Computed tomography in predicting smear-negative pulmonary tuberculosis in AIDS patients.

BACKGROUND: The correct diagnosis of sputum smear-negative pulmonary tuberculosis in AIDS patients is very important to their therapy. We aimed to assess the value of the computed tomography (CT) and clinical findings in predicting smear-negative pulmonary tuberculosis in AIDS patients. METHODS: A total of 121 AIDS patients suspected of smear-negative pulmonary tuberculosis by clinical and radiographic findings were recruited. Pulmonary tuberculosis was diagnosed in 57 (47.1%) patients. The CT and clinical predictors were selected to diagnose AIDS-related pulmonary tuberculosis through univariate and multivariate Logistic analysis. RESULTS: Multivariate analysis showed that five variables, including weight loss, presence of miliary nodules, necrotic lymph node, lobular consolidation, tree-in-bud sign, were independent predictors of pulmonary tuberculosis in AIDS patients. Predicted scores based on the five variables were used to identify pulmonary tuberculosis. If the predicted score of 3 was taken as the ideal cut-off point in the diagnosis of AIDS-related smear-negative pulmonary tuberculosis, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 84.2%, 81.2%, 80.0%, 85.2%, and 82.6%, respectively. CONCLUSION: The prediction method based on five key factors of clinical and CT findings are useful in guiding the diagnosis of smear-negative pulmonary tuberculosis in AIDS patients.