The object of this study was to quantify the incidence of serious upper gastrointestinal complications among nonusers of nonsteroidal anti-inflammatory drugs (NSAIDs). Systematic review of epidemiologic studies published between 1980 and 2000 that provided data on the incidence of upper gastrointestinal bleeding, perforation, or other upper gastrointestinal tract event resulting in death, hospitalization, or visit to a specialist among nonusers of nonsteroidal anti-inflammatory drugs. The authors calculated summary incidence rates and analyzed heterogeneity among results according to outcome definition, population characteristics, and methodology of primary studies. Forty-one population-based studies were reviewed, and 12 were included in the final analysis. Differences in outcome definitions accounted for much of the variability in incidence rate estimates reported in original studies. The pooled incidence rate estimate among nonusers of prescription NSAIDs per 1,000 person-years was 0.1 (95% confidence interval: 0.04-0.23) for perforations alone, 0.8 (0.58-0.68) for bleeding lesions alone, 0.9 (0.66-1.27) for bleeding or perforated lesions, and 1.0 (0.83-1.15) for serious gastrointestinal ulcer (complicated or without bleeding). Rates increased with age, and were approximately twice as high in men than in women. Epidemiologic studies based on automated data may slightly under- or overestimate the true incidence rate among nonusers of NSAIDs. Overall, the incidence rate of serious upper gastrointestinal complications was in the order of 1 per 1,000 person-years among nonusers of prescription NSAIDs.