Study selection and outcome measurement
Two reviewers (Mingjie Lin and Wenqiang Han) independently screened the titles and abstracts according to the designed criteria; and full articles were obtained for all titles potentially meeting inclusion and exclusion criteria. Endnote Version x9.1 (Clarivate Analytics (US) LLC) was used to manage identified studies and recode the decisions. Reviewers were blinded to each other’s decisions. Disagreements between individual judgements were resolved by discussion between themselves and another senior reviewer (Jingquan Zhong). The inclusion criteria were: i) patients with AF who were ≥ 18 years; ii) investigating the effects of OACs on the incidence of dementia; iii) reporting the outcomes as HR, RR, or OR, with 95%CIs; iv) RCTs or observational studies. The exclusion criteria were: i) patients with the history of dementia or moderate to severe cognitive dysfunction; ii) only crude or unadjusted HR, RR, or OR; iii) cross-sectional studies; iv) following types of articles: conference abstracts, editorials, letters, expert opinions, case reports, case series, or reviews. OACs included vitamin K antagonists (VKAs) and non-vitamin K OACs (NOACs); the control, i.e. Non-OACs included antiplatelet agents or no antithrombotic therapy; dementia included Alzheimer’s/Lewy-body/senile/vascular dementia; for overlapping participants, the studies with the longest follow-up and the most detailed information were chosen. The PICOS criteria are presented in Supplementary Table S2.
The primary outcome was the incidence of dementia during the follow-up in AF patients with and without OAC use. The dementia was diagnosed based on patient narrative, collateral information from families, clinical examination, the standardized cognitive tests,2 and diagnostic criteria for dementia including Diagnostic and Statistical Manual of Mental Disorders (DSM)-III, DSM-IV, DSM-V, diagnosis based on international classification of diseases codes (ICD), or United Kingdom (UK) Read codes.9, 23 The additional outcomes were the incidence of dementia in AF patients with: VKAs vs. non-OACs, VKAs vs. NOACs, NOACs vs. non-OACs, and in AF patients with and without OACs use whose CHA2DS2-VASc score was ≥ 2.