This multicentre study sought to estimate the incidence of upper gastrointestinal (UGI) bleeding in haemophiliacs and its relationship to use of non-steroidal anti-inflammatory drugs (NSAIDs). Cox models were used to estimate relative hazards (RH) with 95% confidence intervals (CI) for postulated risk factors. Conditional logistic regression and stored sera were used to assess UGI bleeding risk with Heliobacter pylori seropositivity in cases compared with closely matched controls. During a mean of 17.4 months (range 2-34), 2285 participants, ages 13-89 (mean 36.5) were followed for 3309 person-years (py). Forty-two experienced a UGI bleeding event (incidence 1.3 per 100 py), most from ulcer (11), gastritis (four), varices (five) and Mallory Weiss tears (eight). RH was significantly increased with traditional NSAID use for <1 month (OR: 3.66; 95% CI: 1.1-11.9), but not with coxibs use. RH was significantly and independently increased with age >46 years (3.5; 95% CI: 1.1-10.6) and hepatic decompensation (4.4; 95% CI: 1.7-11.6). Likelihood of bleeding was substantially but not significantly increased (OR: 4.6; 95% CI: 0.3-83.9) with H. pylori seropositivity. These findings suggest that coxibs are a safer alternative than traditional NSAIDs in the treatment of haemophilic arthropathy.