Association between sodium–glucose cotransporter-2 inhibitors and risk of sudden cardiac death or ventricular arrhythmias: a meta-analysis of randomized controlled trials

Abstract

Aims 

Sudden cardiac death (SCD) and ventricular arrhythmias (VAs) are important causes of mortality in patients with type 2 diabetes mellitus (T2DM), heart failure (HF), or chronic kidney disease (CKD). We evaluated the effect of sodium–glucose cotransporter-2 (SGLT2) inhibitors on SCD and VAs in these patients.

Methods and results 

We performed a systematic review and meta-analysis of randomized controlled trials (RCTs) that enrolled patients with T2DM and/or HF and/or CKD comparing SGLT2i and placebo or active control. PubMed and ClinicalTrials.gov were systematically searched until November 2020. A total of 19 RCTs with 55 ,590 participants were included. Sudden cardiac death events were reported in 9 RCTs (48 patients receiving SGLT2i and 57 placebo subjects). There was no significant association between SGLT2i therapy and SCD [risk ratio (RR) 0.74, 95% confidence interval (CI) 0.50–1.08; P = 0.12]. Ventricular arrhythmias were reported in 17 RCTs (126 patients receiving SGLT2i and 134 controls). SGLT2i therapy was not associated with a lower risk of VAs (RR 0.84, 95% CI 0.66–1.06; P = 0.14). Besides the subgroup of low-dosage SGLT2i therapy that demonstrated decreased VAs compared to control (RR 0.45, 95% CI 0.25–0.82; P = 0.009), or to placebo (RR 0.46, 95% CI 0.25–0.85; P = 0.01), further subgroup analysis did not demonstrate any significant differences.

Conclusion 

SGLT2i therapy was not associated with an overall lower risk of SCD or VAs in patients with T2DM and/or HF and/or CKD. However, further research is needed since the number of SCD and VA events were relatively few leading to wide confidence intervals, and the point estimates suggested potential benefits.

Introduction

Sudden cardiac death (SCD), generally defined as unexpected death due to cardiovascular causes occurring within a short period of time after the onset of symptoms, represents a leading cause of mortality in patients with type 2 diabetes mellitus (T2DM),¹ heart failure (HF),² or chronic kidney disease (CKD).³ Sudden cardiac death occurs twice as commonly in patients with T2DM as non-diabetic subjects and accounts for around half of the deaths from cardiovascular causes in this population.¹ Similarly, patients with HF have six to nine times the rate of SCD of the general population, while ∼50% of the deaths in HF are attributable to SCD.² Importantly, most cases of SCD in the aforementioned patient populations are related to ventricular arrhythmias (VAs), such as ventricular tachycardia (VT) or ventricular fibrillation (VF).

Recent clinical trials have reported that sodium–glucose cotransporter-2 (SGLT2) inhibitors have remarkable beneficial cardiovascular effects. In particular, treatment with SGLT2i has been shown to reduce the incidence of cardiovascular death and HF requiring hospitalization in patients with T2DM,^4–6 with the latter benefit extending to patients with prevalent HF with reduced ejection fraction (HFrEF) regardless of the presence or absence of T2DM.^7,8 Of note, experimental and clinical studies have suggested several potential mechanisms mediating SGLT2i beneficial effects, including modulation of adverse ventricular remodelling and fibrosis, amelioration of oxidative stress and inflammation, inhibition of sympathetic nervous system activity, and improvement of renal function and cardiac energy metabolism.⁹ Interestingly, these cardioprotective mechanisms are also implicated in the pathophysiology of SCD and VAs, implying that SGLT2i might reduce their incidence. However, to the best of our knowledge, no randomized controlled trials (RCTs) or meta-analyses have addressed this relationship as a primary endpoint up to date.

We, therefore, conducted a systematic review and meta-analysis of randomized controlled trials examining the association between SGLT2i therapy and incidence of SCD or VAs in patients with T2DM and/or HF and/or CKD. Furthermore, subgroup analyses were conducted to compare the SGLT2i treatment effect between different drug types, study designs, length of follow-up, clinical conditions, and different dosages.

Methods

Search strategy

This meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A systematic literature search of relevant randomized trials was conducted in PubMed and ClinicalTrials.gov until 30 November 2020. We screened published and unpublished RCTs using the following search strategy: (sodium-glucose cotransporter 2 inhibitors OR SGLT2 inhibitors OR SGLT-2 inhibitors OR SGLT 2 inhibitors OR empagliflozin OR dapagliflozin OR canagliflozin OR ertugliflozin OR tofogliflozin OR ipragliflozin OR luseogliflozin OR remogliflozin OR sergliflozin OR sotagliflozin) AND (cardiovascular OR cardiovascular outcomes OR sudden cardiac death OR cardiac sudden death OR arrhythmias OR ventricular arrhythmias). No language restrictions were applied. References listed in the identified studies were carefully checked in order to identify additional studies. In instances of multiple publications originating from the same RCT, the most recently published one was included.

Inclusion and exclusion criteria

Eligible RCTs were supposed to meet the following criteria: (i) patients aged 18 years or older with a diagnosis of T2DM and/or HF and/or CKD; (ii) comparison between SGLT2i and placebo or active control; (iii) reported outcomes of interest. RCTs focused on T1DM or other conditions were excluded. Trial eligibility was confirmed by two independent reviewers (N.Z. and D.S.). Discrepancies between reviewers were resolved by consensus or, if necessary, by a third investigator (T.L.).

Outcomes of interest

Outcomes of interest were reported as serious adverse events according to Medical Dictionary for Drug Regulatory Activities (MedDRA) to reduce potential bias. Primary outcomes were incidence of SCD and VAs. VAs in this meta-analysis included ventricular tachycardia, ventricular fibrillation, ventricular flutter, ventricular extrasystoles, ventricular arrhythmia, and Torsades de Pointes, and the detailed definition of SCD is presented in Supplementary material online, Appendix. Subgroup analyses were conducted to compare the SGLT2i treatment effect between different drug types, study designs, length of follow-up, clinical conditions, and different dosages. Canagliflozin 100 mg and empagliflozin 10 mg once-daily, were considered low dosage, whereas the high dosage was 300 mg and 25 mg, respectively. Subgroup analysis by dosage was not conducted in dapagliflozin group, because all participants were administrated dapagliflozin at a dose of 10 mg once-daily. Only RCTs with consistent dose during the study period were eligible for subgroup analysis by dosage.

Data extraction and quality assessment

Two reviewers (N.Z. and D.S.) independently performed data extraction using a standard form, including the following items: (i) characteristics of studies, sample size, study design, follow-up duration; (ii) baseline characteristics of study population; (iii) interventions, comparisons, and outcomes of interest. Study quality was evaluated using the Cochrane Risk of Bias Tool (performed by N.Z. and D.S.) and disagreements were solved by consensus or adjudication from a third author (T.L.).

Statistical analysis

We used pooled relative risks (RRs) with corresponding 95% confidence intervals (CIs) to present the incidence of SCD or VAs. Heterogeneity was assessed by using the Cochran Q test statistic and Higgins and Thompson I². For the Q test, P value <0.1 was considered statistically significant. Additionally, I² > 50% indicated at least moderate heterogeneity. If significant heterogeneity was found, we used the random-effects (RE) model; otherwise, the fixed-effects (FE) model was applied. The analysis was conducted with Review Manager (RevMan, version 5.3; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).

Results

Characteristics of eligible studies

A total of 2,287 records were identified, of which 19 RCTs were included in the meta-analysis, comprising a total sample size of 55 ,590 participants (30 ,024 patients treated with SGLT2i and 25,566 controls) (Figure 1).^{4–7,10–22} The baseline characteristics of the included studies and patients are summarized in Table 1 and Supplementary material online, Table S1, respectively. Of these, 14 RCTs were placebo-controlled^{4–7,10–17} and 5 were active-controlled (4 with sulfonylureas^18–21 and 1 with semaglutide²²). Except for one unpublished trial (NCT03448419), the other 18 trials have been published. The effects of a particular type of SGLT2 inhibitor were examined in six trials for empagliflozin,^{4,10,16,19,22} six trials for canagliflozin,^{5,11,17,18,21} and seven trials for dapagliflozin.^{6,7,12–15,20} Nine and 17 RCTs reported SCD and VAs, respectively. The mean age of participants was 55.1–69.0 years, and the percentage of males ranged from 45.6% to 76.6%, with a median follow-up between 12 and 296 weeks.

Quality assessment results are presented in Figure 2. Most RCTs were considered at high methodological quality. Although two of them had an open-label design,^21,22 given the well-defined diagnostic criteria and manifestations of the outcomes, we considered them to be at low risk of performance bias. Participants included in the safety analysis set in some RCTs were less than the full analysis set, which yielded a total 22-person difference between baseline population (n = 55, 590) and outcome evaluation population (n = 55, 568). Considering that the gap between the two sets was tiny (0.04% of the total sample size) and unlikely affected the pooled results, a low risk of bias was still considered.

Incidence of sudden cardiac death

There were nine RCTs that involved SCD events,^{4–7,10–13} all of them taking placebo in the control arm. Of the total 105 SCD events, 48 occurred in the SGLT2i group (n = 26, 318), and 57 occurred in the placebo group (n = 22 ,411). After pooling the nine trials there was no significant association between SGLT2i therapies and the incidence of SCD [risk ratio (RR) 0.74, 95% confidence interval (CI) 0.50–1.08; P = 0.12] (Figure 3). In the subgroup analysis by drug types, the results were consistent with the overall effect; empagliflozin (RR 0.71, 95% CI 0.17–2.98; P = 0.64), canagliflozin (RR 0.83, 95% CI 0.34–2.06; P = 0.70), dapagliflozin (RR 0.72, 95% CI 0.46–1.12; P = 0.15); namely none of them showed protective effect on SCD (Supplementary material online, Appendix Figure S1).

When focusing the effects of follow-up period on SCD, neither shorter duration (<2 years; RR 0.63, 95% CI 0.35–1.11; P = 0.11) nor longer duration (>2 years, RR 0.85, 95% CI 0.50–1.43; P = 0.53) showed an effect on SCD incidence (Supplementary material online, Appendix Figure S2). In terms of dosage, both low dosage (RR 0.51, 95% CI 0.18–1.47; P = 0.21) and high dosage of SGLT2i (RR 1.35, 95% CI 0.56–3.27; P = 0.51) appeared to lack a favourable effect on reducing SCD (Supplementary material online, Appendix Figure S3). Given that there were three and two RCTs that limited their population to CKD^10–12 and HFrEF^7,13 settings, respectively, we also conducted subgroup analyses by clinical conditions. It turned out that SGLT2i administration showed no protective effect on SCD in patients with HFrEF (RR 0.65, 95% CI 0.36–1.16; P = 0.15) or CKD (RR 0.60, 95% CI 0.16–2.26; P = 0.45) (Supplementary material online, Appendix Figures S4 and S5). All pooled results above presented low heterogeneity. Visual inspection of the funnel plot suggested little publication bias (Figure 4A).

Incidence of ventricular arrhythmias

Ventricular arrhythmias were reported in 17 RCTs,^{4–7,11,12,14–22} of which 12 compared SGLT2i with placebo,^{4–7,11,12,14–17} 4 with sulfonylureas (3 with glimepiride,^18,19,21 1 with glipizide²⁰), and 1 with semaglutide.²² A total of 126 patients treated with SGLT2i experienced VAs (n = 29, 462), whereas 134 individuals in the control group experienced VAs (n = 25 ,105). The aggregated data indicated that SGLT2i therapy was not associated with a lower risk of VAs compared to control (RR 0.84, 95% CI 0.66–1.06; P = 0.14) (Figure 5); neither VT (RR 0.76, 95% CI 0.56–1.03; P = 0.08) nor VF (RR 0.97, 95% CI 0.57–1.66; P = 0.92) (Supplementary material online, Appendix Figure S6). To identify the specific role of different study designs, subgroup analyses were conducted according to different controls. Both placebo (RR 0.84, 95% CI 0.67–1.07; P = 0.17) and active control (RR 0.73, 95% CI 0.22–2.43; P = 0.61) showed no significant relationship with VAs incidence. Interestingly, when comparing the effect of SGLT2i with sulfonylureas alone, a significant association was still not evident (RR 0.86, 95% CI 0.23–3.22; P = 0.83) (Supplementary material online, Appendix Figure S7).

In subgroup analysis by SGLT2 inhibitor subtype, both empagliflozin (RR 0.53, 95% CI 0.27–1.05; P = 0.07) and dapagliflozin (RR 0.99, 95% CI 0.75–1.31; P = 0.94) showed no difference in reduction of VAs compared to control, while canagliflozin therapy yielded a borderline significance (RR 0.56, 95% CI 0.31–1.01; P = 0.05) compared to control (Supplementary material online, Appendix Figure S8). In order to clarify the effect of canagliflozin on VAs, further subgroup analysis was conducted by comparing its effect with placebo or active control, respectively. However, no difference was identified in both settings, possibly due to the limited number of studies and events. Regarding follow-up duration, we defined three subgroups, shorter duration (<1 year; RR 1.02, 95% CI 0.26–4.06; P = 0.98), median duration (≥1 year and <2 years; RR 0.85, 95% CI 0.61–1.19; P = 0.35), and longer duration (≥2 years; RR 0.82, 95% CI 0.58–1.15; P = 0.24), and none of them affected the incidence of VAs (Supplementary material online, Appendix Figure S9).

When considering the effect yielded by different dosages, low dosage of SGLT2i significantly decreased the incidence of reported VAs (RR 0.45, 95% CI 0.25–0.82; P = 0.009) compared to control (Figure 6). This association remained significant when SGLT2i were compared only with placebo (RR 0.46, 95% CI 0.25–0.85; P = 0.01) (Supplementary material online, Appendix Figure S10). However, when VAs were split into VT and VF, no protective effect was observed in neither arrhythmia. To identify the effect of SGLT2i treatment in HF and CKD patients, we performed corresponding subgroup analyses. However, neither HF (RR 0.83, 95% CI 0.59–1.17; P = 0.29) nor CKD (RR 0.69, 95% CI 0.24–1.93 P = 0.48) showed a significant association (Supplementary material online, Appendix Figures S11 and S12).

All analyses performed above showed low heterogeneity. The results of the funnel plot suggested no significant publication bias for the 17 RCTs (Figure 4B).

Discussion

The present meta-analysis showed that SGLT2i therapy was not associated overall with a lower risk of SCD or VAs in patients with T2DM and/or HF and/or CKD. Apart from the subgroup of low-dosage SGLT2i therapy that exhibited significantly decreased incidence of reported VAs, subgroup analysis did not demonstrate a significant treatment effect in neither of the subgroups of different drug types, study designs, length of follow-up, and different clinical conditions, such as HF or CKD. Of note, when VAs were divided into VT and VF, no protective effect was observed in neither arrhythmia. Additionally, different dosages of SGLT2i therapy appeared to lack of favourable effect on reducing SCD.

Sudden cardiac death in patients with T2DM is partly mediated by the increased presence of coronary heart disease, which facilitates the occurrence of VAs.²³ Additionally, T2DM is associated with microangiopathy, autonomic dysfunction, QTc prolongation, inflammation, oxidative stress, and renal impairment, factors that have the potential to increase the risk of VAs even further.^23,24 Of the subtypes of HF, HFrEF is associated with the highest risk of SCD and VAs, mainly because of the greater extent of adverse ventricular remodelling and fibrosis. SCD in patients with HFrEF most commonly occurs from acute electrical or mechanical failure of the severely remodelled and fibrotic ventricle, manifesting in the first case as VAs or, in the second case, as bradyarrhythmias, asystole, or pulseless electrical activity.²⁵ While some data suggest that acute triggers, such as surges in sympathetic nervous system activity or abrupt electrolyte imbalances, may provoke SCD in patients with HFrEF, many SCD events have no acute precipitant. Indeed, the severely remodelled left ventricle in HFrEF might generate arrhythmias spontaneously and independently.²⁵ As far as HF with preserved ejection fraction (HFpEF) is concerned, the primary mechanism of SCD seems to be myocardial fibrosis that enables re-entry circuits, whereas ischaemia may also contribute to VAs.²⁶ Chronic kidney disease has recently been considered an independent risk factor for SCD, with traditional cardiovascular risk factors being considered to play a lesser role in its occurrence, and CKD-specific risk factors, structural heart disease, and VAs being thought to contribute the most.²⁷

Data from several experimental and human studies have suggested that SGLT2i might exert beneficial effects on SCD and VAs. First and foremost, several studies have demonstrated beneficial effects of SGLT2 inhibition on adverse ventricular remodelling,^28–33 including reports for its prevention and reversal,³² as well as reports for significant favourable effects on ventricular fibrosis^34,35 and hypertrophy.³² Furthermore, SGLT2i have been shown to attenuate inflammation^36–38 and oxidative stress,³⁹ resulting in less cardiac fibrosis and remodelling in T2DM, and less chronic inflammation in HF through the suppression of NLRP3 inflammasome.⁹ In addition to the above mechanisms, a mounting body of data has indicated that SGLT2i might inhibit the sympathetic nervous system either directly or as a result of reduction of renal stress leading to suppression of renal afferent sympathetic activation.⁹ Nevertheless, SGLT2i have been demonstrated to improve renal function which can indirectly improve cardiac function through a reduction in afferent sympathetic nervous system activation, attenuation of inflammation, and amelioration of oxidative stress.⁹ Even further, SGLT2i have been shown to increase circulating ketone body levels which are in turn oxidized by the failing heart resulting in a net increase of adenosine triphosphate production, thereby improving cardiac energy metabolism.^40,41 Finally, SGLT2i have been associated with beneficial effects on reduction of epicardial fat mass, increasing natriuresis/diuresis, prevention of ischaemia/reperfusion injury, reduction of hyperuricaemia, blood pressure lowering, increasing erythropoietin levels, inhibition of the Na⁺/H⁺ exchanger, increasing autophagy and lysosomal degradation, increasing circulating pro-vascular progenitor cells, and improvement of vascular function.⁹

The results of our meta-analysis could indicate that the favourable impact of SGLT2i therapy on the mechanistic pathways of SCD and VAs might be modest or negligible overall and not clinically meaningful. In support of this notion, a recent meta-analysis demonstrated that SGLT2i significantly reduced left ventricular end-diastolic diameter and E/e’ ratio improving diastolic function but did not appear to have an impact on left ventricular end-diastolic volume (LVEDV) and left ventricular ejection fraction (LVEF).⁴² It has been demonstrated that LV volumes (end-diastolic and end-systolic) represent surrogate markers for adverse ventricular remodelling in HF strongly correlating with the impact of a particular drug or device therapy on patient survival.^43,44 Additionally, the odds for neutral or favourable effects in mortality RCTs increase with mean increases in LVEF.⁴⁴ Consequently, it can be speculated that since SGLT2i therapy does not seem to have an impact on LVEDV and LVEF, its effect on adverse ventricular remodelling might be moderate overall and therefore its antiarrhythmic potential on VAs negligible. Another potential explanation of the apparent failure of SGLT2i therapy to decrease SCD may be the fact that although SGLT2i have been shown to reduce the incidence of acute myocardial infarction through a reduction in preload and a resultant decrease in myocardial wall tension and myocardial oxygen demands, there is no evidence that they reduce thrombosis, nor that they improve atherosclerotic plaque stability in a clinically relevant manner.⁴⁵

Even though this meta-analysis did not detect a significant association between SGLT2i therapy and lower overall risk of SCD or VAs, further research is needed before reaching definitive conclusions. Indeed, the number of SCD and VA events in this study were relatively few leading to wide confidence intervals, and the point estimates suggested potential benefits, indicating that these results might have been influenced by low power. Although the differences between relevant groups did not meet the conventional levels of statistical significance, this meta-analysis might have detected a favourable signal towards SCD reduction in the subgroup of patients with HFrEF receiving SGLT2i (P = 0.15). One potential explanation for this might be that the median follow-up time in the trials used in the above-mentioned subgroup analysis, namely DAPA-HF and DEFINE-HF, were only 18.2 months and 13 weeks, respectively, being probably insufficient to determine a potential impact of SGLT2i therapy on the risk of SCD. Furthermore, only four of the included RCTs, namely DAPA-HF, DEFINE-HF, EMPERIAL-Reduced, and CANDLE, included patients with HFrEF in whom the risk of SCD and VAs, as well as the expected benefits, are highest. This under-representation of HFrEF patients together with the short follow-up periods in these trials might have influenced the overall results, preventing the detection of a significant overall association. Even further, as far as the results of our subgroup analysis by dosage are concerned, there is no plausible explanation as to why low-dosage SGLT2i therapy might decrease the incidence of VAs. Therefore, further adequately powered clinical studies with longer follow-up periods are required before reaching definitive conclusions.

Interestingly, recent evidence indicates that SGLT2i therapy significantly decreases incident atrial fibrillation.⁴⁶ However, the long-term effects of SGLT2i on cardiovascular remodelling and on arrhythmia burden have not been well studied.

Limitations

This meta-analysis has several potential limitations. Firstly, it included studies that had multiple types of patients, namely patients with T2DM, HF (including different types of HF such as HFrEF), and CKD, and thus this population heterogeneity may have influenced the combined results. Secondly, the precise estimation of the incidence of SCD is problematic since its definition is diverse and the specific mechanisms by which each death event occurs are often difficult to ascertain. Additionally, the lack of information on the number of patients with implantable cardioverter-defibrillator (ICD) in most of the included studies may have influenced the number of SCD events. Thirdly, the number of SCD and VA events in this study were relatively few leading to wide confidence intervals, and the point estimates suggested potential benefits, indicating that the results might have been influenced by low power. Fourthly, patients with HFrEF in whom the risks of SCD and VAs are the highest were under-represented in this study, a fact that may have prevented the detection of a significant overall association. Fifthly, the median follow-up period of HFrEF patients was short. Sixthly, as far as the results of our subgroup analysis are concerned, there is no plausible explanation as to why low-dosage SGLT2i therapy might decrease the incidence of VAs.

Conclusion

SGLT2i therapy was not associated with an overall lower risk of SCD or VAs in patients with T2DM and/or HF and/or CKD. However, the number of SCD and VA events were relatively few leading to wide confidence intervals, and the point estimates suggested potential benefits. Therefore, further adequately powered clinical studies with longer follow-up periods are required before reaching definitive conclusions.

Supplementary material

Supplementary material is available at Europace online.

Funding

This work was supported by grants from the National Natural Science Foundation of China (grant no. 81970270 to T.L.).

Conflict of interest: G.Y.H.L. has served as consultant for Bayer/Janssen, BMS/Pfizer, Biotronik, Medtronic, Boehringer Ingelheim, Microlife, and Daiichi-Sankyo. Speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Microlife, Roche, and Daiichi-Sankyo. No fees were received personally. Other authors have no disclosures to declare.

Data availability

Data are available from the corresponding author on reasonable request.

References

Author notes