, which was lower than that for the Swiss population (0.42).However, the monthly cost for sacubitril/valsartan in the Dutch and Swiss models was quite similar; the amounts were €161.7 and CHF 176 respectively.In recent years, there have been a number of cost-effectiveness studies undertaken for heart failure patients.The interventions considered included ivabradine, eplerenone, ACEI, and beta blockers.These were compared with placebo or standard of care.In particular, ivabradine was one of the most studied drugs in heart failure patients.Lifetime ICERs ranged from €7634 in Poland [31] to $53 710 in Mexico [32] within these studies [31,[33][34][35][36][37][38].ICERs for sacubitril/valsartan appear to be in a comparable range.

Strengths and limitations
Patient-level data from a large international randomised clinical trial formed the basis of this analysis.Use of data from PARADIGM-HF, a large randomized controlled trial comparing sacubitril/valsartan to a real-world standard of care, allowed for a high level of internal consistency in the model.Multivariate risk equations allowed us to characterise and take into account between-patient heterogeneity.
A relatively novel approach was used to predict quality of life, by extrapolating EQ-5D utility values based on time trends observed in PARADIGM-HF.The use of local data with regards to non-cardiovascular deaths and unit costs allowed for adjustment to the Swiss healthcare environment.Consistency across countries with regards to medical resource use was partially improved by using the Western European part of the PARADIGM-HF population for the calculation of some parameters, such as hospitalisations.However, the transferability of clinical trial results is necessarily affected by simplifications and assumptions, as these are required in all health economic models.
The main limitation was extrapolation of the treatment effect beyond the observation period of the PARADIGM-HF trial.This is a common limitation shared by most economic evaluation studies when the lifetime impact of an emerging treatment is assessed.Assumptions made with regards to the long-term effects of treatment on mortality, healthrelated quality of life and all-cause hospitalisation were addressed by performing state-of-the-art sensitivity and scenario analyses.When these assumptions were changed one at a time, they were found to have a relatively modest impact on the final results.Hence, results derived in the basecase analysis seem to be realistic based upon assumptions in the economic evaluation.The mean age of patients treated in the PARADIGM-HF trial was 64 years, while the mean age of patients in the Swiss population might be higher.This may have led to an over-or underestimation of the true differences between sacubitril/valsartan and standard treatment to be expected for Switzerland, with unclear implications for the ICER.However, in a subgroup analysis including only PARA-DIGM-HF [39] patients aged at least 75 at baseline, the resulting ICER was very similar to the base case ICER.
The model used Swiss input data where relevant and available, but some approximations were required due to lack of data.One related limitation was the lack of information with regards to medical resource use in heart failure patients.This highlights the need for high-quality Swiss data to cover the aspects of resource utilisation.In the absence of such data, we have adopted resource-use estimates from the UK and verified these with the Swiss literature published as far as possible.This approach may have led to an underestimation of the medical resource use of Swiss CHF patients, which is expected to have a relatively unclear impact on this ICER.In addition, we were not able to capture out-of-pocket expenses incurred by patients themselves with regards to heart failure, and this information should be included in future assessments of costs from the perspective of the Swiss healthcare system.

Conclusion
From the perspective of the

Scenario analyses
Basing non-cardiovascular death from Swiss life tables, observations from the PARADIGM-HF trial were used for cardiovascular (CV) mortality.Towards this end, CV mortality parametric survival model was used where an effect of sacubitril/valsartan on CV mortality was considered.Sacubitril/valsartan showed a small positive effect on European Quality of Life-5 Dimensions (EQ-5D) score (0.011, p = 0.001).To test the impact of this assumption, in the scenario analysis this effect was set to zero.The treatment effect of sacubitril/valsartan was applied to heart failure (HF) hospitalisations only, whereas the basecase analyses modelled the observed impact of sacubitril/ valsartan treatment on all cause hospitalisations.Hospitalisation-related utility decrements were set to zero whereas in the base-case analyses, utility decrements for hospitalisation in the previous 30 days and in the previous 30-90 days were incorporated.The median follow-up time in the PARADIGM-HF trial was 27 months.In the absence of long term follow up data, the base-case analyses assumed that the treatment effect of sacubitril/valsartan on mortality, hospitalisations and health-related quality of life (HRQoL) would continue over a lifetime horizon.In scenario analyses, all sacubitril/ valsartan treatment effects were assumed to cease after 5 or 10 years (but the accrual of sacubitril/valsartan treatment costs was assumed to continue).While the base-case analyses included discontinuation as seen in PARADIGM-HF, scenario analysis assumed an exponential survival model of treatment discontinuation, implying a constant rate of discontinuation.Upon discontinuation, costs and efficacy for sacubitril/valsartan patients were assumed to revert to that of angiotensin convertingenzyme inhibitors (ACEIs).This change in efficacy was assumed for all treatment effects, i.e. mortality, hospitalisations, HRQoL and adverse event occurrence.Costs for discontinued ACEI patients were based on angiotensin receptor blocker (ARB) costs, with efficacy assumed to be the same for ACEI and ARBs.(ARBs were shown to have comparable efficacy to ACEI [40].)Another scenario assumed there would be no discontinuation after 3 years.Given geographical proximity, we additionally applied utility estimates based on the French and German EQ-5D value sets.The former was based on a French time tradeoff study by Chevalier et al. [41].This study recruited a total of 452 respondents aged over 18 years who were representative of the French population with regard to age, gender, and socio-professional group [41].Secondly, Greiner et al. provided a German value set for the EQ-5D [42] based on the stated preferences of the German general public.A sample of 339 individuals in northern Germany valued 15 different health states from a sample of 36 states.
Similarly as described for the base-case model, mixedeffects regression models based on patient-level EQ-5D utility values were estimated to predict EQ-5D utility as a function of baseline characteristics (including baseline EQ-5D), hospitalisations, adverse events, treatment arm and time since randomisation [42].
Another scenario analysis assumed that N-terminal probrain natriuretic peptide tests would be routinely performed in heart failure patients.
Findings were similar to the results of three previously published cost-effectiveness analyses in the United States[27][28][29].These studies (Gaziano et al.,King et al. and  Sandhu et al.)  found sacubitril/valsartan to be cost-effective.The first published economic analysis for the US[27] used the same analytical framework over a 30-year time horizon and displayed an ICER of US$45 017 per QALY gained.Differences observed with our study affected costs and quality of life.Incremental costs and effects were higher in the US population.For example, the monthly cost for sacubitril/valsartan in the US was $375, whereas in Switzerland it was CHF 176.The cost of heart failure hospitalisation were $18 158 in the US and CHF 13 599 in Switzerland.Incremental QALYs gained were 0.78 for the US population and 0.42 for the Swiss population.The second set of cost-effectiveness analyses undertaken by King et al.[28] found similar results with an ICER of $50 959 per QALY gained over a lifetime.However, their model included population with NYHA class I[28], whereas NYHA class I population was excluded from the PARADIGM-HF trial.The third economic evaluation by Sadhu et al.[29] displayed a cost per QALY gained of $47 053, and differences with our study were mainly due to modelling techniques and input parameters.Monthly cost for sacubitril/valsartan in the study by Sadhu et al.[29] was assumed to be $380, and heart failure hospitalisation costs were assumed as $11 829.In terms of health outcomes, the study by Sadhu et al.[29] displayed a higher incremental QALY gained of 0.62, as compared to that of the Swiss population (0.42).

Table 1 :
Base case results (all costs are expressed in CHF).

valsartan ACEI Difference Clinical effectiveness parameters (discounted)
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table 2 :
Results of scenario analyses.

Area of uncertainty Scenario ICER (cost per QALY in CHF) Base case (patient level data) 25
Cost of primary therapiesCost of ACEI/ sacubitril/valsartan based on PARADIGM-HF target doses 26
eGFR = estimated glomerular filtration rate; EQ-5D, European Quality of Life-5 Dimensions; HF = heart failure; NT-proBNP = N-terminal pro-brain natriuretic peptide; NYHA II-IV = New York Heart Association class II-IV; SE = standard error † Variable centred on mean ‡ Constant term in Gompertz regression § The ancillary parameter that controls the shape of the baseline hazard

Table S4 :
All-cause mortality of the Swiss population in 2012.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S5 :
Cardiovascular mortality of the Swiss population in 2012.

Table S6 :
Non-cardiovascular mortality of the Swiss population in 2012.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S7 :
Negative binomial regression model for all-cause hospitalisation.

Table S8 :
Occurrence of less serious adverse events in the PARADIGM-HF trial.Absolute number of each adverse event taken from McMurray et al (10) ‡ No treatment or use of antihistamines § Use of catecholamines or glucocorticoids without hospitalisation ¶ Hospitalisation without airway compromise Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S10 :
Swiss drug costs of primary and background therapy for heart failure.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S11 :
Swiss DRG codes for hospitalisation costs (description of surgical procedures).
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S12 :
Parameters used in univariate and probabilistic sensitivity analyses.Cost parameters are in CHF.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table S13 :
Results of subgroup analyses.