Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, with an estimated global prevalence exceeding 52 million individuals as of 2021. ^{1, 2} AF confers a well-established independent risk for ischemic stroke, and AF-related strokes are characteristically more severe, carry higher disability rates, and are associated with greater mortality compared with strokes of other etiologies. ^{3, 4} Despite widespread adoption of oral anticoagulation, the residual risk of stroke recurrence in AF patients remains clinically significant; a recent systematic review reported pooled recurrence rates of 3.75% per year overall and 7.20% per year among anticoagulated patients. ⁵

Prediabetes, defined by hemoglobin A1c (HbA1c) levels of 5.7% to 6.4%, impaired fasting glucose, or impaired glucose tolerance, affects hundreds of millions of adults worldwide and is associated with endothelial dysfunction, chronic systemic inflammation, and a prothrombotic state. ^{6– 9} While overt diabetes mellitus is incorporated into the CHA ₂DS ₂-VASc scoring system, the independent contribution of prediabetes to stroke recurrence in AF patients has not been adequately characterized. ^{10, 11}

Evidence to date has been conflicting. Desai et al. reported a nearly two-fold higher risk of recurrent stroke among prediabetic AF patients using the National Inpatient Sample (NIS), a cross-sectional database limited to in-hospital events during single encounters (N = 480 prediabetes cases). ¹² Conversely, a large meta-analysis by Cai et al. found no significant association between prediabetes and stroke recurrence in patients with established atherosclerotic cardiovascular disease (relative risk 1.05, 95% CI 0.81–1.36). ¹³ No large-scale, propensity-matched study has examined longitudinal time-to-event outcomes for recurrent stroke in AF patients stratified by prediabetes status. Accordingly, we investigated the association of prediabetes with recurrent ischemic stroke, all-cause mortality, gastrointestinal (GI) bleeding, and intracranial hemorrhage (ICH) using the TriNetX US Collaborative Network. ¹⁴

After propensity score matching, 80,335 patients remained in each cohort with excellent covariate balance (all SMDs <0.02; Table 1). The mean age at index was 69.9 ± 10.6 years in the prediabetes cohort and 70.0 ± 11.2 years in the control cohort. Male sex was equally represented at 54.7% in both groups. Comorbidity burden was well balanced across all ICD-10 diagnosis chapter categories.

In this propensity-matched analysis of over 160,000 AF patients with prior ischemic stroke from 67 US healthcare organizations, prediabetes was not associated with increased cumulative recurrent stroke risk. Time-to-recurrence was significantly longer in the prediabetes cohort, and prediabetes was associated with lower all-cause mortality and lower GI bleeding. These findings diverge from prior work and warrant consideration of the methodological factors that may account for the discrepancy.

The contrast with the results of Desai et al. ¹² is likely attributable to several differences in study design. Their analysis used the NIS, an administrative database capturing events during single hospitalizations without longitudinal follow-up. The prediabetes sample was small (N = 480), and the outcome metric (in-hospital stroke recurrence) is susceptible to length-of-stay bias. Our study, by contrast, leveraged a federated EHR platform with longitudinal patient-level data, rigorous one-to-one propensity score matching on demographic and clinical covariates, and Kaplan-Meier survival analysis across a substantially larger sample (80,335 matched pairs). The findings align more closely with the meta-analysis by Cai et al., ¹³ which found no significant increase in stroke recurrence among patients with established cardiovascular disease and prediabetes.

A key observation from this analysis is the discrepancy between cumulative risk and time-to-event results for recurrent stroke. While overall proportions of patients experiencing recurrence were similar between groups, the Kaplan-Meier analysis showed that patients with prediabetes reached this threshold at a measurably slower rate (median difference of approximately 99 days). These two analytic approaches provide complementary information: cumulative risk reflects the proportion who ultimately experience the event, whereas survival analysis captures the tempo of events. The clinical significance of a three-month delay in stroke recurrence, in terms of preserved neurological function and quality-adjusted life years, should not be dismissed. ^{5, 16}

We propose a metabolic surveillance effect as one potential explanation for these findings. Patients diagnosed with prediabetes have, by definition, undergone glycemic screening and entered a framework of heightened medical attention. This surveillance may lead to earlier identification and optimization of modifiable risk factors, including hypertension, dyslipidemia, and anticoagulation management, all of which independently influence stroke recurrence. ^{13, 17} This hypothesis is consistent with the differential effect of prediabetes by baseline cardiovascular risk reported by Cai et al., in which prediabetes conferred excess risk in general populations but not in those with established cardiovascular disease already under active management. ¹³

The lower all-cause mortality observed in the prediabetes cohort (HR 0.891; P < 0.001; median survival difference approximately 380 days) may similarly reflect the downstream benefits of increased clinical contact and preventive care. The lower rate of GI bleeding (HR 0.918; P = 0.003) could stem from more careful anticoagulant selection and dose adjustment in patients under closer metabolic surveillance. The absence of a significant difference in ICH provides reassurance that prediabetes does not confer additional hemorrhagic risk in this population.

In AF patients with prior ischemic stroke, prediabetes was not associated with increased cumulative recurrent stroke risk and was associated with longer time-to-recurrence, lower all-cause mortality, and lower GI bleeding. These results suggest that prediabetes alone should not be considered an additional high-risk marker for stroke recurrence in this population. The observed associations may partly reflect a metabolic surveillance benefit. Prospective studies should evaluate whether structured glycemic monitoring programs can improve long-term cerebrovascular outcomes in AF patients.

This study was granted an institutional review board (IRB) exemption by the Western IRB (WCG IRB), as all data were de-identified electronic health records and the study did not constitute human subjects research under 45 CFR 46. A formal IRB reference or permit number was not assigned, as the study qualified for exemption rather than full board review. The study was performed in accordance with the principles stated in the Declaration of Helsinki ( https://www.wma.net/policies-post/wma-declaration-of-helsinki/). Individual informed consent (written or verbal) was not required and was waived by the IRB, as the study exclusively utilized de-identified electronic health records with no direct patient contact, no access to identifiable patient information, and no prospective enrollment of human participants.

During the preparation of this work, the authors used AI-assisted tools for manuscript formatting and language editing. The authors reviewed and edited all content and take full responsibility for the content of the published article.