The safety and efficacy of oral anticoagulation for prevention of major adverse cardiovascular events in people with atrial fibrillation and spontaneous intracranial haemorrhage are uncertain. We planned to estimate the effects of starting versus avoiding oral anticoagulation in people with spontaneous intracranial haemorrhage and atrial fibrillation. In this prospective meta-analysis, we searched bibliographic databases and trial registries using the strategies of a Cochrane systematic review (CD012144) on June 23, 2023. We included clinical trials if they were registered, randomised, and included participants with spontaneous intracranial haemorrhage and atrial fibrillation who were assigned to either start long-term use of any oral anticoagulant agent or avoid oral anticoagulation (ie, placebo, open control, another antithrombotic agent, or another intervention for the prevention of major adverse cardiovascular events). We assessed eligible trials using the Cochrane Risk of Bias tool. We sought data for individual participants who had not opted out of data sharing from chief investigators of completed trials, pending completion of ongoing trials in 2028. The primary outcome was any stroke or cardiovascular death. We used individual participant data to construct a Cox regression model of the time to the first occurrence of outcome events during follow-up in the intention-to-treat dataset supplied by each trial, followed by meta-analysis using a fixed-effect inverse-variance model to generate a pooled estimate of the hazard ratio (HR) with 95% CI. This study is registered with PROSPERO, CRD42021246133. We identified four eligible trials; three were restricted to participants with atrial fibrillation and intracranial haemorrhage (SoSTART [NCT03153150], with 203 participants) or intracerebral haemorrhage (APACHE-AF [NCT02565693], with 101 participants, and NASPAF-ICH [NCT02998905], with 30 participants), and one included a subgroup of participants with previous intracranial haemorrhage (ELDERCARE-AF [NCT02801669], with 80 participants). After excluding two participants who opted out of data sharing, we included 412 participants (310 [75%] aged 75 years or older, 249 [60%] with CHA2DS2-VASc score ≤4, and 163 [40%] with CHA2DS2-VASc score >4). The intervention was a direct oral anticoagulant in 209 (99%) of 212 participants who were assigned to start oral anticoagulation, and the comparator was antiplatelet monotherapy in 67 (33%) of 200 participants assigned to avoid oral anticoagulation. The primary outcome of any stroke or cardiovascular death occurred in 29 (14%) of 212 participants who started oral anticoagulation versus 43 (22%) of 200 who avoided oral anticoagulation (pooled HR 0·68 [95% CI 0·42-1·10]; I2=0%). Oral anticoagulation reduced the risk of ischaemic major adverse cardiovascular events (nine [4%] of 212 vs 38 [19%] of 200; pooled HR 0·27 [95% CI 0·13-0·56]; I2=0%). There was no significant increase in haemorrhagic major adverse cardiovascular events (15 [7%] of 212 vs nine [5%] of 200; pooled HR 1·80 [95% CI 0·77-4·21]; I2=0%), death from any cause (38 [18%] of 212 vs 29 [15%] of 200; 1·29 [0·78-2·11]; I2=50%), or death or dependence after 1 year (78 [53%] of 147 vs 74 [51%] of 145; pooled odds ratio 1·12 [95% CI 0·70-1·79]; I2=0%). For people with atrial fibrillation and intracranial haemorrhage, oral anticoagulation had uncertain effects on the risk of any stroke or cardiovascular death (both overall and in subgroups), haemorrhagic major adverse cardiovascular events, and functional outcome. Oral anticoagulation reduced the risk of ischaemic major adverse cardiovascular events, which can inform clinical practice. These findings should encourage recruitment to, and completion of, ongoing trials. British Heart Foundation.