What does IGRA testing add to the diagnosis of ocular tuberculosis? A Bayesian latent class analysis

To evaluate the contribution made to the diagnostic work-up for patients with suspected ocular tuberculosis (TB) by QuantiFERON-TB Gold In-Tube (QFT) tests using latent class analysis model.

A single centre retrospective cohort study. A Bayesian latent class model was constructed on the basis of demographics, phenotypes and test results from patients attending a tertiary referral center in the UK. This estimated the probability of ocular TB for each patient in two versions, first with and then without QFT. The estimated probability of ocular TB was compared with treatment failure.

From a database of 365 patients with clinical signs suggestive of ocular TB, 267 patients who had QFT and complete data were evaluated. Mean age was 45.0 ± 15.4 years with 141 (52.9%) male and 148 (50.5%) of Asian ethnicity. QFT was positive in 208 (70.1%) patients and ATT was instituted in 145 (49.5%) patients with 100 (34.1%) patients also having concurrent systemic corticosteroid therapy. The best estimate of a QFT level separating TB-positive and TB-negative patients was extremely low. This weak discrimination between TB and non-TB groups was reflected in poor positive and negative predictive values for treatment failure.

The latent class model did not successfully predict treatment failure, despite taking all variables into account. The threshold between TB and non-TB in QFT values was implausibly low and removing QFT from the model made prediction slightly worse. A larger prospective study is required to establish the role of all tests, demographics and phenotypes in diagnosis.