Utility of high-resolution computed tomography for predicting risk of sputum smear-negative pulmonary tuberculosis

Background: To diagnose sputum smear-negative pulmonary tuberculosis (PTB) is difficult and the ability of high-resolution computed tomography (HRCT) for diagnosing PTB has remained unclear in the sputum smear-negative setting. We retrospectively investigated whether or not this imaging modality can predict risk for sputum smear-negative PTB. Methods: We used HRCT to examine the findings of 116 patients with suspected PTB despite negative sputum smears for acid-fast bacilli (AFB). We investigated their clinical features and HRCT-findings to predict the risk for PTB by multivariate analysis and a combination of HRCT findings by stepwise regression analysis. We then designed provisional HRCT diagnostic criteria based on these results to rank the risk of FTB and blinded observers assessed the validity and reliability of these criteria. Results: A positive tuberculin skin test alone among clinical laboratory findings was significantly associated with an increase of risk of PTB. Multivariate regression analysis showed that large nodules, tree-in-bud appearance, lobular consolidation and the main lesion being located in Si, S2, and S6 were significantly associated with an increased risk of FTB. Stepwise regression analysis showed that coexistence of the above 4 factors was most significantly associated with an increase in the risk for FTB. Ranking of the results using our HRCT diagnostic criteria by blinded observers revealed good utility and agreement for predicting PTB risk. Conclusions: Even in the sputum smear-negative setting, HRCT can predict the risk of PTB with good reproducibility and can select patients having a high probability of PTB. (C) 2008 Elsevier Ireland Ltd. All rights reserved.