Cost-effectiveness of Empagliflozin in Patients With Heart Failure With Preserved Ejection Fraction

This economic evaluation estimates the cost-effectiveness of empagliflozin in patients with heart failure with preserved ejection fraction.

Key Points

Question

Is empagliflozin cost-effective for patients with heart failure with preserved ejection fraction (HFpEF)?

Findings

In this economic evaluation using a Markov model based on the EMPEROR-Preserved trial population of 5988 patients, empagliflozin had an incremental cost-effectiveness ratio of $437 442 per quality-adjusted life-year gained. The results were most sensitive to the monthly cost, quality-of-life benefit, and effect on cardiovascular mortality of empagliflozin.

Meaning

Findings suggest that empagliflozin provides low economic value compared with standard of care for HFpEF, largely due to the lack of benefit on mortality and small benefit on quality of life.

Abstract

Importance

In the Empagliflozin Outcome Trial in Patients With Chronic Heart Failure With Preserved Ejection Fraction (EMPEROR-Preserved), empagliflozin significantly reduced hospitalizations for heart failure while improving patient-reported health status compared with placebo. The long-term cost-effectiveness of empagliflozin among patients who have heart failure with preserved ejection fraction (HFpEF) remains unclear.

Objective

To estimate the cost-effectiveness of empagliflozin in patients with HFpEF.

Design, Setting, and Participants

This cost-effectiveness analysis performed from October 2021 to April 2022 included a Markov model using estimates of treatment efficacy, event probabilities, and utilities from EMPEROR-Preserved and published literature. Costs were derived from national surveys and pricing data sets. Quality of life was imputed from a heart failure–specific quality-of-life measure. Two analyses were performed, with and without a treatment effect on cardiovascular mortality. Subgroup analyses were based on diabetes status, ejection fraction, and health status impairment due to heart failure. The model reproduced the event rates and risk reduction with empagliflozin observed in EMPEROR-Preserved over 26 months of follow-up; future projections extended across the lifetime of patients.

Exposures

Empagliflozin or standard of care.

Main Outcomes and Measures

Hospitalizations for heart failure, life-years, quality-adjusted life-years (QALYs), lifetime costs, and lifetime incremental cost-effectiveness ratio.

Results

A total of 5988 patients were included in the analysis, with a mean age of 72 years, New York Heart Association class II to IV heart failure, and left ventricular ejection fraction greater than 40%. At the Federal Supply Schedule price of $327 per month, empagliflozin yielded 0.06 additional QALYs and $26 257 incremental costs compared with standard of care, producing a cost per QALY gained of $437 442. Incremental costs consisted of total drug costs of $29 586 and savings of $3329 from reduced hospitalizations for heart failure. Cost-effectiveness was similar across subgroups. The results were most sensitive to the monthly cost, quality-of-life benefit, and mortality effect of empagliflozin. A price reduction to $153 per month, incremental utility of 0.02, or 8% reduction in cardiovascular mortality would bring empagliflozin to $180 000 per QALY gained, the threshold for intermediate value. Using Medicare Part D monthly pricing of $375 after rebates and $511 before rebates, empagliflozin would remain low value at $509 636 and $710 825 per QALY gained, respectively. Cost-effectiveness estimates were robust to variation in the frequency and disutility of heart failure hospitalizations.

Conclusions and Relevance

In this economic evaluation, based on current cost-effectiveness benchmarks, empagliflozin provides low economic value compared with standard of care for HFpEF, largely due to its lack of efficacy on mortality and small benefit on quality of life.

Introduction

Heart failure with preserved ejection fraction (HFpEF) continues to cause substantial morbidity as effective therapies remain limited.¹ Patients with HFpEF have similar risk of adverse clinical events and impairment in health-related quality of life as patients with HF with reduced EF.^2,3 Multimorbidity and high burden of illness in HFpEF drive significant costs, particularly around hospital admissions for noncardiovascular causes.⁴

EMPEROR-Preserved (Empagliflozin Outcome Trial in Patients With Chronic Heart Failure With Preserved Ejection Fraction) was a randomized, double-blind trial that compared empagliflozin (10 mg once daily) with placebo in addition to usual therapy in 5988 patients with symptomatic HF and an EF greater than 40% followed up for a median of 26.2 months.⁵ Empagliflozin significantly reduced the primary composite end point of cardiovascular death or hospitalization for HF, mainly driven by a lower risk of HF hospitalization, and improved patient-reported health status.⁶ While empagliflozin improved clinical outcomes among patients with HFpEF, it remains a costly therapy. Given concerns around rising health care costs, understanding the cost-effectiveness of new therapies will be imperative for health systems and payers.

We performed an independent cost-effectiveness analysis of empagliflozin in a cohort of patients with HF and EF greater than 40%, based on the EMPEROR-Preserved trial population. We also conducted subgroup analyses by diabetes status, EF, and health status impairment due to HF.

Methods

Decision Model

We developed a state-transition Markov model to evaluate empagliflozin, a sodium-glucose cotransporter-2 (SGLT2) inhibitor, compared with standard of care (SOC) in patients with a mean age of 72 years, New York Heart Association (NYHA) class II to IV symptoms, and EF greater than 40%.⁵ In the model, patients incurred a monthly risk of hospitalization for HF, hospitalization for other causes, cardiovascular death, noncardiovascular death, and treatment intolerance (eFigure 1 in the Supplement). We performed 3 subgroup analyses in which we separately evaluated (1) patients with and without diabetes, (2) patients with EF less than 50% and 50% or greater, and (3) patients across levels of health status impairment due to HF. We defined health status impairment based on the Kansas City Cardiomyopathy Questionnaire Overall Summary Score (KCCQ-OSS). The KCCQ has demonstrated superior reproducibility, clinical granularity, and prognostic value compared with NYHA class.^7,8,9 EMPEROR-Preserved reported health-related quality-of-life outcomes by baseline KCCQ-OSS tertiles, which correlated with changes in the KCCQ-OSS (+1.94 for OSS <61.2, +1.97 for OSS 61.2-82.3, and +1.20 for OSS ≥82.3 at 52 weeks).⁶ Given similarity of benefit in the lower 2 tertiles, we created 2 strata: moderate (KCCQ-OSS <82.3) and mild (KCCQ-OSS ≥82.3) HF-related impairment of health status.

The model followed up patients over their lifetime with a monthly time cycle and was developed from a health system perspective, capturing all current and future direct health care costs and health benefits.¹⁰ Future costs and utilities were discounted at 3% annually (eMethods 1 in the Supplement). The primary outcome was the incremental cost-effectiveness ratio (ICER), or the incremental cost between the 2 arms divided by the incremental quality-adjusted life-years (QALYs). We used multiple willingness-to-pay thresholds recommended by the cost-effectiveness guidelines of the American College of Cardiology/American Heart Association and World Health Organization: less than $60 000 per QALY as high value, $60 000 to less than $180 000 per QALY as intermediate value, and $180 000 or greater per QALY as low value.^11,12 We used these thresholds, as opposed to conventional thresholds of $50 000/QALY and $150 000/QALY, based on the increase in the US gross domestic product per capita over time. This simulation analysis was exempt from institutional review board approval and patient informed consent as this was a secondary analysis of public use or deidentified data. Adherence to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline is presented in eTable 13 in the Supplement.¹³

Event Probabilities

We estimated the monthly risk of HF hospitalization, non-HF hospitalization, cardiovascular death, and noncardiovascular death for SOC from the EMPEROR-Preserved control arm (eMethods 2 and eTable 1 in the Supplement).⁵ Age-related changes in the risk of each outcome were derived from HFpEF-specific literature estimates (eTable 2 and eFigure 2 in the Supplement).¹⁴ For the empagliflozin arm, the probability of HF hospitalization was calculated by applying the rate ratio from EMPEROR-Preserved to the SOC rate (eMethods 3 and eTable 4 in the Supplement).

EMPEROR-Preserved found a nonsignificant reduction in cardiovascular mortality (hazard ratio [HR], 0.91; 95% CI, 0.76-1.09). Given uncertainty as to whether empagliflozin reduces cardiovascular mortality, we performed a secondary analysis with the observed 9% reduction in cardiovascular death.¹⁵ In analyses without the mortality reduction, we pooled the cardiovascular mortality rates across arms. In both analyses, the non-HF hospitalization and noncardiovascular death rates were also pooled across the trial population for both arms in the model. The treatment intolerance rate was based on the rate of adverse events leading to discontinuation of empagliflozin reported in the CONSORT diagram of EMPEROR-Preserved.⁵ We assumed that patients who did not tolerate empagliflozin discontinued the therapy and incurred no further treatment-related costs or benefit.

Baseline morbidity rates for subgroups were obtained from EMPEROR-Preserved⁵ when available or from the literature (eMethods 4 in the Supplement).^16,17,18 Across diabetes status and HF-related health status subgroups, we assumed equivalent HRs and health status benefit for empagliflozin given nonsignificant heterogeneity of treatment effects. For the EF analysis, we applied different risk reductions for HF hospitalization across subgroups.¹⁶ We calibrated each analysis to match overall trial event rates and treatment effects at 26 months (eMethods 5 and eTable 3 in the Supplement).

Costs

We used the Federal Supply Schedule (FSS) to determine a 2021 monthly cost of empagliflozin ($326.69).¹⁹ The FSS is intended to allow direct federal health care programs to buy brand-name prescription drugs at lower prices negotiated between the Department of Veterans Affairs (VA) and drug manufacturers. Given cost variation across payers, we evaluated a broad range of prices spanning plus or minus 50% of the base case price in sensitivity analysis. We also evaluated several other cost estimates, including prices for the VA, Medicaid, Medicare Part D, and retail pharmacies (eMethods 6 in the Supplement).

Outpatient costs of HF care were derived from the lifetime costs of HF from diagnosis until death or loss to follow-up (eMethods 7 and eTable 6 in the Supplement).^20,21,22 Higher costs were assumed for moderate vs mild health status impairment due to HF.²³ Costs of HF and non-HF hospitalization were obtained from the Agency for Healthcare Research and Quality’s National Inpatient Sample and 2021 Medicare Physician Fee Schedule (eMethods 8 in the Supplement).^24,25,26 Cost of medication intolerance was set to Medicare reimbursements for a primary care appointment for mild adverse events and an average non-HF hospitalization for severe adverse events.^27,28 All costs were inflation-adjusted to 2021 using the Personal Consumption Expenditure price indexes for health care.²⁹

Utilities

For utility estimates, we used a validated algorithm mapping KCCQ-OSSs to EuroQol Health-Related Quality of Life 5-Dimension 3-Level (EQ-5D-3L) scores and subsequent US-based utility estimates.^28,30 This algorithm was developed from 3 patient cohorts (CHAMP-HF, PARTNER 2, and HF-ACTION) with both KCCQ scores (HF-specific health status) and EQ-5D-3L scores (generic health status). The mapping algorithm converts EQ-5D-3L scores into utility scores based on previously published valuations.^31,32 We estimated a mean baseline utility of 0.80 for the EMPEROR-Preserved population from a mean KCCQ-OSS of 72.1 (eMethods 9 and eTable 7 in the Supplement).^6,33 We included an annual decline in utility of 3.41% to account for the effects of aging and disease progression.^34,35,36 For the diabetes status and EF subgroups, we leveraged literature estimates of the relative KCCQ-OSS between subgroups to approximate their baseline utilities.^37,38 For the HF-related health status subgroups, we simulated the KCCQ-OSS distribution based on the mean cohort OSS and standard deviation and mapped these scores to estimated utilities (eMethods 10 in the Supplement).⁶

We also used the differences in KCCQ-OSS across arms to model the quality-of-life benefit of empagliflozin. The adjusted mean difference in KCCQ-OSS at 52 weeks was 1.60 between SOC and empagliflozin arms, which mapped to an annual incremental utility of 0.0079 (eMethods 11 in the Supplement). We modeled a constant incremental utility benefit over time that continued beyond the end of follow-up at 26 months. Three alternate trials of SGLT2 inhibitors in HFpEF have demonstrated larger improvements in KCCQ-OSS.^39,40,41 Therefore, as a sensitivity analysis, we modeled a higher incremental utility of 0.019 based on the estimated utility of pooled trial estimates.

For HF and non-HF hospitalizations, we applied a 19% and 11% decrease from baseline utility for the month of admission, respectively.^42,43,44,45 The disutility for empagliflozin intolerance during the month of occurrence was 0.0029, the weighted average of the disutility for a mild and severe adverse event (eMethods 12 in the Supplement).^28,46 This estimate approximates the monthly average utility decrement of 0.003 commonly cited for ambulatory visits for urinary tract infections.^47,48

Sensitivity Analyses

We used 1-way sensitivity analyses to evaluate the uncertainty across model parameters (Table 1). Parameters were varied across prespecified uncertainty distributions (eTables 1, 5, and 7 in the Supplement). We performed threshold analyses of the cost, incremental utility, and HR for cardiovascular death of empagliflozin at a willingness-to-pay threshold of $180 000 per QALY. Two-way and 3-way sensitivity analyses evaluated variables with potential interactions.

Table 1. Selected Model Inputs.

Model input	Valuea	Source
Transition probabilities, % (range) b
Standard of care
HF hospitalization	0.67 (0.47-0.86)	Anker et al,5 2021
Non-HF hospitalization	2.69 (1.88-3.50)	Anker et al,5 2021
Cardiovascular death	0.32 (0.22-0.41)	Anker et al,5 2021
Noncardiovascular death	0.24 (0.17-0.32)	Anker et al,5 2021
Empagliflozin
Intolerance	0.45 (0.32-0.59)	Anker et al,5 2021
Rate ratios for subgroups, RR (95% CI)c
Diabetes
HF hospitalization	1.77 (1.45-2.16)	Kristensen et al,17 2017
Non-HF hospitalization	1.41 (1.19-1.68)	Kristensen et al,17 2017
Cardiovascular death	1.59 (1.28-1.96)	Kristensen et al,17 2017
Noncardiovascular death	1.60 (1.14-2.25)	Kristensen et al,17 2017
EF <50%
HF hospitalization	1.27 (1.08-1.46)	Anker,16 2021
Cardiovascular death	1.43 (1.21-1.64)	Anker,16 2021
Noncardiovascular death	1.11 (0.94-1.28)	Anker,16 2021
Moderately impaired HF health statusd
HF hospitalization	1.44 (0.85-1.68)	Pitt et al,18 2014
Non-HF hospitalization	1.33 (1.13-1.56)	Pitt et al,18 2014
Cardiovascular death	1.19 (1.10-1.89)	Pitt et al,18 2014
Noncardiovascular death	1.16 (0.78-1.73)	Pitt et al,18 2014
Rate ratios with empagliflozin, RR or HR (95% CI)
HF hospitalization	0.73 (0.61-0.87)	Anker et al,5 2021
EF ≥50%	0.83 (0.66-1.03)	Anker,16 2021
EF <50%	0.56 (0.41-0.77)	Anker,16 2021
Cardiovascular death	0.91 (0.76-1.08)	Anker et al,5 2021
Utilities, value (range)
Baseline
Overall cohort	0.80 (0.68-0.92)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021
Without diabetes	0.82 (0.70-0.94)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021; Yap et al,37 2019
With diabetes	0.79 (0.67-0.90)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021; Yap et al,37 2019
EF ≥50%	0.81 (0.68-0.93)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021; Gold,31 1996; Shah et al,38 2021
EF <50%	0.79 (0.68-0.91)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021; Gold,31 1996; Shah et al,38 2021
Mildly impaired HF health status	0.90 (0.77-1.00)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021
Moderately impaired HF health status	0.73 (0.62-0.84)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021
Empagliflozin	0.0079 (0.0040-0.012)	Anker et al,5 2021; Butler et al,6 2022; Thomas et al,30 2021
Disutility, % baseline utility
HF hospitalization	19.4 (9.7-29.1)	Ambrosy et al,42 2016; Griffiths et al,43 2014; McMurray et al,44 2018
Non-HF hospitalization	11.1 (5.6-16.7)	Mangen et al,45 2017
Monthly costs [2021], value (range), $
Standard of care	0	NA
Empagliflozin	326.69 (163.35-490.04)	Office of Procurement, Acquisition, and Logistics,19 2021
HF hospitalization	14 736 (7369-22 103)	Liang et al,24 2020; CMS,25 2021; AHRQ,26 2018
Non-HF hospitalization	11 577 (5789-17 366)	Liang et al,24 2020; CMS,25 2021; AHRQ,26 2018
Cardiovascular death	70 011 (35 005-105 016)	Liao et al,23 2007; Liang et al,24 2020; CMS,25 2021; Reed et al,49 2012
Noncardiovascular death	89 294 (44 647-133 940)	Liao et al,23 2007; Liang et al,24 2020; CMS,25 2021; Reed et al,49 2012

Abbreviations: AHRQ, Agency for Healthcare Research and Quality; CMS, Centers for Medicare & Medicaid Services; EF, ejection fraction; HF, heart failure; HR, hazard ratio; NA, not applicable; RR, rate ratio.

^a Used in 1-, 2-, and 3-way sensitivity analyses as well as probabilistic sensitivity analysis.

^b Event rates in the model were calibrated to the rates seen during the trial period of 26 months and subsequently converted to monthly probabilities (eMethods 5 in the Supplement).

^c Rate ratios for patients with diabetes are relative to patients without diabetes, EF less than 50% relative to EF of 50% or greater, and moderately impaired HF health status relative to mildly impaired HF health status.

^d Mildly impaired health status due to heart failure is defined as Kansas City Cardiomyopathy Questionnaire Overall Summary Score of 82.3 or greater and moderately impaired as less than 82.3.

Several additional scenarios were evaluated (eMethods 13 in the Supplement). First, we assessed a scenario in which there was no annual utility decrement. Second, we accounted for the influence of generic competition between 2025 and 2029 assuming a price decline to the average of current generic HF therapies.⁵⁰ Third, we evaluated a higher intolerance rate for empagliflozin to model potential safety risks outside a controlled trial setting. Fourth, we reevaluated the HF-related health status subgroup model using differential treatment effects on HF hospitalization and KCCQ-OSS. Fifth, we performed 2 scenarios varying the starting age while assuming different levels of risk reduction for HF hospitalization in patients younger than 70 years and 70 years or older. Finally, we performed probabilistic sensitivity analyses to evaluate model uncertainty (eMethods 14 in the Supplement). Analyses were performed using TreeAge Pro 2021, release 1.2 (TreeAge LLC) and Microsoft Excel 2021, version 16.54 (Microsoft Corporation).

Results

A total of 5988 patients were included in the analysis, with a mean age of 72 years, NYHA class II to IV heart failure, and left ventricular EF greater than 40%. The model predicted a mean of 1.0 HF hospitalizations for patients in the SOC arm compared with 0.8 HF hospitalizations for patients receiving empagliflozin. The average length of survival was estimated at 8.9 life-years. On a discounted basis, the SOC arm had a mean of 4.9 QALYs at a lifetime cost of $171 357. Empagliflozin increased QALYs by 0.06 and average lifetime costs by $26 257, based on total drug costs of $29 586 and savings of $3329 from reduction in HF hospitalizations (11% of the total cost). There were no savings from reduced hospitalizations for non-HF causes, given that rates of non-HF hospitalization did not differ significantly between treatment arms in EMPEROR-Preserved.⁵ The cost per QALY gained with empagliflozin was $437 442. After including a 9% reduction in cardiovascular death, patients in the SOC arm survived an average of 8.8 life-years compared with 9.0 life-years for those receiving empagliflozin. In this secondary analysis, empagliflozin added 0.17 QALYs and $29 745 in costs, amounting to an ICER of $174 053 per QALY gained. Cost savings from the reduction in HF hospitalizations and cardiovascular death were largely offset by increased outpatient costs and non-HF hospitalizations with prolonged life as well as noncardiovascular death.

Subgroup Analyses

Empagliflozin had similar cost-effectiveness among patients with and without diabetes and across EF strata and HF-related health status (Table 2). The cost per QALY was slightly lower for patients with diabetes and EF less than 50% due to higher baseline rates of HF hospitalization. In the scenario with mortality reduction, the mild health status impairment subgroup had a larger improvement in cost-effectiveness compared with the moderate impairment subgroup due to prolonged survival at higher baseline utility.

Table 2. Outcomes of Empagliflozin Treatment vs Standard of Care.

Treatment group	Mortality reduction
	Without CV					With CV
	Mean No. of HFH	Life-yearsa	Mean QALYs	Cost, $	ICER, $/QALY	Life-yearsa	Mean QALYs	Cost, $	Cost/LY gained, $b	ICER, $/QALY
Full cohort
Standard of care	1.0	8.9/7.5	4.9	171 357	NA	8.8/7.4	4.9	169 438	NA	NA
Empagliflozin	0.8		5.0	197 615	437 442	9.0/7.6	5.0	199 183	151 929	174 053
Subgroups
Diabetes
Standard of care	1.1	7.8/6.6	4.4	168 147	NA	7.6/6.5	4.3	166 175	NA	NA
Empagliflozin	0.8		4.4	190 702	419 739	7.9/6.7	4.5	192 287	139 728	162 334
No diabetes
Standard of care	0.9	10.4/8.5	5.6	175 522	NA	10.2/8.4	5.6	173 651	NA	NA
Empagliflozin	0.7		5.7	206 357	454 942	10.5/8.6	5.7	207 916	166 379	188 464
EF <50%
Standard of care	1.0	8.1/6.8	4.5	163 430	NA	7.9/6.7	4.5	161 918	NA	NA
Empagliflozin	0.6		4.6	184 921	378 607	8.2/7.0	4.7	187 005	122 870	143 821
EF ≥50%
Standard of care	1.0	9.4/7.8	5.2	175 944	NA	9.2/7.7	5.1	173 764	NA	NA
Empagliflozin	0.9		5.2	204 917	468 424	9.5/7.9	5.3	206 127	170 399	192 367
Moderately impaired HF health statusc
Standard of care	1.1	8.6/7.2	4.4	176 289	NA	8.5/7.1	4.3	174 245	NA	NA
Empagliflozin	0.8		4.4	201 369	430 888	8.7/7.3	4.5	203 015	148 273	180 749
Mildly impaired HF health statusc
Standard of care	0.9	9.6/8.0	5.9	164 319	NA	9.4/7.8	5.8	162 596	NA	NA
Empagliflozin	0.7		5.9	193 040	451 658	9.7/8.0	6.0	194 488	160 124	169 647

Abbreviations: CV, cardiovascular; EF, ejection fraction; HF, heart failure; HFH, heart failure hospitalization; HR, hazard ratio; ICER, incremental cost-effectiveness ratio; LY, life-year; NA, not applicable; RR, rate ratio; QALY, quality-adjusted life-year.

^a Presented as undiscounted/discounted life-years.

^b Costs per life-year gained uses discounted life-years given costs are discounted in the model.

^c Mildly impaired health status due to heart failure is defined as Kansas City Cardiomyopathy Questionnaire Overall Summary Score of 82.3 or greater and moderately impaired as less than 82.3.

Price Analyses

We then modeled cost variability for empagliflozin to assess the outcome of lower drug prices (Figure 1). To meet a willingness-to-pay threshold of $180 000 per QALY, empagliflozin would need to cost less than $153.56 per month, substantially lower than the base case price of $326.69 per month. Accordingly, the Medicaid price of $121.86 per month yielded an ICER of $132 862 per QALY gained. However, the more broadly available Medicare Part D monthly prices of $510.54 before rebates and $375.24 after rebates amounted to ICERs of $710 825 and $509 636 per QALY gained, respectively. The ICERs for alternative cost estimates, with and without a reduction in cardiovascular mortality, are presented in eTable 8 in the Supplement. Based on upcoming patent expirations, we also evaluated the effect of potential generic equivalents entering the market between 2025 and 2029. Assuming a significant price decline to $5.15 per month, the cost-effectiveness of empagliflozin ranged from $143 533 to $316 748 per QALY gained without mortality reduction. The cost per QALY would be below $180 000 per QALY if empagliflozin became a low-cost generic by 2026.

Figure 1. Cost-effectiveness Based on the Monthly Price of Empagliflozin.

The incremental cost-effectiveness ratio is plotted across a broad range of price points for empagliflozin in 2021 dollars. Scenarios with and without treatment effect on cardiovascular mortality are shown. Light shaded and darker shaded regions indicate intermediate and high value of treatment, respectively. Vertical lines are plotted at different monthly costs associated with empagliflozin: A, Medicaid price of $121.86; B, VA price of $195.69; C, base case of $326.69; D, Medicare Part D price of $375.24 after mean rebate (26.5%); E, wholesale acquisition cost of $548.54; F, retail pharmacy price of $644.50. CV indicates cardiovascular; QALY, quality-adjusted life-year; VA, Department of Veterans Affairs; WTP, willingness to pay.

Additional Scenarios

Several clinically relevant scenarios were modeled (eTable 9 in the Supplement). The substantially higher incremental utility of 0.019 seen in other clinical trials added 0.14 QALYs without mortality reduction and 0.26 QALYs with mortality reduction; the costs per QALY gained were $182 410 and $116 326, respectively. Separately, we assumed no annual decrement in baseline utility, which yielded an ICER of $433 176 per QALY gained without mortality reduction and $137 091 per QALY gained with mortality reduction.

We assessed heterogeneous treatment effects in 2 scenarios (eTable 9 in the Supplement). First, we modeled a differential treatment effect on HF hospitalization and KCCQ change across HF-related health status subgroups using EMPEROR-Preserved effectiveness estimates across KCCQ tertiles. The ICERs remained similar. Second, we modeled potential differences in treatment effectiveness across age based on trial subgroup effectiveness estimates. Assuming a mortality reduction, we estimated a lower cost per QALY among older patients.

Sensitivity Analysis

We performed 1-way sensitivity analyses on all input parameters (Figure 2). In addition to monthly cost, the model was most sensitive to health status benefits as measured by KCCQ changes and treatment effect on cardiovascular death. To meet a willingness-to-pay threshold of $180 000 per QALY, empagliflozin would need to increase utility by 0.020, compared with 0.008 in the base case. Assuming base-case cost and health status benefits, empagliflozin would yield a cost per QALY below $180 000 if it reduced cardiovascular death by 8% (eTable 12 in the Supplement), comparable to the trial-based point estimate of 9%.⁵ Cost-effectiveness estimates were notably insensitive to the costs of hospitalization and death.

Figure 2. Tornado Plot Demonstrating 1-Way Sensitivity Analyses for Relevant Parameters.

The 1-way sensitivity analyses were performed for scenarios with and without treatment effect on cardiovascular mortality. Parameters were tested across 95% CIs where available or across reasonable uncertainty intervals otherwise. CVD indicates cardiovascular death; HFH, heart failure hospitalization; HR, hazard ratio; NCVD, noncardiovascular death; RR, rate ratio; QALY, quality-adjusted life-year.

Figure 3 demonstrates the interactions between incremental utility and monthly cost of empagliflozin and between the frequency and disutility of HF hospitalization. Other 2-way and 3-way sensitivity analyses did not reveal notable interactions (eTables 10 and 11 in the Supplement). In probabilistic sensitivity analysis, the cost per QALY gained was less than $180 000 in 2.7% of simulations without mortality reduction and 57.7% of simulations with mortality reduction (eFigure 3 in the Supplement).

Figure 3. Select 2-Way Sensitivity Analyses.

The 2-way sensitivity analyses were performed for scenarios with and without treatment effect on cardiovascular mortality. Light shaded and darker shaded regions indicate intermediate and high value of treatment, respectively. A and B, Incremental utility (KCCQ change) and monthly cost of empagliflozin are simultaneously varied. The base case quality-of-life benefit of empagliflozin, as well as higher estimates from other trials of sodium-glucose cotransporter-2 inhibitors, are shown as vertical lines. C and D, Monthly rate and duration of disutility of hospitalization for heart failure are simultaneously varied. The base case rate of hospitalization for heart failure is shown as a vertical line. Duration of disutility may also be interpreted as severity of quality-of-life impairment (eg, 2 months of disutility equivalent to 1 month of disutility twice as severe as the base case estimate). CV indicates cardiovascular; HFH, heart failure hospitalization; ICER, incremental cost-effectiveness ratio; KCCQ, Kansas City Cardiomyopathy Questionnaire; RR, rate ratio; QALY, quality-adjusted life-year.

Discussion

In this economic evaluation, we evaluated the cost-effectiveness of empagliflozin in patients with EF greater than 40% from a US health system perspective. We found that empagliflozin provides low value in the treatment of patients with HFpEF, with a cost per QALY gained of $437 442. This is partially explained by its lack of efficacy on noncardiovascular end points such as non-HF hospitalizations and noncardiovascular mortality, which comprise a high proportion of the overall morbidity in HFpEF.⁴

Our primary analysis did not assume any cardiovascular mortality benefit with empagliflozin. The EMPEROR-Preserved trial did not demonstrate, nor was powered for, a significant effect on cardiovascular mortality (HR, 0.91; 95% CI, 0.76-1.09).⁵ In a secondary analysis, a 9% relative reduction in cardiovascular mortality brought the cost per QALY gained to $174 053, just under the threshold for intermediate value ($180 000 per QALY). The DELIVER (Dapagliflozin in Heart Failure With Mildly Reduced or Preserved Ejection Fraction) randomized clinical trial found that dapagliflozin significantly reduced the composite outcome of cardiovascular death or worsening HF (HR, 0.82; 95% CI, 0.73-0.92) among patients with HF with EF greater than 40% without a significant reduction in cardiovascular death (HR, 0.88; 95% CI, 0.74-1.05).⁵¹ A meta-analysis⁵² of EMPEROR-Preserved and DELIVER also did not demonstrate a significant reduction in cardiovascular death (HR, 0.88; 95% CI, 0.77-1.00, excluding unknown deaths). However, including trials of SGLT2 inhibitors that enrolled patients with reduced EF or inpatients with worsening HF yielded significant pooled effects on cardiovascular death (HR, 0.87; 95% CI, 0.79-0.95) and all-cause death (HR, 0.92; 95% CI, 0.86-0.99) without evidence of heterogeneity.⁵² Further evidence supporting the cardiovascular mortality benefit of empagliflozin among patients with HFpEF would improve its economic value.

The results of the current analysis were sensitive to the cost of empagliflozin, which varies widely across health systems and payers. Absent mortality reduction or larger health status benefit, empagliflozin would need to cost less than $153.56 monthly to meet the willingness-to-pay threshold of $180 000 per QALY (Figure 2). Therefore, at the price paid by Medicaid ($122 per month), treatment with empagliflozin may be considered intermediate value. However, at the more broadly available monthly pricing for Medicare Part D of $375 after rebates and $511 before rebates, empagliflozin yields ICERs of $509 636 and $710 825 per QALY gained, respectively, which would remain low value (>$180 000/QALY). Generic competition may also put downward pricing pressure on empagliflozin, but loss of patent exclusivity will not occur in the near term.⁵⁰ Although price discounts improve the cost-effectiveness of empagliflozin, modest reductions in price alone may be inadequate to bolster the economic value of empagliflozin among patients with HFpEF.

Importantly, the incremental utility of empagliflozin, driven by improvements in patient-reported health status, also factors heavily in its cost-effectiveness. Several trials of SGLT2 inhibitors—dapagliflozin in PRESERVED-HF and DETERMINE-Preserved and canagliflozin in CHIEF-HF^39,40,41—have demonstrated changes in KCCQ scores approximately 2 to 3 times the effect observed in EMPEROR-Preserved.⁵ Pooling the incremental utilities from these trials led to costs per QALY gained of $182 410, nearing intermediate value. Additional research may better quantify the magnitude of efficacy that SGLT2 inhibitors have on health status. If future studies replicate the larger health status benefits seen in other trials of SGLT2 inhibitors in HFpEF, empagliflozin may be able to provide intermediate value at prices achieved by many payers, especially considering expected generic entry within 5 to 10 years (Figure 3).

The cost-effectiveness estimate for empagliflozin was relatively robust to changes in the rate and disutility of HF hospitalizations. This study illustrates that reduction in HF hospitalization alone generally has limited effect on the economic value of a novel HF therapy in contrast to mortality reduction or substantial health status benefit. In addition, the efficacy of empagliflozin on non-HF or noncardiac hospitalizations, a source of substantial morbidity in HFpEF, was not demonstrated in EMPEROR-Preserved.⁵ Nevertheless, Figure 3 indicates that empagliflozin could provide intermediate value for super-utilizers—patients with the highest admission rates, longer hospital stays, and often disproportionate burden of illness due to medical and/or social complexity.^53,54 When selecting patients for treatment, remembering the trial’s inclusion and exclusion criteria will be important to ensure appropriate use of therapy, as only 0.3% of EMPEROR-Preserved participants had NYHA class IV heart failure.⁵⁵

Cost-effectiveness is only one input to coverage decisions. For HFpEF, the high disease morbidity and lack of alternate effective therapies should be considered. Empagliflozin has the potential to change the SOC and improve cardiovascular outcomes for patients with HFpEF. However, as the population ages and prevalence of cardiovascular disease rises,^56,57 the budget impact of empagliflozin will likely grow. Widespread adoption of empagliflozin may lead to high out-of-pocket costs, which typically widen disparities in care. Health systems and payers therefore face important decisions around access to empagliflozin. Based on the findings of the present analysis and current available data, empagliflozin provides low value at the population level, but certain high-risk subgroups may experience additional value from treatment with empagliflozin. The clinical heterogeneity of HFpEF underscores the need for microsimulation cost-effectiveness modeling to accommodate patient-level complexity across clinical history, health status, and outcomes.

Limitations

The findings of this study should be interpreted in the context of the following limitations. First, EMPEROR-Preserved⁵ had a median follow-up of 26 months. We assumed efficacy and quality-of-life benefits with empagliflozin would extend beyond this period. Given that the value of empagliflozin would be lower if benefits do not persist after 26 months, our analyses should be updated when extended follow-up data become available. Second, EQ-5D-3L measurements were not directly reported in EMPEROR-Preserved, so we used published KCCQ-OSS estimates to calculate QALYs. The lack of direct utility mapping for patient-reported health status measures, such as the KCCQ, is a notable limitation. While the KCCQ is not an exact estimate of health state utilities, we minimized inaccuracy by bootstrapping conversions of KCCQ-OSS to EQ-5D-3L scores using a validated algorithm.³⁰ Furthermore, our utility estimate of 0.0079 from a mean KCCQ-OSS difference of 1.6 is comparable to the mean changes in EQ-5D-3L of 0.0084 and in KCCQ-OSS of 2.3 reported in PARADIGM-HF.⁵⁸ Third, we did not account for any renoprotective benefits of empagliflozin because the effect size for the composite renal outcome was small and nonsignificant.⁵ However, there was less worsening of estimated glomerular filtration rate among patients treated with empagliflozin. This may improve long-term outcomes, independent of benefits observed in the trial. Fourth, our treatment cost estimates across payers are based on average rebates and discounts for top-selling brand-name therapies.⁵⁹ Actual net prices for specific drugs are frequently unavailable or confidential. Fifth, willingness-to-pay thresholds are inherently uncertain, likely to vary across health systems, and should be viewed accordingly as a rough guide for interpreting cost-effectiveness estimates. Finally, although other SGLT2 inhibitors are being studied for HFpEF, results of the current analysis do not represent the cost-effectiveness of those agents, as model assumptions will differ.

Conclusions

In this economic evaluation, findings suggest that empagliflozin currently provides low economic value for patients with NYHA class II to IV heart failure and EF greater than 40% based on its effect on quality of life and HF hospitalizations alone. These results were robust to sensitivity analyses that addressed the uncertainty in model parameters. The cost-effectiveness of empagliflozin should be reevaluated if future data or market developments demonstrate substantial survival benefit, improved quality of life, or broad availability at lower prices. Further research may investigate the economic value of empagliflozin in subgroups of HFpEF that potentially derive larger benefits from treatment.

Supplement.

eMethods 1. Technical Details

eMethods 2. Standard of Care Arm Event Rates

eMethods 3. Empagliflozin Arm Event Rates

eMethods 4. Event Rates by Subgroups

eMethods 5. Calibration of Event Rates and Rate Ratios

eMethods 6. Treatment Costs

eMethods 7. Heart Failure Costs

eMethods 8. Additional Costs

eMethods 9. Utility of Chronic Heart Failure

eMethods 10. Utility by Subgroup

eMethods 11. Incremental Utility of Empagliflozin Treatment

eMethods 12. Disutility of Adverse Events

eMethods 13. Additional Sensitivity Analyses

eMethods 14. Probabilistic Sensitivity Analysis

eTable 1. Transition Probabilities (Monthly) and Rate Ratios

eTable 2. Monthly Probabilities and Rate Ratios of Cardiovascular Mortality, Non-cardiovascular Mortality, and HF Hospitalizations by Age

eTable 3. Comparison of Model Output to Trial Data over Median Trial Follow-up of 26 Months

eTable 4. Event Probabilities and Rate Ratios Before and After Calibration

eTable 5. Cost Inputs

eTable 6. Outpatient Costs by Age and Cohort

eTable 7. Utility Inputs

eTable 8. Incremental Cost-Effectiveness Ratio with Alternate Cost Estimates for Empagliflozin

eTable 9. Additional Scenario Analyses

eTable 10. Additional 2-Way Sensitivity Analysis Results

eTable 11. Additional 3-Way Sensitivity Analysis Results

eTable 12. Threshold Analyses for a Willingness-to-pay of $180,000 / QALY

eTable 13. CHEERS Checklist

eFigure 1. Model Summary

eFigure 2. Age-related Variability in Cardiovascular Mortality, Non-cardiovascular Mortality, and Heart Failure Hospitalization Probabilities

eFigure 3. Probabilistic Sensitivity Analysis Scatter Plot

eReferences.