SGLT2 inhibitors for patients with heart failure with preserved ejection fraction in China: a cost-effectiveness...

Background: The potential benefits of intervention with empagliflozin or dapagliflozin for patients with heart failure with preserved ejection fraction (HFpEF) were first demonstrated in the EMPEROR-Preserved and DELIVER studies. However, the cost-effectiveness of this intervention (empagliflozin or dapagliflozin) is yet to be established. Methods: In the context of Chinese healthcare, a Markov model was proposed, which incorporates clinical outcomes from the EMPEROR-Preserved and DELIVER studies, to predict the utility and costs over a lifetime. The time horizon was 20 years, and a 5% discount rate was applied to the costs and utilities. The incremental cost-effectiveness ratio (ICER) threshold against willingness to pay (WTP) was set as the primary outcome. The robustness of the decision was evaluated using sensitivity analyses. Results: After a simulated 20-year lifetime, a 72-year-old patient with HFpEF in the intervention group (empagliflozin) showed an increase of 0.44 quality-adjusted life years (QALYs) and $1,623.58 with an ICER of $3,691.56 per QALY, which was lower than the WTP threshold of $12,032.10 per QALY. A 72-year-old patient with HFpEF in the intervention group (dapagliflozin) showed an increase of 0.34 QALYs and $2,002.13 with an ICER of $5,907.79 per QALY, which was lower than the WTP threshold of $12,032.10 per QALY. One-way sensitivity analyses showed that cardiovascular (CV) mortality in the intervention and comparator groups was the most sensitive to the decision. Cost-effectiveness was demonstrated in the intervention group (empagliflozin or dapagliflozin) in 67.9% or 62.2% of 1000 Monte Carlo simulations, respectively. Conclusion: In Chinese healthcare, the interventions (empagliflozin or dapagliflozin) for HFpEF were more cost-effective than the comparators. Our study has provided a quantitative evaluation of the costs and benefits of such interventions for a lifetime using the model.