Publication bias and strength of the body of evidence
The funnel plot presented a relative symmetry (Supplementary Figure S3). There was no obvious publication bias evaluated by Begg’s test (P=0.348) and Egger’s test (P =0.770), and the pooled RR was unchanged with the trim-and-fill method (RR [95%CI] =0.72 [0.60, 0.86], P =0.000).
As listed in Supplementary Table S6, with the GRADE system, we graded SOE for all investigated outcomes as low to moderate, because the plausible confounding factors may exist for the observational design although the effect was fully adjusted. The SOE for outcomes between OAC/VKA/NOAC use and no-OAC use was graded as low due to the inconsistency and indirectness.