Eradication of systemic tuberculosis (TB) has been limited by neglected populations and the HIV pandemic. Whereas ocular TB often presents as uveitis without any prior evidence of systemic TB, the existing uncertainty in the diagnosis of TB uveitis may perpetuate missed opportunities to address systemic TB.

To examine the clinical features of TB uveitis and the associations with response to antitubercular therapy (ATT).

This retrospective multinational cohort study included patients from 25 ophthalmology referral centers diagnosed with TB uveitis and treated with ATT from January 1, 2004, through December 31, 2014, with a minimum follow-up of 1 year.

Treatment failure, defined as a persistence or recurrence of inflammation within 6 months of completing ATT, inability to taper oral corticosteroids to less than 10 mg/d or topical corticosteroid drops to less than 2 drops daily, and/or recalcitrant inflammation necessitating corticosteroid-sparing immunosuppressive therapy.

A total of 801 patients (1272 eyes) were studied (mean [SD] age, 40.5 [14.8] years; 413 [51.6%] male and 388 [48.4%] female; 577 [73.6%] Asian). Most patients had no known history (498 of 661 [75.3%]) of systemic TB. Most patients had bilateral involvement (471 of 801 [58.8%]). Common clinical signs reported include vitreous haze (523 of 1153 [45.4%]), retinal vasculitis (374 of 874 [42.8%]), and choroidal involvement (419 of 651 [64.4%]). Treatment failure developed in 102 of the 801 patients (12.7%). On univariate regression analysis, the hazard ratios (HRs) associated with intermediate uveitis (HR, 2.21; 95% CI, 1.07-4.55; P = .03), anterior uveitis (HR, 2.68; 95% CI, 1.32-2.35; P = .006), and panuveitis (HR, 3.28; 95% CI, 1.89-5.67; P < .001) were significantly higher compared with posterior distribution. The presence of vitreous haze had a statistically significant association (HR, 1.95; 95% CI, 1.26-3.02; P = .003) compared with absence of vitreous haze. Bilaterality had an associated HR of 1.50 (95% CI, 0.96-2.35) compared with unilaterality (HR, 1 [reference]), although this finding was not statistically significant (P = .07). On multivariate Cox proportional hazards regression analysis, the presence of vitreous haze had an adjusted HR of 2.98 (95% CI, 1.50-5.94; P = .002), presence of snow banking had an adjusted HR of 3.71 (95% CI, 1.18-11.62; P = .02), and presence of choroidal involvement had an adjusted HR of 2.88 (95% CI, 1.22-6.78; P = .02).

A low treatment failure rate occurred in patients with TB uveitis treated with ATT. Phenotypes and test results are studied whereby patients with panuveitis having vitreous and choroidal involvement had a higher risk of treatment failure. These findings are limited by retrospective methods. A prospectively derived composite clinical risk score might address this diagnostic uncertainty through holistic and standardized assessment of the combinations of clinical features and investigation results that may warrant diagnosis of TB uveitis and treatment with ATT.