Nonsteroidal anti-inflammatory drugs (NSAIDs) are responsible for substantial morbidity and mortality as a result of the complications associated with gastroduodenal ulcers, such as perforation and bleeding. The central mechanism leading to the gastroduodenal toxicity of NSAIDs is their ability to inhibit mucosal prostaglandin synthesis. Recent recognition that there are 2 isoforms of the enzyme cyclooxygenase (COX) responsible for prostaglandin synthesis has enabled the development of drugs capable of sparing the gastric mucosa. The inducible COX-2 enzyme is responsible for some aspects of pain and inflammation in arthritis while the constitutive COX-1 enzyme appears responsible for most of the gastro-protective prostaglandin synthesis in the stomach and duodenum. Drugs selective for COX-2 probably act by binding to a pocket in the enzyme that is present in COX-2 but not in COX-1. As a result, drugs that have little or no COX-1 activity across their therapeutic dosage range have been developed. Two drugs that are claimed to be highly selective or specific in their ability to inhibit COX-2, rofecoxib and celecoxib, are now available on prescription in the US and rofecoxib is available in Europe. Short term volunteer studies of 7 days' duration and patient studies of 6 months' duration have shown these drugs to have a level of gastroduodenal injury that is similar or equivalent to that seen with placebo, whereas high rates of damage and ulceration are seen with nonselective NSAIDs. In addition, there appear to have been fewer perforations, clinical ulcers and bleeds in the phase III clinical trials of these agents, compared with nonselective NSAIDS. However, more experience will be needed before this promise can be confirmed. In addition, COX-2 inhibitors share the adverse effects of NSAIDs outside the gastrointestinal tract that are dependent on COX-2 inhibition.