Heart failure (HF) is a growing global pandemic, affecting over 64.3 million individuals worldwide in 2017 [1]. The improved survival rates among patients with pre-existing HF and increased availability of evidence-based treatments, along with longer life expectancy, are anticipated to elevate HF prevalence [1]. The burden of mortality and hospitalization of HF remains substantial, with a global 5-year mortality rate of 50–75% [1]. Despite the ongoing efforts to treat and manage HF, mortality and hospitalization rates continue to persist largely unchanged [2]. In Hong Kong, the rate of new HF hospitalizations was 0.59 per 1000 population, while the all-cause mortality rates reached 19.5% at 1 year and 54% at 5 years. Cardiovascular mortality rates were 7.2% at 1 year and 20.7% at 5 years. This placed immense pressure on the local healthcare system. The estimated average direct medical cost in 2015 was USD$ 19,969 per patient-year [3]. Therefore, it is necessary to investigate more effective treatment options to reduce the hospitalization and mortality rates caused by HF and alleviate the burden on the local healthcare system.

HF is defined as a clinical syndrome with signs and symptoms caused by a structural and/or functional cardiac abnormality [4]. The American Heart Association has classified HF into three subtypes based on the left ventricular ejection fraction (LVEF): heart failure with preserved ejection fraction (HFpEF) for patients with LVEF greater than 50%, heart failure with mid-range ejection fraction (HFmrEF) for those with LVEF between 40–49%, and heart failure with reduced ejection fraction (HFrEF) for those with LVEF less than 40% [5]. In 2016, the proportion of HF patients with HFpEF and HFrEF was roughly the same (52.3% and 47.7%, respectively) in Hong Kong [6].

Type 2 diabetes mellitus (T2DM) is a major risk factor of HF, and the two conditions frequently coexist. The presence of T2DM not only doubles the risk of developing HF, but also leads to poorer cardiovascular outcomes and prognosis compared to those without T2DM [7]. In patients with HFpEF, T2DM is particularly prevalent, affecting 20–40% of individuals [8]. Clinical trial data have demonstrated that patients with comorbid T2DM and HFpEF experience more severe in-patient and post-discharge morbidity, including a prolonged length of stay, reduced likelihood of discharge, and elevated rates of all-cause and HF readmissions within 30 days [9]. These findings emphasize the importance of identifying and managing T2DM in patients with HFpEF to improve outcomes and reduce the burden of disease.

Despite the availability of multiple treatment options for HFrEF, including angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, beta-blockers, mineralocorticoid receptor antagonists, angiotensin receptor blocker-neprilysin inhibitor, ivabradine, and sodium-glucose co-transporter 2 (SGLT-2) inhibitors, therapeutic approaches for HFpEF remain limited [5]. The current focus of HFpEF management is on symptom relief through decongestion and control of risk factors only, which underscores the need for more effective treatments [5]. While international guidelines recommend the use of SGLT-2 inhibitors in HFpEF with less enthusiasm than in HFrEF, recent landmark trials (i.e. EMPEROR-Preserved and DELIVER) have provided compelling evidence for the use of both empagliflozin and dapagliflozin in reducing the composite outcome of cardiovascular death or first hospitalization for heart failure [10,11]. However, the results of the DELIVER trial, when stratified by race, suggest that dapagliflozin (hazard ratio [HR], 0.92; 95% confidence interval [CI], 0.69–1.22) may not be as effective as empagliflozin (HR, 0.65; 95% CI, 0.46–0.92) in Asian patients with HFpEF [10,11]. This raises concerns about the comparative efficacy of these two agents in this population. Therefore, this study aims to compare and address the disparities in the cardiovascular outcomes of dapagliflozin and empagliflozin in a real-world setting, with a focus on Asian patients with HFpEF.

Data Source

Data were obtained from the Clinical Data Analysis and Reporting System (CDARS) and the Electronic Patient Record system. CDARS is a clinical database managed by the Hong Kong Hospital Authority, which is a statutory body that provides healthcare services to over seven million people in Hong Kong [12]. It has been used in pharmacoepidemiology studies to provide territory-wide data with high positive predictive values. The database captures more than 90% of Hong Kong population, which serves as an excellent data source for population-based observational studies [13-16]. The Electronic Patient Record system will be used in adjunct with CDARS to analyse patients’ medical profiles.

Study Site

This retrospective study was conducted at Princess Margaret Hospital, an acute care center located in the Kowloon West district of Hong Kong. The hospital serves as a territory-wide tertiary referral center for specialized medical services. Princess Margaret Hospital accounts for 8.5% of the total medical specialist out-patient clinic services and in-patient discharges in Hong Kong [17]. The study retrospectively included patients who had received medical care from the Medicine and Geriatric Department at the hospital.

Inclusion and Exclusion Criteria

The study population comprised adult patients who were newly prescribed dapagliflozin or empagliflozin from August 1st, 2016 to March 31st, 2022 during their hospital stay or outpatient visit. These patients had a prior diagnosis of comorbid HF and T2DM, as identified by the International Classification of Diseases, Ninth Revision (ICD-9) codes 428 and 250, respectively [18,19]. Only individuals aged 18 years or older at the time of their first prescription of SGLT-2 inhibitors and with a LVEF greater than 40% obtained three months after the initiation of SGLT-2 inhibitors, as determined by echocardiography or invasive angiography, were included in the study. Patients were excluded if they (1) had a life expectancy of less than 1 year as deemed by the investigator due to the presence of any disease other than HF; (2) lacked an LVEF reading within 3 years before to 3 months after the initiation of SGLT-2 inhibitors; (3) underwent heart transplant during the follow-up period; (4) had complex heart disease, such as complex congenital heart disease, infiltrative cardiomyopathy, active myocarditis, constrictive pericarditis, or cardiac tamponade; or (5) had used other SGLT-2 inhibitors or combined SGLT-1 and 2 inhibitors different from the one prescribed by the Hospital Authority within four weeks before or during the initiation of SGLT-2 inhibitor (Fig. 1).

Figure 1. Patient selection process.

Primary, Secondary and Safety Outcomes

The primary outcome is a composite of cardiovascular mortality and hospitalization due to HF. All-cause mortality was also investigated as secondary outcomes. The incidence of adverse events, including amputation, diabetic ketoacidosis, major hypoglycaemic events, and urinary tract infection, were also reported.

HF hospitalization is defined as: (1) the patient is admitted to any public hospital with a primary diagnosis of HF; (2) the length of stay is equal to or greater than 24 hours or spans across two calendar days; and (3) the patient receives at least one of the following HF-specific treatments: (i) initiation of maintenance loop diuretic therapy, or (ii) initiation of combination diuretic therapy, or (iii) initiation of intravenous diuretic or vasoactive agent (such as inotrope, vasopressor, vasodilator).

Ethics Approval

This study is approved by the Hong Kong Hospital Authority Kowloon West Cluster Research Ethics Committee (Application number: KW/EX-23-006 [180]).

Statistical Tests and Software

Continuous data were reported in mean±standard deviation while categorial data were presented in number and percentage. To evaluate the composite outcome of hospitalization due to HF and cardiovascular mortality, as well as all-cause mortality, a Cox proportional hazards model was utilized. Patients were followed from the initiation of SGLT-2 inhibitor treatment until the occurrence of the outcome, death, or the end of the study period (April 17th, 2023), whichever came first. The results were reported as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). Descriptive statistics were employed to summarize the number and percentage of patients in each treatment group who experienced adverse events of interest. Furthermore, subgroup analysis was conducted to determine differences in outcomes by stratifying across age, sex, history of hospitalization for heart failure in previous 12 months, baseline LVEF, and creatinine clearance (<30 mL/min vs. ≥30 mL/min). The study employed IBM SPSS Statistics version 26 for data manipulation and analysis.

Baseline Characteristics

From August 1st, 2016 to March 31st, 2022, a total of 546 patients were screened for eligibility. Following the application of inclusion and exclusion criteria, 85 patients with comorbid HFpEF and T2DM were identified. The demographic data of all included patients were stratified by multiple covariates and are presented in Table 1. Most of the patients were ethnic Chinese (except 1 Pakistani). Patient characteristics were similar between the two treatment groups.

Table 1 . Characteristics of the patients at baseline^†.

	Empagliflozin	Dapagliflozin
All patients	59	26
Age as of SGLT2 inhibitor started– yr (mean)	71.4	69.3
Female sex – no. (%)	28 (47.5%)	14 (53.8%)
Race – no. (%)
Chinese	59 (100%)	25 (96.2%)
Pakistanis	0 (0%)	1 (3.8%)
Left ventricular ejection fraction
Measured by Echocardiogram	52 (88.1%)	18 (69.2%)
Measured by invasive angiogram	7 (11.9%)	8 (30.8%)
Mean left ventricular ejection fraction – %	55.9±10.8	56.9±11.3
Left ventricular ejection fraction ≥40% to <50% – no. (%)	23 (39.0%)	10 (38.5%)
Left ventricular ejection fraction ≥50% to <60% – no. (%)	16 (27.1%)	5 (19.2%)
Left ventricular ejection fraction ≥60% – no. (%)	20 (33.9%)	11 (42.3%)
Cardiovascular history – no. (%)
Hospitalization for heart failure during previous 12 months	18 (30.5%)	10 (38.5%)
Atrial fibrillation	6 (10.2%)	5 (19.2%)
Mean CrCl – mL/min§	40.4±16.7	43.5±23.5
CrCl <30 mL/min – no. (%)§	19 (32.2%)	7 (26.9%)
Baseline medications (%)
Angiotensin-converting enzyme inhibitors	20 (33.9%)	8 (30.8%)
Angiotensin receptor blockers	23 (39.0%)	12 (46.2%)
Angiotensin receptor blocker-neprilysin inhibitor	4 (6.8%)	1 (3.8%)
Beta-blockers	47 (79.7%)	17 (65.4%)
Ivabradine	0 (0%)	2 (7.7%)
Mineralocorticoid receptor antagonists	14 (23.7%)	5 (19.2%)
Thiazide diuretics	2 (3.4%)	1 (3.8%)
Loop diuretics	40 (67.8%)	13 (50.0%)
Hydralazine	7 (11.9%)	0 (0%)
Nitrate	20 (33.9%)	8 (30.8%)
Aspirin	37 (62.7%)	14 (53.8%)
Statin	52 (88.1%)	20 (76.9%)
Calcium channel blockers	23 (39.0%)	8 (30.8%)

^†Percentages may not total 100 due to rounding..

± are means plus or minus standard deviation..

^§The abbreviation CrCl denotes creatinine clearance calculated by the Cockcroft-Gault Equation, assuming the body weight of male is 60 kg and female is 50 kg. Since eGFR is often not available in the Electronic Patient Record system, CrCl is used to measure baseline renal function using the serum creatinine measured in renal function test..

Efficacy

In the study cohort, the primary composite outcome was observed in 25 (39.0%) and 11 (42.3%) patients on empagliflozin and dapagliflozin, respectively. Analysis of patient-year at risk revealed that empagliflozin and dapagliflozin were associated with incidence rates of 24.9 and 29.9 events per 100 patient-years, respectively, with a non-significant hazard ratio of 0.83 (95% CI, 0.40–1.70; p=0.604) (Table 2, Fig. 2). Evaluation of hospitalization rates for HF as a primary diagnosis since initiation of SGLT-2 inhibitors indicated incidence rates of 18.9 and 21.8 events per 100 patient-years for empagliflozin and dapagliflozin, respectively, with no significant difference between the two groups (HR, 0.92; 95% CI, 0.40–2.12; p=0.844). Similarly, no significant intergroup difference was observed for cardiovascular death (HR, 0.46; 95% CI, 0.14–1.5; p=0.199) or all-cause mortality (HR, 0.72; 95% CI, 0.29–1.81; p=0.487).

Table 2 . Efficacy and safety outcomes.

Variable	Empagliflozin (n=59)		Dapagliflozin (n=26)		Hazard ratio or difference (95% CI)	p-value
	Events per 100 patient-year			Events per 100 patient-year
Primary composite outcome — no. (%)	25 (39.0%)	24.9	11 (42.3%)	29.9	0.83 (0.40–1.70)	0.604
Hospitalization for HF	19 (32.2%)	18.9	8 (30.8%)	21.8	0.92 (0.40–2.12)	0.844
Cardiovascular death	6 (10.2%)	4.5	5 (19.2%)	9.9	0.46 (0.14–1.51)	0.199
Death from any cause — no. (%)	13 (22.0%)	9.8	7 (26.9%)	13.8	0.72 (0.29–1.81)	0.487
Safety outcomes — no./total no. (%)‡		-		-	-	-
Urinary tract infection	7 (11.9%)	-	2 (7.7%)	-	-	-
Any amputation	1 (1.7%)	-	0	-	-	-
Any major hypoglycemic event	0	-	1 (3.8%)	-	-	-
Any definite or probable diabetic ketoacidosis	0	-	0	-	-	-

^‡Safety outcome is counted as any event of interest endorsed in the electronic patient record system after the initiation of SGLT-2 inhibitor..

Figure 2. Kaplan Meier curves for primary outcomes against the time since started SGLT-2 inhibitors.

Safety

Empagliflozin and dapagliflozin are associated with adverse events, including urinary tract infections, amputations, and major hypoglycaemic events, as outlined in Table 2. Urinary tract infections were the most frequently reported adverse event, with 7 cases observed in the empagliflozin arm and 2 cases in the dapagliflozin arm. Notably, 1 case of severe hypotension and 1 case of amputation were also documented.

Sub-Group Analysis

The number of patients reaching the primary composite outcome is stratified into different subgroups for comparison. Among patients without a history of hospitalization for heart failure during the previous 12 months, those receiving empagliflozin showed a significantly lower risk of the primary composite outcome compared to those taking dapagliflozin (HR, 0.30; 95% CI, 0.12–0.76). In contrast, among patients with a history of hospitalization for heart failure during the previous 12 months, those receiving dapagliflozin showed a lower risk of the primary composite outcome compared to those taking empagliflozin (HR, 4.55; 95% CI, 1.02–20.3).

Our study has revealed a notably higher incidence rate of cardiovascular and mortality outcomes compared to the rates reported in current trials. The number of events per 100 patient-years in our study was nearly triple that observed in the landmark trials (Table 2) [10,11]. These findings stand in contrast to the epidemiology of HF in Hong Kong, where the prevalence and 1-year mortality for HF are much lower than in most other countries worldwide [20]. Several factors may contribute to this observation. Firstly, the detection of primary composite outcomes in large prospective trials is often more stringent [10,11]. In contrast, our retrospective study, due to the limited availability of data in the electronic database, employed a slightly more lenient definition. Secondly, the accessibility of medical services differs between geographical locations, and Hong Kong’s efficient public hospital system allows timely detection and admission of patients to the hospital for treatment.

Empagliflozin demonstrated superiority over dapagliflozin in patients without a history of HF-caused hospitalization within the past 12 months (HR, 0.30; 95% CI, 0.12–0.76). Conversely, our study found that empagliflozin was inferior to dapagliflozin in patients with a recent history of HF hospitalization within the past 12 months (HR, 4.55; 95% CI, 1.02–20.3) (Table 3). These findings are consistent with a retrospective Korean study, which showed that dapagliflozin produced significantly better outcomes compared to empagliflozin for patients with a history of heart failure across the spectrum (HR, 0.80; 95% CI, 0.67–0.96) [21].

Table 3 . Sub-group analysis for empagliflozin and dapagliflozin in reaching the primary composite outcome.

Subgroup	Empagliflozin No. of patients with events/total no.	Dapagliflozin No. of patients with events/total no.	Hazard ratio (95% CI)
All Patients	25/59	11/26	0.83 (0.40–1.70)
Age
<70	28/59	9/26	0.96 (0.31–2.98)
≥70 yr	31/59	17/26	1.43 (0.55–3.70)
Gender
Female	28/59	14/26	0.69 (0.25–1.94)
Male	31/59	12/26	0.94 (0.33–2.67)
LVEF at enrolment
≥40% to <50%	23/59	10/26	0.77 (0.26–2.26)
≥50% to <60%	16/59	5/26	0.70 (0.13–3.84)
≥60%	20/59	11/26	0.92 (0.27–3.16)
Hospitalization for heart failure during previous 12 months
Yes	18/59	10/26	4.55 (1.02–20.3)
No	41/59	16/26	0.30 (0.12–0.76)
CrCl at baseline§
<30 mL/min	19/59	7/26	3.44 (0.43–27.53)
≥30 mL/min	40/59	19/26	0.56 (0.25–1.26)

^§The abbreviation CrCl denotes creatinine clearance calculated by the Cockcroft-Gault Equation, assuming the body weight of male is 60kg and female is 50 kg..

Several potential mechanisms have been proposed to account for the superiority of dapagliflozin in patients with a history of cardiovascular disease. Firstly, the differential effects of empagliflozin and dapagliflozin on the SGLT-2:SGLT-1 inhibition ratio may play a role. Dapagliflozin exhibits a lower selectivity ratio for SGLT-2:SGLT-1 (1200-fold) compared to empagliflozin (2500-fold) [22]. This distinction is important because SGLT-1 receptors are primarily present in the human intestine. The higher selectivity of SGLT-1 receptors can reduce postprandial blood glucose fluctuations, potentially lowering the risk of heart failure [23]. Secondly, dapagliflozin has a slower kidney excretion, which translates to a more stable and prolonged osmotic diuretic effect compared to empagliflozin [24]. This increased stability leads to a reduction in variability in systolic blood pressure, which is associated with a lower risk of cardiovascular diseases [25]. However, further evidence is necessary to fully understand why dapagliflozin is superior to empagliflozin only in patients with a recent history of HF but not consistently in those without. Additionally, it is also important to acknowledge the modest sample size of this study may affect the precision of the HR measurements.

Our study identified an elevated rate of initial decline in survival and freedom from the primary composite outcome, as well as hospitalization due to HF during the first year of SGLT-2 inhibitor treatment. The steeper slope in the first year, as depicted in our study, is not observed in the DELIVER and EMPEROR-Preserved trials, where the cumulative incidence of these outcomes increases steadily following randomization. In both the EMPEROR-Preserved and DELIVER trials, the cumulative incidence of the primary composite outcome and hospitalization for HF at 1 year after starting SGLT-2 inhibitors was less than 10% [10,11]. However, the cumulative incidence for the primary composite outcome and hospitalization for HF reached around 20% at 1 year for patients using these two drugs in our study. These findings suggested a unique characteristic of local patients with T2DM and HFpEF using SGLT-2 inhibitors, which may be attributed to the prescribing restrictions under the Hong Kong Hospital Authority. These prescribing restrictions may prompt physicians to consider SGLT-2 inhibitors at a later stage of the disease when patients are more likely to experience disease progression and hospitalization [1]. The higher initial incidence highlights the importance of close follow-up during the initiation phase of treatments. Pharmacist-led HF clinics may play a role in these scenarios by timely and accurate protocol-driven dose titration, laboratory monitoring and patient education by pharmacists [26-30]. Further research is warranted to determine the optimal approach for managing these patients and improving outcomes.

This is the first local study that head-to-head compares the two SGLT-2 inhibitors, as well as the first study analysing the cardiovascular outcome among the local population. Also, the use of a territory-wide healthcare database in HK has been recognized to provide high-quality data for pharmacoepidemiology studies. Despite its contributions, this study has several limitations. First, the data used in this study was obtained solely from one public hospitals and may not be generalizable to other populations. Second, this study was conducted in a single centre, where the small sample size limited the significance of the effects detected by this study. Nevertheless, this study can serve as a pioneering pilot study. By expanding the scope of this research to encompass the entire territory and include individuals without diabetes mellitus, the study is poised to attain enhanced overall significance. Lastly, residual confounding cannot be ruled out due to the observational nature of the study.

This retrospective head-to-head comparison study aimed to assess the differences between dapagliflozin and empagliflozin in terms of hospitalization for heart failure, cardiovascular death, and all-cause mortality. The study did not identify any significant differences between the two medications in these outcomes. However, it is noteworthy that the incidence rates of cardiovascular and mortality outcomes observed in this study were higher than those reported in current trials. Subgroup analysis also revealed the history of heart failure in the 12 months preceding the initiation of SGLT-2 inhibitors may impact on which SGLT-2 inhibitors are more efficacious. Additionally, the study observed an elevated rate of initial decline in survival from the primary composite outcome, particularly during the first year following SGLT-2 inhibitor initiation. These findings highlight the need for further research to expand the study’s population across multiple hospitals and establish the optimal clinical management approach for patients with HFpEF.

None.

None.

No potential conflict of interest relevant to this article was reported.