This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
OBJECTIVE: Prior metaanalyses of the risk of upper gastrointestinal (GI) complications associated with nonsteroidal antiinflammatory drugs (NSAID) have focused on the published English language epidemiologic literature and/or only a portion of the relevant evidence, restrictions that are now known to be associated with bias in metaanalysis. We synthesized the published and unpublished evidence to determine the least biased estimates of the risks of perforations, ulcers, and bleeds (PUB) associated with NSAID use from all study designs and all languages. METHODS: Data sources: Using MEDLINE, EMBASE, HEALTHSTAR, and BIOSIS, we searched for English and non-English language studies of NSAID from 1966-1998 reporting primary data on GI complications. We obtained unpublished data from the US Food and Drug Administration (FDA) new drug application (NDA) reviews. NDA were hand searched to identify unpublished studies with inclusion criteria identical to those used for published reports. Study selection: Studies had to assess the use of oral NSAID for more than 4 days duration in subjects > 18 years of age and report on the clinically relevant upper GI outcomes of PUB. RESULTS: Two reviewers evaluated 4881 published titles and identified 13 NSAID versus placebo randomized clinical trials and 3 previously unpublished FDA placebo controlled randomized controlled trials, 9 cohort studies, and 23 case control studies sufficiently clinically homogeneous to pool. Two reviewers extracted data about study characteristics and study quality. Data synthesis: The majority of clinical trials were of good quality, but observational studies had methodologic limitations. The pooled odds ratio (OR) from 16 NSAID versus placebo clinical trials, comprising 4431 patients, was 5.36 (95% CI: 1.79, 16.1). The pooled relative risk of PUB from 9 cohort studies comprising over 750,000 person-years of exposure was 2.7 (95% CI: 2.1, 3.5). The pooled OR of PUB from 23 case control studies using age and sex matching, representing 25,732 patients, was 3.0 (95% CI: 2.5, 3.7). Data were insufficient to justify subgroup analyses stratified by age, comorbid conditions, drug, or dose. CONCLUSION: These data support an association between the use of NSAID and serious upper GI complications, including estimates from different study designs. Prior pooled estimates about the effect of patient and drug variables on increased risk must be viewed with caution.