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

Key Points Question Is the addition of dapagliflozin to guideline-directed medical therapy cost-effective for the treatment of heart failure with reduced ejection fraction? Findings In this economic evaluation of a simulated cohort of US adults with heart failure, adding dapagliflozin to guideline-directed medical therapy was projected to prolong survival by 0.63 quality-adjusted life-years while increasing lifetime costs by $42 800, producing an incremental cost-effectiveness ratio of $68 300 per quality-adjusted life-year. Results in individuals with and without diabetes were similar. Meaning These results suggest that widespread uptake of dapagliflozin for the treatment of heart failure with reduced ejection fraction has the potential to improve long-term clinical outcomes and is likely to meet conventional cost-effectiveness thresholds.

In particular, the probability of heart failure readmissions was derived from published literature. In an analysis of Medicare data by Wadhera et.al, the 30-day heart failure readmission after a heart failure hospitalization was approximately 20%. 7 Krumholz and colleagues note that 37% of readmissions that occur within 30 days of discharge after a heart failure hospitalization are for a heart failure exacerbation. 15 Our model used these inputs to simulate the 30-day readmissions for heart failure.

c. Survival Model
The survival model was developed using the following non-parametric approach: i. In a subgroup analysis of the DAPA-HF trial, Petrie et al. reported the trial outcomes stratified by diabetes status at baseline. 4 We first digitized the Kaplan Meier curves for all-cause mortality using WebPlotDigitizer 4.3 (Pacifica, CA). These data were used to estimate the monthly rate of allcause mortality in the control arm, separately for patients without and with diabetes at baseline. In eAppendix Figure 1 ii. We computed the mortality rate of patients in the control arm during the last 6 months of the trial (0.039 in patients without and 0.069 in patients with diabetes at baseline).
iii. We estimated the mortality rate for the age-matched US population using the US Life Tables published by the National Center for Health Statistics (for ages 67.5 to 68 years). 16 In eAppendix Figure 1,  v. We assumed that this increased mortality would be sustained beyond the duration of the trial period. We applied the rate ratio computed in step iv above to the survival in the general US population (eAppendix Figure 1 vi. In each case, all-cause mortality in the intervention arm was estimated by applying the rate ratio for all-cause mortality for dapagliflozin 0.78 for patients with diabetes, and 0.88 for patients without diabetes) to the corresponding age-matched mortality rate in the control arm.
vii. In sensitivity analyses, we modeled a more optimistic estimate of longterm survival based on a recently published analysis that pooled trial-level data from 3 contemporary trials of HFrEF. 3,[17][18][19] In this study by Dr.
Vaduganathan and colleagues, the investigators projected absolute survival gains with comprehensive disease-modifying pharmacological therapy if applied long term, compared with conventional therapies. 17 Patients in the control arm of the studies were assumed to receive at least and angiotensin-converting-enzyme inhibitor, or an angiotensin-receptor blocker, or and a beta-blocker at the recommended or maximal tolerated dose. We used these data to model survival in the control arm in a scenario analysis. In eAppendix Figure 1, panel E, the solid red line shows the survival in patients with diabetes at baseline and solid green line shows the survival in patients without diabetes at baseline, using data from the study by Vaduganathan and colleagues.

d. Cost of Dapagliflozin
In the U.S. pharmaceutical market, the actual drug price faced by payors is obscured by confidential discounts and rebates offered by manufacturers. As a result, the Second Panel of Cost-Effectiveness in Health and Medicine recommended that the reference case assume a drug price equivalent to that in publicly available Federal Supply Schedule. 20 In the base case, we assumed an annual cost of dapagliflozin therapy of $4,192 per the August 2020 Federal Supply Schedule (FSS) "Big 4" pricing. This is the price paid by the four largest federal purchasers of drugs -the Department of Veterans Affairs, the Department of Defense, the Public Health Service, and the Coast Guard in August 2020). 21, 22 We varied this input widely in sensitivity analyses from $953, a highly discounted net price available in some US markets according to SSR Health, a source of discount and net pricing data used in the drug pricing literature, to $6,188, the list price in August 2020. 23,24 We also computed the cost at which the addition of dapagliflozin to GDMT would become cost-effective at cost-effectiveness thresholds of $50,000, The cost of an urgent visit for heart failure is not well defined in the literature. We estimated the mean cost of an urgent heart failure visit at the Direct Access Cardiac Care Unit in our institution. We assumed that every urgent heart failure visit included a venipuncture, one set of laboratory studies (complete blood count, blood chemistry with electrolytes, kidney function tests, and brain natriuretic peptide), one electrocardiogram, and professional charges. Based on expert input, we assumed that 80% of visits require an intravenous infusion of furosemide after the initial bolus dose; the remainder have an adequate response to the intravenous bolus of furosemide alone. Based on expert input and our own institutional experience over the past one year, we assumed that 50% of patients undergo a chest x-ray, and 20% undergo transthoracic echocardiography during that visit. We collected the mean charges for services typically provided during an urgent care visit and applied a cost-center-specific charge-to-payment ratio of 39% from an ambulatory cost center to estimate the cost of the urgent care visit. eAppendix Table 1 shows detailed costs of charges included in the calculation of the urgent heart failure visit cost. g.

Quality-of-Life Parameters
The Kansas City Cardiomyopathy Questionnaire overall summary score (KCCQ-OSS) was used in DAPA-HF to measure heart failure-specific health status at baseline and at 4 and 8-month follow-up. 3,4,10,27 The KCCQ-OSS is a self-administered, 23-item questionnaire that quantifies physical limitations, symptoms, self-efficacy, social interference, and quality of life. The overall summary score ranges from 0 to 100 with higher scores indicating fewer symptoms, and a change of 5 or more points is considered clinically meaningful. 28 The mean (± standard deviation) baseline KCCQ-OSS was  Table 2). 11 eAppendix Table 3  i.

Sensitivity Analyses
As would be expected, secondary analyses of the DAPA-HF study cohort stratified by diabetes status yielded estimates that were less precise than those derived from the primary analysis that included the entire study cohort. Although formal tests of heterogeneity by diabetes did not suggest differences in effect size by diabetes status, when an outcome did not achieve statistical significance in stratified analyses but did so in the entire trial cohort (e.g., all-cause mortality in patients without diabetes at baseline and urgent heart failure visits in patients with diabetes at baseline), we performed additional deterministic sensitivity analyses that examined the effect of a null-or harmful effect of dapagliflozin on these outcomes. 4 © 2021 Isaza N et al. JAMA Network Open.

II. Model Calibration
For model calibration, we performed 10,000 first-order microsimulations of the model to estimate event rates at 18 months and compared absolute rates of events (in the control and dapagliflozin arms) to published results from the DAPA-HF trial. We compared results for the entire study cohort as well as stratified by diabetes status at baseline.
For instance, all-cause mortality among patients receiving GDMT was 13.0% at 18 months in the model compared with 13.1% in the DAPA-HF trial. 3 The rate ratio for mortality among patients

a. Sensitivity of the ICER to Changes in Selected Variables
The ICER in our model was most sensitive to the variation in the annual cost of dapagliflozin cost, the effect of dapagliflozin on the risk of death in both patients without and with diabetes, and the incidence rate of diabetes among individuals without diabetes at baseline. This is depicted in the tornado plot included in Figure2 If dapagliflozin were assumed to have no effect on the risk of incident diabetes, the ICER for dapagliflozin and GDMT compared with GDMT alone in patients without diabetes at baseline increased from $68,300 per QALY gained to $73,500 per QALY gained.
d. Sensitivity Analysis of Declining Effectiveness of Dapagliflozin eAppendix Table 5 shows the results of a sensitivity analysis in which we assumed that the effectiveness of dapagliflozin on all-cause mortality would wean linearly for 5 years after trial completion. In this context, the ICER for dapagliflozin and GDMT compared with GDMT alone increased from $68,300 per QALY gained to $89,300 per QALY gained.
e. Sensitivity Analysis of Alternative Survival eAppendix Table 6 shows the results for a sensitivity analysis in which we used an