Diagnosis of thoracic complications in AIDS: accuracy of CT

The purpose of this study was to assess the accuracy of CT interpretation in the diagnosis of thoracic complications of AIDS. If CT is accurate, more invasive diagnostic procedures can be avoided.

We reviewed the CT scans of 102 patients who had AIDS with proved thoracic complications and 20 HIV-positive patients without active intrathoracic disease. The scans were independently assessed by two observers from a different institution who did not know the clinical or pathologic data. The observers listed the three most likely diagnoses in order of probability and recorded the degree of confidence in their first-choice diagnosis. Imaging diagnoses were compared with final diagnoses based on results of sputum cultures, bronchoscopy, fine-needle aspiration, mediastinoscopy, open lung biopsy, autopsy, gallium scans, other clinical data, and extensive follow-up.

Nineteen of the 20 cases in which no active disease was present were correctly identified by one observer, and 18 were correctly identified by the other. All 102 cases of active disease were correctly identified as abnormal by one observer, and 101 cases were identified as abnormal by the second observer. The correct first-choice diagnosis, regardless of the degree of confidence, was made in 66% of the cases. A confident diagnosis was made in 48% of all cases, and the observers were correct in 92% of those cases. There was good agreement between the two observers in the confident diagnosis based on CT (kappa statistic = 0.71). The interpretations of CT scans were most often accurate in the confident diagnosis of pneumocystis pneumonia (94%) and Kaposi's sarcoma (90%) and in the exclusion of active thoracic disease (93%).

Certain findings on chest CT scans allow confident diagnoses of specific complications in patients with AIDS. CT-based diagnosis may preclude more invasive diagnostic procedures in selected cases.