GLP-1RAs for Ischemic Stroke Prevention in Patients With Type 2 Diabetes Without Established Atherosclerotic Cardiovascular Disease

Although stroke mortality has decreased worldwide in the past two decades, the overall global burden of stroke is still increasing (1). In most Asian countries, stroke is a major cause of mortality (2) and is the third leading cause of death in Taiwan (3). Compared with the population without diabetes, the risk of stroke increases in patients with type 2 diabetes (T2D) (3,4). Despite the usefulness of therapies and lifestyle modifications following an ischemic stroke, a substantial residual lifetime risk of stroke recurrence and cardiovascular morbidity remains.

Several stroke-prevention medications, including antithrombotic, lipid-lowering, anti-inflammatory, and antidiabetic drugs, are available. Antidiabetic drugs, such as thiazolidinedione (5) and glucagon-like peptide 1 receptor agonists (GLP-1 RAs) (6), have had encouraging results in preventing stroke in patients with established atherosclerotic cardiovascular disease (ASCVD). However, evidence to support prevention of stroke in those without ASCVD is still lacking. GLP-1 RAs are a class of medications used in the treatment of T2D, including exenatide, lixisenatide, liraglutide, albiglutide, dulaglutide, and semaglutide. Exenatide, liraglutide, and dulaglutide were approved in Taiwan in 2009, 2011, and 2015, respectively. GLP-1 is an incretin hormone that stimulates insulin secretion after an oral glucose load through the incretin effect. In T2D, this process can become blunted or absent (7). In patients with a pharmacological level of GLP-1, it could potentiate glucose-stimulated insulin secretion (8). GLP-1RAs also appear to promote neuronal survival, attenuate apoptosis and oxidative stress in the brain (9,10), and prevent or mitigate the progression of carotid intima-media thickness, a recognized marker of subclinical atherosclerosis (11,12). Thus, the literature suggests GLP-1RAs is a promising drug in the treatment of conditions other than T2D. If GLP-1RAs promote clinically meaningful antiatherosclerotic activity in humans, one potential application may be reducing the burden of certain neurodegenerative disorders, such as ischemic stroke.

Since 2018, seven large cardiovascular outcome trials (CVOTs) have assessed the effect of GLP-1 RAs on cardiovascular outcomes in people with T2D (13). Of these seven trials, four reported that GLP-1RAs (liraglutide, semaglutide, albiglutide, and dulaglutide) significantly and safely reduced the primary composite cardiovascular outcome of the occurrence of nonfatal myocardial infarction, nonfatal stroke, or death resulting from cardiovascular problems (14–17). Meta-analysis of these seven CVOTs has indicated that treatment with GLP-1RAs was associated with a 15% decrease in nonfatal stroke but no significant difference in fatal stroke, and only the SUSTAIN 6 and REWIND studies demonstrated a statistically significant reduction in stroke occurrence by 39% and 24%, respectively (15,17).

These CVOTs have included a large population of patients at high cardiovascular risk or established ASCVD. Only the REWIND study included large primary prevention cohorts; only 31% of the patients had ASCVD. Furthermore, although all of the CVOTs included a variety of people, Asians accounted for <10%. Finally, stroke was a secondary end point in all trials. Therefore, to assess whether the results of CVOTs are generalizable to the whole population with diabetes, an analysis of real-world cohorts of patients treated in daily practice is necessary. Whether treatment with GLP-1RAs results in protection against stroke in those without ASCVD remains unknown, and the potential neuroprotective effects associated with GLP-1 RA treatment, as suggested by experimental models, remain to be tested. Results from population-based cohort studies of GLP-1RAs are limited. In this study, we assessed the effect of GLP-1RAs on hospitalization for ischemic stroke prevention in the Asian population with T2D without established ASCVD.

The data used in this retrospective cohort study were extracted from the Taiwan National Health Insurance Research Database (NHIRD). The NHIRD is derived from the single-payer National Health Insurance (NHI) program launched in Taiwan in 1995, and >99% of the population who reside in Taiwan are enrolled in this program. To ensure the privacy of medical information, the database uses encryption. The database contains information regarding dates of clinic visits and hospitalizations, diagnosis records of diseases coded according to the ICD-9 and ICD-10, Clinical Modification (CM), and drug and prescription records, including drug codes and duration of prescription supplies. The study was approved by the Research Ethics Committee of China Medical University and Hospital, Taichung, Taiwan (CMUH109-REC2–031).

Patients with at least two outpatient visits or one hospital admission of T2D (ICD-9-CM: 250, except 250.x1 and 250.x3; ICD-10-CM: E08–E13, except E10) between 1998 and 2018 were selected from the NHIRD. GLP-1RA users were defined as patients with T2D who had received exenatide (Anatomic Therapeutic Chemical [ATC] code: A10BJ01), liraglutide (ATC code: A10BJ02), and dulaglutide (ATC code: A10BJ05) at least once between 2011 and 2017. Patients with T2D who did not qualify as GLP-1RA users were identified as GLP-1RA nonusers. The censoring of the study included withdrawal from the NHI program, death, or the end of the study (31 December 2018). The date of the first prescription record after T2D diagnosis was set as the index date for GLP-1RA users. The outcome of interest in this study was the primary diagnosis attributed to hospitalization for ischemic stroke (ICD-9-CM: 43–435; ICD-10-CM: G45.0–G45.2, G45.8, G45.9, G46.0–G46.2, I63, I65, I66, and I67.84). The follow-up period for each patient started on the index date and continued until the onset of the outcome or censoring. The exclusion criteria were as follows: 1) patients with coronary artery disease (ICD-9-CM: 410–414; ICD-10-CM: I10–I16), peripheral artery occlusive disease (ICD-9-CM: 440.0, 440.2, 440.3, 440.8, 440.9, 443, 444.0, 444.22, 444.8, 447.8, and 447.9; ICD-10-CM: I70.0, I70.2–I70.9, I73, I74.0, I74.3-I74.5, I74.8, I75.0, I75.89, I77.3, I77.89, I77.9, I79.1, and I79.8), cerebrovascular disease (ICD-9-CM: 430–438; ICD-10-CM: G45.0–G45.2, G45.4, G45.8, G45.9, G46, and I60–I69), or gestational diabetes mellitus (ICD-9-CM: 648.83; ICD-10-CM: O24.410, O24.414, O24.419, and O99.810) with at least two outpatient visits or one hospital admission at baseline; 2) patients aged <20 years; 3) patients whose index dates out of range from diagnosis date of T2D to censoring time; 4) patients whose index dates were not between 2011 and 2017; 5) patients with invalid sex categories; and 6) patients with ischemic stroke before index date. Hypertension (ICD-9-CM: 401–405; ICD-10-CM: I10-I16), dyslipidemia (ICD-9-CM: 272; ICD-10-CM: E78), smoking (ICD-9-CM: 305.1, 649.0, V15.82; ICD-10-CM: F17.2, Z87.891), obesity (ICD-9-CM: 278, 649.2, 783.1, V45.86, V77.8, V85.2-V85.4; ICD-10-CM: E65-E68, R63.5, Z13.89), chronic kidney disease stage higher than 3 (ICD-9-CM: 585; ICD-10-CM: N18.3-N18.6), and family history of cardiovascular disease (coronary artery disease) (ICD-9-CM: V17.3, V17.4; ICD-10-CM: Z82.41 Z82.49) at baseline were included as comorbidities. Oral antidiabetic drugs (OADs) and insulins prescribed during the period from the diagnosis date of T2D to censoring time were identified as concomitant medications. OADs included metformin, sulfonylureas, thiazolidinediones, and sodium–glucose cotransporter 2 inhibitors. Other concomitant medications, such as statins, antihypertensive drugs, anticoagulants, and antiplatelets, prescribed at baseline were identified as well. Antihypertensive drugs included ACE inhibitors, angiotensin II receptor blockers, calcium channel blockers, β-blockers, and thiazides. Compliance rate for GLP-1RA users was calculated through the number of GLP-1RA prescription refill records divided by the number of outpatient visits. To minimize the influence of confounders, propensity score matching was used to balance the potential confounders between GLP-1RA users and GLP-1RA nonusers at a 1:1 ratio. The enrollment process is displayed in Fig. 1.

Categorical variables are presented as numbers and percentages, and χ² tests of homogeneity were used to determine whether frequencies of categorical data were evenly distributed across multiple groups. When count values in levels of categorical variables were less than three, they should be combined with count values in other levels to protect the privacy of beneficiaries. Some categorical variables were grouped based on the first, second (median), and third quartiles. Continuous variables are presented as means and SDs, and Student t tests were used to determine whether the means of two groups differed significantly. The incidence rate (IR) in this study was calculated by the equation: IR = the number of outcomes during the follow-up interval/the summed person-years of the population during the follow-up interval. The estimated cumulative incidence of the outcome was obtained using the Kaplan–Meier approach, and the log-rank test was used to compare the survival distributions of independent groups. Time-to-event analyses were performed using univariate and multivariate Cox proportional hazard models to estimate the crude and adjusted hazard ratios (HRs) and the corresponding 95% CIs of outcomes between the two groups. Statistical interaction analyses between the exposure of interest and confounders were performed to explore the magnitude of the effect of the exposure among strata of confounders. Univariate and multivariate logistic models with the outcome as the dependent variable and exposure or other factors as independent variables were also performed. The type I error rate was set at 0.05, and SAS 9.4 software (SAS Institute Inc., Cary, NC) was used for all analyses.

Both the GLP-1RA and non–GLP-1RA user groups included 6,534 patients. Table 1 displays the baseline characteristics of patients with T2D with and without GLP-1RA use after 1:1 matching, showing minimal differences in baseline clinical characteristics. The two groups were balanced on measured covariates. Mean age (SD) was 49 (12) years, 53% of patients were women, 67% had hypertension, 86% had dyslipidemia, and 6% had chronic kidney disease (stage higher than 3). The use of OADs, insulin, antihypertensive drugs, antiplatelets, and anticoagulants did not differ between groups. There was only modest difference in the average number of outpatient visits for T2D was 24.46 ± 19.97 for GLP-1RA nonusers and 26.89 ± 21.13 for GLP-1RA users (P < 0.0001). The average number of GLP-1RA prescription refill records was 6.89 ± 6.19. The overall average compliance rate for GLP-1RAs was 40.32 ± 51.10%, increasing to 65% for those with >251 days of GLP-1RAs. Supplementary Table 1 shows baseline demographics of patients with T2D with and without GLP-1RAs stratified by days of supplies of GLP-1RA compared with GLP-1RA nonusers.

Table 2 presents the IR and crude and adjusted HRs for hospitalization for ischemic stroke between patients with T2D with and without GLP-1RA use. A total of 113 patients were hospitalized because of ischemic stroke, which consisted of 67 GLP-1RA nonusers and 46 GLP-1RA users. The IR of hospitalization for ischemic stroke was 2.96/1,000 person-years in GLP-1RA nonusers and 2.01 in GLP-1RA users. The overall risk of hospitalization for ischemic stroke for GLP-1RA users was not significantly lower than that for GLP-1RA nonusers (adjusted HR 0.69 [95% CI 0.47–1.00]; P = 0.0506). When we stratified GLP-1RA users by days of supply, we found those with a >251 days of supply during the study period had a significantly lower risk of ischemic stroke hospitalization than did GLP-1RA nonusers (adjusted HR 0.28 [95% CI 0.11–0.71]).

The IR and adjusted HRs for hospitalization for ischemic stroke between patients with and without GLP-1RA use stratified by baseline characteristics are shown in Table 3. There were no statistical interactions between GLP-1RA use and baseline characteristics subgroups. HRs for hospitalization for ischemic stroke were lower in male patients, patients aged 40–49 years, and patients receiving antihypertensive drugs. Supplementary Fig. 1 displays the forest plot of Table 3. Logistic regression analysis for risk of hospitalization for ischemic stroke associated with GLP-1RA use and days of supply of GLP-1RAs; baseline characteristics are shown in Supplementary Table 2. GLP-1RA users with >251-day supplies of GLP-1RAs during the study period had a significantly decreased risk of hospital-admitted ischemic stroke than GLP-1RA nonusers (adjusted odds ratio [OR] 0.28 [95% CI 0.11–0.70]). Patients with hypertension (adjusted OR 3.66 [95% CI 1.74–7.71]) and receiving insulins (adjusted OR 4.00 [95% CI 1.84–8.70]) had an increased risk of hospital-admitted ischemic stroke.

Figure 2 displays the cumulative incidence of hospitalization for ischemic stroke among patients with T2D with and without GLP-1RAs using the Kaplan–Meier approach. Patients with T2D taking GLP-1RAs for >251 days had a lower cumulative incidence of ischemic stroke hospitalization than those not taking GLP-1RAs or taking GLP-1RAs for ≤251 days (P = 0.0176).

The treatment dosage effect of different drugs belonging to the same class of GLP-1RA was shown in Supplementary Table 3. Higher cumulative dose of GLP-1RAs (>1,784 mg) was associated with significantly lower risk of ischemic stroke hospitalization (adjusted HR 0.30 [95% CI 0.12–0.75] for the cumulative doses of 1,785–6,047 mg of GLP-1RAs and adjusted HR 0.46 [95% CI 0.25–0.85] for the cumulative doses of >6,047 mg of GLP-1RAs).

Our findings revealed that the use of GLP-1 RAs (exenatide, liraglutide, and dulaglutide) was associated with a decreased risk of hospitalization for ischemic stroke among Asian patients with T2D. GLP-1RA users with a >251-day supply during the study period had a significantly lower risk of ischemic stroke hospitalization than GLP-1RA nonusers did.

Data from a real-world study assessing the effect of GLP-1 RAs in the Asian population without established ASCVD are lacking. The only published research including data on the Asian population is a retrospective observational cohort study from Taiwan’s Health and Welfare Database provided by the Health and Welfare Science Center. That study evaluated the risk of cardiovascular events in high-cardiovascular-risk patients with T2D who initially used a GLP-1RA (liraglutide) compared with those who initially used basal insulin. The liraglutide group exhibited a lower risk of nonfatal stroke (HR 0.54 [95% CI 0.34–0.87]; P = 0.01) compared with the basal insulin group (18). Another retrospective database analysis of 39,275 patients with diabetes from the U.S. demonstrated that treatment with exenatide decreased the risk of incident cardiovascular and cerebrovascular disease by 19% (HR 0.81 [95% CI 0.68–0.95]; P = 0.01) compared with other glucose-lowering modalities (19). An Italian observational cohort study reported that first-time use of GLP-1RAs was associated with lower rates of cerebrovascular disease and ischemic stroke (HR 0.70 [95% CI 0.63–0.79]; and HR 0.72 [95% CI 0.60–0.87]) (20). All of these studies included a substantial proportion of patients with cardiovascular comorbidities. Our study excluded patients with coronary artery disease, peripheral artery occlusive disease, and cerebrovascular disease. To the best of our knowledge, this is the first study to investigate the effect of GLP-1RAs on stroke prevention in an Asian population without ASCVD.

A meta-analysis assessing the cardiovascular safety of glucose-lowering drugs on the basis of 301 randomized controlled trials (RCTs) of ≥24 weeks’ duration confirmed that some of these RCTs have included patients without ASCVD. However, mono- or dual therapy with GLP-1RAs did not reduce the risk of stroke compared with other classes of antidiabetic drugs (21). Another meta-analysis of 33 trials did not find any additional benefit from treatment with GLP-1RAs against stroke (OR 0.87 [95% CI 0.37–2.05]; P = 0.75) (22). A possible reason for the inconsistent results in stroke reduction is the follow-up time. Most studies were short-term, and longer follow-up is required to detect stroke incidence in such relatively lower-risk patients.

In an exploratory analysis of patients with T2D in the REWIND study (23), in which 69% of patients did not have ASCVD but did have additional cardiovascular risk factors and subclinical atherosclerosis, the rate of fatal or nonfatal stroke events was significantly lower in the dulaglutide group than the placebo group (HR 0.76 [0.62–0.94]; P = 0.0096); the median follow-up time was ∼5 years. Although our study only included those without ASCVD, most of the patients in this cohort had two crucial risk factors for stroke: dyslipidemia and hypertension. The median follow-up time was 3 years. Although the overall risk of ischemic stroke hospitalization for GLP-1RA users was not significantly lower than that for GLP-1RA nonusers, GLP-1RA users with a >251-day supply during the study period had a significantly lower risk of ischemic stroke hospitalization than did GLP-1RA nonusers (adjusted HR 0.28 [95% CI 0.11–0.71]). The current results also highlight that the continual use of these drugs can diminish the occurrence of hospitalization for ischemic stroke. Our study involved a population-based cohort, with clinical characteristics highly similar to patients treated in daily practice, to assess whether the results of RCTs with GLP-1RAs are generalizable to the whole population with diabetes. This study reflects GLP-1RA clinical use in this nationwide population study; GLP-1RA was used in a relative younger patient population without prior evident established cardiovascular disease. Prior studies of GLP-1 analogs had included older patients and required a prior vascular outcome or target-organ damage at enrollment. The results of this study provide additional information of the role of GLP-1RAs in prevention of hospitalization for ischemic stroke. GLP-1RAs may be a treatment option for Asian patients with T2D without prior established ASCVD but with dyslipidemia and hypertension. Identifying this group of patients with a different probability of benefitting from GLP-1RA treatment represents an essential clinical need.

Our study has some limitations. First, because this was an observational study, it may be affected by bias and possible confounding factors. However, propensity score matching was used to match baseline characteristics and adjust for potential confounders during the analysis of the risk of hospitalization for ischemic stroke between GLP-1RA users and non–GLP-1RA users. Moreover, the use of an administrative database prevented the underreporting of medical visits. Because this is a national population-based study, it is highly representative of the general population with diabetes. Second, the identities of patients were encrypted for privacy and data security reasons. As a result, we could not contact patients to discuss their adherence to GLP-1RA treatment; however, treatment adherence could be determined according to diabetes outpatient clinic visits and prescription refill records. Compliance rate for GLP-1RA users could be calculated through the number of GLP-1RA prescription refill records divided by the number of outpatient visits. For those who continue use, the refill is >50% during the follow-up period. Third, several potential confounding factors, such as blood pressure, serum glucose level, lipid panel, family history, and subclinical atherosclerosis status, were not included in the database because these data are not available from Taiwan’s NHI claims data. The reliability of the prevalence of hypertension and dyslipidemia in the study population may be compromised. Antihypertensive drugs and antilipid drugs were added to increase the reliability of the prevalence of hypertension and dyslipidemia in the study population. Fourth, diabetes and hypertension are well-known risk factors for lacunar stroke, an often-unrecognized condition that patients are frequently unaware of. The outcome of interest in this study was attributed to the diagnosis of ischemic stroke during hospitalization, but subclinical stroke (which did not require hospitalization or was not recognized) could be overlooked. Finally, because our study included only Asian-Taiwanese patients, our results may not be applicable to other populations.

A longer duration and higher cumulative dose of GLP-1 RA use was associated with a decreased risk of hospitalization for ischemic stroke among Asian patients who had T2D and dyslipidemia or hypertension but without established ASCVD. Further studies are necessary to determine the clinical relevance of GLP-1 RAs in the prevention of stroke or subclinical stroke in patients with T2D and various risk factors.

Funding. This study is supported in part by the Taiwan Ministry of Health and Welfare Clinical Trial Center (MOHW110-TDU-B-212-124004), Ministry of Science and Technology (MOST 110-2321-B-039-003), and China Medical University Hospital (DMR-111-105). We are grateful to Health Data Science Center, China Medical University Hospital, for providing administrative, technical, and funding support. The funders had no role in the study design, data collection and analysis, the decision to publish, or preparation of the manuscript. No additional external funding was received for this study.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.-S.Y. and C.-H.K. were responsible for conception and design. C.-H.K. provided administrative support. All authors were responsible for collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of the manuscript. C.-H.K. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.