Introduction
Atrial fibrillation (AF) and atrial flutter (AFL) are leading causes of mortality worldwide that frequently result in cerebrovascular events [
1]. It is generally acknowledged that type 2 diabetes is significantly associated with an increased risk of developing AF/AFL [
2,
3]. The underlying mechanisms can be attributed to insulin resistance that engenders myocardial remodelling, or expansion of epicardial adipose tissue which leads to inflammation-related cardiac fibrosis and the change of atrial electrical properties [
4‐
6]. However, whether hypoglycaemic agents alter the risk of AF/AFL is incompletely understood [
3].
Sodium-glucose cotransporter 2 inhibitors (SGLT2i), a novel class of oral hypoglycaemic medication, have been demonstrated to potentially reduce the risk of cardiovascular outcomes, especially heart failure (HF) and all-cause mortality, in several large placebo-controlled randomized controlled trials (RCTs) and meta-analyses [
7‐
10]. The key mechanisms may be explained by the potentially direct myocardial effects of SGLT2 inhibitors on diuresis and myocardial metabolism [
11]. Furthermore, HF and AF/AFL are closely aligned and share risk factors such as diabetes, obesity and hypertension [
12,
13]. Atrial structural and neurohormonal alterations in HF are extremely likely to promote the development and progression of AF/AFL [
12]. Thus, we hypothesize that at the same time, pharmacologic therapies for HF could contribute to the reduced AF/AFL risk. Additionally, SGLT2 inhibitors have been reported to decrease glycated haemoglobin (HbA1c), body weight and blood pressure [
14]. Such findings also imply crucial roles of SGLT2 inhibitors in AF/AFL improvement. Nonetheless, no RCTs to date, other than a post hoc analysis of the DECLARE-TIMI 58 trial [
15] address the relationship between SGLT2 inhibitors and the risk of AF/AFL in the population. Hence, the overarching purpose of the present meta-analysis was to pool data from all placebo-controlled RCTs that evaluated AF/AFL outcomes of SGLT2 inhibitors, from which we gained more reliable assessments of the efficacy and safety of specific results overall and in relevant subgroups.
Methods
Data sources and search strategy, data acquisition, inclusion and exclusion criteria, outcome measurements, quality assessment and statistical methods in the present report were performed in accord with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISRM) guidelines [
16].
Data sources and search strategy
An article search was carried out in January 2020 without restrictions on publication types, regions, sample sizes or languages. The main data sources were PubMed, Embase, and ClinicalTrials.gov. We screened unpublished and published RCTs through MeSH terms and their combinations related to AF with a free text search for SGLT2 inhibitors from inception of the above electronic databases. The detailed search algorithm is presented Additional file
1. References listed in identified studies and function of the related articles were also assessed to broaden the scope of search. When there were multiple RCTs with the same patient cohorts, the most recently published trial was included.
Inclusion and exclusion criteria
A pair of 2 independent reviewers (WJL and XQC) identified the titles, abstracts and full-texts of all citations. Independent reviewers (WJL and XQC, LLX and YQL), again working in pairs, identified the full-text version of qualified references. Any disagreement was resolved by the third senior author (BHL).
We included studies if: they were RCTs published in English, enrolled patients aged 18 years or older who had type 2 diabetes mellitus, were performed in out- or inpatient-settings, compared SGLT2 inhibitors with placebo, and reported outcomes of interest. We excluded those focused on patients with type 1 diabetes mellitus or malignant tumours. Letters to the editor, editorials, case reports, review articles, and animal model literature were eliminated.
Outcomes of interest
The pre-set overarching outcome of interest was the incidence of AF/AFL (the composite of new-onset and recurrent AF/AFL). We also examined all-cause mortality, HF, cerebrovascular events, and myocardial infarction as the primary outcomes. The secondary outcomes were other safety endpoints, including urinary tract infections, and the effects of SGLT2 inhibitors on changes in HbA1c, body weight loss, systolic blood pressure (SBP) and diastolic blood pressure (DBP). Cerebrovascular events were defined as the combination of cerebral haemorrhage and ischaemic stroke.
Prespecified subgroup analyses were conducted to compare the SGLT2 inhibitor treatment effect between different agent types, the proportion of subjects by sex or duration of treatments. We predefined high doses of 10 mg for dapagliflozin, 300 mg for canagliflozin, and 25 mg for empagliflozin and low doses of 2.5 mg or 5 mg for dapagliflozin, 100 mg for canagliflozin, and 10 mg for empagliflozin. Subgroups of interest were also present in AF/AFL, including age (< 60 vs. ≥ 60 years), body weight (< 90 vs. ≥ 90 kg), HbA1c (< 8.1 vs. ≥ 8.1%), and SBP (< 135 vs. ≥ 135 mmHg).
Data extraction and quality assessment
Using a standardized form, 2 independent reviewers (WJL and XQC) manually extracted information from the included studies as follows: (1) study design, patient characteristics, follow-up durations; (2) comparisons, interventions, background intervention, outcomes at different time points or with different agent types; and (3) related items for outcomes of interest. We also tried to contact authors of the trials screened in our search by email in order to obtain additional data where necessary. Any missing data were found in ClinicalTrials.gov for RCTs.
The Cochrane risk-of-bias tool was applied to assess the methodological quality of the RCTs [
17]. Two reviewers (WJL and XQC) independently assessed the risk of bias of the included studies at the study, intervention and outcome levels. Risk-of-bias assessments with disagreement were reanalysed and discussed until a consensus was reached.
Grading the strength of evidence
We evaluated the applicability of the analysis outcomes via the Agency for Healthcare Research and Quality (AHRQ) criteria by using the Strength of Evidence (SOE). Two reviewers (WJL and LLX) independently graded the SOE for each of the outcomes of interest as low, moderate or high. Any conflict was resolved by consulting a third reviewer (BHL).
Statistical analysis
All meta-analyses were performed using RevMan 5.3 and Stata 14.0. We used pooled relative risks (RRs) with corresponding 95% confidence intervals (CIs) for the incidence of AF/AFL and predefined safety endpoints in patients with type 2 diabetes mellitus who received standard treatment with or without SGLT2 inhibitors. Weighed mean differences (WMDs) with standard deviation (SD) were applied for continuous variables.
To assess the extent to which the outcomes of the included studies are consistent. Heterogeneity was assessed by using the Cochrane Q test and Higgins and Thompsons’ I
2 before the meta-analysis. For the Q test, we determined that a threshold
p value < 0.1 was statistically significant. Additionally, heterogeneity was deemed to be low if I
2 was < 50%; otherwise, it was high if I
2 was > 50% [
17]. If high heterogeneity between studies was found, we used the random-effects (RE) model; otherwise, the fixed-effects (FE) model was applied. We further investigated heterogeneity among studies by conducting subgroup analyses, meta-regression and sensitivity analyses. Interaction terms were used to evaluate whether the occurrence of AF/AFL would change with different factors across subgroups throughout Revman 5.3 software. A p-value < 0.05 was set as significant factors that were associated with the occurrence of AF/AFL. Publication bias was investigated by the use of funnel plots and Egger’s test. P-values < 0.05 (two-sided) were considered statistically significant and we did not adjust for multiple testing.
To investigate heterogeneity, univariable meta-regression was carried out if more than 10 trials were included in the meta-analysis. We considered a p-value less than 0.05 to be statistically significant. The Monte Carlo permutation test (5001 permutations, p < 0.1) was subsequently used to calculate p values and decrease the false-positive/negative findings in the meta-regression. The following covariates were investigated: sample size; follow-up of trials, and proportion of females (%). A P value less than 0.1 was set as the criterion for heterogeneity source. Finally, sensitivity analyses were performed to assess the robustness of the outcomes by removing each included study individually to explore the remaining overall estimates of AF/AFL events.
Discussion
To the best of our knowledge, this is the first attempt to systematically evaluate the relationship between SGLT2 inhibitors and AF/AFL in type 2 diabetes, and the totality of the present findings emphasises several patterns. First, SGLT2 inhibitors demonstrated great benefits in reducing the relative risk of AF/AFL in type 2 diabetes. This effect was consistent regardless of age, HbA1c, blood pressure and body weight and was more obvious with a long duration of SGLT2 inhibitor treatment. Second, for the specific results the clinical effects of SGLT2 inhibitors depended on the patient population to which they are applied. The reduction in all-cause mortality was apparent in patients treated with SGLT2 inhibitors, especially in the younger population. Third, the use of SGLT2 inhibitors did not induce an extra risk of cerebrovascular events and myocardial infarction. This is especially remarkable because no individual trial has evaluated these findings, and it was revealed only after analysing the results. Largely, there were no obvious heterogeneity in the above analyses from the results of heterogeneity tests, suggesting the reliability of the present outcomes.
The present meta-analysis outlined that SGLT2 inhibitors had great benefit in reducing the risk of AF/AFL. Notably, the reduction in AF/AFL was strongly supported by the recent large-scale trials, NCT01730534 [
26] and NCT01032629 [
19], whose results were different from those of other early trials. The inconsistent results may be derived from the small sample sizes and short follow-up periods of previous studies, which might limit the ability to identify AF/AFL-reduction effects and lead to false-negative findings. Despite extensive exploratory studies, the mechanisms of action of SGLT2 inhibitors in AF/AFL treatment remain unclear. The present incorporated data demonstrated favourable effects on reducing HbA1c, blood pressure and body weight in patients randomized to SGLT2 inhibitors, which might be of importance in attenuating AF/AFL progression. On the one hand, increased glucose excretion essentially leads to additional osmotic diuresis, which in turn will cause a reduction in arterial blood pressure and retard myocardial structural remodelling followed by atrial fibrosis [
32]. On the other hand, glucose lost in the urine with diuresis may contribute to weight loss and be maintained for a long time, thus reducing atrial dilation and the occurrence of AF [
33‐
35]. Many AF/AFL risk factors are potentially reversible [
36]. In addition, our further analysis found that AF/AFL reductions were consistent across subgroups including age, HbA1c, blood pressure and body weight. Therefore, natriuresis followed by reduction in blood pressure and body weight induced by SGLT2 inhibitors may confer a specific AF/AFL-reduction benefit in this susceptible population.
Likewise, atrial fibrosis has central roles in pathogenic remodelling in HF [
37]. We confirmed that HF events were significantly decreased in patients randomized to SGLT2 inhibitors. Reductions in both HF and AF/AFL and their concomitant interventions and downstream complications might subsequently reduce the risk of all-cause mortality in our present work [
3,
38]. In particular, we found that this effect could be enhanced with therapy of a longer duration. The underlying cause may be the negative correlation between the duration of therapy with SGLT2 inhibitors and HbA1c levels [
39]. Our further analysis found that all-cause mortality reductions slightly differed between age groups and appeared to be more distinct in younger patients. This finding may be attributed to vulnerability to comorbidities and competing risks from other diseases in the elderly population [
40].
Numerous investigations in diabetic animal models have revealed that SGLT2 inhibitors directly target the amelioration of cardiac fibrosis. Shao et al. [
41] found that glycemic control with SGLT2 inhibitors notably mitigated atrial remodelling and cardiac fibrosis through the improvement of mitochondrial function. Habibi et al. [
42] also discovered that the SGLT2 inhibitor empagliflozin could improve pro-fibrosis signalling and related interstitial fibrosis. Of note, the pharmacological effects on ameliorating cardiac fibrosis caused by AF/AFL appear to be different from those in HF. In our meta regression, no clinical correlation between AF/AFL and HF was observed. This finding implied that the AF/AFL-reduction effects of SGLT2 inhibitors may be partly independent of HF improvement. Similarly, the analysis of the DECLARE-TIMI 58 trial suggested that dapagliflozin could lower the AF/AFL risk in type 2 diabetes patients irrespective of history of HF [
15]. More large-scare investigations are required to corroborate this finding.
Overall, SGLT2 inhibitors are well tolerated and generally safe agents. Because of the higher levels of urine glucose caused by the glycosidic effect of SGLT2 inhibitors [
43], relatively high prevalence of urinary tract infections was noted in our analysis. Of note, in the present study the effect of dapagliflozin on urinary tract infections was more manifest than that of empagliflozin or canagliflozin. The same conclusion was also drawn in the meta-analysis by Liu et al. [
44], who elucidated that dapagliflozin alone is related to a significantly higher risk for urinary tract infections. However, such infections are easy to manage and rarely recur [
45]. Additionally, initial concerns on safety signals for cerebrovascular risk were not supported in the present analysis [
46]. The drug effect on myocardial infarction is still a topic of meaningful investigation [
14].
The data presented herein indicated that SGLT2 inhibitors should be considered in populations with type 2 diabetes for AF/AFL prevention, given that they safely reduced HbA1c, body weight, and blood pressure and widely reduced the risk of HF across the spectrum of these patients. These risk factors did not modify AF/AFL reductions. Moreover, such effects appeared to be more evident with longer durations of SGLT2 inhibitor therapy. Reductions in all-cause mortality could also be expected, which slightly differed in magnitude based on baseline age characteristics, but were present throughout the age range. Patients with diabetes are a particularly susceptible population at increased risk of AF/AFL and HF [
3,
47]. A considerable body of large-scale placebo-controlled trials in populations with type 2 diabetes are needed to clarify whether SGLT2 inhibitors exhibit beneficial effects in reducing AF/AFL.
Although this meta-analysis provides the first evidence for a favourable effect of SGLT2 inhibitors on reducing AF/AFL risk, several limitations in our work should be emphasized. First, several small-sized studies did not report AF outcomes. However, the results of our meta-analysis are expected to be statistically stable and robust based on the large sample size. Second, the data that we used lacks information at the individual level, and we failed to identify new and recurrent AF/AFL. Even though we have identified the specific AF/AFL-reduction benefit of SGLT2 inhibitors, further investigations are required to explore the role of SGLT2 inhibitors in both new and recurrent AF/AFL. Third, each patient in the included studies was given background hypoglycaemic therapy, which may influence cardiovascular outcomes to some extent. The results of our work showed that patients with type 2 diabetes had reductions in AF/AFL incidence from the addition of SGLT2 inhibitors to guideline-directed medical therapy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.