Prevention of nonsteroidal anti-inflammatory drug-induced gastrointestinal mucosal injury. A meta-analysis of randomized controlled clinical trials.

BACKGROUND The policy of prevention of nonsteroidal anti-inflammatory drug (NSAID)-induced gastrointestinal mucosal injury is still a matter of discussion. Indeed, no consensus exists as to whether cotherapy with histamine type 2 (H2) blockers or misoprostol is cost-effective. METHODS Placebo-controlled randomized clinical trials on the use of H2 blockers or misoprostol, as preventive agents (published between) January 1970 and December 1994), were identified through MEDLINE and reference lists from literature reviews. Crude rates of endoscopic lesions with short-term (< 2 weeks) and long-term (> 4 weeks) NSAID treatment were systematically assessed by 3 independent observers based on the intention-to-treat principle. The method of DerSimonian and Laird was used for pooling data. Heterogeneity was evaluated by using the Q statistic and the plots described by L'Abbe and colleagues. RESULTS Twenty-four trials met the criteria for entry into the study. Gastric ulcer was found to be significantly reduced by misoprostol-both in short-term (pooled rate difference [RD], -13%, 95% confidence interval [CI], -26% to -1%) and long-term (RD, -8%; 95% CI, -18% to -1%) NSAID treatment-but not by H2 blockers. The risk for duodenal ulcer was significantly reduced by H2 blockers (RD, -2%; 95% CI, -5% to -0.2%) and by misoprostol (RD, -3%; 95% CI, -6% to -0.1%) in long-term but not in short-term administration. CONCLUSIONS The use of misoprostol, but no that of H2 blockers, was beneficial in the prevention of NSAID-induced gastric ulcers. The number of patients to be treated to prevent 1 gastric ulcer with short- and long-term NSAID treatment is 11 and 15, respectively, for an intermediate baseline risk of 10%. Misoprostol and H2 blockers were beneficial in the long-term prevention of duodenal ulcers; misoprostol or H2 blockers in the short-term prevention of duodenal ulcers remains to be confirmed.