Study Characteristics and Quality
The baseline characteristics of the included studies are illustrated inTable 1 . Among the 9 included observational studies, 3 were from the UK[29], Germany[8], and Japan [30], the other 6 were derived from nationwide or health insurance claims databases in the United States[15, 23-27]. Of the 3 included post-hoc analyses of RCTs[9, 10, 28], all were multicenter large-scale randomized clinical trials. The mean age of patients ranged from 60.1 to 83.0 years, and the sample size was from 1,558 to 188,863. Across studies, the study populations in the NOACs group were administrated with dabigatran, apixaban, rivaroxaban, and edoxaban. Supplementary Table Ⅱ shows the clinical outcomes of the included articles and the adjustment for confounding factors of the outcomes. Risk of bias evaluation was performed, shown in Supplementary Table Ⅲ . All the studies had a NOS of ≥6 points suggesting moderate-to-high quality.
Polypharmacy definition
There was a slight variation in the definition of polypharmacy used across the studies included in our meta-analysis. We defined polypharmacy as a discrete definition. When dividing the boundaries of non-polypharmacy, moderate polypharmacy, and severe polypharmacy, the threshold used by the study author was used. Specifically, 5 studies [10, 15, 23-25] included articles defined moderate polypharmacy as the use of 5-9 drugs, 3 studies [8, 28, 30] included articles defined moderate polypharmacy use as the use of 5-8 drugs, and 2 studies [9, 29] included articles defined moderate polypharmacy use as 6-8 drug use, and 1 article[27] defined moderate polydrug use as 4-8 drug use. Correspondingly, 6 articles [8, 9, 27-30] defined severe polypharmacy as the use of ≥9 drugs, and 6 articles [10, 15, 23-26] defined severe polypharmacy as the use of ≥10 drugs. It should be noted that the article by Martinez et al. [26] defines polypharmacy as the use of 5 or more drugs and explores the grouping of 10 or more drugs in the secondary analysis. Therefore, we did not classify its data into the moderate polypharmacy group but used its secondary analysis data as the severe polypharmacy group. Detailed classification information is shown in Table 1 .