Comparative Effectiveness of Sodium-Glucose Cotransporter 2 Inhibitors vs the Other Hypoglycemic Drugs on Major Adverse Cardiovascular Events and Hospitalization for Heart Failure in Patients with Type 2 Diabetes Mellitus and Atrial Fibrillation

Background: In the treatment of type 2 diabetes mellitus (T2DM) in atrial �brillation (AF) patients, evidence of the comparative effectiveness of sodium-glucose cotransporter 2 (SGLT2) inhibitors vs the other hypoglycemic drugs on cardiovascular morbidity and mortality is lacking. This study aimed to evaluate the comparative effectiveness of SGLT2 inhibitors vs the other hypoglycemic drugs on major adverse cardiovascular events and heart failure for hospitalization in patients with T2DM and AF. Methods: A cohort study used data from the Korean National Health Insurance Service database compared the use of SGLT2 inhibitors vs the other hypoglycemic drugs for treatment of T2DM in AF patients. We identi�ed 40,268 patients with T2DM and AF who were newly prescribed oral hypoglycemic drugs (2,977 individualst with SGLT2 inhibitors and 37,291 individuals without SGLT2 inhibitors) between 2014 and 2018. After 1:4 propensity score matching, patients who received SGLT2 inhibitors (n=2,958) and those who did not receive SGLT2 inhibitors (n=10,691) were enrolled, and followed up until December 31, 2018. The primary study outcome was major adverse cardiovascular events (MACE), a composite of cardiovascular mortality, myocardial infarction, or ischemic stroke. Secondary outcoms were hospitalization for heart failure and all-cause mortality. The risks of study outcomes were assessed using a Cox proportional hazard model. Results: Mean age of all participants was 59.7 ± 11.9 years and 8.023 [58%] were male. During a mean follow-up duration of 2.1 ± 1.4 years, SGLT2 inhibitors were not signi�cantly associated with reduced risk of MACE(hazard ratio [HR], 0.96; 95% con�dence interval [CI], 0.76-1.21) compared with the other hypoglycemic drugs. However, patients who received SGLT2 inhibitors had signi�cantly lower risks of hospitalization for heart failure (HR, 0.70; 95% CI, 0.53-0.93) and all-cause mortality (HR, 0.74; 95% CI, 0.


Background
Sodium-glucose cotransporter 2 (SGLT2) inhibitors have been recently developed as oral agents for lowering glucose.Unlike insulin secretagogues and insulin sensitizers, SGLT2 inhibitors are independent of pancreatic ß cells because they block glucose and sodium reabsorption in the proximal tubule of the kidney and thereby lower glucose in patients with type 2 diabetes mellitus (T2DM).[1] SGLT2 inhibitors have shown varying degrees of protective effects on cardiovascular mortality and morbidity in T2DM patients with high cardiovascular risks.[2][3][4][5] Empagli ozin and canagli ozin were shown to decrease the incidence of cardiovascular events including cardiovascular death, myocardial infarction, and stroke.[2,3] Dapagli ozin did not reduce the risk of major adverse cardiovascular events (MACE) including cardiovascular mortality, myocardial infarction, or stroke, but resulted in a lower rate of cardiovascular death or hospitalization for heart failure.[4] Ertugli ozin was not superior to placebo with respect to cardiovascular mortality or morbidity.[5] In terms of hospitalization for heart failure, all SGLT2 inhibitors have shown protective effects in patients with T2DM.In a meta-analysis of three randomized trials including empagli ozin, canagli ozin, and dapagli ozin, SGLT2 inhibitors showed robust bene ts in reducing hospitalization for heart failure regardless of existing atherosclerotic cardiovascular disease or a history of heart failure.[6,7] Atrial brillation (AF) is the most common cardiac arrhythmia in adults.[8]AF is associated with increased risks of stroke, heart failure, and death, thus imposing a signi cant burden on patients, societal health, and health economy.[9] Population-based studies have shown that AF was associated with 1.5-and 2.2-fold increases in the risk of mortality in men and women, respectively.[10,11] Whereas the mechanistic explanation for the increases in the mortality risk in AF patients is multifaceted, associated comorbidities, especially heart failure, play an important role.[12][13][14] Aside from reducing the risk of hospitalization for heart failure,[6, 7] SGLT2 inhibitors also decrease the incidence of AF in T2DM patients.[15,16] However, there is a lack of evidence about the effectiveness of SGLT2 inhibitors on MACE or hospitalization for heart failure in patients with T2DM and AF.Therefore, we aimed to evaluate the comparative effectiveness of SGLT2 inhibitors vs the other hypoglycemic drugs on MACE and hospitalization for heart failure in patients with T2DM and AF by analyzing a nationwide population-based cohort.

Data sources
This study used the National Health Information Database produced by the National Health Insurance Service (NHIS) of South Korea.The authors declare no con ict of interest with NHIS.The NHIS provides mandatory health insurance for all South Korean citizens, covering 97% of the national population; [17] the remaining 3% of the population with low income is covered by the Medical Aid program, which has been incorporated into the single NHIS database.The NHIS claims database includes data on demographic characteristics, diagnoses, prescriptions, and deaths.The details of the database are described elsewhere.
[18] All diagnoses are recorded in the NHIS database using the International Classi cation of Diseases, Tenth Revision (ICD-10) codes.The authors are restricted from sharing the data used for this study because the Korean NHIS owns the data.This study was approved by the Institutional Review Board of Asan Medical Center (2019-0917).An anonymized dataset was provided by the NHIS for this study and the requirement for informed consent was waived.

Study population
Initially, we identi ed patients with T2DM (ICD-10 codes E11-E14) [19] and AF (ICD-10 codes I48) who were prescribed hypoglycemic drugs between January 2014 and December 2018.We then extracted patients who were newly prescribed hypoglycemic drugs since January 2014 because the Korean government insurance began reimbursing SGLT2 inhibitors as secondary hypoglycemic drugs for T2DM from September 2014.We then identi ed patients who were diagnosed with AF before the prescription of hypoglycemic drugs.Among them, we excluded patients who met the following exclusion criteria: (1) less than 19 years of age; (2) use of hypoglycemic drugs for <90 days; (3) history of end-stage renal disease and/or dialysis; (4) history of cancer; and (5) missing data.Finally, we included a total of 40,268 patients in this analysis.We divided the remaining patients into two groups according to the use of SGLT2 inhibitors.The process of patient selection ow is described in Figure 1.The index date was de ned as the date of the rst prescription of SGLT2 inhibitors or other hypoglycemic drugs from January 2014.

Clinical variables
The patient's baseline comorbidities and risk factors were de ned as the ICD-10 codes assigned within 5 years of cohort entry.Supplementary Table 1 shows the details of the comorbidity de nitions.Charlson's comorbidity index and CHA 2 DS 2 -VASc score were calculated to evaluate the general comorbidity status and thromboembolic risk, respectively.The CHA 2 DS 2 -VASc score was de ned as a composite score of congestive heart failure, hypertension, age ≥ 75 (doubled), diabetes mellitus, prior stroke or transient ischemic accidents (doubled), vascular disease, age 65-74, sex category (female sex).[20] Previous cardiovascular disease was de ned as a composite of heart failure, stable angina, unstable angina, myocardial infarction, ischemic stroke, transient ischemic accidents, or systemic embolism.Concomitant cardiovascular therapies were de ned when cardiovascular medications were used continuously from baseline for at least 30 days alongside the index drug.

Study outcomes
The primary study outcome was de ned as MACE, a composite of cardiovascular mortality, myocardial infarction, or ischemic stroke.The secondary study outcomes were hospitalization for heart failure and allcause mortality.Vital status and the cause of death were retrieved from the mortality records of the National Statistical O ce of Korea.Cardiovascular death was de ned according to the ICD-10 codes I00.X-I99.X or R96, R98, and R99.Events of myocardial infarction (ICD-10 codes I21-I22) or ischemic stroke (ICD-10 codes I63-I64) were identi ed as a primary or secondary diagnosis at admission.Relevant procedural codes were additionally considered to enhance the validity of myocardial infarction or ischemic stroke episodes.[21] Hospitalization for heart failure was de ned as an admission (≥1 day) with a main or subsidiary diagnosis of heart failure (ICD-10 code I50).Detailed de nitions of the study outcomes are described in Supplementary Table 2.The censoring date was the earliest of the following: date of discontinuation of index drug, date of death, date of study outcomes, or the end date of the study period (December 31, 2018).

Statistical analysis
Baseline characteristics are presented as mean ± standard deviation for continuous variables and as frequencies with percentages for categorical variables.The differences between the treatment groups were evaluated using analysis of variance for continuous variables and either the chi-squared or Fisher's exact test for categorical variables.To reduce the effect of selection bias, we performed a propensity score matching analysis for the comparison between the SGLT2 inhibitors group and No SGLT2 inhibitors group.[22] In addition, inverse probability of treatment weighting based on propensity score modeling was used for evaluating robustness of the ndings.[23] We quanti ed as the likelihood of receiving SGLT2 inhibitors in the year before the index date by multiple logistic regression analysis.In the model, the use of SGLT2 inhibitors was used as a dependent variable, and all measured baseline characteristics as listed above, and clinically plausible interaction terms were included in the analysis.Model discrimination was assessed with C statistics (0.755), and model calibration was assessed with Hosmer-Lemeshow statistics (χ 2 =7.633; df=8; P=0.4701).After calculating the predicted probabilities, we matched each patient in the SGLT2 inhibitors group to those in the No SGLT2 inhibitors group at a 1:4 ratio using the Greedy algorithm with calipers of width equal to 0.2 of the standard deviation of the logit of the propensity score.The balance of baseline characteristics between the SGLT2 inhibitors group and the No SGLT2 inhibitors group was evaluated using the standardized mean difference, and a standardized mean difference of <0.1 indicated a negligible difference.
For the study outcome analysis, incidence rates were estimated using the total number of study outcomes during the follow-up period divided by 100 person-years at risk.The cumulative rates of the study outcomes were evaluated using Kaplan-Meier analyses, and P values were estimated by Cox proportional hazard model that accounted for the matched nature.The risks of study outcomes were assessed using a Cox proportional hazard model and are presented as hazard ratios (HRs) P values <0.05 were considered statistically signi cant.R software version 4.0.3(R Foundation for Statistical Computing, Vienna, Austria) and SAS Enterprise Guide software version 7.1 (SAS Institute Inc., Cary, NC, USA) were used for statistical analyses.

Baseline characteristics
A total of 40,268 patients with AF and T2DM who were newly prescribed hypoglycemic drugs with SGLT2 inhibitors (n=2,977) or without SGLT2 inhibitors (n=37,291) were included in the study.Before propensity score matching, patients who received SGLT2 inhibitors were signi cantly younger and had lower CHA 2 DS 2 -VASc scores (Table 1).After 1:4 propensity score matching, a total of 13,649 patients were included in the nal analysis (SGLT2 inhibitors group, n=2,958; No SGLT2 inhibitors group, n=10,691) (Supplementary Figure 1).Of the matched patients as a whole (male, n=8,023 [58.7%]), the mean age was 59.7 ± 11.9 years and the mean CHA 2 DS 2 -VASc score was 3.0 ± 1.3.Baseline characteristics of the patients according to the use of SGLT2 inhibitors after propensity score matching are summarized in Table 1.
Overall, the two matched cohorts were well-balanced (standardized mean difference <0.1).
The incidences of each component of MACE were not signi cantly different between the two groups (Table 2).The overall cumulative incidence curves of MACE and its components did not show signi cant differences between the two groups as well (Figure 2A-D).
The incidence of hospitalization for heart failure was signi cantly lower in the SGLT2 inhibitors group (1.29 per 100 person-years; adjusted HR, 0.70; 95% CI, 0.53 to 0.93; P=0.012) compared with the No SGLT2 inhibitors group (1.67 per 100 person-years) (Table 2).Kaplan-Meier curves also showed that the rate of hospitalization for heart failure was signi cantly lower in the SGLT2 inhibitors group than in the No SGLT2 inhibitors group (P=0.012)(Figure 3A).
The incidence of all-cause mortality was signi cantly lower in the SGLT2 inhibitors group (1.28 per personyears; adjusted HR, 0.74; 95% CI, 0.56 to 0.98; P=0.034)) than in the No SGLT2 inhibitors group (1.90 per person-years) (Table 2).Kaplan-Meier curves also showed that the rate of all-cause mortality was signi cantly lower in the SGLT2 inhibitors group than in the No SGLT2 inhibitors group (P=0.034)(Figure 3B).
In Cox proportional hazard model using the inverse probability of treatment weighting in initial cohort (2,977 in SGLT2 inhibitors vs 37,291 in No SGLT2 inhibitors), use of SGLT2 inhibitors was signi cantly associated with reduced risk of hospitalization for heart failure compared with the other hypoglycemic drugs (Table 3).However, there was no signi cant difference of the incidences of MACE or all-cause mortality between SGLT2 inhibitors and No SGLT2 inhibitors groups.

Subgroup analysis according to baseline characteristics
In a subgroup analysis according to baseline characteristics, the bene t of SGLT2 inhibitors in terms of MACE and cardiovascular mortality was not evident in most subgroups (Supplementary Figure 1 and 2).
For hospitalization for heart failure, there were no signi cant interactions between SGLT2 inhibitors and all subgroups, except in the subgroups strati ed by CHA 2 DS 2 -VASc scores and insulin use (both P for interaction<0.001;Supplementary Figure 3).SGLT2 inhibitors signi cantly reduced the risk of hospitalization for heart failure among male patients, those with previous cardiovascular disease, previous heart failure, previous myocardial infarction, hypertension, CHA 2 DS 2 -VASc scores ≥3, renin-angiotensin receptor blocker use, beta-blocker use, diuretics use, or anticoagulant use, and those without previous ischemic stroke, chronic kidney disease, or insulin use.
The effect of SGLT2 inhibitors in terms of all-cause mortality was shown in Supplementary Figure 4. Use of SGLT2 inhibitors tended to have bene t for all-cause mortality in almost subgroups.

Discussion
To the best of our knowledge, this study is the rst investigation of the effectiveness of SGLT2 inhibitors in real-world patients with T2DM and AF.The main ndings of this study are as follows: (1) patients treated with SGLT2 inhibitors did not have a signi cantly lower risk of MACE and its components compared with those treated without SGLT2 inhibitors; (2) the use of SGLT2 inhibitors was signi cantly associated with a lower risk of hospitalization for heart failure and all-cause mortality than the other hypoglycemic drugs; and (3) the bene ts of SGLT2 inhibitors for hospitalization for heart failure were consistent across high-risk subgroups, including those with previous cardiovascular disease, previous heart failure, previous myocardial infarction, or high CHA 2 DS 2 -VASc scores (≥3).
In randomized cardiovascular outcome trials of SGLTs inhibitors, results regarding cardiovascular mortality and all-cause mortality were different depending on the type of SGLT2 inhibitor used.[2-5]A recent metaanalysis of three randomized trials using empagli ozin, canagli ozin, and dapagli ozin reported that SGLT2 inhibitors reduced the risks of cardiovascular mortality and all-cause mortality speci cally in patients with established atherosclerotic cardiovascular disease.[6]In a real-world cohort study, the use of SGLT2 inhibitors was associated with reduced cardiovascular mortality and all-cause mortality.[24,25] In the present study, SGLT2 inhibitors use was associated with a signi cant reduction in all-cause mortality but not in cardiovascular mortality.Recently, two studies on Korean T2DM patients also reported that new users of SGLT2 inhibitors had a lower risk of all-cause mortality than those who received other hypoglycemic drugs; [19,26] however, these studies did not evaluate the risk of cardiovascular mortality.[19,26] In the present study, we found a discrepancy in the effect of SLGT2 inhibitors for cardiovascular and allcause mortalities.Unfortunately, we cannot explain the exact reason for this discrepancy, but we assume that the following mechanisms may be involved.First, the study patients were younger compared with those included in previous randomized trials on SGLT2 inhibitors[2-5] (59 years vs. 63-64 years) and had fewer cardiovascular therapeutics including antiplatelet agents, angiotensin-converting enzyme inhibitors, angiotensin-receptor blocker, beta-blocker, statin, and diuretics even though the prevalence of cardiovascular disease was not overtly different (53% vs. 40%-99%).Thus, the cardiovascular effect of SGLT2 inhibitors might be diminished in these patients with relatively low cardiovascular risks.Second, the number of cardiovascular deaths in the propensity score-matched cohort was too small to produce a statistically signi cant difference between treatment groups.Moreover, during the propensity score matching process, relatively fewer patients with cardiovascular risks were selected among the SGLT2 inhibitors group (54.5% vs. 51.1%),which might have led to the non-signi cant results for cardiovascular mortality.Finally, the pleiotropic effects of SGLT2 inhibitors on energy metabolism, lipid pro le, uric acid, blood pressure, hematocrit, renin-angiotensin-aldosterone system, in ammation, and nitric oxide as well as hypoglycemia and weight reduction may have contributed to the reduced risk of all-cause mortality compared with other hypoglycemic drugs.[27,28] In the current study, the bene cial effect of SGLT2 inhibitors for hospitalization for heart failure was consistent with those reported in previous randomized trials and large population-based cohort studies.[2-4, 24, 25] [29] The incidences of hospitalization for heart failure were 0.6-0.9 per 100 patient-years in the SGLT2 inhibitor groups in previous randomized trials, [2][3][4][5] while it was 1.3 per 100 patient-years in this study.This difference might be due to the fact that we de ned hospitalization for heart failure with a main or subsidiary diagnosis of heart failure, which led to a higher incidence of hospitalization for heart failure than those in previous randomized trials.When we de ned the hospitalization for heart failure with the main diagnosis of heart failure, its incidence was 0.6 and 1.0 per 100 patient-years (HR, 0.56; 95% CI, 0.38 to 0.84; P=0.004) in the SGLT2 inhibitors group and the No SGLT2 inhibitors group, respectively.AF and heart failure share common risk factors and pathophysiology that lead to the development of the other condition, [30,31]and patients with AF and heart failure have a poor prognosis.[13] Individuals with T2DM showed a 2-to 4-fold increased risk of heart failure compared with those without in observational studies.[32] Therefore, our results support the effectiveness of SGLT2 inhibitors for preventing hospitalization for heart failure in patients with T2DM and AF.The bene cial effect of SGLT2 inhibitors for hospitalization for heart failure was most evident in those treated with dapagli ozin, which is in line with the results of previous studies.[4,33,34] The effect of other SGLT-2 inhibitors was not evident in this study and should be further assessed in larger prospective studies.

Limitations
This study has several limitations and our results should be interpreted with caution.First, because this was a retrospective observational study using the NHIS database, we could not ascertain speci c clinical data such as the duration, glucose control status, or severity of T2DM,the duration, type (paroxysmal, persistent, or permanent), or rhythm status of AF as well as laboratory data such as HbA1c, brain natriuretic peptide and echocardiographic data.Second, there could be selection bias and residual confounding factors, but we tried to overcome these limitations by using propensity score matching patients with SGLT2 inhibitors and without SGLT2 inhibitors.Thirdthe study population was a homogeneous group of Korean ethnicity, so caution is required when generalizing the ndings of this study to other ethnic groups.Fouth, as we de ned cardiovascular outcomes using ICD-10 codes, there is a possibility of outcome misclassi cation.However, a recent study comparing diagnoses from claims databases with medical records showed that the overall accuracy rates were 92.0% for myocardial infarction and 90.5% for ischemic stroke.[35] Thus, we believe that misclassi cation of study outcomes might be low and would not have signi cantly affected the main results.Fifth, we were not allowed to evaluate the speci c prescription or dose of medication including SGLT2 inhibitors and other hypoglycemic drugs because NHIS policy has changed not to provide an information about speci c brand name drug.So, we could not compare the effectiveness of each SGLT2 inhibitors vs other hypoglycemic drugs.And also, we did not have information about use of metformin for reasons mentioned above.However, there would be no signi cant difference of metformin prescription between two study groups because SGLT2 inhibitors or the other hypoglycemic drugs had health insurance coverage as secondary hypoglycemic drugs after metformin for T2DM patients.In addition, we could not incorporate the effect of medication compliance in our analysis.Finally, this study had a relatively short follow-up duration (2.1 ± 1.4 years).Thus, further studies are needed to evaluate the long-term effects of SGLT2 inhibitors in patients with T2DM and AF.Despite these limitations, this study has the strength of being a nationwide study with large sample size.Furthermore, to the best of our knowledge, this is the rst and the largest Asian study to report the cardiovascular effects of SGLT2 inhibitors in real-world patients with T2DM and AF.

Conclusion
In a real-world clinical practice in Asian patients with T2DM and AF, new users of SGLT2 inhibitors had a lower risk of hospitalization for heart failure and all-cause mortality compared with those without SGLT2 inhibitors.

Figure 1 Flow chart of the study patients Figure 2 Kaplan-
Figure 1

Table 2 .
Risk of study outcomes according to sodium-glucose cotransporter 2 inhibitors use in propensity MACE was de ned as a composite of cardiovascular mortality, myocardial infarction, or ischemic stroke.Hazards ratio was estimated by Cox proportional hazard model that accounted for the matched nature.

Table 3 .
Risk of study outcomes according to sodium-glucose cotransporter 2 inhibitors use in initial cohort MACE was de ned as a composite of cardiovascular mortality, myocardial infarction, or ischemic stroke.Hazards ratio was estimated by Cox proportional hazard model using the inverse probability of treatment weighting.