In this study, we used data of the Clinical Research Center for Stroke-5 (CRCS-5) registry, which is a prospective registry of patients with acute ischemic stroke admitted to 15 academic centers in Korea. The CRCS-5 prospective registry was approved by the Institutional Review Boards of all participating centers (http://www.stroke-crc.or.kr). Detailed information on the CRCS-5 registry have been published previously [15,17]. Because the purposes of the registry were to monitor and improve the quality of stroke care in Korea and a computer-assisted de-identification system ensured the anonymity of individual patients during enrollment, the informed consent from individual patients or their legally authorized representatives was waived by the relevant institutional review boards. However, this study was additionally approved by the institutional review board of Inje University Ilsan Paik Hospital where the principal investigator of this study is affiliated.

Of 27,511 patients with acute ischemic stroke enrolled in the CRCS-5 registry between October 2009 and November 2014, 2,611 patients were treated with IV-TPA. We excluded the following patients; (1) missing information on the prestroke antiplatelet use, (2) prestroke warfarin therapy with international normalized ratio (INR) >1.7 or non-vitamin K antagonist oral anticoagulants, (3) endovascular recanalization therapy added to IV-TPA therapy, (4) prestroke modified Rankin Scale (mRS) score >1, (5) IV-TPA treatment at outside hospital, (6) IV-TPA treatment beyond 4.5 hours, and (7) unavailability of discharge mRS score.

Prestroke antiplatelet use was defined as taking antiplatelet within 7 days before the onset of index ischemic stroke. At the inception of the CRCS-5 registry on April 2008, we prospectively compiled data of the use of prestroke antiplatelet including aspirin, clopidogrel, cliostazol, triflusal, dipyridamole, ticlopidine, other antiplatelet, and any combination of antiplatelets. Additionally, we prospectively collected data regarding demographics, vascular risk factors, prestroke medications, initial stroke severity and ischemic stroke subtypes, vital signs, laboratory findings, SICH, and discharge mRS score. The primary safety outcome was SICH, which was defined as any hemorrhagic transformation associated with National Institutes of Health Stroke Scale (NIHSS) score worsening ≥4 points. Secondary outcomes were discharge mRS score and in-hospital mortality.

For missing variables, variables with missing data of more than 10% of total observation were excluded from the analysis. Simple imputation with median values was used for continuous variables with missing data of less than 10% of the total observations. For descriptive statistics, continuous variables were summarized as means (standard deviations) or median (interquartile range), and categorical variables as numbers with percentages. Baseline characteristics between prestroke antiplatelet users and non-users were compared with Student t-test for continuous variables with a normal distribution, Wilcoxon rank-sum test for continuous variables without a normal distribution, or the chi-square test for categorical variables.

To compare the risks of SICH and mortality between the two groups, we used Chi-square test or Fisher’s exact test for unadjusted analysis and binary logistic regression analysis for covariates adjustment. To estimate absolute difference of SICH between two groups after multivariable adjustment, we used the method described by Ciolino et al. [18]. For mRS outcomes, we compared the overall distribution of discharge mRS scores and dichotomized mRS outcomes (mRS 0-1 vs. 2-6 and mRS 0-2 vs. 3-6) between the two groups. The odds ratio (OR) with 95% confidence interval (CI) was calculated to estimate the probability of favorable shift on the mRS score with prestroke antiplatelet use. The Cochran-Mantel-Haenszel test was used for unadjusted shift analysis, and the ordinal logistic regression analysis was used to adjust covariates.

In addition to multivariable analyses, we performed propensity score (PS)-matched analysis because the baseline characteristics of patients taking prestroke antiplatelet and those without prestroke antiplatelet were expected to differ significantly. To estimate PS, the binary logistic regression with all relevant characteristics related to the prestroke antiplatelet was used. The PS model included all the variables shown in Table 1 except for onset to arrival time, and TPA dose. Greedy-matching algorithm was employed with width of calipers between 0.2 and 0.6 standard deviations of the logit of the estimated PS. To check the baseline imbalances between the two groups after the PS matching, the absolute standardized difference (ASD) was computed for each covariate. We performed outcome analysis using ordinal or logistic regression analysis based on the generalized estimation equation method after the PS matching. Covariates with an ASD >0.1 were further adjusted in PS-matched analysis. However, we included age, sex, initial NIHSS score, onset to treatment time, and TPA dose in the final model as clinically important covariates regardless of their ASD values. As a sensitivity test, we repeated the same analyses with mRS outcome at 3 months after the stroke as a dependent variable. In all analyses, a P value <0.05 was considered statistically significant. SAS computer software (Version 9.3, SAS Institute, Cary, NC, USA) was used for the statistical analyses.

Of 2,611 patients treated with IV-TPA, we excluded 896 patients in the following order: (1) missing information on the prestroke antiplatelet use (16 patients), (2) prestroke warfarin therapy with INR >1.7 or non-vitamin K antagonist oral anticoagulants (4 patients), (3) prestroke mRS score >1 (420 patients), (4) IV-TPA treatment at outside hospital (325 patients), (5) treated with IV-TPA beyond 4.5 hours (128 patients), and (6) unavailability of discharge mRS score (3 patients) (Figure 1). As a result, 1,715 patients were included in this study, and 441 (25.7%) were prestroke antiplatelet users. Of them, 353 patients were taking single antiplatelet therapy (aspirin, 279 patients; clopidogreal, 59 patients; 15 other monotherapy), 87 patients were taking dual antiplatelet therapy (aspirin plus clopidogrel, 69 patients; other dual therapy, 19 patients) and one patient was taking triple antiplatelet therapy. Table 1 presents demographic and clinical characteristics of prestroke antiplatelet users and non-users before and after PS matching. Before PS matching, prestroke antiplatelet users versus non-users were more likely to be older, to have prestroke disability, hypertension, diabetes, dyslipidemia, atrial fibrillation, prior stroke, and coronary heart disease, to take statins, and to have lower levels of blood pressure, total and low-density lipoprotein (LDL) cholesterol, and hemoglobin. They were also more likely to have higher NIHSS score at baseline, and to have stroke subtype of cardioembolism. After PS matching, except for hypertension and TPA dose, the baseline characteristics were well balanced between the two groups (Supplemental Figure 1). Among 324 patients treated with prestroke antiplatelet therapy, 273 (84.3%) were taking single antiplatelet therapy and 51 (15.7%) were taking dual antiplatelet therapy.

Overall, SICH occurred in 64 patients (3.7%) after IV-TPA therapy. Before PS matching, prestroke antiplatelet users versus non-users had a higher rate of SICH before adjustment (5.9% vs. 3.0%; unadjusted OR 2.04 [1.22-3.40]), and the increased risk remained significant after multivariable adjustment (OR 1.79 [1.05-3.04]). The estimated absolute difference after adjustment was 1.8%. In particular, prestroke dual antiplatelet use was associated with a greater risk of SICH (unadjusted OR 3.29 [1.28-7.48]) compared to single antiplatelet use (unadjusted OR 1.75 [0.93-3.19]).

In PS matching analysis, there was a trend of increased risk of SICH with prestroke antiplatelet use, but it did not reach statistical significance (OR 1.94 [0.96-3.93]) (Table 2).

In subgroup analyses, the increased risk of SICH with prestroke antiplatelet use did not differ among subgroups stratified by age, sex, diabetes mellitus, IV-TPA treatment window, TPA dose, baseline NIHSS score, and prestroke statin use (Table 3 and Figure 2).

At discharge the two groups did not differ in mRS distribution (adjusted OR 0.90 [0.72-1.14]; adjusted OR after PS matching 0.94 [0.71-1.23]), mRS 0-1 outcome (34.2% vs. 33.7%; adjusted OR 1.27 [0.94-1.14]; adjusted OR after PS matching 1.24 [0.87-1.79]), mRS 0-2 outcome (52.4% vs. 52.9%; adjusted OR 1.21 [0.90-1.63]; adjusted OR after PS matching 1.20 [0.84-1.70]), and in-hospital mortality (6.1% vs. 4.2%; adjusted OR 1.19 [0.71-2.01]; adjusted OR after PS matching 1.26 [0.60-2.65]) in both multivariable and PS matching analyses (Table 4 and Figure 3). The mRS score at 3 months after the stroke was available for 1,432 patients (83.5%), and prestroke antiplatelet users versus non-users had comparable mRS distribution or dichotomized mRS outcomes in multivariable and PS matching analysis (Table 5).