Background: The aim of this study was to prospectively compare the diagnostic accuracy of videoendoscopy, with and without Lugol staining, for the detection of esophageal cancer in alcoholic or smoking patients older than 40 years. Methods: Daily alcohol and tobacco consumption and overt and latent symptoms were noted. The 158 patients included were examined by videoendoscopy and with Lugol dye. Results: The mean consumption of alcohol and tobacco was 86 ± 49 gm/day for 26 ± 11 years, and 30 ± 18 pack-years, respectively. Twenty-five patients had no symptoms. Before Lugol staining, 12 patients had endoscopically identified cancerous lesions. After Lugol staining, 13 patients had 17 esophageal cancers, 3 of which were high-grade dysplasia. The prevalence of esophageal cancer was 8.2%: 95% CI [4,14]. Dye-free surfaces were significantly larger than the endoscopic patterns observed before Lugol staining (11.6 ± 9.2 cm 2 vs 1.4 ± 1.7 cm 2; p < 0.02). Conclusions: In an alcoholic smoking population, the prevalence of esophageal cancer detected by endoscopy is high and not related to symptoms described by patients. Lugol staining only moderately improves the diagnostic accuracy of videoendoscopy; its main advantage is the assessment of the mucosal extension of esophageal cancer. (Gastrointest Endosc 1997;45:480-4.)