BackgroundAs a common sequela after stroke, cognitive impairment negatively impacts patients' activities of daily living and overall rehabilitation. Non-pharmacological therapies have recently drawn widespread attention for their potential in improving cognitive function. However, the optimal choice of non-pharmacological therapies for post-stroke cognitive impairment (PSCI) is still unclear. Hence, in this study, we compared and ranked 5 non-pharmacological therapies for PSCI with a Bayesian Network Meta-analysis (NMA), to offer a foundation for clinical treatment decision-making.MethodsPubMed, EMBASE, Web of Science, Cochrane Central Register of Controlled Trials, Chinese Biomedical Medicine, China National Knowledge Infrastructure, Wangfang Database, and China Science and Technology Journal Database were searched from database inception to December 31, 2021, to collect Randomized Controlled Trials for PSCI. All of the studies were assessed (according to Cochrane Handbook for Systematic Reviews) and then data were extracted by two researchers separately. Pairwise meta-analysis for direct comparisons was performed using Revman. NMA of Bayesian hierarchical model was performed by WinBUGS and ADDIS. STATA was used to construct network evidence plots and funnel plots.ResultsA total of 55 trials (53 Two-arm trials and 2 Three-arm trials) with 3,092 individuals were included in this study. In the pair-wise meta-analysis, Transcranial Magnetic Stimulation (TMS), Virtual Reality Exposure Therapy (VR), Computer-assisted cognitive rehabilitation (CA), Transcranial Direct Current Stimulation (tDCS), and Acupuncture were superior to normal cognition training in terms of MoCA, MMSE, and BI outcomes. Bayesian NMA showed that the MoCA outcome ranked Acupuncture (84.7%) as the best therapy and TMS (79.7%) as the second. The MMSE outcome ranked TMS (76.1%) as the best therapy and Acupuncture as the second (72.1%). For BI outcome, TMS (89.1%) ranked the best.ConclusionsTMS and Acupuncture had a better effect on improving cognitive function in post-stroke patients according to our Bayesian NMA. However, this conclusion still needs to be confirmed with large sample size and high-quality randomized controlled trials.Registrationhttps://inplasy.com (No. INPLASY202260036).