AIMS: chronic gastrointestinal bleeding is the most common cause of iron deficiency anemia (IDA) in the general population. The objectives of this study were to determine the most frequent gastrointestinal lesions in IDA, the frequency and localization of potentially bleeding lesions, the value of the clinical history in diagnosis, the value of fecal occult blood testing, and the most appropriate diagnostic procedure for these patients. METHODS: we prospectively studied 80 patients older than 40 years with IDA, using upper gastrointestinal tract (GI) endoscopy and colonoscopy, beginning with the former (group A) or the latter (group B) depending on the clinical findings. Barium enema was done when colonoscopy was incomplete or unsatisfactory. If all these tests were negative, conventional barium contrast study of the small intestine and arteriography were done, if necessary. RESULTS: upper GI endoscopy found at least one lesion in 50 patients (72%), 13 in association with a colonic lesion (26%). Colonoscopy detected at least one lesion in 31 patients (45%), among whom 11 had another upper GI lesion (35.5%). Barium enema was positive in 4 out of 24 patients (17%). Barium contrast study of the small intestine detected lesions in 1 out of 7 patients (14%), and arteriography in 1 out of 4 patients (25%). The most common upper GI lesions were of peptic origin (esophagitis in 10, gastroduodenal erosions in 10, and peptic ulcer in 8). Neoplasms (17 cancers and 3 polyps) were the most common colonic lesion. Thirteen out of 38 patients (34%) with a potentially bleeding benign upper GI lesion had another lesion in the colon. The fecal occult blood test was positive in 9 out of 10 patients with colonic cancer and in 5 out of 9 with gastric cancer (74% positive predictive value). Nonsteroid antiinflammatory drug use did not correlate with the presence, location or type of lesion. The reliability of the clinically suspected origin of bleeding was 96% sensitivity, 43% specificity and 74% positive predictive value in group A, and 34%, 93% and 80% respectively in group B. CONCLUSIONS: lesions that cause chronic bleeding were more frequently located in the upper digestive tract than in the colon. There was a high prevalence of neoplasms in patients with IDA. One-third of the patients with a potentially bleeding benign lesion in the upper digestive tract had another lesion in the colon. A positive fecal occult blood test correlated highly with neoplastic lesions, and the presence of blood in the stool did not indicate whether bleeding originated in the upper or lower GI tract. Clinical history was of limited value in predicting the location of a bleeding lesion, but can be suggestive of a prior upper GI tract exploration. These patients need a complete study of both the upper and lower GI tracts. In patients in whom the aforementioned explorations are negative, the small bowel should be studied.