The benefits of sodium-glucose cotransporter 2 inhibitors (SGLT2i) in acute ischemic stroke (AIS) have been demonstrated preclinically [1,2]. However, clinical studies on the association between SGLT2i and stroke have been mainly focused on determining whether SGLT2i increase the risk of incident ischemic stroke in the strokenaïve population [3]. Moreover, given the results of clinical trials suggested that SGLT2i use may be less beneficial with respect to stroke compared to other cardiovascular outcomes [4,5], physicians may raise concerns that SGLT2i could elevate the vulnerability to ischemic stroke. From this perspective, SGLT2i might potentially lead to early neurological deterioration (END) in patients with AIS, resulting in poor neurological outcome. Due to the controversial nature of this issue, the role of SGLT2i in cerebral ischemia remains unclear in patients with ischemic stroke. We aimed to explore the association between SGLT2i treatment and stroke outcomes, including neurological deterioration and recovery up to 3 months after stroke.

This retrospective observational study reviewed consecutive diabetic patients with AIS enrolled in Seoul National University Hospital (SNUH) stroke registry from January 2018 to June 2022. Patients were grouped into SGLT2i and control groups according to their prescription of SGLT2i at admission. To investigate the mechanistic implications, we used data from another cohort where neurologically stabilized participants with AIS were prospectively enrolled and underwent ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG-PET), to predict stroke recurrence. The Institutional Review Board of SNUH approved this study (No. 1009-062-332). The need for informed consent was waived for the SNUH stroke registry, owing to the retrospective nature of the study. For those included in the FDG-PET cohort, informed consent was obtained (No. 2105-013-1217).

Clinical data were collected upon admission from all patients. The decision to prescribe SGLT2i to diabetic patients with stroke was at the physician’s discretion. Following 1:4 propensity score matching, binary and ordinal logistic regression analyses were performed to evaluate the association between SGLT2i use and clinical outcomes including END during admission, National Institutes of Health Stroke Scale (NIHSS) score at discharge, and modified Rankin Scale (mRS) scores at discharge and at 3 months. Using the FDG-PET cohort, we compared metabolic activity of various organs based on SGLT2i use, represented as the target-to-background ratio (Supplementary Methods).

Among the 806 eligible patients, 88 (10.9%) were prescribed SGLT2i on admission, and 74 continued the prescription at discharge. Two hundred fifty-nine controls were propensity score-matched to the 71 patients in the SGLT2i group (32 with dapagliflozin and with 39 empagliflozin) and were well-balanced (Supplementary Figure 1, Supplementary Table 1, and Table 1). During the total admission period of 7 [5-11] days, SGLT2i were prescribed for a median duration of 6 [4.5-10] days. Information regarding antidiabetic medications other than SGLT2i was summarized in Supplementary Table 2.

The SGLT2i group demonstrated a generally favorable functional outcome compared to the control group, both before and after propensity score matching, particularly at 3 months (Supplementary Figure 2). Binary logistic regression analyses of the matched population showed significantly higher odds of achieving excellent 3-month functional outcomes and a trend toward favorable outcomes in the SGLT2i group. Ordinal logistic regression analysis indicated significantly higher odds of a better 3-month mRS score with SGLT2i use during admission. No significant differences in END occurrence, discharge NIHSS score, or discharge mRS score were found between the two groups (Table 2). There were no significant interactions between SGLT2i use and various clinical factors for excellent outcomes (Supplementary Figure 3).

In the SGLT2i group, 32 were pre-stroke SGLT2i users, while the remaining 39 were first prescribed SGLT2i during their admission. Clinical outcomes were comparable regardless of pre-stroke SGLT2i use (Supplementary Table 3). The patients discharged with SGLT2i had more favorable outcomes at discharge and at 3 months compared to those without SGLT2i prescription at discharge (Supplementary Table 4). In the sensitivity analysis encompassing all patients prescribed SGLT2i at admission as the SGLT2i group, consistent with the main analyses, the SGLT2i group had significantly higher odds of excellent outcomes at 3 months (Supplementary Tables 5 and 6).

Fourteen diabetic and 26 nondiabetic stroke survivors participated in the FDG-PET cohort, and six of diabetic patients were prescribed SGLT2i at admission. Diabetic patients using SGLT2i exhibited elevated metabolism in supraclavicular brown and subcutaneous adipose tissues compared to diabetic patients not using SGLT2i (Figure 1 and Supplementary Table 7).

Our study showed that SGLT2i may be safely used without increasing END after AIS. Remarkably, acute phase SGLT2i use was significantly associated with better neurological outcomes at 3 months, while acute outcomes at discharge were not affected. Continuing SGLT2i use beyond discharge, rather than transient use during admission, was associated with a more favorable 3-month functional outcome. Given that SGLT2i enhanced synaptic function in diabetic animal models and ameliorated neuronal loss in neurodegenerative animal models in previous studies [6,7], our findings may imply a potential role of SGLT2i in post-stroke functional recovery as well as in neuroprotection against acute ischemia. The FDG-PET cohort findings indicate that brown adipose tissue and beige adipocytes, which reportedly benefit metabolic health and reduce arterial inflammation [8,9], might have some role in the association between SGLT2i and favorable outcomes. However, the conclusions are tentative due to the small sample size, necessitating further research for validation.

This study had several limitations. Firstly, its retrospective design introduces the possibility of unmeasured confounders affecting the outcomes, such as the unavailability of post-discharge drug compliance data. Secondly, the limited generalizability due to the predominantly minor neurological deficits among study subjects and the small sample size, particularly in the SGLT2i group, calls for further validation in larger, more diverse cohorts. Finally, this study only included ischemic stroke patients with diabetes, and the neuroprotective effect of SGLT2i in non-diabetic stroke patients remains uncertain.

In conclusion, our findings suggest that SGLT2i may be a priority for diabetic patients with AIS because of its potential benefits on neurological outcomes. Dedicated studies are needed to validate whether SGLT2i have neuroprotective effects in ischemic stroke and explore its potential mechanism.