Background: Establishing a diagnosis of viral esophagitis in patients with human immunodeficiency virus (HIV) infection has important clinical relevance. However, the number of biopsies required to diagnose viral esophagitis is currently unknown. Methods: Over a 34-month period, all HIV- infected patients with esophageal ulcer underwent 10 biopsies of the largest and/or most accessible lesion, primarily from the ulcer base. The first 3 specimens were placed in one formalin container, the second 3 in another, and 4 additional specimens in the third. Standard histopathologic methods were employed, as well as in situ hybridization or immunohistochemical studies in most patients, and viral cytopathic effect was defined using previously proposed criteria. Patients were then treated on the basis of the results of the initial biopsy specimens with both clinical and endoscopic follow-up. Results: One hundred HIV-infected patients with esophageal ulcer were studied. Cytomegalovirus (CMV) was considered etiologic in 50 patients. Of these 50 patients, the first three biopsy specimens were sufficient to diagnosis CMV in 40 (80%). In 5 patients (10%), the first two sets were negative with only the third set of biopsies positive. Similarly, of the 4 patients with simultaneous CMV and herpes simplex virus (HSV) esophagitis, three sets of biopsy specimens were required for diagnosis of both agents in 3 patients. HSV esophagitis alone was found in 2 patients; diagnostic viral inclusions were present in the first 3 biopsies in each patient. Thirty-five patients had HIV-associated idiopathic esophageal ulcer; only one of these patients was misdiagnosed. Conclusions: At least 10 biopsies may be required to exclude viral esophagitis in HIV-infected patients. If biopsy specimens are adequate and no evidence of viral cytopathic effect has been found, the patient may be treated on the basis of the results of the initial clinical, endoscopic, and pathologic findings with close clinical follow-up rather than repeat endoscopy.