As the numbers of patients receiving antiretroviral therapy (ART) increases worldwide, it is increasingly important to monitor its associated risks. An important and common complication of ART is drug-induced hepatotoxicity. Retrospective studies have shown overall incidence rates of ART-related severe hepatotoxicity from 8.5% to 23%,1-7 and serious life-threatening liver events among patients on ART occur at a rate of 2.6 per 100 person-years.4
The identification of pretreatment risk factors for severe hepatotoxicity (SH) becomes important to prevent unnecessary morbidity and mortality from ART-induced liver injury. Reported risk factors for the development of ART-related hepatotoxicity have included elevated baseline liver function tests, coinfection with hepatitis B and/or C virus (HBV and HCV, respectively), and the use of nevirapine (NVP) and protease inhibitors (PIs), including high-dose ritonavir.4,6-9 To understand predictors of SH better, a retrospective analysis of 16 United States AIDS Clinical Trials Group (ACTG) studies, the largest HIV cohort studied to date, was performed.
The database consisted of 8851 subjects who were enrolled in 16 ACTG ART trials between October 1989 and June 1999. This was a subset of studies reported in a preliminary analysis.5 Subjects from each study were grouped into 4 specific treatment categories according to their initial treatment assignment: a single nucleoside reverse transcriptase inhibitor (NRTI) (including zidovudine [ZDV], didanosine [ddI], zalcitabine [ddC], stavudine [d4T], and lamivudine [3TC]), NRTI(s) and a nonnucleoside reverse transcriptase inhibitor (NNRTI) (NVP or delavirdine [DLV]), and NRTIs and a PI (indinavir [IDV]).
Hepatotoxicity, Mortality, and Hepatic-Related Death
SH was defined using the 1992 Division of AIDS (DAIDS) Table for Adult Adverse Experiences (National Institute of Allergy and Infectious Diseases, Bethesda, MD) as grade ≥3 elevation in liver function tests: aspartate aminotransferase (AST) or alanine aminotransferase (ALT) >5.0 times the upper limit of normal (ULN) or total bilirubin >2.5 times ULN. SH occurring in the first 6 months of ART was further classified as early hepatotoxicity. SH occurring during the following 6 months was classified as late hepatotoxicity. Abnormal baseline laboratory values were based on the DAIDS toxicity criteria: creatinine >1.5 times ULN, glucose <55 mg/dL, AST and ALT >2.5 times ULN, total bilirubin >1.0 times ULN, triglycerides >400 mg/dL, white blood cell (WBC) count <3900 cells/mm3, and platelets <75,000/μL.
Death, ascertained by searching the death and adverse events reporting forms, was considered to be hepatic related if the following words were listed as the primary or contributing cause of death: liver or hepatic failure, liver problems, liver disease, cirrhosis, fatty liver, hepatitis, hepatic encephalopathy, hepatocellular carcinoma, or hepatorenal syndrome.
Baseline HBV and HBC serologies were not available; therefore, a case-control study was performed to determine the relative risk of HBV or HCV coinfection for SH during the first 6 months of ART. Cases, defined as subjects with SH during months 1 through 6, and controls, defined as subjects without SH during this interval, were matched at a 1:1 ratio by gender, ACTG study, treatment arm, and age (within 3 years). Positive hepatitis B surface antigen (HBsAg) or anti-HCV antibody results were confirmed with the Amplicor HCV RNA qualitative assay (lower limit of detection <50 IU/mL; Roche Diagnostics, Branchburg, NJ). Active HCV infection was defined as a positive test result for anti-HCV antibody and qualitative HCV RNA.
Univariate and Multivariate Analyses
Covariates considered for the analysis are listed in Table 1. Associations between SH and categoric covariates were analyzed using a 2-sided Pearson χ2 test (data not shown). Baseline predictors of SH were determined by stepwise multicovariate logistic regression. Conditional odds ratios (ORs) and 95% Wald confidence intervals (CIs) are reported where appropriate. Analyses were performed in SAS (SAS Institute, Cary, NC). All P values were 2-sided, and 0.05 was used as the threshold for statistical significance.
The association of HCV and HBV with case-control status was evaluated with matched logistic regression. A total of 132 case-control pairs were used for the final 1:1 matched analysis. Assuming a 16% HCV prevalence, there was 80% power (2-sided α = 0.05) to detect a relative risk of 2.3.
Characteristics of Subjects, Hepatotoxicity, and Mortality
In our cohort of 8851 subjects, there were 1260 (14.2%) women and 7591 (85.8%) men, with a median age 35.96 years. The distribution of ethnicity was as follows: 5787 (65.4%) white/non-Hispanic, 1746 (19.7%) black/non-Hispanic, 1154 (13%) Hispanic, 70 (0.79%) Asian/Pacific Islander, 37 (0.42%) American Indian/Alaskan native, and 57 (0.64%) other/unknown.
Among 8851 subjects, 824 (9.3%) developed SH during the first 0 to 12 months of the study ART regimen, 613 (6.92%) during the first 6 months, and 211 (2.38%) during the following 6 months. Excluding subjects on IDV-based regimens who had hyperbilirubinemia alone (n = 56), 768 (8.7%) developed SH during the first 0 to 12 months of the study ART regimen. A total of 1137 deaths (12.8%) were recorded, 79 (0.9%) of which were hepatic related.
Early Hepatotoxicity: Months 0 Through 6
Among 1958 subjects on single NRTIs, abnormal baseline AST (OR = 2.64, 95% CI: 1.9 to 3.7), ALT (OR = 1.61, 95% CI: 1.1 to 2.3), or total bilirubin (OR = 2.55, 95% CI: 1.4 to 4.6) and ddI-containing regimens (OR = 1.62, 95% CI: 1.1 to 2.4) were associated with early SH (see Table 1). A baseline WBC count <3900/mm3 (OR = 0.65, 95% CI: 0.45 to 0.95) was associated with a reduced risk of early SH. Among 2464 subjects on multiple NRTIs, a history of Mycobacterium avium intracellulare infection (MAI; OR = 0.21, 95% CI: 0.05 to 0.86) and 3TC (OR = 0.37, 95 % CI: 0.20 to 0.68) was associated with a reduced risk of early SH. Elevated baseline AST (OR = 2.38, 95% CI: 1.8 to 3.1), creatinine (OR = 3.81, 95% CI: 1.8 to 8.2), and platelets (OR = 2.43, 95% CI: 1.3 to 4.5) were significantly associated with an increased risk of early SH. Among the 872 subjects on NNRTI-containing regimens, abnormal baseline AST (OR = 2.63, 95% CI: 1.7 to 4.1) was significantly associated with an increased risk of SH. Among 774 subjects on IDV-containing regimens, treatment with d4T (OR = 1.94, 95% CI: 1.1 to 3.3) and abnormal baseline AST (OR = 2.64, 95% CI: 1.8 to 4.0) and total bilirubin (OR = 4.27, 95% CI: 2.4 to 7.6) were significantly associated with an increased risk of early SH.
Late Hepatotoxicity: Months 7 Through 12
Among subjects on single NRTIs, abnormal baseline ALT (OR = 1.86, 95% CI: 1.2 to 2.8) (see Table 1) was significantly associated with late SH. Among subjects on multiple NRTI regimens, abnormal baseline ALT (OR = 2.08, 95% CI: 1.5 to 2.9) and concomitant hepatotoxic medications (HEPTOX) (OR = 1.44, 95% CI: 1.0 to 2.0) were independently associated with an increased risk of late SH. Among subjects on NNRTI-containing regimens, NVP-containing regimens (OR = 3.54, 95% CI: 1.2 to 11.0) and abnormal baseline ALT (OR = 2.58, 95% CI: 1.4 to 4.9) were associated with an increased risk of late SH. Among subjects on IDV-containing regimens, HEPTOX-DC (OR = 4.20, 95% CI: 1.3 to 13.4), injection drug use (IDU) (OR = 4.22, 95% CI: 1.5 to 11.9), and abnormal baseline total bilirubin (OR = 4.1, 95% CI: 1.3 to 12.6) were significantly associated with late SH.
Hepatitis B Virus and Hepatitis C Virus Coinfection
Among the 132 case-control pairs analyzed, 30% of cases and 14% of controls had evidence of active HCV infection (positive anti-HCV antibody and qualitative HCV RNA). The matched pairs OR was 2.7 (P < 0.003) suggesting that HCV coinfection was associated with a statistically significant increased risk of SH among ACTG participants during the first year. Only 11% of cases and 8% of controls were HBsAg-positive. The matched pairs OR was 1.44; however, this did not reach statistical significance (P = 0.40).
This analysis of 16 ACTG studies combining data on 8851 subjects is the largest published analysis of HIV-positive subjects enrolled into ART clinical trials. This study confirms that abnormal baseline aminotransferases are reliable predictors of ART-related hepatotoxicity over the first year of ART.
HEPTOX was a risk factor for late SH for subjects on multiple NRTI regimens, perhaps reflecting the importance of cumulative exposure to ART and HEPTOX. Medications with potential hepatotoxicity included antifungals (fluconazole), antibiotics (eg, erythromycin, amoxicillin-clavulanate, dapsone, trimethoprim sulfamethoxazole), antitubercular drugs (eg, ethambutol, isoniazid, pyrazinamide, rifampin, rifabutin, clofazimine), and anticonvulsants (eg, phenytoin, carbamazepine). The association between HEPTOX and SH underscores the importance of a comprehensive survey of all medications to identify agents that may predispose patients to hepatotoxicity, especially among individuals with other risk factors for SH.
This analysis also revealed risk factors for SH not previously reported in the literature: creatinine >1.5 times ULN and platelets <75,000 cells/mm3. Elevated baseline creatinine may be associated with decreased clearance of some hepatotoxic agents. Also, decreased renal clearance may itself be a marker of chronic liver disease. In addition, thrombocytopenia may be a surrogate for chronic liver disease with portal hypertension.
The results of our case-control analysis confirm that coinfection with HCV is a risk factor for the development of SH. The incidence of active HBV was low in this population; as a result, the case-control study was not adequately powered for HBV. Therefore, the existence of a smaller but nonetheless important association with HBV, as has been cited in other studies,6,7,9 cannot be excluded.
This study has several limitations. Comparisons across treatment categories are limited, because subjects in each treatment group have been pooled from studies that accrued at different points in time, from 1989 through 1999, resulting in heterogeneous populations. In addition, this analysis included older treatment regimens, such as single- and dual-NRTI regimens, that are rarely used in the United States today. This did allow the evaluation of antiretrovirals when they are used singly and in pairs to isolate hepatotoxic effects, however.
This analysis included combinations that are currently less commonly used in the developed world, such as NVP- and/or d4T-containing regimens. Because of the availability of generics and combination antiretroviral formulations, however, they are commonly prescribed in resource-limited environments. Because routine monitoring for SH is not currently feasible in many of these areas, it is critical to identify patients who are predisposed to SH before the initiation of ART so as to prevent significant morbidity and mortality from liver disease.
In summary, this study confirms the importance of evaluating HIV-infected patients for their risk of developing SH during the first year of ART. Ideally, initial pre-ART screening should include baseline liver function testing and assessment of hepatitis coinfection status, particularly HCV. Specific ART agents, ddI, and NVP, may carry higher risks for SH, and alternative regimens should be considered in some cases. The presence of newly defined risk factors for SH, such as the hepatotoxic potential of concomitantly administered medications, renal insufficiency, and thrombocytopenia, should also be factored into treatment monitoring. Although further study is necessary to confirm these findings, the presence of these risk factors in patients initiating ART should prompt more vigilant monitoring for ART-related hepatotoxicity. Indeed, in resource-limited settings, selection of persons with preexisting risk factors for SH to undergo routine liver test monitoring should be a practical application of these findings.