We prospectively evaluated the diagnostic value of blind brushing of the esophagus via nasogastric tube in 66 patients with human immunodeficiency virus (HIV) infection [acquired immune deficiency syndrome (AIDS) (N = 59), or AIDS-related complex (ARC), (N = 7)] complaining of odynophagia and/or dysphagia. Brushings were obtained between 20 and 35 cm from the incisors. Patients then underwent upper endoscopy with directed brushings and biopsies; esophageal lavage was also done in the first 40 patients. Candida esophagitis was defined as an abnormal appearance of the esophageal mucosa, together with microscopic evidence of pseudohyphae in the endoscopic brushings or invasive candidiasis on biopsy. The presence of oral thrush was also recorded. Candida esophagitis was present in 28 (42%) of the 66 patients. Blind brushings diagnosed candidiasis in 27/28 cases and produced five false positives (sensitivity 96%, specificity 87%). Blind brushing of the esophagus was significantly more sensitive than the presence of oral thrush for the diagnosis of esophageal candidiasis (p = 0.02). Oral thrush was found in only 20/28 cases of Candida esophagitis and in eight patients without Candida (sensitivity 71%, specificity 79%). Esophageal lavage yielded Candida in all cases (sensitivity 100%) but had a poor specificity (64%). We conclude that blind brushing of the esophagus is a rapid, safe, and economical way to diagnose Candida esophagitis in patients with AIDS. This procedure can be performed by primary care physicians with minimal loss of sensitivity and specificity as compared to endoscopy.