Background Tuberculosis-associated immune reconstitution inflammatory syndrome (TB-IRIS) is an important early complication of antiretroviral therapy (ART) in countries with high rates of endemic TB, but data from South-East Asia are incomplete. Identification of prevalence, risk factors and treatment outcomes of TB-IRIS in Malaysia was sought. Methods: A 3-year retrospective study was conducted among TB-HIV co-infected patients treated at the University of Malaya Medical Centre. Simple and adjusted logistic regressions were used to identify the predictors for TB-IRIS while Cox regression was used to assess the influence of TB-IRIS on long-term CD4 T-cell recovery. Results: One hundred and fifty-three TB-HIV patients were enrolled, of whom 106 had received both anti-TB treatment (ATT) and ART. The median (IQR) baseline CD4 T-cell count was 52 cells L-1 (13-130 cells L-1). Nine of 96 patients (9.4 ) developed paradoxical TB-IRIS and eight developed unmasking TB-IRIS, at a median (IQR) time of 27 (12-64) and 19 (14-65) days, respectively. In adjusted logistic regression analysis, only disseminated TB was predictive of TB-IRIS [OR: 10.7 (95 CI: 1.2-94.3), P=0.032]. Mortality rates were similar for TB-IRIS (n=1, 5.9 ) and non-TB-IRIS (n=5, 5.7 ) patients and CD4 T-cell recovery post-ART was not different between the two groups (P=0.363). Conclusion: Disseminated TB was a strong independent predictor of TB-IRIS in Malaysian HIV-TB patients after commencing ART. This finding underscores the role of a high pathogen load in the pathogenesis of TB-IRIS; so interventions that reduce pathogen load before ART may benefit HIV patients with disseminated TB.