Objective: To evaluate the International Network for the Study of HIV-associated IRIS (INSHI) case definitions for tuberculosis (TB)-associated immune reconstitution inflammatory syndrome (IRIS) in a South African cohort.Methods: Prospective cohort of 498 adult HIV-infected patients initiating antiretroviral therapy. Patients were followed LIP for 24 weeks and all clinical events were recorded. Events with TB-IRIS as possible Cause were assessed by consensus expert opinion and INSHI case definition. Positive, negative, and chance-corrected agreement (kappa) were calculated, and reasons for disagreement were assessed.Results: One hundred and two (20%) patients were receiving TB therapy at antiretroviral therapy initiation. Three hundred and thirty-three events were evaluated (74 potential paradoxical IRIS, 259 potential unmasking IRIS). Based on expert opinion, there were 18 cases of paradoxical IRIS associated with TB and/or other opportunistic disease. The INSHI criteria for TB-IRIS agreed in 13 paradoxical cases, giving positive agreement of 72.2%, negative agreement in 52/56 non-TB-IRIS events (92.9%), and kappa of 0.66. There were 19 unmasking TB-IRIS cases based on expert opinion, of which 12 were considered IRIS using the INSHI definition (positive agreement 63.2%). There was agreement in all 240 non-TB-IRIS events (negative agreement 100%) and kappa was 0.76.Conclusion: There was good agreement between the INSHI case definition for both paradoxical and unmasking TB-IRIS and consensus expert opinion. These results Support the use of this definition in clinical and research practice, with minor caveats in its application. (C) 2010 Wolters Kluwer Health vertical bar Lippincott Williams & Wilkins