Cost-effectiveness of Dapagliflozin for the Treatment of Heart Failure With Reduced Ejection Fraction.

Importance: Heart failure with reduced ejection fraction produces substantial morbidity, mortality, and health care costs. Dapagliflozin is the first sodium-glucose cotransporter 2 inhibitor approved for the treatment of heart failure with reduced ejection fraction. Objective: To examine the cost-effectiveness of adding dapagliflozin to guideline-directed medical therapy for heart failure with reduced ejection fraction in patients with or without diabetes. Design, Setting, and Participants: This economic evaluation developed and used a Markov cohort model that compared dapagliflozin and guideline-directed medical therapy with guideline-directed medical therapy alone in a hypothetical cohort of US adults with similar clinical characteristics as participants of the Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction (DAPA-HF) trial. Dapagliflozin was assumed to cost $4192 annually. Nonparametric modeling was used to estimate long-term survival. Deterministic and probabilistic sensitivity analyses examined the impact of parameter uncertainty. Data were analyzed between September 2019 and January 2021. Main Outcomes and Measures: Lifetime incremental cost-effectiveness ratio in 2020 US dollars per quality-adjusted life-year (QALY) gained.
Results: The simulated cohort had a starting age of 66 years, and 41.8% had diabetes at baseline. Median (interquartile range) survival in the guideline-directed medical therapy arm was 6.8 (3.5-11.3) years. Dapagliflozin was projected to add 0.63 (95% uncertainty interval [UI], 0.25-1.15) QALYs at an incremental lifetime cost of $42 800 (95% UI, $37 100-$50 300), for an incremental cost-effectiveness ratio of $68 300 per QALY gained (95% UI, $54 600-$117 600 per QALY gained; cost-effective in 94% of probabilistic simulations at a threshold of $100 000 per QALY gained). Findings were similar in individuals with or without diabetes but were sensitive to drug cost. Conclusions and Relevance: In this study, adding dapagliflozin to guideline-directed medical therapy was projected to improve long-term clinical outcomes in patients with heart failure with reduced ejection fraction and be cost-effective at current US prices. Scalable strategies for improving uptake of dapagliflozin may improve long-term outcomes in patients with heart failure with reduced ejection fraction.

Introduction

Sodium glucose cotransporter 2 (SGLT2) inhibitors were developed for the treatment of diabetes but were incidentally noted to improve cardiovascular outcomes.[1,2,3,4,5] In the Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction (DAPA-HF) trial, adding dapagliflozin to guideline-directed medical therapy (GDMT) for heart failure with reduced ejection fraction (HFrEF) reduced the risk of cardiovascular death or heart failure hospitalization by 26% compared with GDMT alone, regardless of the presence or absence of diabetes.[4] On May 5, 2020, dapagliflozin became the first SGLT2 inhibitor to receive approval from the US Food and Drug Administration (FDA) for the treatment of HFrEF, including among patients without diabetes.[6]

The arrival of a new class of therapeutics for HFrEF is a welcome development, as heart failure remains a prevalent disease that produces substantial morbidity and mortality and generates enormous health care costs. As heart failure is a leading cause of hospitalization in the US, dapagliflozin may improve health outcomes and reduce health care costs by averting heart failure hospitalizations downstream, but widespread uptake at a list price of $6188 per year of treatment could produce a large increase in pharmaceutical expenditures. A systematic cost-effectiveness evaluation from a US health care sector perspective would quantify the trade-off between increased health care costs and improved health outcomes and therefore help inform pricing and adoption strategies. This is particularly important for SGLT2 inhibitors because diabetes therapies have historically seen large manufacturer discounts, suggesting that actual prices faced by payers may be highly variable and subject to negotiation with manufacturers.[7] We therefore evaluated the cost-effectiveness of adding dapagliflozin to GDMT for the management of HFrEF in patients with or without diabetes at baseline from a US health care sector perspective and a lifetime analytic horizon.

Methods

Model Overview

We developed a Markov cohort model to compare 2 treatment strategies among patients with HFrEF: (1) GDMT, comprised of an angiotensin converting enzyme inhibitor, an angiotensin-receptor blocker, or an angiotensin receptor neprilysin inhibitor, in addition to a β-blocker and a mineralocorticoid receptor antagonist[8,9]; and (2) dapagliflozin (10 mg once daily) added to GDMT. The analysis adopted a US health care sector perspective and a lifetime analytic horizon. Future costs and outcomes were discounted at 3% annually.[10] We adhered to the recommendations of the Second Panel on Cost-Effectiveness in Health and Medicine to the extent feasible.[10] Given that the study relied on publicly available data sets, this was deemed to not be human participants research and institutional review board approval was not required per the Common Rule. This study was conducted independently from the commercial sponsor of the DAPA-HF trial.

Model creation and analyses were performed using TreeAge Pro Healthcare 2020 (TreeAge Software), and Microsoft Excel version 16 (Microsoft Corporation). Additional modeling details are provided in eMethods in the Supplement.

Simulated Population

We simulated a hypothetical cohort with characteristics similar to the participants of the DAPA-HF trial, which was a phase 3, placebo-controlled, double-masked, randomized trial that enrolled patients with HFrEF who had New York Heart Association class II, III, or IV symptoms and a left ventricular ejection fraction of 40% or less. Patients with an estimated glomerular filtration rate less than 30 mL/min/1.73 m2 of body surface were excluded. The cohort was stratified by baseline diabetes status to capture the heterogeneity of outcomes by diabetes status in the control arm of the DAPA-HF trial (Figure 1).[11]

Figure 1.

Model Structure

The Markov cohort model used in this study simulated a hypothetical cohort of patients with heart failure with reduced ejection fraction (HFrEF) with clinical characteristics similar to the participants of the Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction trial. The model compared guideline-directed medical therapy (GDMT) in the control arm with dapagliflozin added to GDMT in the intervention arm. In monthly cycles, patients could continue to live with HF, experience an urgent HF visit, experience a HF hospitalization (with or without a readmission in the first 30 days after the index hospitalization), develop incident diabetes, or die from any cause.

Model Inputs

Transitional Probabilities

Key model inputs were derived from published primary and secondary analyses of the DAPA-HF, publicly available US regulatory review documents, published clinical and epidemiologic studies, national health care claims data, and the Medical Expenditure Panel Survey (Table 1).[4,11,12,17,18,19,21,22,23,24,25,26] In monthly cycles, patients could experience a fatal or nonfatal heart failure hospitalization, a heart failure–related urgent care visit, or death from any cause. Patients who were discharged after a heart failure hospitalization in 1 cycle were at increased risk of readmission during the following cycle.[12,13] Patients with diabetes experienced higher rates of heart failure hospitalization and all-cause mortality than patients without diabetes at baseline. Patients without diabetes at baseline could develop diabetes during follow-up (with the attendant increase in risk of adverse events).[26]

Table 1.

Input Parameters

Parameter	Base-case value (range in sensitivity analyses)	Distribution for probabilistic analyses	Source
Transitional probabilities for the GDMT (control) arm
Rate of HF hospitalizations, per person year			Petrie et al,[11] 2020
No diabetes	0.080 (0.071 to 0.089)	β
Diabetes	0.122 (0.111 to 0.133)	β
Proportion of HF hospitalizations that are fatal	0.0954 (0.035 to 0.105)	β	Wadhera et al,[12] 2018
Probability of 30-d readmission after a HF hospitalization	0.20	β	Wadhera et al,[12] 2018
Proportion of HF-specific readmissions	0.37	β	Krumholz,[13] 2013
Rate of urgent HF visits, per person year			Petrie et al,[11] 2020
No diabetes	0.006 (0.003 to 0.009)	β
Diabetes	0.007 (0.004 to 0.010)	β
Rate of incident diabetes, per person year	0.037 (0.030 to 0.044)	β	Inzucchi et al,[14] 2020
All-cause mortality in first 24 mo, per person yeara			Petrie et al,[11] 2020
No diabetes	0.078 (0.069 to 0.087)	β
Diabetes	0.117 (0.106 to 0.128)	β
Death from any cause (beyond 24 mo)	Ratio for all-cause mortality comparing the control arm to the US general population (see eMethods in the Supplement)
Effectiveness of dapagliflozin (intervention arm)
HR for HF hospitalizations, compared with GDMT			Petrie et al,[11] 2020
No diabetes	0.63 (0.48 to 0.81)	Log normal
Diabetes	0.76 (0.61 to 0.95)	Log normal
HR for urgent HF visits, compared with GDMT			Petrie et al,[11] 2020
No diabetes	0.25 (0.07 to 0.89)	Log normal
Diabetes	0.62 (0.24 to 1.59)	Log normal
HR for death from any cause compared with GDMT (first 24 mo)			Petrie et al,[11] 2020
No diabetes	0.88 (0.70 to 1.12)	Log normal
Diabetes	0.78 (0.63 to 0.97)	Log normal
HR for incident diabetes compared with GDMT	0.68 (0.50 to 0.94)	Log normal	Inzucchi et al,[14] 2020
Costs
Dapagliflozin therapy, $/y	4192 (953 to 6188)	Log normal	Base case: FSS-Big 4; lower bound: heavily discounted price; upper bound: wholesale acquisition (all prices estimated August 2020)[15,16]
Background health care costs, $/y			HCUP[17]; Peterson et al,[18] 2015; Kazi et al,[19] 2020
No diabetes
Age, y
<75	20 629 (16 503 to 24 755)	Log normal
75-85	22 512 (18 010 to 27 015)	Log normal
>85	30 811 (24 648 to 36 973)	Log normal
Diabetes
Age, y
<75	28 923 (23 139 to 34 708)	Log normal
75-85	26 430 (21 144 to 31 716)	Log normal
>85	34 249 (27 400 to 41 099)	Log normal
HF hospitalization costs, $	11 827 (8899 to 15 591)	Log normal	Medicare Provider Utilization and Payment Data 2017[20]
Urgent HF visit cost, $	807 (646 to 968)	Log normal	Charges for services provided during an urgent care visit and a cost-center–specific charge-to-payment ratio
Quality of life
Baseline KCCQ-OSS in the GDMT (control) arm	68.6 (68.1 to 69.1)	Normal	Kosiborod et al,[21] 2020
Baseline KCCQ-OSS in the dapagliflozin (intervention) arm	68.4 (68.1 to 69.1)	Normal	Kosiborod et al,[21] 2020
KCCQ-OSS in the GDMT (control) arm at 8 mo	72.7 (72.0 to 73.2)	Normal	Kosiborod et al,[21] 2020
KCCQ-OSS in the dapagliflozin (intervention) arm at 8 mo	75.0 (74.4 to 75.4)	Normal	Kosiborod et al,[21] 2020
Quality-of-life penalty applied for diagnosis of diabetes	−0.0351 (−0.0350 to −0.0352)	Normal	Sullivan et al,[22] 2006
Quality-of-life penalty applied for HF hospitalization	−0.0066 (−0.0135 to 0)	Normal	Jaagosild et al,[23] 1998
Quality-of-life penalty applied for urgent HF visit	−0.0045 (−0.009 to 0)	Normal	Jaagosild et al,[23] 1998

Abbreviations: FSS, Federal Supply Schedule; GDMT, guideline-directed medical therapy; HCUP, Healthcare Costs and Utilization Project; HF, heart failure; HR, hazard ratio; KCCQ-OSS, Kansas City Cardiomyopathy Questionnaire–Overall Summary Score.

Ratio for all-cause mortality comparing the control arm with the US general population beyond 24 mo available in the eMethods in the Supplement.

The model was calibrated to reproduce rates of 18-month survival, urgent heart failure visits, and HFrEF hospitalizations observed in the DAPA-HF trial. Therefore, hospitalizations for HFrEF exacerbations were more frequent in patients with diabetes compared with patients without diabetes, but rates of urgent care visits for heart failure were similar in the 2 groups (Table 1).[11]

Effectiveness of Dapagliflozin

We assumed that the use of dapagliflozin would reduce HFrEF hospitalization, urgent care visits, and all-cause mortality as observed in the DAPA-HF trial (with the effect size stratified by patients’ diabetes status) (Table 1).[11] The base case assumed that the effectiveness of dapagliflozin would be sustained over the lifetime of the patients, but a sensitivity analysis assumed a linear decline in effectiveness of dapagliflozin beyond the trial duration such that dapagliflozin would become ineffective 5 years after trial completion. The base case assumed that dapagliflozin would reduce the risk of incident diabetes among individuals without diabetes at baseline by 32%[26]; a sensitivity analysis assumed no reduction in the risk of incident diabetes.

Safety, Pill-Related Disutility, and Treatment Discontinuation

As there were no observed differences in safety outcomes in the DAPA-HF trial,[4,11] we did not model any additional costs or quality-of-life penalties caused by adverse drug events. As patients in the GDMT arm would already be receiving pills on a daily basis (either several times or twice daily), we did not model any additional pill-related disutility (ie, decrement in quality-of-life from taking a daily pill of dapagliflozin).[27] The base case assumed that patients on dapagliflozin would be adherent to the therapy; a sensitivity analysis examined the effect of nonadherence by incorporating a 0.27% monthly probability of discontinuing dapagliflozin for the first 2 years (in order to replicate the 4.7% discontinuation observed in the DAPA-HF trial).[4]

Survival

The model incorporated a nonparametric survival model as follows. Over the first 18 months, the control arms replicated the survival observed in the DAPA HF trial (separately for patients with and without diabetes at baseline). Next, we compared the observed mortality rate in the control arms in the last 6 months of the trial with the mortality rate in the age-matched US general population, yielding a mortality rate ratio of 8.66 in patients with HFrEF and diabetes and 4.88 in patients with HFrEF but without diabetes at baseline.[28] This rate ratio was applied to the age-specific mortality in the general US population to estimate the age-specific survival in the control arms beyond trial completion (eFigure 1 in the Supplement). A sensitivity analysis assumed more favorable long-term survival in the control arm based on a recently published analysis that pooled the control arms of several contemporary HFrEF trials (eFigure 1 in the Supplement).[29]

The age-specific mortality rate in the intervention arms was estimated by applying the hazard ratio for all-cause mortality observed in the DAPA-HF trial to the control survival curves (separately for patients with or without diabetes). Additional modeling details are available in eMethods in the Supplement.

Costs

Manufacturer discounts and rebates for diabetes therapies such as SGLT2 inhibitors have been particularly large in recent years,[7] so that the net price paid by payers is substantially lower than the list price. As recommended by the Second Panel on Cost-Effectiveness, our base case incorporated the drug cost reported in the Federal Supply Schedule ($4192 for a year’s supply in August 2020).[10,15] In deterministic sensitivity analyses, we varied the annual cost of dapagliflozin from $953 (a heavily discounted net price at which dapagliflozin is available to some US payers) to $6188 (the wholesale acquisition cost).[16] We also identified the cost at which the addition of dapagliflozin to GDMT would become cost-effective at thresholds of $50 000, $100 000, and $150 000 per quality-adjusted life-year (QALY) gained (additional details regarding the estimation of costs are provided in the eMethods and eTable 1 in the Supplement). Patients with diabetes were assumed to have higher background health care costs than patients without diabetes. All costs were inflated to 2020 US dollars using the medical component of the Personal Consumption Expenditure Index.[10,30,31]

Quality-of-Life Estimates

The Kansas City Cardiomyopathy Questionnaire overall summary score (KCCQ-OSS) was used in the DAPA-HF trial to measure the heart failure–specific health status of patients at baseline and at 4-month and 8-month follow-up.[4,11,21,32] To translate KCCQ-OSS to quality-of-life weights, we used an algorithm developed by Spertus et al[19] that maps individual-level KCCQ scores to EQ-5D–based health-related quality-of-life estimates (eTables 2 and 3 in the Supplement). The base case replicated quality-of-life estimates reported in the DAPA-HF trial and then modeled an age-adjusted decline beyond the end of the trial period based on the community-based preference scores derived from the Medical Expenditure Panel Survey.[22] The model applied short-term quality-of-life tolls for heart failure hospitalizations and urgent heart failure visits, and a sustained penalty for incident diabetes.[22,23,33]

Statistical Analysis

Our primary outcome was the incremental cost-effectiveness ratio (ICER) of adding dapagliflozin to GDMT compared with GDMT alone (in US dollars per life-year gained and US dollars per QALY gained), for the entire cohort and stratified by baseline diabetes status. We ran 10 000 simulations using input parameter values randomly drawn from prespecified statistical distributions to generate 95% uncertainty intervals (UI) for key outcome measures and to determine the proportion of simulations in which dapagliflozin would be cost-effective at varying willingness-to-pay thresholds. We assumed a cost-effectiveness threshold of $100 000 per QALY gained,[34] but also examined the value of the intervention per the cost/value methodology recommendations of the American College of Cardiology and American Heart Association (high-value, less than $50 000 per QALY gained; intermediate-value, $50 000 or more to less than $150 000 per QALY gained; and low-value, $150 000 or more per QALY gained).[35] In addition to the probabilistic analyses described above, we performed deterministic sensitivity analyses to reflect uncertainty in key input parameters by varying 1 input at a time across the range specified in Table 1 while holding all other parameters at their base-case values.

Results

The simulated cohort in the model had a starting age of 66 years, and 41.8% of patients in the simulation had diabetes at baseline. Median (interquartile range) undiscounted survival in the GDMT arm was 6.8 (3.5-11.3) years (patients without diabetes, 7.6 [3.9-12.3] years; patients with diabetes, 5.7 [3.0-9.9] years). The model replicated the rates of all-cause mortality, HFrEF hospitalization, and urgent heart failure visits observed in the DAPA-HF trial (eTable 4 in the Supplement).

Base Case

Adding dapagliflozin to GDMT in patients with HFrEF was projected to lower the rate of HFrEF hospitalizations from 0.10 (95% CI, 0.09-0.11) to 0.07 (95% CI, 0.06-0.08) per person-year and improve quality-adjusted survival by 0.63 (95% uncertainty interval [UI], 0.25-0.94) QALYs (Table 2). Patients receiving dapagliflozin incurred $27 700 (95% UI, $25 700-$29 800) in lifetime spending on dapagliflozin, which was only partially offset by savings resulting from reduced HFrEF hospitalizations. After accounting for increased health care costs related to prolonged survival, the intervention arm had a net increase in lifetime health care costs of $42 800 (95% UI, $37 100-$50 300). As a result, adding dapagliflozin to GDMT had an ICER of $68 300 per QALY gained (95% UI, $54 600-$117 600 per QALY gained) compared with GDMT alone, and was cost-effective in 94% of 10 000 probabilistic simulations (Table 2 and Figure 2).[36] Subgroup analyses demonstrated similar ICERs among patients without or with diabetes (without diabetes: $69 600 [95% UI, $50 700-$445 700] per QALY gained; cost-effective in 84% of 10 000 probabilistic simulations; with diabetes: $66 800 [95% UI, $53 800-$116 600] per QALY gained; cost-effective in 95% of 10 000 probabilistic simulations) (Table 2).

Table 2.

Base Case Results

Characteristic	All patients		No diabetes		Diabetes
	GDMT	Dapagliflozin	GDMT	Dapagliflozin	GDMT	Dapagliflozin
Health outcomes
Survival, life-years (undiscounted)	6.82 (6.77-6.86)	7.73 (7.10-7.76)	7.60 (7.51-7.68)	8.42 (7.57-9.28)	5.73 (5.38-6.13)	6.77 (5.59-8.18)
Survival, life-years (discounted)	5.91 (5.87-5.91)	6.6 (6.13-7.10)	6.52 (6.46-6.57)	7.12 (6.48-7.77)	5.07 (4.79-5.39)	5.88 (4.96-6.96)
Incremental life years (discounted)	[Reference]	0.64 (0.21-1.11)	[Reference]	0.61 (0.18-1.13)	[Reference]	0.81 (0.22-1.45)
QALYs (discounted)	4.73 (4.69-4.76)	5.36 (4.98-5.76)	5.28 (5.23-5.33)	5.86 (5.33-6.38)	3.96 (3.77-4.24)	4.66 (3.96-5.56)
Incremental QALYs (discounted)	[Reference]	0.63 (0.25-0.94)	[Reference]	0.58 (0.21-0.98)	[Reference]	0.70 (0.23-1.20)
evLYGa	[Reference]	0.76 (0.29-1.26)	[Reference]	0.69 (0.04-1.31)	[Reference]	0.87 (0.28-1.51)
Direct health care costs, $b
Lifetime health care costs (discounted)	150 600 (131 300-172 200)	193 400 (168 400-222 500)	148 900 (126 000-174 700)	189 000 (158 800-224 400)	152 900 (126 000-174 700)	199 400 (163 400-224 400)
Spending on dapagliflozin	NA	27 700 (25 700-29 800)	NA	29 900 (27 200-32 600)	NA	24 700 (22 200-27 300)
Spending on HF hospitalizations	6900 (5200-9200)	5400 (3900-7500)	6600 (4900-8800)	4600 (3200-6600)	7300 (5500-9800)	6500 (4500-9300)
Incremental health costs (discounted)	[Reference]	42 800 (37 100-50 300)	[Reference]	40 100 (32 700-49 700)	[Reference]	46 500 (32 700-49 700)
ICER, $
Per life-year gained	[Reference]	61 800 (47 500 131 700)	[Reference]	66 200 (45 000-dominated)	[Reference]	57 300 (44 800-123 800)
Per QALY gained	[Reference]	68 300 (54 600-117 600)	[Reference]	69 600 (50 700-445 700)	[Reference]	66 800 (53 800-116 600)
Per evLYGa	[Reference]	56 100 (44 700-100 700)	[Reference]	58 500 (42 300-315 600)	[Reference]	53 400 (42 800-97 200)

Abbreviations: evLYG, equal value of life-years gained; GDMT, guideline-directed medical therapy; HF, heart failure; ICER, incremental cost-effectiveness ratio; NA, not applicable; QALY, quality-adjusted life-year; UI, uncertainty interval.

As the use of QALYs may undervalue prolonged survival among individuals with imperfect quality-of-life at baseline, we also computed the incremental cost evLYG, an approach that assumes that any extension of life has a perfect quality-of-life.[36]

In 2020 US dollars.

Figure 2.

Sensitivity Analyses

In panel A, the blue dashed line indicates the assumed cost-effectiveness threshold of $100 000 per quality-adjusted life years (QALYs) gained (blue dashed line). Adding dapagliflozin to GDMT was cost-effective in 94% of 10 000 probabilistic simulations.

Sensitivity Analyses

The cost-effectiveness of adding dapagliflozin to GDMT compared with GDMT alone was sensitive to the annual cost of dapagliflozin (Figure 3), the association of dapagliflozin with the risk of developing diabetes among individuals without diabetes at baseline, and the durability of effectiveness (Figure 2). For instance, if the annual cost of dapagliflozin were as low as $500, the ICER for dapagliflozin compared with GDMT alone would decline to $29 400 per QALY gained. If dapagliflozin had no effect on the rate of incident diabetes among individuals with diabetes at baseline, the ICER would increase to $73 500 per QALY gained. If the association of dapagliflozin with all-cause mortality were to reduce linearly for 5 years after trial completion, the ICER would increase to $89 300 per QALY gained. The use of dapagliflozin would be cost-effective at $50 000, $100 000, and $150 000 per QALY gained at annual costs of $2500 (40.4% lower than base-case price and 59.6% lower than list price), $7200 (72% higher than base-case price and 16% higher than list price), and $11 900 (183% higher than base-case price and 92% higher than list price), respectively (Figure 3). Additional results from sensitivity analyses are presented in the eMethods, eTables 5 and 6, and eFigure 2 in the Supplement.

Figure 3.

Incremental Cost-effectiveness of Dapagliflozin in Treatment of Heart Failure With Reduced Ejection Fraction

We varied the annual cost of dapagliflozin, holding all other input parameters at their base-case value. The different annual costs analyzed included: (1) Federal Supply Schedule (base case, $4192); (2) wholesale acquisition price ($6188); (3) the price obtained if the wholesale acquisition were discounted by 40% (average value of rebates and discounts on diabetes pharmaceuticals) ($3713); and (4) a heavily discounted net price at which dapagliflozin is available in some US markets ($953). The color quadrants indicate the annual cost at which adding dapagliflozin to guideline-directed medical therapy would be cost-effective relative to guideline-directed medical therapy alone at thresholds of $50 000, $100 000, and $150 000 per quality-adjusted life year (QALY) gained. These results permit health systems and clinicians to estimate the incremental cost-effectiveness of dapagliflozin for HFrEF in the context of their cost of a year’s supply of dapagliflozin.

Discussion

In a simulation model of HFrEF in the US, we found that the addition of dapagliflozin to GDMT was projected to improve quality-adjusted survival in patients with HFrEF, and its use would be cost-effective relative to GDMT alone at 2020 Federal Supply Schedule prices. The lifetime ICER of $68 300 per QALY gained compared with GDMT alone would make the use of dapagliflozin for the treatment of HFrEF an “intermediate-value” intervention per the American College of Cardiology/American Heart Association Cost/Value framework.[35] Remarkably, the ICER was similar in patients with or without diabetes at baseline. This is because, compared with patients without diabetes, patients with diabetes receive a larger clinical benefit (given their higher baseline risk of major adverse cardiovascular events) but also accrue higher health care costs for every year of prolonged survival. The observation that dapagliflozin therapy was cost-effective in 94% of 10 000 probabilistic simulations suggests that our findings are robust across a wide range of estimates of key model parameters.

To our knowledge, this is the first systematic cost-effectiveness analysis of dapagliflozin for patients with HFrEF from a US health care sector perspective. Prior analyses performed from the perspective of other health systems (including UK, Germany, Spain, China, and Australia) also projected dapagliflozin to be cost-effective in those health systems, but substantial cross-country differences in drug and health care costs and HFrEF outcomes preclude direct comparisons.[37,38,39] For instance, our estimates of long-term effectiveness of dapagliflozin (a gain of 0.63 QALYs) is comparable to the gain of 0.48 to 0.50 QALYs in the British and European models, but somewhat higher than the gains in the Chinese (0.38) and Australian (0.29) models. The annual cost of dapagliflozin is substantially higher in the US compared with other health systems ($4192 in our base case, compared with the equivalent of $302-$780 in Europe, $257 in China, and $563 in Australia per the published analyses).[37,38,39] Furthermore, we included all background health care costs (in line with current recommendations) whereas the European analysis only included costs related to diabetes and heart failure.[10] As a result, despite comparable health gains, the ICER was substantially higher in our analysis ($54 600-$117 600 per QALY gained compared with $6343-$11 091 per QALY gained in Europe, $3828 per QALY gained in China, and $9550 per QALY gained in Australia).

Cost-effectiveness analysis provides a formal mechanism to quantify the gains or losses in population health as a result of a particular intervention.[40] Treatments for HFrEF may potentially have an outsized impact on health care budgets because a large number of patients may be eligible for lifelong therapy, which necessitates the prioritization of cost-effective therapies. However, the slow initial uptake of sacubitril-valsartan suggests that evidence of cost-effectiveness alone cannot surmount the constellation of patient-, physician-, and health system–related barriers that can impede widespread adoption of new therapies.[5,41,42] Key among them is the fragmented US health care system, where financial incentives for various stakeholders are frequently misaligned. For instance, pharmaceutical and medical coverage may be offered by different payers, so that the pharmaceutical plan does not have the incentive to pay for high-cost medications that prevent hospitalizations, the savings of which accrue to the medical plan. Even when insurers are responsible for both medical and pharmaceutical benefits, as with many commercial insurers and Medicare Advantage plans, payers may not benefit from savings that occur several years in the future, at which point the patient may have switched to a different plan. As a result, payers frequently impose prior authorization requirements or high copayments to restrict uptake of high-cost therapies. Copayments are typically based on the list price of the drug before any manufacturer discounts and vary from month-to-month based on the benefit structure of the insurance plan[7,43]; they therefore pose a substantial financial burden on patients and adversely affect persistence. Until policy reforms address the misalignment of financial incentives for payers in the US health system and cap or eliminate copayments for cost-effective therapies (as in recent proposals to expand access to insulin),[44] clinicians should discuss medication affordability with their patients at regular intervals in order to preempt cost-related nonadherence.

Although total and out-of-pocket costs have slowed uptake of novel cardiovascular therapies in the recent past,[19,43,45,46] 3 factors justify our cautious optimism that the experience with SGLT2 inhibitors will be different. First, diabetes therapies, including SGLT2 inhibitors, receive some of the largest manufacturer discounts observed in the US pharmaceutical market, and these discounts have fully offset increases in list prices in recent years.[7] Second, the availability of several SGLT2 inhibitors with proven benefit in HFrEF treatment may allow payers to negotiate better prices in return for preferential formulary placement.[5,47] As a result, the majority of US payers will be able to continue to access at least 1 SGLT2 inhibitor at a reasonable price. Finally, recently proposed policy reforms would force payers to pass on at least some of the manufacturer discounts to patients.[7] Collectively, these changes could substantially lower total costs for payers and out-of-pocket costs for patients, and expand access to this cost-effective therapy.

Limitations

This study had several limitations. The efficacy and safety of dapagliflozin were estimated from a single randomized clinical trial with a mean follow-up of 18 months. We estimated long-term survival based on a combination of trial and vital statistics data, and examined alternative survival models in sensitivity analyses, but our results should be updated when data from longer follow-up become available. We did not examine heterogeneity other than by diabetes status (eg, by New York Heart Association symptom class). However, the goal of this study was to evaluate population-level effects on health and health care spending, which are adequately reflected in our use of population-level estimates of risk, benefit, and cost. If future studies demonstrate that treatment with dapagliflozin reduces the need for kidney replacement therapy in patients with HFrEF, our model would have underestimated the economic benefits of SGLT2 inhibitor therapy.[48] The price of dapagliflozin is likely to decline after expiration of the market exclusivity period and entry of multiple generics. We modeled contemporary US prices so that our findings could inform current pricing and uptake. We explored a wide range of drug prices in sensitivity analyses, but our findings should be updated when new data on pricing, effectiveness, or safety become available. We used a novel approach to convert KCCQ-OSS to health-related quality-of-life. However, any bias in the conversion would not have materially affected our findings as it would affect both the control and intervention arms, and the health gains in the dapagliflozin arm were primarily because of improved survival (rather than from improved quality of life).

Conclusions

In a simulation model calibrated to the results of the DAPA-HF trial, adding dapagliflozin to GDMT was projected to produce substantial clinical gains in patients with HFrEF, with an acceptable increase in associated costs. At a willingness-to-pay threshold of $100 000 per QALY gained, treatment with dapagliflozin, at an annual cost of $4192, would be cost-effective therapy in patients with HFrEF regardless of whether treated patients have diabetes. Scalable strategies to ensure affordable access to and widespread uptake of dapagliflozin are urgently needed.