Design: Retrospective data analysis of human immunodeficiency virus (HIV) infected patients attending an HIV clinic (referral hospital), Cambodia. Chest X-rays (CXRs) were read independently by onsite and offsite physicians.
Results: Data on 881 patients were analyzed (smear-negative = 776, smear-positive = 105). Overall, the prevalence of culture-confirmed pulmonary tuberculosis (PTB) was 17% (150/881, smear-negative = 62/150). For those with any positive culture, a smear-negative case was four times more likely to be mycobacteria other than tuberculosis (MOTT) than Mycobacterium tuberculosis (P = 0.001). Median CD4 count was higher in smear-negative than smear-positive PTB patients (92.5 vs. 42, P = 0.24). Age, symptoms (cough >3 weeks or hemoptysis or fever >1 month) (aOR 2.6, P = 0.02) and an abnormal CXR (offsite reading) (aOR 4.9, P < 0.001) were significant predictors of smear-negative PTB. CXR was no longer significant in the model using the onsite reading (aOR 1.6, P = 0.11). The combination of age >or=30 years plus symptoms had a sensitivity of 100% but a positive predictive value (PPV) of 9%. CXR (offsite), as the next diagnostic test, had a sensitivity of 50% and specificity of 83%. The sensitivity of smear microscopy was 59% and its specificity 97%.
Conclusions: While age and symptoms are useful both in screening smear-negative PTB suspects and in predicting smear-negative PTB cases, they have limited PPV. Given the limitations of smear microscopy, culture is required to diagnose smear-negative disease. Where culture is unavailable, CXR is an important adjunct for diagnosis. However, inaccurate CXR interpretation can impact case detection.