OBJECTIVES: To investigate the benefits and risks of using midazolam for sedation during upper gastrointestinal endoscopic procedures in older persons. DESIGN: Randomized, double-blind, placebo-controlled study. SETTING: A 304-bed geriatric university hospital. PATIENTS: Sixty-five geriatric inpatients (mean age 84 +/- 7) undergoing gastroscopy. INTERVENTION: Sedation with either midazolam (30 microg/kg IV) or saline (placebo). All patients received supplemental oxygen during the procedure (2 L/minute). MEASUREMENTS AND RESULTS: Patients' recall of their tolerance to the exam (categorical scale) and pain score were significantly in favor of midazolam at 2 and 24 hours. Multivariate analysis at 2 hours showed that midazolam increased the probability of good tolerance (odds ratio (OR) = 19.3; 95% confidence interval (CI) 2.2-170.4, P = .008). Circumstantial amnesia occurred at 24 hours in 84% (midazolam) versus 27% (placebo) (P < .001). With midazolam, mean sedation time was 83 +/- 13 minutes and mean arterial pressure (MAP) was about 10 mm Hg lower without clinically significant hypotension. Hypoxemia (SaO2 < 92%) was more frequent in the midazolam group after endoscopy (44% vs. 18%, P = .033), but no major desaturation was observed. Cognitive function (Mini-Mental State Exam, MMSE) was similar before and 2 and 24 hours after the exam in both groups. Acute confusion was observed in two patients (1 midazolam, 1 placebo). In multivariate analysis, midazolam was associated with a higher risk of hypoxemia after endoscopy (OR = 3.5; 95% CI 1.1-10.8, P = .029) but not of confusion. CONCLUSIONS: Under adequate surveillance, the benefits in terms of tolerance to the procedure of low-dose midazolam for upper gastrointestinal endoscopic sedation outweigh the risks in older people.