Abstract
BACKGROUND
Tuberculosis (TB) and sarcoidosis commonly present with pulmonary and ocular involvement. Routine chest radiography (CXR) is recommended in the workup for suspected intraocular TB (IOTB) or intraocular sarcoidosis (IOS); however, data on the utility of CXR in this setting are lacking.
METHODS
A post-hoc analysis was performed of a prospectively collected data set comprising 104 patients with uveitis of unknown cause. A pulmonologist and thoracic radiologist, blinded to the final diagnosis, independently reported these CXRs as being in keeping with TB or sarcoidosis.
RESULTS
CXRs were reported as normal/indeterminate (n = 88), probable/previous TB (n = 9) or possible/probable sarcoidosis (n = 8), with a 96% inter-observer concordance. CXRs were more often abnormal in IOS than in IOTB (5/8 vs. 5/34, P = 0.01). CXR had a sensitivity of 14.7%, specificity of 94.3%, positive predictive value (PPV) of 55.6% and negative predictive value (NPV) of 69.5% for IOTB, compared with a sensitivity of 62.5%, specificity of 96.9%, PPV of 62.5% and NPV of 96.9% for IOS. Overall diagnostic accuracy was 54.5% (58.1% in human immunodeficiency virus [HIV] positive participants) in the case of IOTB and 79.9% for IOS.
CONCLUSION
CXR had high specificity and NPV for IOS, and poor overall diagnostic accuracy for IOTB, including in the HIV-positive population.
Collapse