Background: Chest computed tomography (CT) screening of patients with blunt trauma for thoracic aortic injury is controversial. This study was undertaken to determine whether CT could exclude aortic injury and be used to select patients for aortography. Methods: Computed tomography and aortography were used to evaluate 155 patients with blunt trauma. Computed tomography scans were reviewed separately by four attending radiologists who were unaware of the patients' clinical course and angiographic findings. Results: Eight of 155 patients had aortic injuries requiring operation. Computed tomography scans in five patients were read as positive by all reviewers. One scan was read as positive by three reviewers and as negative by one. Two scans were read as positive by two radiologists and as negative by two. After poor scans were excluded, the combined sensitivity of CT for detecting aortic injury was 88%, specificity was 54%, positive predictive value was 9%, and negative predictive value 99%. Conclusions: The sensitivity of CT scan for indicating the need for aortography is observer dependent. As CT manifestations of aortic injury are often subtle, CT does not reliably exclude aortic injury.