Background: Endoscopically applied methylene blue selectively stains specialized columnar epithelium in Barrett's esophagus. Methods: The diagnostic yield and cost of cancer surveillance in patients with Barrett's esophagus using methylene blue–directed biopsies (MBDB) were compared with surveillance using a “jumbo” random biopsy technique in a prospective, sequential, controlled trial. Esophagogastroduodenoscopy was performed with either MBDB or random biopsy in a randomized sequence. The proportions of various types of epithelia in each biopsy were estimated and dysplasia was graded in a blinded fashion. Results: Forty-three patients with short- (n = 8), limited- (n = 10), and long-segment (n = 25) Barrett's esophagus were studied. Using MBDB technique, the average number of biopsies obtained per patient was significantly lower and the proportion of specialized columnar epithelium in each specimen was significantly higher compared with random biopsy. Dysplasia or cancer was diagnosed in significantly more MBDB specimens (12% vs. 6%, p = 0.004). Despite fewer biopsies per patient using MBDB, dysplasia or cancer was diagnosed in significantly more patients (44% vs. 28%, p = 0.03) than by random biopsy technique. MBDB cost less and detected more cancers than random biopsy. Conclusions: MBDB is a more accurate and cost-effective technique than random biopsy for diagnosing specialized columnar epithelium and dysplasia/cancer, particularly in long-segment Barrett's esophagus. (Gastrointest Endosc 2000;51:560-8.)