Background:  The lack of a widely accepted, standardized case definition is a major impediment to research on the pathogenesis and management of immune reconstitution inflammatory syndrome (IRIS). Our objective was to evaluate the level of agreement and reasons for disagreement between consensus expert opinion and 2 published IRIS case definitions in a large urban patient cohort initiating ART in Durban, South Africa.

Methods:  We prospectively enrolled and monitored monthly for 6 months, 381 patients initiating ART between December 2006 and August 2007 at 2 hospital-based clinics in Durban. All new or worsening clinical events were assessed by ≥2 physicians and graded 0 to 4 according to likelihood of IRIS (expert opinion). This was compared with 2 existing case definitions, CD1 (French et al, AIDS 2004) and CD2 (Robertson et al, Clin Infect Dis 2006), using percentage positive and negative agreement and κ values to adjust for chance agreement.

Results:  As of September 2007, 437 events were reported in 381 patients. Of these, expert opinion categorized 15 (3%) as definite IRIS; 94 (22%) as probable; 91 (21%) as possible; 143 (33%) as unlikely; and 94 (22%) as definite non-IRIS events. We combined definite and probable as positive IRIS cases (n = 109) and all other grades as negative:  49% were dermatological and 21% genitourinary. Based on CD1 and CD2, there were 54 and 148 IRIS events, respectively. Positive and negative agreement between expert opinion and CD1 was 37% and 96% (κ = 0.39), and with CD2 was 86% and 84% (κ = 0.62). κ for agreement between CD1 and CD2 was 0.25. Positive and negative agreement for all 3 occurred in 35 (8%) and 264 (60%) events respectively. Disagreement between expert opinion and CD1 was greatest for unmasking IRIS (31% positive agreement) and dermatological and genitourinary events (both 26%) compared to paradoxical reactions (49%) and other systems (61%).

Conclusions: This is the first formal evaluation of the utility of existing case definitions for IRIS in a resource-poor setting. There was only modest agreement with existing CD (CD2 >CD1). Existing generic definitions work best for serious opportunistic infections and paradoxical IRIS, but perform poorly for non-serious and dermatological and genitourinary events. The additional requirement of a reduction in viral load (CD1 and 2) or an increase in pathogen-specific immune responses (CD1) further limits their application in resource-poor settings. These findings form the basis for the development of pathogen-specific IRIS case definitions.