Background:  Endoscopic databases are increasingly used for clinical research, but their validity as research instruments has not been assessed. We compared the accuracy of endoscopic indications recorded in an endoscopic database with patient symptom questionnaires. Methods:  All patients infected with the human immunodeficiency virus referred to the outpatient gastroenterology practice were prospectively evaluated using recognized symptom questionnaires. For patients undergoing esophagogastroduodenoscopy, the procedure indications recorded in the endoscopic database and the patient’s self-reported symptom scores were compared. Results: Ninety-three patients were evaluated. The symptoms of nausea/vomiting, diarrhea, and anorexia were highly predictive for the presence of these symptoms on the patient questionnaires. The symptoms of dyspepsia/abdominal pain did not predict well the presence of these symptoms on the questionnaire. Patients reported frequent and severe symptoms that were not recorded as indications for the procedures. The overall agreement (kappa statistic) was highly variable, from slight (κ = 0.07 for anorexia) to moderate (κ = 0.44 for diarrhea). Conclusions: Endoscopic indications are variably associated with self-reported symptom scores. These findings raise concerns about using some endoscopic database indications as accurate representations of patients’ symptoms. Until performance characteristics of a given database are known, symptom-oriented research should use validated questionnaires whenever possible. (Gastrointest Endosc 2000;51:129-33.)