Recent placebo-controlled clinical trials suggest excessive adverse cardiovascular (CV) events associated with cyclooxygenase (COX)-2 selective nonsteroidal anti-inflammatory drugs (NSAIDs), developed to reduce gastrointestinal irritation associated with nonselective NSAIDs. Subsequent retrospective analyses of observational series and non-CV clinical trials suggest that CV events may occur with modest excess with all NSAIDs, nonselective and COX-2 selective, compared with nonuse of these drugs, may be dose related, and do not differ substantially in frequency among various NSAIDs. However, inadequacy of study designs, controls, and events has precluded definition of the risk-benefit relationship of COX-2 selective and nonselective NSAIDs. Resolution of this problem requires several different types of studies, necessarily including appropriately designed randomized, controlled trials comparing commonly employed nonselective and COX-2 selective NSAIDs in patients expected to benefit (ie, those with symptomatically severe arthritis) who also have coronary occlusive disease so that achievable noninferiority trial size has power sufficient to resolve relatively small differences in adverse CV (and gastrointestinal) event rates. This article explores the goals and possible designs of trials appropriate for defining risk-benefit relationship that must be known for optimal application of NSAID therapy.