What this is
- This study compares the risks of stroke and () between two diabetes medications: sodium-glucose cotransporter 2 inhibitors (SGLT2i) and glucagon-like peptide-1 receptor agonists (GLP-1RA).
- Using a population-based cohort of Chinese patients with type 2 diabetes, the study evaluates stroke subtypes and incident over a median follow-up of 17 months.
- Findings indicate that GLP-1RA is associated with a lower risk of , while SGLT2i users have a lower risk of incident .
Essence
- GLP-1RA use is linked to a lower risk of compared to SGLT2i, which is associated with a lower risk of incident . No significant difference in risk was found.
Key takeaways
- GLP-1RA users experienced a lower risk of compared to SGLT2i users, with a hazard ratio of 1.53 (95% CI 1.01–2.33, p=0.044). This finding suggests that GLP-1RA may be more beneficial for patients at risk of .
- SGLT2i users had a significantly lower risk of incident (HR 0.43, 95% CI 0.23–0.79, p=0.006) compared to GLP-1RA users. This indicates a potential advantage of SGLT2i in reducing risk.
- No significant difference was observed in the risk of between the two groups, with a hazard ratio of 1.29 (95% CI 0.53–3.14, p=0.582) for SGLT2i vs. GLP-1RA.
Caveats
- The study's follow-up duration of 17 months may limit the number of observed events, particularly in subgroup analyses. Longer follow-up could provide more robust data.
- The cohort consisted only of Chinese patients, which may affect the generalizability of the findings to other populations.
- Potential confounding factors, such as lifestyle choices and drug adherence, could not be fully accounted for in this observational study.
Definitions
- Atrial Fibrillation (AF): An irregular and often rapid heart rate that can lead to blood clots, stroke, and other heart-related complications.
- Ischemic Stroke: A type of stroke that occurs when a blood vessel supplying blood to the brain is obstructed.
- Hemorrhagic Stroke: A type of stroke caused by bleeding into the brain.
AI simplified
Background
Both sodium-glucose cotransporter 2 inhibitors (SGLT2i) and glucagon-like peptide-1 receptor agonists (GLP-1RA) are newer classes of anti-diabetic agents which have shown cardiorenal benefits in landmark trials [1]. These agents have expanded the armamentarium in the management of type 2 diabetes [2], providing benefits beyond glycemic control. Head-to-head comparisons between SGLT2i and GLP-1RA may provide further insights in the selection of appropriate anti-diabetic agents, currently limited in the literature especially regarding stroke risk. A recent network meta-analysis of cardiovascular outcome trials involving SGLT2i and GLP-1RA revealed that GLP-1RA statistically significantly reduced risk of non-fatal stroke by 16% compared with placebo, while GLP-1RA had a 10% lower risk of non-fatal stroke compared with SGLT2i but not reaching statistical significance [3]. Nonetheless, this network meta-analysis represents an indirect comparison of the cardiovascular outcome trials which requires cautions in data interpretation.
Few studies have evaluated the risks of other subtypes of stroke among SGLT2i and GLP-1RA users. Two meta-analyses of large randomized controlled trials of SGLT2i suggested that SGLT2i use was associated with reduced risk of hemorrhagic stroke but neutral on ischemic stroke compared with placebo [4, 5]. Regarding GLP-1RA, exploratory analysis of REWIND trial suggested that dulaglutide was associated with lower risk of ischemic stroke but neutral on hemorrhagic stroke compared with placebo [6]. Furthermore, SGLT2i was shown to be associated with lower risk of incident atrial fibrillation (AF) [7], but not for GLP-1RA [8]. Whether this may imply differential benefits on the risk of cardioembolic stroke between SGLT2i and GLP1-RA users remains to be elucidated.
We carried out this real-world population-based study of Chinese patients with type 2 diabetes who were started on SGLT2i or GLP-1RA to compare their risks of stroke with its subtypes and incident AF.
Methods
Study design and data source
This retrospective population-wide cohort study utilized electronic medical records from the Hong Kong Hospital Authority database. The Hospital Authority is the statutory body that provides public healthcare services in outpatient clinics and all public hospitals in Hong Kong, covering 90% of all secondary and tertiary care in the territory [9]. The service is available to all Hong Kong residents (> 7.3 million) [10]. Electronic medical records of patients include demographics, registered death records, laboratory tests, disease diagnoses, procedures, medication dispensing records, hospital admissions, outpatient and emergency department visits. Data from this database have been commonly used in diabetes-related research [10, 11]. Patients with type 2 diabetes managed in the Hospital Authority public health clinics receive regular diabetic complication screening. During each diabetic complication screening session, patients are assessed clinically and have laboratory investigations to determine their control of diabetes, its related cardiovascular risk factors and the presence of diabetic complications [10, 12]. Comorbidities are also recorded by clinicians as diagnosis codes of International Classification of Primary Care, second edition (ICPC-2) or the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) when patients attend any clinic visits or are hospitalized. The use of medications refers to the dispensing records under the Hospital Authority when patients attend any clinic visits or are hospitalized.
Subject's inclusion and exclusion criteria
The diagnosis of type 2 diabetes was defined by the ICPC-2 codes 'T90', the ICD-9-CM codes '250.x0' and '250.x2', or dispensing record of oral anti-diabetes medications.
The exposure of interest was the use of SGLT2i versus GLP-1RA. Patients with at least one dispensing record of SGLT2i or GLP-1RA were included. They were classified into either SGLT2i or GLP-1RA group. Patients who had ever received SGLT2i and GLP-1RA were assigned to the group of the novel anti-diabetic medication that were initiated earlier. In this study, active-comparator new-user design was adopted, and hence, no placebo group was included. The index date of each patient was set at the date of SGLT2i or GLP-1RA initiation. Patients who were aged < 18 years, had end-stage kidney disease (estimated glomerular filtration rate [eGFR] < 15 mL/min/1.73m2, dialysis or kidney transplantation) on or before the index date, or had co-initiation of SGLT2i and GLP-1RA at index date were excluded from the study.
After applying the inclusion and exclusion criteria, a total of 38,543 eligible patients were identified and assigned to SGLT2i (n = 35,546) and GLP-1RA (n = 2997) groups.
Flowchart of subject inclusion. The index date was defined as the first date of receiving SGLT2i or GLP-1RA. SGLT2i, sodium–glucose cotransporter-2 inhibitors; GLP-1RA, glucagon-like peptide-1 receptor agonist
Outcome measures
The primary outcome was all events of stroke, consisting of hemorrhagic and ischemic stroke. Secondary outcomes included hemorrhagic stroke, ischemic stroke, incident AF, and cardioembolic stroke.
Hemorrhagic, ischemic stroke, and AF were identified by ICD-9-CM codes '430' to '432.x', '433.x' to '438.x', and '427.31', respectively. Cardioembolic stroke was identified among ischemic stroke, with the help of the TOAST cardioembolic features described in the study by Guan et al. using electronic health database [13]. For patients who had history of stroke before the index date, incident stroke was identified if the relevant diagnosis codes were listed as the primary diagnosis of the hospitalization records.
Definitions of variables
Variables for confounding adjustment included demographics, clinical parameters, comorbidities, and use of medications in the six months preceding the index date. These were included in view of their clinical relevance. Clinical parameters included hemoglobin A1c (HbA1c), fasting glucose, systolic and diastolic blood pressure, low-density lipoprotein cholesterol, total cholesterol to high-density lipoprotein cholesterol ratio, triglyceride, body mass index (BMI), eGFR (≥ 60 or < 60 mL/min/1.73m2), albuminuria (normal if urine albumin creatinine ratio [ACR] < 3.4 mg/mmol Cr; microalbuminuria if urine ACR ≥ 3.4 to < 34 mg/mmol Cr; and macroalbuminuria if urine ACR ≥ 34 mg/mmol Cr), and duration of diabetes. Comorbidities comprised congestive heart failure, hemorrhagic and ischemic stroke, chronic obstructive pulmonary disease, liver disease, severe hypoglycemia, AF, vascular disease (including coronary heart disease and peripheral vascular disease), diabetic retinopathy and diabetic neuropathy. Medication use included anti-diabetic, anti-hypertensive, and lipid-lowering agents, antiplatelets, and anticoagulants.
Complete data on demographics, comorbidity status, and prescription records of medications were extracted, while the completion rate of laboratory parameters ranged from 62.9 to 99.2%. The data completion rates of baseline covariates between treatment groups are presented in Additional file: Table S1. 1
Statistical analyses
Regarding baseline characteristics of the patients, continuous variables were reported as mean with standard deviation (SD) while categorical variables were reported as number with percentage. Multiple imputation by chained equation was used to impute missing baseline covariates [14, 15]. Each missing datum was imputed using known or imputed variables by five times. Five complete datasets were generated and analyzed independently, and the analyzed results were pooled to an overall estimate by using the Rubin's rule [16]. One-to-one propensity score matching was performed to compare patients who had initiated SGLT2i or GLP-1RA at clinical equipoise. A logistic regression model was fitted to estimate the propensity score of patients by using the baseline covariates. Patients in SGLT2i and GLP-1RA groups were pairwise matched using the propensity score with a caliper of 0.001. An absolute standardized mean difference (ASMD) ≤ 0.1 of baseline covariate indicated the covariate balance between the two groups [17].
Each included patient had complete follow-up and was observed from the index date until any of the following: (i) the occurrence of events; (ii) death; (iii) addition or switching of treatments to another study exposures (e.g., patients who were initiated on SGLT2i first and switched to GLP-1RA were censored at the start date of GLP-1RA, and vice versa); or (iv) the date of the end of study (i.e., 31st December 2020), whichever was earlier. Cox proportion hazard regression models were constructed to estimate hazard ratio (HR) and 95% confidence interval (CI) of each outcome between the two groups. In addition, incidence rate ratio (IRR) for each outcome was estimated by Poisson regression models.
Subgroup analyses were performed by stratification according to sex (male and female), age group (< 60 and ≥ 60 years), HbA1c (< 8 and ≥ 8%), history of stroke, history of AF, and duration of diabetes (< 10 and ≥ 10 years). Tests for interaction between the novel anti-diabetic medications and subgroups were conducted.
Sensitivity analyses were conducted to check whether consistent results were obtained with different analytic approaches: (i) analysis without censoring on switching treatments; (ii) inclusion of patients with follow-up duration ≥ 1 year; (iii) inclusion of patients with at least two dispensing records within 12 months; (iv) inclusion of only patients who were initiated on SGLT2i or GLP-1RA in or after 2015; (v) 'as-treated' analysis, censoring when patients stopped being dispensed the index treatment; (vi) analysis using regression adjustment to control for confounders instead of propensity score matching; and (vii) analysis without adjustment of covariates using either propensity score matching or regression adjustment.
All statistical analyses were performed using Stata version 14.0 (StataCorp, College Station, Texas). p values < 0.05 were considered statistically significant. The current study was reported according to the guideline of STrengthening the Reporting of OBservational studies in Epidemiology (STROBE).
Results
| Factors, mean (standard deviation) or % (n) | All patients (N = 5840) | SGLT2i (N = 2920) | GLP-1RA (N = 2920) | ASMD |
|---|---|---|---|---|
| Sex | 0.01 | |||
| Male | 56.1% (3274) | 55.8% (1630) | 56.3% (1644) | |
| Female | 43.9% (2566) | 44.2% (1290) | 43.7% (1276) | |
| Age, years | 55.5 (12.5) | 55.4 (12.4) | 55.6 (12.6) | 0.017 |
| Clinical and laboratory parameters | ||||
| Hemoglobin A1c, % | 8.9 (1.7) | 8.9 (1.7) | 8.9 (1.7) | 0.029 |
| Fasting glucose, mmol/L | 9.5 (3.6) | 9.5 (3.6) | 9.5 (3.6) | 0.01 |
| Systolic blood pressure, mmHg | 137.4 (23.0) | 137.1 (22.1) | 137.7 (22.1) | 0.028 |
| Diastolic blood pressure, mmHg | 79.0 (14.5) | 78.9 (14.9) | 79.2 (14.1) | 0.017 |
| Low-density lipoprotein cholesterol, mmol/L | 2.3 (0.8) | 2.3 (0.8) | 2.2 (0.8) | 0.009 |
| Total cholesterol to high-density lipoprotein cholesterol ratio | 4.0 (1.3) | 4.0 (1.4) | 4.0 (1.2) | 0.019 |
| Triglyceride, mmol/L | 2.0 (1.4) | 2.0 (1.5) | 2.0 (1.4) | 0.046 |
| Body mass index, kg/m2 | 30.2 (6.6) | 29.9 (6.7) | 30.4 (6.5) | 0.087 |
| Estimated glomerular filtration rate | 0.02 | |||
| ≥ 60 mL/min/1.73m2 | 72.2% (4217) | 72.6% (2121) | 71.8% (2095) | |
| < 60 mL/min/1.73m2 | 27.8% (1623) | 27.4% (799) | 28.2% (825) | |
| Status of microalbuminuria/macroalbuminuria | 0.016 | |||
| Normal | 44.4% (2591) | 44.2% (1290) | 44.6% (1,301) | |
| Microalbuminuria | 34.6% (2019) | 35.0% (1021) | 34.2% (999) | |
| Macroalbuminuria | 21.1% (1229) | 20.9% (609) | 21.2% (620) | |
| Duration of diabetes, year | 13.7 (8.8) | 13.7 (8.7) | 13.8 (9.0) | 0.01 |
| Comorbidities | ||||
| Atrial fibrillation | 2.2% (131) | 2.1% (62) | 2.4% (69) | 0.016 |
| Congestive heart failure | 4.4% (259) | 4.3% (127) | 4.5% (132) | 0.008 |
| Stroke | 6.0% (351) | 5.7% (166) | 6.3% (185) | 0.027 |
| Hemorrhagic stroke | 1.0% (58) | 0.9% (25) | 1.1% (33) | 0.028 |
| Ischemic stroke | 5.4% (317) | 5.2% (151) | 5.7% (166) | 0.023 |
| Chronic obstructive pulmonary disease | 3.3% (191) | 3.4% (100) | 3.1% (91) | 0.017 |
| Liver disease | 0.9% (52) | 0.9% (27) | 0.9% (25) | 0.007 |
| Severe hypoglycemia | 4.9% (288) | 4.9% (143) | 5.0% (145) | 0.003 |
| Vascular disease | 13.4% (783) | 13.2% (384) | 13.7% (399) | 0.015 |
| Diabetic retinopathy | 15.2% (890) | 15.3% (448) | 15.1% (442) | 0.006 |
| Diabetic neuropathy | 3.7% (216) | 3.8% (110) | 3.6% (106) | 0.007 |
| Baseline medications | ||||
| Anti-diabetic medications | ||||
| Insulin | 71.8% (4192) | 73.0% (2131) | 70.6% (2061) | 0.053 |
| Metformin | 84.2% (4918) | 84.7% (2474) | 83.7% (2444) | 0.028 |
| Sulfonylurea | 43.9% (2564) | 43.5% (1271) | 44.3% (1293) | 0.015 |
| Thiazolidinedione | 23.5% (1372) | 24.0% (701) | 23.0% (671) | 0.024 |
| Dipeptidyl peptidase 4 inhibitors | 37.7% (2201) | 37.8% (1103) | 37.6% (1098) | 0.004 |
| Alpha-glucosidase inhibitors | 1.7% (98) | 1.4% (42) | 1.9% (56) | 0.037 |
| Anti-hypertensive medications | ||||
| Angiotensin-converting enzyme inhibitors/angiotensin receptor blockers | 74.9% (4372) | 74.5% (2174) | 75.3% (2198) | 0.019 |
| Beta blockers | 39.3% (2297) | 39.0% (1139) | 39.7% (1158) | 0.013 |
| Calcium channel blockers | 57.3% (3348) | 57.4% (1675) | 57.3% (1673) | 0.001 |
| Diuretics | 21.1% (1233) | 20.9% (610) | 21.3% (623) | 0.011 |
| Other anti-hypertensive medications | 10.1% (588) | 9.8% (286) | 10.3% (302) | 0.018 |
| Lipid-lowering agents | 77.9% (4550) | 77.7% (2270) | 78.1% (2280) | 0.008 |
| Antiplatelets | 28.0% (1636) | 27.4% (801) | 28.6% (835) | 0.026 |
| Anticoagulants | 1.6% (93) | 1.5% (43) | 1.7% (50) | 0.019 |
| Type of SGLT2i | ||||
| Canagliflozin | NA | 0.3% (9) | NA | NA |
| Dapagliflozin | 36.6% (1068) | |||
| Empagliflozin | 63.3% (1847) | |||
| Type of GLP-1RA | ||||
| Dulaglutide | NA | NA | 30.5% (892) | NA |
| Exenatide | 18.9% (553) | |||
| Liraglutide | 46.5% (1359) | |||
| Lixisenatide | 4.0% (117) | |||
Primary outcomes
| Number of case (%) | Incidence rate per 100 person-years (95% CI) | Person-years | Median follow-up month (IQR) | Incidence rate ratio (95% CI) | P-value | Hazard ratio (95% CI) | P-value | |
|---|---|---|---|---|---|---|---|---|
| All stroke | ||||||||
| All patients | 111 (1.9%) | 0.94 (0.78,1.14) | 11,771 | 17 (8–34) | ||||
| SGLT2i | 62 (2.1%) | 1.10 (0.86,1.41) | 5621 | 19 (8–35) | 1.38 (0.95,2.01) | 0.089 | 1.46 (0.99,2.17) | 0.058 |
| GLP-1RA | 49 (1.7%) | 0.80 (0.60,1.05) | 6150 | 17 (8–34) | 1 (Reference) | 1 (Reference) | ||
| Hemorrhagic stroke | ||||||||
| All patients | 23 (0.4%) | 0.19 (0.13,0.29) | 11,918 | 17 (8–34) | ||||
| SGLT2i | 11 (0.4%) | 0.19 (0.11,0.35) | 5683 | 19 (9–35) | 1.01 (0.44,2.28) | 0.989 | 1.29 (0.53,3.14) | 0.582 |
| GLP-1RA | 12 (0.4%) | 0.19 (0.11,0.34) | 6236 | 16 (7–32) | 1 (Reference) | 1 (Reference) | ||
| Ischemic stroke | ||||||||
| All patients | 98 (1.7%) | 0.83 (0.68,1.01) | 11,780 | 17 (8–34) | ||||
| SGLT2i | 57 (2.0%) | 1.01 (0.78,1.31) | 5628 | 19 (8–35) | 1.52 (1.02,2.27) | 0.041* | 1.53 (1.01,2.33) | 0.044* |
| GLP-1RA | 41 (1.4%) | 0.67 (0.49,0.91) | 6152 | 16 (7–32) | 1 (Reference) | 1 (Reference) | ||
| Cardioembolic stroke | ||||||||
| All patients | 25 (0.4%) | 0.21 (0.14,0.31) | 11,906 | 17 (8–34) | ||||
| SGLT2i | 14 (0.5%) | 0.25 (0.15,0.42) | 5676 | 19 (9–35) | 1.40 (0.63,3.08) | 0.407 | 1.40 (0.62,3.17) | 0.414 |
| GLP-1RA | 11 (0.4%) | 0.18 (0.10,0.32) | 6230 | 16 (7–32) | 1 (Reference) | 1 (Reference) | ||
| Incident atrial fibrillation | ||||||||
| All patients | 50 (0.9%) | 0.43 (0.33,0.57) | 11,648 | 17 (8–34) | ||||
| SGLT2i | 15 (0.5%) | 0.27 (0.16,0.45) | 5574 | 19 (9–35) | 0.47 (0.26,0.86) | 0.014* | 0.43 (0.23,0.79) | 0.006* |
| GLP-1RA | 35 (1.2%) | 0.58 (0.41,0.80) | 6074 | 16 (7–32) | 1 (Reference) | 1 (Reference) | ||
Secondary outcomes
When analyzed according to subtypes of stroke, we observed a statistically significantly higher risk of ischemic stroke among SGLT2i users compared with GLP-1RA users (HR 1.53, 95% CI 1.01–2.33, p = 0.044; IRR 1.52, 95% CI 1.02–2.27, p = 0.041), whereas the risk of hemorrhagic stroke (HR 1.29, 95% CI 0.53–3.14, p = 0.582; IRR 1.01, 95% CI 0.44–2.28, p = 0.989) and cardioembolic stroke (HR 1.40, 95% CI 0.62–3.17, p = 0.414; IRR 1.40, 95% CI 0.63–3.08, p = 0.407) was comparable between groups. The incidence rate of AF in the SGLT2i group and GLP-1RA were 0.27 and 0.58 cases/100 person-years respectively. The risk of incident AF was statistically significantly lower among SGLT2i users (HR 0.43, 95% CI 0.23–0.79, p = 0.006; IRR 0.47, 95% CI 0.26–0.86, p = 0.014).
Subgroup analyses
We performed subgroup analyses on the primary outcome and secondary outcome of ischemic stroke, with results shown in Additional file: Tables S2 and S3 respectively. Regarding all stroke, no significant interaction was observed between subgroups by age, sex, glycemic control and history of AF. However, patients with history of stroke appeared to benefit more from GLP-1RA use (p for interaction = 0.043) than those without history of stroke. Moreover, patients with shorter duration of diabetes (< 10 years) also appeared to benefit more from GLP-1RA use than those with longer duration of diabetes (p for interaction = 0.017). When focusing on ischemic stroke, we observed no interaction across subgroups by age, sex, glycemic control, history of stroke and history of AF. Again, patients with shorter duration of diabetes (< 10 years) appeared to benefit more from GLP-1RA than those with longer duration of diabetes (p for interaction = 0.009). 1
Sensitivity analyses
Main results were consistent to those obtained from analyses using different analytic approaches, namely: (i) analysis without censoring on switching treatments; (ii) inclusion of patients with follow-up duration ≥ 1 year; (iii) inclusion of patients with at least two dispensing records within 12 months; (iv) inclusion of only patients who were initiated on SGLT2i or GLP-1RA in or after 2015; (v) 'as-treated' analysis, censoring when patients stopped being dispensed the index treatment; (vi) analysis using regression adjustment to control for confounders instead of propensity score matching; and (vii) analysis without adjustment of covariates using either propensity score matching or regression adjustments (Additional file: Table S4). 1
Discussion
Our real-world population-based propensity-score matched cohort study of Chinese patients with type 2 diabetes was among the few head-to-head comparisons dedicated to risks of stroke between SGLT2i and GLP-1RA use. Upon a median follow-up of 17 months, there was no significant difference in the risk of all stroke between SGLT2i and GLP-1RA users. Regarding the subtype of stroke, GLP-1RA use was associated with lower risk of ischemic stroke compared with SGLT2i use. On the other hand, there was no significant difference in the risk of hemorrhagic stroke between SGLT2i and GLP-1RA. SGLT2i use was associated with lower risk of incident AF compared with GLP-1RA use. Our results suggested that, among the newer classes of anti-diabetic agents, GLP-1RA may be preferred for patients with type 2 diabetes at risk of ischemic stroke.
Head-to-head comparison of the risk of stroke between SGLT2i and GLP-1RA users was performed only in a few real-world studies. A recent summary of real-world studies suggested that SGLT2i and GLP-1RA initiators were at a similar risk of stroke (HR 1.01, 95% CI 0.93–1.10) [18]. However, these had a relatively short follow-up time of < 1 year. Patorno et al. [19] evaluated using United States claims database and showed no significant difference in risk of stroke (composite of ischemic and hemorrhagic) between the two groups of users. There was also no heterogeneity when analyzed according to history of cardiovascular diseases. The same group also reported comparable risks of stroke between SGLT2i and GLP-1RA users among older individuals with type 2 diabetes [20]. An earlier study of a smaller scale using US claims database showed similar risks of ischemic stroke between the two groups [21]. DeRemer et al. reported that addition of SGLT2i or GLP-1RA to metformin resulted in similar risks of stroke (subtype not specified) [22]. Studies from Danish nationwide database [23] and Italian electronic health database [24] also demonstrated similar stroke risk (subtype not specified) respectively. Nonetheless, important laboratory results regarding glycemic control and renal function were missing in some of the datasets, which may modify the comparative risk of stroke. Lugner et al. evaluated using Swedish national database with more complete laboratory parameters showed that SGLT2i use might be inferior to GLP-1RA regarding risks of stroke (composite of ischemic and hemorrhagic stroke) (HR 1.44, 95% CI 0.99–2.08, p = 0.056) [25]. Lately, two real-world studies with longer follow-up time of around 1.5 years demonstrated higher risk of ischemic stroke with SGLT2i use compared with GLP-1RA use in the Swedish database (HR for SGLT2i use vs GLP-1RA = 1.71, 95% CI 1.14–2.59) [26] and the Scandinavian nationwide database (HR 1.14, 95% CI 1.03–1.26) [27]. With a median follow-up of 17 months, we showed that there might be benefit in all stroke with GLP-1RA use compared with SGLT2i in the context of secondary prevention as revealed in the subgroup analysis where GLP-1RA use was associated with statistically significant lower risk of all stroke compared with SGLT2i use in patients with history of stroke. Regarding the specific subtype of stroke, GLP-1RA use showed superiority over SGLT2i in terms of ischemic stroke. Interestingly, subgroup analysis revealed that the superiority with GLP-1RA use in stroke risk (all stroke and ischemic stroke) was more prominent in patients with shorter duration of diabetes, which may suggest the benefit in early initiation of GLP-1RA to derive the maximal protection from the risk of stroke. Taken together, our results may support the preferential use of GLP-1RA in patients with type 2 diabetes and history of stroke.
In contrast to ischemic stroke, there was less extensive information on hemorrhagic stroke risk for the novel anti-diabetic agents. Some evidence showed the benefits of SGLT2i on hemorrhagic stroke in comparison to placebo. In two meta-analyses [4, 5], SGLT2i use was associated with a 50% reduction in the risk of hemorrhagic stroke, postulated to be related blood pressure lowering effect of SGLT2i, which has a more prominent benefit in hemorrhagic stroke [28]. With regard to the potential impact of GLP-1RA on hemorrhagic stroke, pre-clinical studies showed the potential role of GLP-1/GLP-1R axis in reducing hemorrhagic transformation after ischemic stroke [29]. However, post-hoc analyses of randomized controlled trials of semaglutide [30] and dulaglutide [6] did not demonstrate benefit with GLP-1RA use in hemorrhagic stroke compared with placebo. Our current head-to-head real-world study showed that SGLT2i and GLP-1RA were comparable in terms of the risk of hemorrhagic stroke.
We observed that SGLT2i use was associated with a lower risk of incident AF compared with GLP-1RA use, consistent with the existing literature [31]. SGLT2i use was shown to be associated with lower risk of incident AF compared with placebo [7]. In contrast, results from both meta-analyses of randomized controlled trials and observational studies showed no significant benefits with GLP-1RA use on reducing incident AF compared with placebo [8]. As AF is closely linked to cardioembolic stroke [32], whether this benefit in the reduction of incident AF translates into reduction of cardioembolic stroke has been evaluated as well. A meta-analysis of 6 trials reporting cardioembolic stroke as an outcome showed that SGLT2i use was associated with a lower risk of cardioembolic stroke compared with placebo [33]. Among GLP-1RA, this has been evaluated in the post-hoc analysis of randomized controlled trials of semaglutide, revealing only 11 events of cardioembolic stroke among 6,480 participants, not affected by semaglutide use [30]. Our current study did not reveal difference in the events of cardioembolic stroke between SGLT2i and GLP-1RA users upon a median follow-up of 17 months, which might be due to the relatively small number of events. Longer follow-up with larger cohorts may be necessary to clarify the comparative risk of cardioembolic stroke between SGLT2i and GLP-1RA users.
Our results should be interpreted bearing certain limitations. First, the duration of follow-up was relatively short, such that the number of events was relatively small for subgroup analyses. Second, all subjects in our cohort were Chinese, and our study did not cover all agents within SGLT2i and GLP-1RA classes (predominantly empagliflozin and dapagliflozin in the SGLT2i group, and liraglutide and dulaglutide in the GLP-1RA group). These limit the generalizability of results. Third, like all large-scale pharmacovigilance studies using electronic medical record databases, drug adherence could not be ascertained. Fourth, lifestyle factors related to cardiovascular diseases, such as dietary habits and physical activity, were not available in the database. Fifth, the current study design could not directly examine the statistical relationship between the incidence of stroke and the incidence of AF [32]. Last but not least, despite our attempts in balancing a range of patient characteristics by propensity score matching, in common with all epidemiological studies, retrospective database analysis cannot exclude residual confounders and infer causation.
Conclusions
This population-based propensity-score matched observational cohort study of Chinese patients with type 2 diabetes demonstrated that GLP-1RA use was associated with a lower risk of ischemic stroke compared with SGLT2i use, whereas SGLT2i use was associated with a lower risk of incident AF compared with GLP-1RA use. On the other hand, there was no significant difference observed in the risk of hemorrhagic stroke between SGLT2i and GLP-1RA use. Our results suggested that GLP-1RA may be the preferred agent over SGLT2i for patients with type 2 diabetes at risk of ischemic stroke.
Supplementary Information
Additional file 1: Table S1. Data complete rate of baseline covariates. Table S2. Subgroup analysis of risk of all stroke between SGLT2i and GLP-1RA groups. Table S3. Subgroup analysis of risk of ischemic stroke between SGLT2i and GLP-1RA groups. Table S4. Sensitivity analyses of stroke risk between SGLT2i and GLP-1RA groups.