OBJECTIVE. To make recommendations for the long-term use of nonsteroidal anti-inflammatory drugs (NSAIDs) in primary care practice, particularly for patients at high risk for NSAID-induced complications. OPTIONS. The use of misoprostol to prevent gastrointestinal ulceration and other unwanted NSAIDs effects was considered. The role of cyclooxygenase-2 (COX-2) versus COX-1 inhibiting agents was also examined. OUTCOMES. Reduction of complications associated with long-term use of NSAIDs. EVIDENCE. Evidence was gathered in late 1995 from published research studies and reviews. Position papers were prepared by faculty and advisory board members and discussed at the Canadian NSAID Consensus Symposium in Cambridge, Ont., Jan. 26 and 27, 1996. VALUES. Recommendations were based on randomized, placebo-controlled clinical trials (level I evidence) and case-control studies (level II evidence) involving NSAID use when such evidence was available. When the scientific literature was incomplete or inconsistent in a particular area, recommendations reflect the consensus of the participants at the symposium (level III evidence). Physicians were recruited from across Canada for their expertise in rheumatology, gastroenterology, epidemiology, gerontology, family practice, and clinical and basic scientific research. BENEFITS, HARMS AND COSTS. Although a reduction in complications due to inappropriate NSAID use should reduce costs of additional investigations, admissions to hospital and time lost from work, definitive cost analysis studies are not yet available. RECOMMENDATIONS. Currently, no NSAID is available that lacks potential for serious toxicity; therefore, long-term use of NSAIDs should be avoided whenever possible, particularly in high-risk patients (e.g., those who are elderly, suffer from hypertension, congestive heart failure, renal or hepatic impairment or volume depletion, take certain concomitant medications or have a history of peptic ulcer disease) (level I evidence). If NSAIDs are to be used in patients with gastric or nephrotoxic risk factors, the lowest effective dose of NSAID should be used (level III evidence); NSAIDs that are weak COX-1 inhibitors may be preferred (level II evidence). In addition, concomitant administration of misoprostol is recommended in patients at increased risk for upper gastrointestinal complications (level I evidence). However, the clinical judgement of the practising clinician must always be part of any therapeutic decision. VALIDATION. These recommendations are based on the consensus of Canadian experts in rheumatology, gastroenterology and epidemiology, and have been subjected to external peer review.