|Home | About | Journals | Submit | Contact Us | Français|
Background.The incidence of immune reconstitution inflammatory syndrome (IRIS) when antiretroviral therapy (ART) is initiated after an AIDS-defining opportunistic infection (OI) is uncertain and understudied for the most common OIs.
Methods.We examined patients in the University of Washington Human Immunodeficiency Virus Cohort initiating potent ART subsequent to an AIDS-defining OI. IRIS was determined through retrospective medical record review and adjudication using a standardized data collection process and clinical case definition. We compared demographic and clinical characteristics, and immunologic changes in patients with and without IRIS.
Results.Among 196 patients with 260 OIs, 21 (11%; 95% confidence interval, 7%–16%) developed paradoxical IRIS in the first year on ART. The 3 most common OIs among study patients were Pneumocystis pneumonia (PCP, 28%), Candida esophagitis (23%), and Kaposi sarcoma (KS, 16%). Cumulative 1-year incidence of IRIS was 29% (12/41) for KS, 16% (4/25) for tuberculosis, 14% (1/7) for Cryptococcus, 10% (1/10) for Mycobacterium avium complex, and 4% (3/72) for PCP. Morbidity and mortality were highest in those with visceral KS-IRIS compared with other types of IRIS (100% [6/6] vs 7% [1/15], P<.01). Patients with mucocutaneous KS and tuberculosis-IRIS experienced greater median increase in CD4+ cell count during the first 6 months of ART compared with those without IRIS (+158 vs +53cells/μL, P=.04, mucocutaneous KS; +261 vs +113, P=.04, tuberculosis).
Conclusions.Cumulative incidence and features of IRIS varied depending on the OI. IRIS occurred in >10% of patients with KS, tuberculosis, or Cryptococcus. Visceral KS-IRIS led to considerable morbidity and mortality.
Potent combination antiretroviral therapy (ART) has dramatically decreased mortality and the incidence of AIDS-defining opportunistic infections (OIs) among persons infected with human immunodeficiency virus (HIV) [1, 2]. Nevertheless, OIs continue to occur, particularly in patients who present late for medical care or delay ART initiation [3, 4]. Profoundly immunodeficient individuals who develop OIs and then initiate ART may experience a paradoxical worsening or recurrence of OI symptoms, termed the immune reconstitution inflammatory syndrome (IRIS). IRIS has been described as an intense inflammatory reaction to foreign antigens after rapid immunologic recovery  that is associated with a greater burden of disease, a low CD4+ T-cell nadir, and, for Mycobacterium tuberculosis, a shorter time period between OI diagnosis and initiation of ART [6–17]. Reports of the occurrence rates for IRIS range from 8% to over 30%; however, most previous studies have examined IRIS related to specific OIs [7–9, 11, 16, 18, 19]. Thus, it is unclear what proportion of patients with a broad range of OIs who are started on ART go on to develop IRIS.
IRIS has been described in association with tuberculosis, Mycobacterium avium complex (MAC), cytomegalovirus (CMV) retinitis, and disseminated Cryptococcus. Little is known about the clinical characteristics and pathogenesis of IRIS among patients with other common OIs such as Candida esophagitis, Pneumocystis pneumonia (PCP), and Kaposi sarcoma (KS) . Only a few studies have directly compared clinical characteristics and outcomes of IRIS across more than 1 AIDS-defining OI [6, 8, 10, 12, 13, 17]. We conducted this study to better understand predictors, outcomes, and immunologic changes associated with IRIS among a cohort of patients with a full spectrum of prior AIDS-defining OIs.
This retrospective cohort study was conducted on the University of Washington (UW) HIV Cohort, a longitudinal observational study of HIV-infected patients receiving care in the UW Harborview Medical Center HIV Clinic and the UW Medical Center Virology Clinic from 1995 to the present, using standardized data collection methods as previously described . We examined all patients in the UW HIV Cohort who: (1) had an initial visit after 1 January 1996, (2) initiated their first ART regimen (3 drugs containing at least 1 nonnucleoside reverse transcriptase inhibitor or protease inhibitor), (3) had HIV viral load suppression of at least 1 log10 copies/mL within the first 6 months after initiating ART, (4) were diagnosed with at least 1 Centers for Disease Control and Prevention AIDS-defining OI, and (5) exhibited stability or improvement in the course of their OI prior to initiation of ART to exclude patients with worsening OIs due to treatment failure or resistance. AIDS-defining diagnoses of recurrent pneumonia, salmonella septicemia, and chronic herpes simplex virus were excluded because of the difficulty in distinguishing IRIS from recurrent disease.
The primary investigator (C. A.) reviewed comprehensive medical records for all patients meeting study criteria, and identified those who experienced clinical worsening or recurrence of a prior OI during the first year after ART initiation, or until discontinuation of therapy. For each case patient, a set of documents from the medical record was compiled for adjudication that contained the following: (1) clinician progress notes, including the 2 most recent visits prior to the initiation of ART and all visits during the worsening or recurrent OI event; and (2) laboratory data (excluding CD4+ cell counts and HIV RNA levels); endoscopic, radiographic, and pathologic reports; and physical exam findings performed during the worsening or recurrent OI event.
The clinical case definition for IRIS included at least a 1 log10 drop in HIV viral load as a measure of adherence and response to ART, similar to recently validated criteria , but different from older studies in not requiring a rise in CD4+ cell count. The clinical criteria required to meet the definition of paradoxical IRIS are presented in Table 1. Cases of unmasking IRIS (underlying OI revealed and diagnosed with IRIS event) were excluded.
Two experienced HIV physicians (R. H. and S. D.) reviewed the set of deidentified medical record documents for each case to determine whether clinical worsening or recurrence of the OI was due to IRIS. If the 2 reviewers did not concur, then a third independent reviewer (H. C.) evaluated the document set using the same criteria to make the final determination. Definitive determination of a case of paradoxical IRIS required the agreement of 2 reviewers. Reviewers were blinded to CD4+/CD8+ cell counts, HIV RNA levels, date of prior OI diagnoses, and any documentation suggesting a diagnosis of IRIS in the medical record.
We calculated cumulative incidence and 95% confidence intervals (CIs) as a percentage of all patients who developed IRIS over the first year of ART. Bivariate comparisons of dichotomous demographic, clinical, and treatment characteristics were performed using the Wilcoxon signed rank test for continuous variables and Fisher exact test for categorical variables. Adjusted analysis was performed using logistic regression. Factors included in the multivariable analysis were type of OI, multiple or widespread OIs, nadir CD4+ cell count, and CD4+ cell count increase in the first 6 months of ART. The Kaplan-Meier survival curves were used to display the time from starting ART to IRIS. Test results were considered statistically significant for a 2-sided P value of <.05.
A total of 196 patients had an improving or stable OI, initiated ART, and then had at least 1 log10 copies/mL drop in HIV RNA (Figure 1). Of the 260 OIs diagnosed in these 196 patients, the 3 most common were PCP (28%), Candida esophagitis (23%), and KS (16%) (Table 2). Within the first year of ART, 50 of 196 (26%) patients had clinical symptoms of OI worsening or recurrence (33/50 [66%] and 17/50 [34%], respectively). After independent case review, the 2 primary reviewers were in agreement regarding whether or not a patient had IRIS as per study criteria in 41/50 (82%) of cases. The remaining 9 cases required a third reviewer to make a final IRIS determination. IRIS was confirmed in 21 of the 50 reviewed cases of OI worsening or recurrence; 11% (95% CI, 7%–16%) of the entire cohort of 196 patients. For the 29 cases not classified as IRIS, reviewers determined that 11 (38%) had a different infection or disease (including 1 unmasking IRIS), 9 (31%) were of unknown etiology or lacked sufficient information to confirm IRIS, 8 (28%) had progression of their prior OI without an atypical or exaggerated inflammatory syndrome, and 1 (3%) was due to medication toxicity.
Paradoxical IRIS occurred in the first year of ART among patients with KS (12/41 [29%]; 95% CI, 16%–46%), tuberculosis (4/25 [16%]; 95% CI, 5%–36%), disseminated Cryptococcus (1/7 [14%]; 95% CI, 0.4%–58%), disseminated MAC (1/10 [10%]; 95 % CI, 0.3%–45%), and PCP (3/72 [4%]; 95% CI, 0.9%–12%) (Table 2). Individual IRIS cases are presented and summarized in Table 3. As observed in Figure 2, two-thirds (14/21) of IRIS cases occurred during the first 3 months of ART. Median time from OI diagnosis to initiation of ART was longer for mucocutaneous KS-IRIS cases than for other types of IRIS (144 vs 36 days, P=.47); however, this difference was not significant. Patients with Tuberculosis-IRIS developed IRIS symptoms after a shorter median time on ART compared with other types of IRIS (15 vs 73 days, P=.07), and Tuberculosis-IRIS cases were also more often treated with either nonsteroidal anti-inflammatory drugs or corticosteroids compared to other IRIS cases (3/4 [75%] vs 2/17 [12%], P =.01). Fewer patients with mucocutaneous KS-IRIS required hospitalization compared to those with other types of IRIS (1/6 [17%] vs 6/15 [40%], P=.31). Among those with visceral KS-IRIS, the median duration of IRIS symptoms (206 vs 96 days, P=.21), and the proportion of cases with IRIS-related morbidity or mortality (6/6 [100%] vs 1/15 [7%], P<.01) were higher compared with all other types of IRIS.
Baseline characteristics of study patients and bivariate associations among those with and without paradoxical IRIS are shown in Table 4. In unadjusted analyses, there was no difference in risk of IRIS according to age, sex, race, HIV transmission risk factor, year or type of ART initiated, median time from OI diagnosis to ART initiation, median nadir CD4+ cell count, or median baseline HIV RNA level. Risk of IRIS was significantly increased for patients with KS (OR 6.7 [95% CI 2.6–17.4, P<.01]) and those with multiple or widespread OIs (odds ratio [OR] 2.7 [95% CI 1.1–6.7, P = .05]). Risk of IRIS was significantly lower for patients with PCP (OR 0.3 [95% CI 0.1–0.99, P=.04]) and Candida esophagitis (0/21 [0%] vs 57/175 [33%], P<.01) when compared with all other OIs.
In adjusted analysis, we observed a significantly higher risk of IRIS only for patients with KS compared with all other OIs (OR 7.5 [95% CI 2.6–21.5, P<.01]).
Median CD4+ and CD8+ cell count increases and plots of CD4+ cell count measurements in the first 6 months of ART are presented in Table 5 and Figure 3. We found higher median increases in CD4+ (+158 vs +53 cells/μL, P=.04) and CD8+ (+693 ves +265 cells/μL, P=.04) cell counts for patients with mucocutaneous KS who developed KS-IRIS compared with those who did not. No significant differences were found for patients with visceral KS; lower median CD4+ (+15 vs +39 cells/μL, P=.25) and CD8+ (+47 vs +229 cells/μL, P=.42) cell count increases occurred in those with KS-IRIS compared to those without. Similar to patients with mucocutaneous KS-IRIS, patients with Tuberculosis-IRIS also experienced greater median increases in CD4+ cell count (+261 vs +113, P=.04), but no difference in CD8+ cell count increase (+221 vs +281, P=.72). Among patients with PCP, there was no association between increases in either CD4+ cell count (+88 vs +133, P=1.0) or CD8+ cell count (+372 vs +431, P=.98) and the development of IRIS.
We found a cumulative incidence of paradoxical IRIS within the first year of ART of 11% (95% CI, 7%–16%), similar to that reported in prospective studies [11, 13, 14, 17, 18]. Distinct from previous studies, we examined the risk of IRIS across many AIDS-defining OIs, including common conditions such as Candida esophagitis, PCP, and KS. The incidence of IRIS for individual OIs varied greatly and often differed from that reported in previous investigations. We observed no cases of IRIS due to CMV disease or Candida esophagitis, whereas 29% of patients with KS developed IRIS in the first year of ART, in contrast to previous studies that reported a combined incidence of immune recovery CMV uveitis of 37.7% and only 6.4% for KS-IRIS . These conflicting findings likely reflect lack of a standard definition of IRIS, temporal and regional differences in OI and HIV treatments, and OI severity, as well as variability in study size, demographics, and follow-up.
We observed a remarkably high rate of IRIS in patients with KS (29%; 95% CI 16%–46%), similar to that reported in a study from sub–Saharan Africa . The higher cumulative incidence of IRIS in patients with visceral KS compared with patients with mucocutaneous KS (6/14 [43%; 95% CI 18%–71%] vs 6/27 [22%; 95% CI 9%–42%], respectively) is consistent with the hypothesis that higher OI disease burden increases the risk of IRIS [7, 10, 18, 22, 23]. An explanation for the high incidence of KS-IRIS might be that worsening or spreading KS lesions are easier to detect clinically and that our case definition was more sensitive than those used in previous studies. It is also possible that some cases identified as KS-IRIS represented KS disease progression. This highlights the need for a standardized definition of IRIS and KS-IRIS in particular.
KS-IRIS was associated with significant morbidity in patients with visceral disease. Three out of 6 patients with visceral KS-IRIS died within 6 months and the other 3 had persistent lymphedema. Several earlier studies have also reported substantial morbidity and mortality associated with KS-IRIS, including persistent mucocutaneous lesions, lymphedema, and death [24–28]. Recent research presentations suggest that IRIS related to KS may be particularly aggressive and respond poorly to standard therapy. The traditional approach to severe IRIS often includes administration of systemic corticosteroids, which may be contraindicated in KS-IRIS given the finding that corticosteroids are associated with progression of the tumor [29–31]. Additional studies are needed to characterize the frequency, natural history, and prognosis of KS-associated IRIS.
Among patients in our study with OIs other than visceral KS, there were no IRIS-related deaths and only a single patient developed long-term sequelae (tuberculosis-IRIS leading to chronic dyspnea). Prospective observational cohorts examining tuberculosis and MAC IRIS in various settings have also reported no excess mortality and minimal morbidity due to IRIS [8, 9, 13, 32]. Unfortunately, our sample size did not allow us to assess whether these observations were due to HIV disease severity or treatment differences.
We observed a remarkably low frequency of IRIS in patients with Candida esophagitis (0%). Despite reports of increased Candida delayed-type hypersensitivity responses and in vitro lymphocyte proliferation early in ART [33, 34], IRIS due to Candida esophagitis has not been reported in the literature. In contrast, IRIS has been observed in patients with Candida meningitis [12, 35], angular cheilitis  and in HIV-uninfected leukemic children with hepatosplenic candidiasis . It is possible that the absence of observed cases of Candida esophagitis IRIS is due to the ease with which the infection is treated and the rapid clearance of antigens that serve as a target for the recovering immune system.
We also observed a low incidence of PCP-related IRIS (4%; 95% CI, 0.9%–12%), similar to that reported for the AIDS Clinical Trials Group (ACTG) 5164 trial (2% [4/171]) . Two possible explanations for this finding are that patients with PCP remain on secondary prophylaxis, leading to declining numbers of organisms in the early phase of ART, and that patients with moderate-to-severe PCP are treated with corticosteroids in addition to antibiotics to prevent paradoxical worsening of the infection early in treatment. Investigators for ACTG 5164 concluded that receipt of corticosteroids during acute OI treatment did not decrease IRIS incidence, but may have led to a delay in IRIS presentation. Our findings also suggest that corticosteroid use at the time of ART initiation does not fully explain this low rate. Nearly all of the patients we studied (67/72 [93%]) initiated ART several days to weeks after completion of their PCP treatment course when they were no longer on corticosteroids (median time from PCP diagnosis and treatment to ART initiation was 66 days [range, 14–1040]).
Our finding of an earlier presentation of IRIS in the setting of tuberculosis compared with other OIs is consistent with previous studies of tuberculosis in which the majority of IRIS events occurred within the first 3 months of ART [8, 11, 13]. In contrast, KS-IRIS and PCP-IRIS events were dispersed throughout the first year, as seen in Figure 2. In some cases, KS-IRIS might be delayed relative to Tuberculosis-IRIS due to an immunomodulary effect of chemotherapy used to treat the KS. However, 1 study of KS patients treated with ART and liposomal anthracycline chemotherapy did not find a significant drop in CD4+ cell counts in the year following chemotherapy , suggesting that if chemotherapy attenuated the immune response, it did so independent of T-cell number.
Finally, even among a highly immune-suppressed group of patients, CD4+ cell reconstitution for those with and without IRIS differed depending on the type of OI. Patients with tuberculosis, Cryptococcus, and mucocutaneous KS-IRIS experienced greater increases in CD4+ cell counts during the first 6 months of ART compared with those without IRIS. Other studies of IRIS in patients with tuberculosis and Cryptococcus also noted greater CD4+ cell count increases in patients who developed IRIS [8, 23, 39]; however, ours is the first to report this finding in patients with mucocutaneous KS-IRIS. Notably, peak elevation in CD4+ cell count often occurs long after the IRIS event (see Figure 3A and 3C), and therefore, may not be helpful in the diagnosis of the majority of IRIS cases. Total peripheral blood CD4+ or CD8+ cell counts, as measured in our study, are nonspecific immunologic markers that may not reflect pathogen-specific immune activity or regional immune responses.
Strengths of our study include a standardized approach to evaluating IRIS and use of a validated clinical definition along with independent event adjudication. Several limitations deserve mention: (1) retrospective review by clinicians not involved in the care and diagnosis of IRIS in these patients could have led to misclassification, as reviewers may have had insufficient information to make appropriate clinical judgments; (2) determination of IRIS was particularly challenging in cases of KS recrudescence; (3) the small number of IRIS cases in our analysis limited examination of complex associations; and (4) limiting our analysis to patients with paradoxical IRIS after certain AIDS-defining diagnoses underestimated the true incidence of all types of IRIS, including “unmasking” versions of this syndrome.
In conclusion, we demonstrated that the risk of developing IRIS varies widely and is dependent on the underlying OI. Patients with profound immune deficiency and AIDS-defining Candida esophagitis or PCP are unlikely to develop IRIS in settings where patients are promptly treated or receive steroids to prevent paradoxical worsening (PCP), whereas over 10% of those with KS, tuberculosis, or Cryptococcus experienced this syndrome. We found that paradoxical visceral KS-IRIS occurred frequently, despite only modest increases in CD4+ cell counts, and was associated with high morbidity and mortality. Our study suggests that additional research is needed to understand and prevent KS-IRIS, particularly in countries with a high burden of this malignancy.
We would like to acknowledge Margaret Griffith, Thomas Davis, Christine Bagley, and Laurie Smith for their significant contributions to this study.
This work was supported by the University of Washington AIDS and STD Research Training Grant (National Institutes of Health T32 AI07140-31).
All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.