Background: Although the embolic stroke of undetermined source (ESUS) construct was proposed as a unifying concept for embolic-appearing strokes without an identified cause, it includes patients with diverse and often unrecognized underlying mechanisms. This study investigates the differential rate of recurrent stroke according to the completeness of neurodiagnostic testing in a diverse multicenter cohort. Methods: This was a retrospective analysis of a multicenter observational cohort study (n=27 sites). Patients with ESUS were categorized into three groups: (1) established mechanism after further work up (ESUS-E), (2) patients who underwent a comprehensive diagnostic workup without an identified etiology (ESUS 2.0), and (3) no advanced workup (ESUS-NAW). The primary outcome of recurrent ischemic stroke was assessed by comparing ESUS 2.0 to both ESUS-E and ESUS-NAW groups using adjusted Cox proportional hazards models. Results: Of the 2,281 patients included, 694 (30%) were classified as ESUS-E, 344 (15%) as ESUS 2.0, and 1,443 (55%) as ESUS-NAW. The ESUS 2.0 group had a significantly lower annualized recurrence rate (1.98 per 100 person-years) compared to both the ESUS-NAW group (7.7 per 100 person-years; aHR 0.27, 95% CI: 0.11–0.62; p=0.002) and the ESUS-E group (6.10 per 100 person-years; aHR = 0.35; 95% CI 0.15–0.85, p=0.020). Within the ESUS-E group, cancer (aHR 5.33; 95% CI: 2.09–13.63; p < 0.001) and non-stenosing atherosclerotic disease or arch atheroma (aHR 3.40; 95% CI: 1.18–9.83; p = 0.023) were associated with significantly increased risk of recurrence when compared to ESUS 2.0. Conclusion: These findings demonstrate a lower risk of recurrent stroke in patients classified as ESUS 2.0, diverging from the 5–6% annual recurrence rates observed in prior ESUS trials. These findings suggest that patients who receive comprehensive diagnostic evaluation without identification of a specific etiology have a lower risk of recurrent stroke, an important consideration for the design of future ESUS studies.
Khan et al. (2026) studied this question.