The introduction of combination antiretroviral therapy (ART) has dramatically improved the survival of HIV-infected patients in developed countries.1–3 However, as individuals with HIV live longer, long-term side effects of HIV are increasingly common, including effects on the liver.4–7 Previously reported common etiologies for hepatic dysfunction in the HIV-infected population include coinfection with chronic hepatitis B or C, alcohol and drug abuse, ART use, diseases such as nonalcoholic steatohepatitis, nonalcoholic fatty liver disease, and opportunistic infections.8–11 However, data are lacking on hepatic dysfunction or hepatic dysfunction–related death when compared with HIV-uninfected individuals of similar age and sex. Such data is important to assess the effect of chronic HIV infection on the liver.
As an integrated health care system, Kaiser Permanente (KP) provides nearly all care to patients within KP, an environment ideal for conducting such research. In addition, KP maintains a comprehensive HIV registry, facilitating comparative analyses for many outcomes between HIV-infected and HIV-uninfected individuals. The aims of this analysis were to compare the incidence of hepatic dysfunction or hepatic dysfunction–related death between a large cohort of HIV-infected and age-matched and sex-matched HIV-uninfected individuals and to examine factors, including degree of immunosuppression, HIV RNA levels, and long-term use of ART, that may affect risk.
We conducted a cohort study from 1996 to 2008 of adult HIV-infected and HIV-uninfected individuals within Kaiser Permanente Southern California (KPSC) and Kaiser Permanente Northern California (KPNC), both of which are large integrated healthcare delivery systems providing comprehensive medical care to more than 6 million members who are demographically representative of all insured Californians.12 December 2008 was chosen as the endpoint as this represented the latest date that California state mortality data were available at the time of this analysis.
HIV-infected individuals were identified from KP-maintained HIV registries, which include all known cases of HIV infection dating back to the early 1980s for KPNC and 2000 for KPSC. HIV-infected individuals are initially identified for inclusion in the registries by a positive HIV antibody test, detectable HIV viral RNA, prescription for an HIV antiretroviral, HIV/AIDS-related diagnosis, or other evidence of HIV infection from electronic sources. Confirmation of cases is done by medical chart review and comparisons of case lists with KP HIV clinics.
The HIV-infected study population consisted of all individuals included in the HIV registries who met the following 3 inclusion criteria: (1) KP member after January 1996, in KPNC and after January 2000, in KPSC (2000 in KPSC represents availability of electronic data required for ascertainment of HIV-infected cases); (2) ≥18 years of age; and (3) identified as “in-care”, defined as having received ≥1 CD4 test. A total of 4% of eligible HIV-infected patients were excluded as not being “in care”. The index date for HIV-infected individuals was the earliest date a member meets all 3 criteria.
The eligible HIV-uninfected study population consisted of all individuals meeting the following 2 inclusion criteria: (1) KP member after January 1996, in KPNC and after January 2000, in KPSC; and (2) ≥18 years of age. HIV-uninfected individuals were excluded if they were ever included in the HIV registry through 2008. HIV-uninfected individuals were frequency matched 10:1 by age (5-year age groups), sex, KP medical center (to control for socioeconomic factors), and initial year of follow-up for HIV-infected individuals (HIV-uninfected individuals were selected from the pool of active health plan members during the year of initial follow-up of HIV-infected persons). The index date for eligible HIV-uninfected individuals was the earliest date an individual was ≥18 years of age during the year for which they were selected to be a part of the HIV-uninfected cohort.
To examine the effect of ART class [either protease inhibitor (PI), or nonnucleoside reverse transcriptase inhibitor (NNRTI) based] and cumulative ART duration on the rate of hepatic dysfunction and hepatic dysfunction–related death, a subgroup of HIV-infected individuals was created in which the cumulative duration of medication exposure could be calculated (treatment analysis cohort). For patients in KPNC, previous chart reviews had established a complete antiretroviral history with duration of therapy for all individuals, including the use before KP membership; in KPSC, such a history was not available, and so individuals selected were those who had Kaiser membership for at least 6 months with no ART prescribed. We defined a simultaneous ART regimen among patients as ≥3 antiretroviral drugs used in combination during the same month; ritonavir at doses ≤400 mg/day was not considered an active drug in the regimen. Total cumulative therapy duration (including for those individuals who received no therapy, who comprised the reference group for this cohort) could thus be calculated in this group of patients by using pharmacy prescription fill and refill data to examine the class-specific cumulative role of ART on the risk for hepatic dysfunction or hepatic dysfunction–related death.
HIV-infected and matched HIV-uninfected KP individuals were followed until the earliest of death, hepatic-related event, health plan disenrollment, or December 31, 2008.
KP has developed and maintained a large number of administrative and clinical electronic data systems that routinely collect health-related information on its members. Linkage across these systems is feasible because members are assigned a unique lifetime medical record number. The primary data sources for this study were electronic medical records including pharmacy utilization, inpatient and outpatient diagnosis and utilization, laboratory records, administrative files, and mortality databases.
We defined hepatic dysfunction as an inpatient or outpatient diagnosis of hepatic failure (ICD-9: 570.xx; ICD-10: K71.1, K72, K72.0, K72.1, K72.9), hepatic encephalopathy (ICD-9: 572.2; ICD-10: included under K72), or esophageal varices with bleeding (ICD-9: 456.0; ICD-10: I85.0). Although a formal validation study was not undertaken to assess the accuracy of this method of diagnosing all cases of hepatic failure, it should in fact capture the majority of cases, and any errors should be nondifferential with regard to HIV status.
In addition, we identified hepatic dysfunction by abnormal labs as individuals with an elevated value of aspartate aminotransferase or alanine transaminase greater than 5 times the upper limit of normal AND either an elevated ammonia (greater than the upper limit of normal), or elevated international normalized ratio greater than the upper limit of normal (patients were defined by an elevated international normalized ratio and not by ammonia 91% of the time). All qualifying laboratory tests had to be performed within 7 days of each other. All patients with prevalent hepatic dysfunction events before the index date were removed from the analysis (0.6% for the case group and 0.1% for the control group). A hepatic dysfunction–related death was defined as any death preceded within 60 days by a diagnosis of hepatic dysfunction as defined above and/or hepatic failure listed as the cause of death (underlying or secondary) on the death certificate.
Initial analyses focused on differences in the incidence of hepatic dysfunction and hepatic dysfunction–related deaths comparing HIV-infected with HIV-uninfected individuals. We first computed incidence and mortality rates per 100,000 person-years by HIV infection status. Multivariable Poisson models compared the risk of hepatic dysfunction and hepatic dysfunction–related death by HIV status with adjustment for age, sex, race/ethnicity, smoking, alcohol abuse diagnosis, drug abuse diagnosis, hepatitis B or C coinfection (obtained by ICD-9 codes and laboratory evidence of infection), hypertension diagnosis, diabetes diagnosis, lipid-lowering medication use (given their known hepatic side effects, records from Kaiser pharmacy system), and calendar year. To examine factors of HIV disease that may affect risk of hepatic dysfunction, we also compared risk in HIV-uninfected persons with HIV-infected individuals stratified by ART use, CD4+ T-cell counts (≤200, 201–499, ≥500 cells/uL), and HIV RNA levels (≤500, 501–9999, 10,000–99,999 and ≥100,000 copies/mL).
Subsequent analyses were restricted to HIV-infected individuals. To examine specific risk factors in HIV-infected individuals, multivariable Poisson models were constructed with terms for age, sex, race/ethnicity, CD4 cell count (both recent and lowest KP recorded CD4, framed in 6 month intervals), recent HIV RNA level (framed in 6 month intervals), any use of ART, alcohol or drug abuse diagnoses, smoking, hepatitis B or C coinfection, hypertension, diabetes, and lipid-lowering medication usage.
We further examined ART class and cumulative ART duration on the rate of hepatic dysfunction and hepatic dysfunction–related death. To do this, the treatment analysis cohort (as described above) was utilized. Multivariable Poisson models were adjusted with terms for age, sex, race/ethnicity, alcohol or drug abuse diagnoses, smoking, hepatitis B or C coinfection, hypertension, diabetes, lipid-lowering medication usage, calendar year, and years known HIV infected.
All analyses with performed with SAS (Version 9.1; Cary, NC), using proc GENMOD for Poisson regression. The institutional review boards at KPSC and KPNC approved this study and provided waivers of informed consent.
We identified 20,775 HIV-infected individuals contributing 85,309 person-years (mean 4.4 years per person) and 215,158 HIV-uninfected individuals contributing 1,013,645 person-years (mean 5.3 years per person; Table 1). The HIV-infected group was predominantly male (91%), white (51%), and consisted mostly of men who have sex with men (59%), and had a mean age at start of follow-up of 41 years. Patients in the HIV-uninfected cohort were less likely to be coinfected with hepatitis than those in the HIV-infected cohort. Patients in KPNC and KPSC were demographically similar with the exception of a larger proportion of Hispanic patients in KPSC (data not shown). A total of 12,872 patients were eligible for the treatment analysis cohort; the demographics of this cohort closely mirrored the larger HIV-infected cohort (Table 1).
Among HIV-infected individuals, 437 met the criteria for hepatic dysfunction (491 per 100,000 person years, CI: 445.3 to 537.5), compared with 755 in the HIV-uninfected group (67 per 100,000 person-years, CI: 62.1 to 71.6), corresponding to an adjusted rate ratio (RR) of 3.5 (95% CI: 3.0 to 4.0, P < 0.001). One-hundred seventy HIV-infected individuals experienced a hepatic dysfunction–related death (188 per 100,000 person-years, CI: 159.9 to 216.5), whereas 192 HIV-uninfected individuals experienced a hepatic dysfunction–related death (17 per 100,000 person-years, CI: 14.6 to 19.4), corresponding to an adjusted RR 5.9 (95% CI: 4.7 to 7.4, P < 0.001). In the 170 cases of hepatic dysfunction deaths seen in the HIV-infected group, only 1 death was listed as primary hepatic failure related. The most common primary causes noted when hepatic failure was listed as a secondary or underlying cause were HIV/AIDS (69%) or other liver-related causes (19%). RRs were similar by sex and geographic location (KPNC vs. KPSC) for both hepatic dysfunction and hepatic dysfunction–related death, and when looking at patients only after the year 2000, when both KPNC and KPSC contributed data. Risk was lower in the modern era (2001–2008) than in the earlier era (1996–2000) and remained elevated even when excluding those patients with known hepatitis B or C coinfection (Table 2).
As shown in Table 3, at all levels of immunodeficiency, HIV-infected individuals had a greater risk of hepatic dysfunction–related death or hepatic dysfunction compared with HIV-uninfected individuals. The event rates were lowest among those individuals with CD4 ≥500 cells per microliter on ART (adjusted RRs for hepatic dysfunction–related death and hepatic dysfunction 1.8 and 1.6, respectively, P = 0.025 and P = 0.002 compared with HIV uninfected), and the highest were seen in individuals with CD4 ≤200 cells per microliter not on ART (adjusted RRs for hepatic dysfunction–related death and hepatic dysfunction 59.4 and 15.7, respectively, P < 0.001 for both when compared with HIV uninfected, Table 3).
Compared with HIV-uninfected individuals, HIV-infected individuals had a higher RR of hepatic dysfunction–related death and hepatic dysfunction with higher levels of HIV RNA (≥100,000 copies/mL) than lower (≤500 copies/mL), with a linear relationship observed between increasing risk and increasing RNA levels. This relationship was consistent whether patients were receiving ART (adjusted RRs for hepatic dysfunction–related death and hepatic dysfunction 3.5 and 2.5 when HIV RNA ≤500 copies/mL vs. 18.4 and 8.9 when HIV RNA ≥100,000 copies/mL) or not receiving ART (adjusted RRs for hepatic dysfunction–related death and hepatic dysfunction 4.3 and 1.7 when HIV RNA ≤500 copies/mL vs. 35.7 and 10.0 when HIV RNA ≥100,000 copies/mL) (Table 3).
Next, we analyzed our HIV-infected only model that considered ART, CD4, and HIV RNA together and other potential clinical and demographic risk factors for hepatic dysfunction (Table 4). The strongest risk factor for hepatic dysfunction was a history of hepatitis B or C coinfection (adjusted RR 5.3, 95% CI: 4.3 to 6.4, P < 0.001). Other risk factors significantly associated with hepatic dysfunction included recent CD4 ≤200 versus >200, lowest recorded CD4 at KP ≤200 versus >200, HIV RNA ≥500 versus <500, alcohol or drug use, and diabetes. Factors not associated with increased risk of hepatic dysfunction included ART use, sex, race/ethnicity, hypertension, and lipid-lowering drug use.
Finally, we evaluated the possible effect of cumulative therapy duration of PI-based or NNRTI-based ART regimens on the risk for hepatic dysfunction or hepatic dysfunction–related death among individuals with known duration of ART use, using the previously described treatment analysis cohort. A total of 8449 individuals in this cohort were exposed to simultaneous 3-drug ART (32,368 person-years, mean 3.8 years per person). Total PI exposure was 21,249 person-years among 5858 people exposed (mean 3.6 years per person); total NNRTI exposure was 15,643 person-years among 5662 people exposed (mean 2.8 years per person). As was observed in the full HIV-infected cohort, the use of ART was not associated with an increased risk for hepatic dysfunction, with an adjusted RR per additional year of ART use of 1.02 (95% CI: 0.96 to 1.08, P = 0.49). The lack of an effect of ART use on hepatic dysfunction was observed for both PI-based ART and NNRTI-based ART (Fig. 1, panel A). For hepatic dysfunction–related death, there was in fact a protective effect of ART, with an adjusted RR per year additional ART use of 0.83 (95% CI: 0.72 to 0.94, P = 0.004). As was seen for hepatic dysfunction, the magnitude of this effect was similar by ART class (Fig. 1, panel B).
HIV-infected individuals have an increased risk of hepatic dysfunction and hepatic dysfunction–related death compared with HIV-uninfected individuals, even with adjustment for demographic and clinical risk factors. Risk was increased regardless of sex, KP geographic location, or time frame, although risk was slightly reduced in the modern era. Although risk remained elevated for HIV-infected individuals at all levels of immunodeficiency, the greatest risk was observed for those not on ART with CD4 ≤200 cells per microliter, and the lowest risk was in individuals on ART with CD4 ≥500 cells per microliter. Risk was also noted to increase with higher levels of HIV RNA. Independent predictors for hepatic dysfunction among HIV-infected individuals included low CD4 cell count (both recent and lowest recorded at KP), high HIV RNA levels, alcohol or drug abuse, hepatitis B or C coinfection, and diabetes. The long-term cumulative use of ART did not increase risk for either hepatic dysfunction or hepatic dysfunction–related death, regardless of class used.
The explanation for the higher observed rate of hepatic dysfunction and hepatic dysfunction–related death was not immediately evident from our study. Although HIV-infected individuals frequently have comorbidities such as viral hepatitis, these factors were adjusted for in our models and did not completely explain the higher risk for hepatic dysfunction and hepatic dysfunction–related death. Similarly, severe immunosuppression alone was not a sufficient explanation for the excess morbidity and mortality because even individuals with CD4 cell counts greater than 500 cells per microliter had an increased risk, albeit at a lower rate than those with lower CD4 cell counts. And our study results clearly did not point to ART as contributing to risk; in fact risk was reduced with longer duration.
It may be possible that HIV itself is leading to the higher rates of hepatic dysfunction we observed. There have been reports of chronic immune activation in HIV, which may lead to chronic stimulation of the inflammatory system,13 and it has been hypothesized that this ongoing activation may lead to many of the long-term complications (cardiac, brain) seen in chronically HIV-infected individuals.14 It is possible that HIV infection leads to activation of the inflammatory cascade in the liver, leading to the greater rate of hepatic dysfunction observed. Such an explanation would be consistent with the observation in our study that patients with high HIV RNA levels had the greatest risk for hepatic dysfunction or hepatic dysfunction–related death. Indeed, other viral infections such as hepatitis C exert their adverse effects on the liver via inflammatory effects leading to liver fibrosis.15,16
The finding of higher risk for hepatic-related death at lower CD4 cell count has been previously reported.4 In the EuroSIDA cohort, a 50% increase in CD4 cell count was associated with a RR for hepatic-related death of 0.77 (P < 0.0001) in a multivariate adjusted model. Similar to the relationship seen in our study, the risk of hepatic-related death decreased with increasing levels of CD4; CD4 cell counts <50 cells per microliter had a death rate of 14.1 per 1000 person-years of follow-up compared with 1.6 per 1000 person-years of follow-up in those patients with CD4 cell counts >200 cells per microliter. However, risk was not compared to a similar HIV-uninfected cohort. And unlike the present study, the EuroSIDA cohort did not examine risk for hepatic dysfunction without death.
Our finding that cumulative ART does not increase risk for hepatic dysfunction or hepatic dysfunction–related death is reassuring, and clinically of great importance in light of calls from the World Health Organization and the United States National HIV/AIDS Strategy to increase the number of HIV-infected individuals on ART,17,18 and the results of the recent HIV Prevention Trials Network 052 study, which demonstrated a reduced rate of sexual transmission of HIV in patients receiving ART.19 Earlier studies have suggested toxicity with nucleoside reverse transcriptase inhibitor therapy (mitochondrial toxicity),20 NNRTI therapy,21 and PI therapy.22 However, unlike our study which measured hepatic dysfunction as defined by impaired hepatic synthetic function or hepatic-related complications such as encephalopathy or esophageal varices, many of these prior studies have used only transaminase elevation as a marker for liver injury. Although the lack of an effect on hepatic dysfunction in ART-treated patients in our study may reflect that fact that patients are being adequately monitored (and quickly taken off therapy), and hence no effect is observed, it may also be possible that many of the more minor transaminase elevations seen in clinical care may not predispose patients to hepatic failure. And finally, although this project did not break down ART use to specific agents, the difference in risk seen between the 2 eras examined does suggest that newer ARTs may have less effect on the liver than older ones.
We acknowledge certain limitations with our study. As no standard definition based on laboratory values exists for mild to moderate hepatic dysfunction, our definitions were chosen to best represent clinical hepatic synthetic dysfunction. The Model for End-Stage Liver Disease, or MELD score, although accurately predicting mortality from advanced liver failure, is not ideal for quantifying more mild disease.23 Our definitions were chosen to represent hepatic dysfunction at a level above that seen in simple transaminase elevation, although at a level lower than end-stage liver disease. Moreover, they were equally applied to HIV-infected and HIV-uninfected individuals, and thus any misclassification is anticipated to be nondifferential with respect to HIV status. Another limitation was that race/ethnicity was missing for 43% of HIV-uninfected individuals. However, in the HIV-infected cohort, we did have data on race for 94% of patients, and this factor was not found to be predictive for hepatic dysfunction or hepatic dysfunction–related death. In the HIV-uninfected cohort, it may be possible that hepatitis B, hepatitis C, and HIV may be undiagnosed. Thus we could not verify with certainty that the HIV-uninfected group was uninfected; instead patients were presumed uninfected by lack of an HIV diagnosis. We also did not include information on specific hepatitis genotypes or other hepatitis markers in our analysis. Finally, although nucleoside reverse transcriptase inhibitors were not examined as a separate class, over the time period examined, this class consistently remained the backbone of ART with the major difference in treatment being the choice of PI versus NNRTI.
The major strength of our study is the use of a large well-characterized population of HIV-infected and matched HIV-uninfected individuals from the same integrated health care system. Our study results have application to those utilizing health care, as data indicates that KP members are very similar to the California state-wide population with regards to age, sex, and race/ethnicity12; moreover, demographics of our HIV-infected population are very similar to reported AIDS cases in California.24
In summary, we found that when compared with age-matched and sex-matched HIV-uninfected controls, an increased risk for hepatic dysfunction and hepatic dysfunction–related death exists for HIV-infected individuals. ART use did not increase risk, and we found no effect by ART class or cumulative duration of therapy. These findings suggest that early diagnosis of HIV and subsequent initiation of ART before significant CD4 cell decline may partially mitigate the risk of hepatic dysfunction or hepatic dysfunction–related death.
2. Public Health Agency of Canada. HIV/AIDS Epi Updates, November 2007. Ottawa, Canada: Surveillance and Risk Assessment Division, Centre for Infectious Disease Prevention and Control, Public Health Agency of Canada; 2007.
3. Mocroft A, Katlama C, Johnson A, et al.. AIDS across Europe; 1994–1998: The EuroSIDA study. Lancet. 2000;356:291–296.
4. Mocroft A, Soriano V, Rockstroh J, et al.. Is there evidence for an increase in the death rate from liver-related disease in patients with HIV? AIDS. 2005;19:2117–2125.
5. Castellares C, Barreiro P, Martin-Carbonero L, et al.. Liver cirrhosis in HIV-infected patients: prevalence, aetiology and clinical outcomes. J Virol Hepat. 2008;15:165–172.
6. Louie J, Hsu L, Osmond D, et al.. Trends in causes of death among persons with acquired immunodeficiency syndrome in the era of highly active antiretroviral therapy, San Francisco, 1994–1998. J Infect Dis. 2002;186:1023–1027.
7. Mocroft A, Brettle R, Kirk O, et al.. Changes in the cause of death among HIV positive patients across Europe: results from the EuroSIDA study. AIDS. 2002;16:1663–1671.
8. Maida I, Nunez M, Rios M, et al.. Severe liver disease associated with prolonged exposure to antiretroviral drugs. J Acquir Immune Defic Syndr. 2006;42:177–182.
9. Ristig M, Dreschsler H, Powderly W. Hepatic steatosis and HIV infection. AIDS Patient Care STDS. 2005;19:356–365.
10. Crum-Cianflone N. Nonalcoholic fatty liver disease: an increasingly common cause of liver disease among HIV-infected persons? AIDS Read. 2007;10:513–518.
11. Guaraldi G, Squillance N, Stentarelli C, et al.. Nonalcoholic fatty liver disease in HIV-infected patients referred to a metabolic clinic: prevalence, characteristics and predictors. Clin Infect Dis. 2008;47:250–257.
12. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health. 1992;82:703–710.
13. Kuller L, Tracy R, Belloso W, et al.. Inflammatory and coagulation biomarkers and mortality in patients with HIV infection. PLoS Med. 2008;5:e203.
14. Deeks S. Immune dysfunction, inflammation, and accelerated aging in patients on antiretroviral therapy. Top HIV Med. 2009;17:118–123.
15. Bartosch B, Thimme R, Blum H, et al.. Hepatitis C virus-induced hepatocarcinogenesis. J Hepatol. 2009;51:810–820.
16. Zoulim F, Chevallier M, Maynard M, et al.. Clinical consequences of hepatitis C virus infection. Rev Med Virol. 2003;13:57–68.
19. Cohen M, Chen Y, McCauley M, et al.. Prevention of HIV-1 Infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
20. Nunez M. Hepatotoxicity of antiretroviral: incidence, mechanisms and management. J Hepatol. 2006;44:S132–S139.
21. Dieterich D, Robinson P, Love J, et al.. Drug-induced liver toxicity associated with the use of nonnucleoside reverse-transcriptase inhibitors. Clin Infect Dis. 2004;38:S80–S89.
22. Sulkowski M. Drug-induced liver injury associated with antiretroviral therapy that includes HIV-1 protease inhibitors. Clin Infect Dis. 2004;38:S90–S97.
23. Kamath P, Wiesner R, Malinchoc M, et al.. A model to predict survival in patients with end-stage liver disease. Hepatology. 2001;33:464–470.
24. California Department of Health Services, Office of AIDS. California AIDS Surveillance Report, Cumulative Cases as of 31 December 2005. Sacremento, CA: California Office of AIDS.