Background Nonsteroidal anti-inflammatory drugs (NSAIDs) have been associated with clinically significant decreases in haemoglobin dependent and independent of acute bleeding events. Aim To evaluate the incidence and time to a clinically meaningful decrease in haemoglobin in two double-blind, prospective randomised clinical trials comparing NSAIDs in patients with osteoarthritis (OA) or rheumatoid arthritis (RA). Methods In CLASS, patients with OA/RA who were aged ≥18 years and required continuous NSAID treatment were included; patients who were Helicobacter pylori positive and/or using aspirin were not excluded. In contrast, in the CONDOR trial, comparing celecoxib alone to diclofenac sustained release (plus omeprazole), patients were aged ≥60 years or ≥18 years with a history of gastroduodenal ulcer and were H. pylori negative; aspirin or other anti-platelet users were excluded. To make a parallel post hoc analysis we limited our study to 6 months and the populations to only the non-aspirin users in CLASS and those patients receiving either celecoxib or diclofenac. A decrease in haemoglobin of ≥2 g/dL defined the primary end point. Results At 6 months, in the CLASS and CONDOR trials, 1.9% and 2.0% of patients treated with celecoxib and 3.3% and 5.7% of patients treated with diclofenac developed a ≥2 g/dL decrease in haemoglobin, respectively, [CLASS: odds ratio (OR) 1.80 (95% confidence interval (CI), 1.22-2.65) and CONDOR: OR 2.93 (95% CI, 2.06-4.15), respectively]. Conclusion In these two large, independent trials, clinically-meaningful decreases in haemoglobin ≥2 g/dL occurred in a relatively similar fashion over time despite differences in trial designs. © 2011 Blackwell Publishing Ltd.