Study design and setting
This was a retrospective cohort study undertaken at the Stroke Center of Seirei Hamamatsu General Hospital (Hamamatsu, Japan), in which medical records between January 2012 and December 2014 were reviewed. A total of 779 ischemic stroke patients were admitted to the hospital within that period. The study was carried out in accordance with the Declaration of Helsinki, and the protocol was approved by the Institutional Review Board of Seirei Hamamatsu General Hospital. The patients were identified based on the registered disease name in the medical records, and all patients were examined to determine whether they met the ESUS criteria.
ESUS criteria
Ischemic stroke patients with a documented acute lesion on brain imaging and hospitalized at the stroke center were included. Patients who were excluded were those who did not satisfy the ESUS criteria, [1] except for extracranial atherosclerosis: (a) stroke detected by magnetic resonance imaging (MRI) that was not lacunar (defined as a subcortical infarct ≤2.0 cm on MRI diffusion images in the largest dimension, and in the distribution of the small, penetrating cerebral arteries); (b) absence of intracranial atherosclerosis causing ≥50% luminal stenosis in arteries supplying the area of ischemia; (c) no major risk of a cardioembolic source of embolism (permanent or paroxysmal atrial fibrillation, sustained atrial flutter, intracardiac thrombus, prosthetic cardiac valve, atrial myxoma or other cardiac tumors, mitral stenosis, <4 weeks after myocardial infarction, left ventricular ejection fraction <30%, valvular vegetations, or infective endocarditis); (d) no other specific cause of stroke identified (arteritis, dissection, migraine/vasospasm, drug misuse); and (e) minimum diagnostic assessment performed (brain MRI, 12-lead electrocardiogram, precordial echocardiography, cardiac monitoring for ≥24 h with automated rhythm detection, MRI or computed tomography (CT) angiography imaging of intracranial arteries supplying the area of brain ischemia and neck echo screening. Patients who were dead on discharge, those taking both anticoagulants and antiplatelets after discharge, and those taking neither anticoagulants nor antiplatelets after discharge were excluded.
Patients’ demographic data
The following demographic data were collected referring to a previous cryptogenic stroke study[12] as variables: age, sex, stroke recurrence, period to recurrence, secondary prevention drug, estimated glomerular filtration rate (eGFR), CHADS2 score,[13] modified Rankin Scale (mRS) score at discharge, death during hospitalization, and medical history of smoking, diabetes mellitus, hypertension, dyslipidemia, stroke, chronic heart failure, and acute coronary syndrome. The datasets were created by reviewing the patients’ electronic medical records. The data were double-checked by two of the authors (Y.S. and T.A.).
Data of the exposed group/non-exposed group
The ESUS patients were divided into two groups: the antiplatelet (AP) group, who took antiplatelets as secondary prevention for ischemic stroke when they were discharged; and the anticoagulant (AC) group who took anticoagulants. The drug choice depended on the judgment of the doctor in charge. Antiplatelets included aspirin, clopidogrel sulfate, and cilostazol. Anticoagulants included warfarin and DOACs. DOACs were used only for deep venous thrombosis. These patients were followed for three years from ESUS onset.
Outcome measures
The primary outcome was recurrence-free survival, with recurrence defined as acute ischemic stroke including transient ischemic attacks (TIAs) evaluated by a stroke specialist regardless of positive or negative neuroimaging. The outcome measure time frame was 3 years.
Patients’ overall survival and major bleeding according to the International Society on Thrombosis and Haemostasis (ISTH) criteria[14] were investigated as secondary outcomes.
Statistical analysis
The outcome variable was event-free survival time, and age, sex, CHADS2 score, and mRS score at discharge were used as covariates. Missing or abnormal data were checked by re-reviewing the medical charts (Y.H.). The means (standard deviation; SD) for continuous variables and frequencies (%) for categorical variables were calculated as summary statistics.
The event-free survival rate was calculated using the Kaplan-Meier method. Crude and multivariate adjusted hazard ratios and their 95% confidence intervals were estimated by Cox proportional hazard models with the maximum likelihood method. Score test statistics were also obtained. P-values < 0.05 on two-sided tests were considered significant. SAS University (SAS Institute, Cary NC, USA) was used for statistical calculations.
The sample size was estimated based on a previous Japanese study report.[15] With a 5% alpha error and 80% power, a sample size of 27 patients in each group was needed. A pilot study was performed with approximately 100 patients to facilitate the sample-size calculation. Consequently, the research period was set to 3 years, since this allowed for seasonal variability in disease prevalence.