Background:  Immune reconstitution inflammatory syndrome (IRIS) complicates ART initiation in 10 to 40% of patients. Despite ART, patients with advanced HIV remain at increased risk for mortality. Few prospective studies have evaluated risk factors for IRIS and mortality in this setting.

Methods:  A5164 randomized subjects with opportunistic infections (excluding TB) to ART at entry (E-ART) vs ART delayed ≥28 days from entry (D-ART). Associations between IRIS and baseline characteristics, entry opportunistic infections, and laboratory values were evaluated using Wilcoxon and Fisher's tests. Multivariate analyses of the time from ART initiation to IRIS were performed using Cox models with time-varying covariates. Risk factors for mortality were determined using Cox models.

Results:  Of 282 subjects (median CD4 29 cells/µL, HIV viral load 5.1 log copies/mL) enrolled, 262 initiated ART; 23 (8.2%) died. Of the remaining 262 subjects, 20 (7.6%; 95%CI 4.7 to 11.5%) developed IRIS after a median of 33 days (IQR 26 to 72) on ART, 8 of 135 on E-ART and 12 of 127 on D-ART (p = 0.35). No baseline variables were associated with IRIS. There was no difference in IRIS between subjects who received corticosteroids during their opportunistic infections and those who did not—9 of 150 (6.0%) vs 11 of 112 (9.8%), p = 0.35. However, no subjects developed IRIS while still receiving corticosteroids. Log viral load decline at 4 weeks was associated with IRIS (–2.48 vs –2.07, p = 0.04) but change in CD4 was not. In Cox models controlling for viral load change, subjects with fungal infections had a hazard ratio (HR) of 2.9 for developing IRIS (p = 0.02). In univariate analyses, entry mycobacterial infections (HR = 5.9, p <0.01), opportunistic infection number (HR = 2.2 for each additional opportunistic infection, p <0.01), hospitalization (HR = 3.7, p <0.01), low albumin (HR = 3.6, p = 0.02), low hemoglobin (HR = 2.8, p = 0.03), and low CD4 (HR=1.3 for each 10 cells/µL decrement, p=0.02) were associated with mortality. In multivariate analysis, entry mycobacterial infection (HR = 4.6, p <0.01), hospitalization (HR = 3.2, p <0.01), and low CD4 (HR = 1.2 for each 10 cells/µL decrement, p = 0.04) predicted mortality.

Conclusions:  Predicting who will develop IRIS at ART initiation is difficult. E-ART does not lead to an increase in IRIS in non-TB opportunistic infections. Change in viral load, but not CD4, predicted IRIS. Corticosteroids, as used in this study, do not appear to prevent IRIS but may delay its presentation. Risk factors for increased mortality in the pre-ART era including low albumin, hemoglobin and CD4 remain important in patients with an opportunistic infection initiating ART. Mycobacterial infections are associated with high rates of mortality despite ART.