Esophageal cancer is a common cause of dysphagia and upper endoscopy is the accepted standard for making the diagnosis; however, the accuracy of endoscopy is not known. The purpose of this study is to determine the sensitivity of upper endoscopy in making the diagnosis of esophageal cancer in clinical practice. All patients with a new diagnosis of esophageal cancer from 1997 to 2001 in the Tumor Registry of Wake Forest University Baptist Medical Center were identified. The medical records were reviewed to identify all patients who had undergone a previous endoscopy within two years that failed to diagnose esophageal cancer. The reason for failure was recorded. One hundred ten new cases of esophageal carcinoma were identified, and ten patients had had a previous false-negative endoscopy within two years for a sensitivity of 90.9% in clinical practice. The reasons for the failure of endoscopy were (1) lesion not seen in seven patients, (2) lesion seen and biopsied with benign histology in two patients, and (3) lesion seen but felt to be benign and not biopsied in one patient. Given the uncertain natural history of esophageal cancer, the data were also analyzed using a one-year window; this resulted in a sensitivity of 94.5%. The sensitivity of upper endoscopy in the diagnosis of esophageal cancer in general clinical practice is 90.9% using a two-year window and 94.5% using a one-year window. Understanding the reasons for the failure of endoscopy may allow us to improve the practice of endoscopy.