BACKGROUND: Upon initiation of antiretroviral therapy (ART), 15.7% [95% confidence interval (CI): 9.7% to 24.5%] of tuberculosis (TB)-HIV-coinfected individuals experience paradoxical worsening of their clinical status with exuberant inflammation consistent with immune reconstitution inflammatory syndrome (IRIS). We investigated whether a positive urinary TB lipoarabinomannan (LAM) antigen enzyme-linked immunosorbent assay test before ART initiation was associated with development of paradoxical TB-IRIS. METHODS: In a prospective observational cohort in Mulago Hospital, Kampala, Uganda, we measured pre-ART urinary LAM concentrations in HIV-infected patients on TB treatment. Patients who developed TB-IRIS (according to the International Network for the Study of HIV-associated IRIS case definition) were compared with patients who remained IRIS free for at least 3 months. RESULTS: Twenty-six individuals with TB-IRIS and 64 without IRIS were included in the analysis. The median time to TB-IRIS was 14 days (interquartile range: 11-14 days). Univariate analysis showed that a positive pre-ART urinary LAM test [OR: 4.6 (95% CI: 1.5 to 13.8), P = 0.006] and a CD4 count <50 cells/mL [OR: 21 (95% CI: 2.6 to 169.4), P = 0.004] were associated with an increased risk of TB-IRIS. In multivariate analysis, only a baseline CD4 T-cell count <50 cells/mL was predictive of IRIS (P < 0.004). Sensitivity and specificity of a positive pre-ART urinary LAM test to diagnose IRIS were 80.8% (95% CI: 60.6 to 93.4) and 52.4% (95% CI: 39.4 to 65.1), respectively. CONCLUSIONS: If CD4 T-cell count testing is available, a pre-highly active antiretroviral therapy urinary LAM test has no added value to predict TB-IRIS. When CD4 T-cell count is not available, a positive LAM test could identify patients at increased risk of TB-IRIS.