A formal statistical overview of all truly randomised trials was undertaken to determine whether antithrombotic therapy is effective and safe in the early treatment of patients with acute stroke. There were 15 completed randomised controlled trials of the value of early antithrombotic treatment in patients with acute stroke. The regimes tested in acute presumed or confirmed ischaemic stroke were: heparin, 10 trials with 1047 patients: oral anticoagulants, one trial with 51 patients: antiplatelet therapy, three trials with 103 patients. Heparin was tested in one trial with 46 patients with acute haemorrhagic stroke. Outcome measures were deep venous thrombosis (confirmed by I125 scanning or venography), pulmonary embolism, death from all causes, haemorrhagic transformation of cerebral infarction, level of disability in survivors. In patients with acute ischaemic stroke, allocation to heparin was associated with a highly significant 81% (SD 8, 2p < 0.00001) reduction in deep venous thrombosis detected by I125 fibrinogen scanning or venogram. Only three trials systematically identified pulmonary emboli, which occurred in 6/106 (5.7%) allocated control vs 3/132 (2.3%) allocated heparin, a non-significant 58% reduction (SD 45.7, 2p > 0.1). There were relatively few deaths in the trials in patients with presumed ischaemic stroke: 94/485 (19.4%) among patients allocated to the control group vs 79/497 (15.9%) among patients who were allocated heparin. The observed 18% (SD 16) reduction in the odds of death was not statistically significant. The least biased estimated of the effect of treatment on haemorrhagic transformation of the cerebral infarct (HTI) comes from trials where all patients were scanned at the end of treatment, irrespective of clinical deterioration; using this analysis, haemorrhagic transformation occurred in 7/102 (6.9%) control vs 8/106 (7.5%) treated, a non-significant 12% increase (SD 56, 2p > 0.1). These data cannot exclude the possibility that heparin substantially increases the risks of HTI. No data on disability in survivors could be obtained. Early heparin treatment might be associated with substantial reductions in deep venous thrombosis (and probably also pulmonary embolism) and possibly a one fifth reduction in mortality (equivalent to the avoidance of 20-40 early deaths per thousand patients treated.) However, the data were wholly inadequate on safety, particularly on the risk of haemorrhagic transformation of the infarct and on the hazards of heparin therapy in patients with known intracerebral haemorrhage. The trials of oral anticoagulants (15 deaths among 57 patients) and antiplatelet therapy (two deaths among 103 patients) were too small to be informative. Much larger randomized trials-comparing aspirin, heparin and the combination of both drugs against control-in patients with acute ischaemic stroke are justified (and several are now planned or underway).