As part of the approval process, new drugs are first studied in controlled clinical trials with carefully selected patients. After gaining approval, however, these drugs are often used in the general community under widely different circumstances. This report explores the effects of such differences on the frequency and severity of side effects using the nonsteroidal anti-inflammatory drugs as a prototype. A review of clinical trials is compared with the accumulated experience at the Stanford and Phoenix Arthritis Centers. Statistical analysis of patient records was accomplished using the American Rheumatism Association Medical Information System Computer Data-Base System. To overcome potential biases in different methods of detecting side effects, a questionnaire was mailed directly to 390 patients, including subjects who were and were not participants in formal drug studies. Agreement between experiment and experience was generally quite good. However, a tendency for severe side effects to occur more frequently in community use than would be predicted from clinical trials was noted.