Sixteen trials were eligible for inclusion in the review. The trials included 42,983 patients (sample sizes ranging from 115 to 18,113) representing 93,131 person-years of follow-up. The median follow-up time was two years. One trial was of lower quality (Jadad score of 2); all others were of moderate or good quality (Jadad score of 3 or more).
The incidence rate of major gastrointestinal bleeding ranged from 0 to 8.9% per 1,000 person-years, or 0 to 1.4% if trials allowing off-protocol aspirin use were excluded.
Adjusted-dose vitamin K antagonist treatment was associated with a higher incidence of major gastrointestinal bleeding than placebo (OR 3.21, 95% CI 1.32 to 7.82; four trials) or aspirin (OR 1.92, 95% CI 1.08 to 3.41; seven trials).
Adjusted-dose vitamin K antagonist plus aspirin was associated with a higher incidence of bleeding than vitamin K antagonists alone (OR 2.66, 95% CI 1.05 to 6.74; two trials) or aspirin alone (OR 4.72, 95% CI 1.35 to 16.49; one trial).
Aspirin was associated with a higher incidence of bleeding than placebo but the result was not statistically significant (OR 3.23, 95%CI 0.56 to 18.66; three trials).
Dabigatran was found to be worse (increased chance of bleeding) than adjusted-dose vitamin K antagonists (OR 1.30, 95% CI 1.06 to 1.59) and aspirin plus clopidogrel worse than aspirin alone (OR 1.93 95% CI 1.46 to 2.56), but these results were both based on one trial.
No other significant relationships were found in the other six drug comparisons considered.
Only one analysis had substantial heterogeneity (Ι²>50%).
No evidence of publication bias was identified.