Comparison of Two Diabetes Drugs and Their Link to New Atrial Fibrillation, Stroke, and Death Risks in Hong Kong Patients

This study obtained ethics approval from the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster and The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee. This was a retrospective, territory-wide cohort study of type 2 diabetes mellitus patients with SGLT2I/DPP4I use between January 1, 2015, and December 31, 2020, in Hong Kong. Patients with any SGLT2I/DPP4I use during the aforementioned period were enrolled and followed up until December 31, 2020, or until death. Patients with both DPP4I and SGLT2I use, with less than 1 month of SGLT2I or DPP4I exposure or with prior AF diagnosis were excluded. The patients were identified from the Clinical Data Analysis and Reporting System (CDARS), a citywide database that centralizes patient information from individual local hospitals to establish comprehensive medical data, including clinical characteristics, disease diagnosis, laboratory results and drug treatment details. The system has been previously used by both our team and other teams in Hong Kong on epidemiological research [18, 19], including those on diabetes [20].

Clinical and biochemical data were extracted for the present study. Patients’ demographics include gender and age of initial SGLT2I/DPP4I use. Prior comorbidities were extracted based on the International Classification of Diseases Ninth Revision (ICD-9) codes (Supplementary Table). The Charlson comorbidity index was also calculated. Mortality was recorded using the International Classification of Diseases Tenth Revision (ICD-10) coding. ICD-10 codes I00–I09, I11, I13 and I20–I51 were used to identify cardiovascular mortality outcomes. Medication histories were also extracted, including the use of metformin, sulfonylurea, insulin, acarbose, thiazolidinedione, glucagon-like peptide-1 receptor agonists, glimepiride, glibenclamide, gliclazide, glipizide, anticoagulants, statins and fibrates. Baseline laboratory data, including complete blood count, biochemical tests, glycemic and lipid profiles were extracted. 1

The study outcomes were new-onset ischemic stroke/transient ischemic attack, new-onset AF, all-cause and cardiovascular mortality. Mortality data were obtained from the Hong Kong Death Registry, a population-based official government registry with the registered death records of all Hong Kong citizens linked to CDARS. Descriptive statistics are used to summarize baseline clinical and biochemical characteristics of patients with SGLT2I and DPP4I use. For baseline clinical characteristics, the continuous variables were presented as median (95% confidence interval [CI]/interquartile range [IQR]) or mean (standard deviation [SD]) and the categorical variables were presented as total number (percentage). Continuous variables were compared using the two-tailed Mann-Whitney U test, whilst the two-tailed Chi-square test with Yates’ correction was used to test 2 × 2 contingency data.

Propensity score matching with 1:1 ratio between SGLT2I and DPP4I users based on demographics, CHA2DS2-VASc score, Charlson comorbidity index, prior comorbidities, use of different medication classes (including other antidiabetic drugs), and baseline hemoglobin A1c (HbA1c) and fasting glucose tests were performed using the nearest neighbor search strategy with the caliper as 0.1. For the missing baseline covariates upon admission, multiple imputation by chained equations (MICE) was employed in the treatment and control groups. Each missing value of laboratory data was imputed 20 times using other variables that might have an impact on the study outcomes. Univariable and multivariable Cox regression models were used to identify significant risk predictors for the study outcomes. Competing risk analysis models (cause-specific and sub-distribution) were considered. A standardized mean difference (SMD) of less than 0.2 between the treatment groups post-weighting was considered an adequate balance. The hazard ratio (HR), 95% CI and P value were reported. Cumulative incidence curves were used to illustrate the difference in the time-to-adverse event in the SGLT2I and DPP4I groups visually. Statistical significance is defined as a P value <0.05. Statistical analyses were performed with RStudio software (version: 1.1.456) and Python (version: 3.6). Stata software (version 13.0) was used for propensity score matching.

This study cohort included 61,233 patients with type 2 diabetes mellitus (55.37% males, median age: 62.7 years [interquartile range (IQR): 54.56–71.92 years]). Of these, 21,713 were SGLT2I users (35.46%) and 39,510 were DPP4I users (64.54%). The breakdown of the number of patients on individual SGLT2Is was as follows: 12,824 (29.53%) on dapagliflozin, 4600 (10.59%) on empagliflozin, 4994 (11.50%) on canagliflozin and 2510 (5.77%) on ertugliflozin. Over a median follow-up duration of 2030 (IQR: 1912–2117) days, 2496 patients (IR: 4.07%) developed new-onset AF, 2179 patients (IR: 3.55%) developed stroke/transient ischemic attack, and 6607 patients (IR: 10.79%) died from any cause, among which 1963 (IR: 3.20%) died from cardiovascular diseases (Fig. 1). The cumulative incidence of the adverse outcomes compared between SGLT2I and DPP4I users after propensity score matching is summarized by cumulative incidence curves in Fig. 2.

The baseline demographics and clinical characteristics of DPP4I and SGLT2I users before and after 1:1 propensity score matching are shown in Table 1. The medications and laboratory tests at baseline are detailed in Supplementary Table 2. The distributions of density as a function of the propensity score before and after matching using a caliper of 0.1 are shown in Supplementary Fig. 1. Both before and after matching, SGLT2I users showed lower incidence of new-onset AF (before: 1.96% vs. 5.23%, SMD = 0.18; after: 1.96% vs. 2.78%, SMD = 0.05), stroke (before: 1.80% vs. 4.52%, SMD = 0.16; after: 1.80% vs. 3.52%, SMD = 0.11), all-cause mortality (before: 2.59% vs. 15.29%, SMD = 0.46; after: 2.59% vs. 7.47%, SMD = 0.22) and cardiovascular mortality (before: 0.47% vs. 4.70%, SMD = 0.27; after: 0.47% vs. 1.56%, SMD = 0.11) compared to DPP4I users.

The detailed results of univariable Cox regression in the unmatched and matched cohorts are shown in Supplementary Table 3 (AF and stroke/transient ischemic attack) and Supplementary Table 4 (cardiovascular and all-cause mortality). Compared to DPP4I users, SGLT2I users showed lower risks of new-onset AF (HR: 0.68, 95% CI: [0.56, 0.83], P = 0.0001), stroke (HR: 0.64, 95% CI: [0.53, 0.79], P < 0.0001), all-cause mortality (HR: 0.44, 95% CI: [0.37, 0.51], P < 0.0001) and cardiovascular mortality (HR: 0.39, 95% CI: [0.27, 0.56], P < 0.0001) after adjusting for significant demographics, past comorbidities, medications and biomarkers (Table 2). These findings were confirmed by competing risk analyses using cause-specific and sub-distribution hazard models (Table 3). Sensitivity analyses using a one-year lag time (Supplementary Table 5) and approaches based on the propensity score (Supplementary Table 6) were performed. All of these analyses demonstrated the association between SGLT2I use and lower risks of incident AF, stroke/transient ischemic attack, cardiovascular mortality, and all-cause mortality compared to DPP4I use.

The number of patients on individual SGLT2Is on study outcomes in the matched cohort is summarized in Table 4. Due to changes in medication switching, some patients were exposed to more than one SGLT2I. Exposure to both empagliflozin and canagliflozin (n = 1596) was the most common, followed by canagliflozin and ertugliflozin (n = 646), dapagliflozin and empagliflozin (n = 575), dapagliflozin and canagliflozin (n = 252), dapagliflozin and ertugliflozin (n = 140), and dapagliflozin, canagliflozin and ertugliflozin (n = 3).

Several limitations should be noted for the present study. First of all, given its observational nature, there is inherent information bias due to under-coding, coding errors and missing data. Also, patients’ drug compliance can only be assessed indirectly through prescription refills, which are ultimately not a direct measurement of drug exposure. Secondly, residual and post-baseline confounding may be present despite robust propensity-matching, particularly with the unavailability of information on lifestyle cardiovascular risk factors. Patients’ drug exposure duration has not been controlled, which may affect their risk against the study outcomes. Moreover, data on blood glucose and HbA1c were largely missing, and thus the glycemic control in the patients cannot be adequately assessed. Additionally, the occurrence of AF out of the hospital is not accounted for, though non-sustained AF that does not result in hospital admission can be considered subclinical.

This study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster and from The Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee.

The authors consent to the publication of manuscript.

None.

Deidentified data are available from the corresponding authors.