Sample Size Estimation

We estimated that a sample size of 200 patients would provide 80% (1 − β) power at a two-tailed 5% (type I error rate, α) significance level to detect a difference between the modification and no modification groups, similar to that in the literature.

Based on a cohort study of patients with symptomatic atherothrombotic intracranial stenosis, the incidence of recurrent stroke was 25% [20, 21], which was adopted in the study. The contemporary treatment strategies included antiplatelets, statins, stent implantation, and/or internal carotid endarterectomy for prevention of recurrent ischemic stroke. The estimation assumed an endpoint rate of 9% [22] of the study population (treatment group, 9%; natural history group, 25%; sampling ratio, 1:1; superiority margin, 0.08). The sample size was estimated as 56 (each group, n = 28).

In a real-world study, all patients received clopidogrel. To test the difference between the modification and no modification groups, we speculated a 1:1 ratio since the maximum variability for CR was 50%. According to the literature, the incidences of ischemic stroke with dual and mono antiplatelet therapies were 9.1% and 27.9% [22], respectively; hence, the sample size was calculated to be 100 (superiority margin, 0.10; each group, n = 50).

A recently published study regarding scheduled neurointervention for unruptured intracranial aneurysms reported that the incidence of thromboembolic complications in the tailored and nontailored groups was 6.6% and 16% [17], respectively; according to the outcome rate of this study, the sample size was calculated to be 296 (superiority margin, 0.05; each group, n = 148).

The estimation was entered into an online sample size calculator (http://www.powerandsamplesize.com/), and the relevant parameters were set. Considering a shedding rate of less than 20%, the trial was designed to enroll at least 120 patients; 356 patients at most should be included. As a result of the disparity in the diseases from the collected studies, we selected 200 patients as the sample size.