Survival in patients with HIV infection and viral hepatitis B or C: a cohort study
AIDS: Volume 18(15) 21 October 2004
Bonacini, Maurizioa,d; Louie, Stanb; Bzowej, Natalied; Wohl, Amy Rockc
From the aDepartment of Medicine, University of Southern California School of Medicine, the bUniversity of Southern California School of Pharmacy, the cHIV Epidemiology Program, Los Angeles County Department of Health Services and the dLos Angeles, Department of Transplantation, California Pacific Medical Center, San Francisco, California, USA.
"Liver deaths were significantly more common in patients co-infected with either hepatitis B or C compared with patients that had HIV alone".
"Co-infection with hepatitis B is associated with hepatic outcomes similar to hepatitis C'
note from editor: at the 2004 DDW meeting Norbert Brau (Bronx VA) reported from study that for coinfected patients with undetectable HIV RNA hepatitis C disease did not progress faster than for monoinfected patients.
Aim: To assess survival in patients with HIV and viral hepatitis co-infection.
Methods: A prospective university clinic cohort of 472 patients with HIV infection who were followed for 8343 patient-months. The outcome measures were the survival from HIV or liver disease assessed by the Kaplan-Meier method. Multivariable analysis using a Cox regression model identified variables associated with mortality.
Results: Patients were divided into four subgroups: HIV/hepatitis B virus (HBV) (n = 72), HIV/hepatitis C virus (HCV) (n = 256), multiple hepatitides (n = 18) and HIV alone (n = 126). One hundred and thirty-four patients (28.4%) died during follow-up. Liver mortality was noted in 55 patients, representing 12% of the cohort and 41% of the total mortality. Survival curves were similar in patients with HIV alone and those with any viral hepatitis co-infection. Liver deaths were more common in patients with multiple hepatitides (28%) HIV/HBV (15%), HIV/HCV co-infection (13%) versus HIV alone (6%). Liver mortality was comparable in HIV/HBV as in HIV/HCV co-infected patients and was not associated with gender, ethnicity, age, or mode of infection. HIV deaths were similar in patients co-infected with viral hepatitis compared with those with HIV alone. In patients with viral hepatitis co-infection, initial CD4 cell count > 200 × 106 cells/l and use of highly active antiretroviral therapy (HAART) were associated with significantly reduced liver mortality.
Conclusions: Patients with HIV and viral hepatitis had greater liver mortality than patients with HIV alone, but had comparable HIV mortality. Co-infection with hepatitis B is associated with hepatic outcomes similar to hepatitis C. Control of immunosuppression with HAART and CD4 counts > 200 × 106 cells/l are associated with better hepatic outcomes and should be the first priority in patients with HIV and viral hepatitis.
The HIV, hepatitis B (HBV) and C (HCV) viruses can all be transmitted via contaminated body fluids, in association with a number of high risk behaviors. Therefore, it is not uncommon to find HIV seropositive patients co-infected with HBV (approximately 9%) or HCV (approximately 30%).
Available data indicate that HBV and HCV are not generally cytopathic viruses but induce liver disease by eliciting a vigorous peripheral and intrahepatic immune response. Thus, it follows that immunosuppression would protect the liver from immune attack and result in slower liver fibrosis progression. This has been reported in one study of hepatitis B and HIV co-infection. Recently however, hepatitis B co-infection was associated with a poor overall survival in patients with HIV.
Recurrent hepatitis C results in more aggressive liver disease in some post-transplant patients, and HIV/HCV co-infected patients have more severe liver fibrosis at later stages of HIV disease. Because of the positive impact of highly active antiretroviral therapy (HAART) on immunosuppression-related mortality, liver disease is emerging as an important cause of morbidity and mortality in patients with HIV. Furthermore, hepatitis C has been associated with CD4 counts recovery of lower magnitude in some but not all studies. Thus, HBV and HCV have the potential to complicate the prognosis of both HIV and liver disease in co-infected patients.
The purpose of this study was to prospectively compare the survival of HIV patients co-infected with viral hepatitis B and/or C with those infected with HIV alone.
A total of 472 HIV-positive patients were evaluated. Based on hepatic serology 126 patients had no hepatitis B or C markers (HIV alone) and 346 patients had HIV with viral hepatitis co-infection: 72 patients had HIV and hepatitis B (HIV/HBV), 256 patients had HIV and HCV infection (HIV/HCV), and 18 patients had multiple viruses (i.e. HCV/HBV or HBV/delta or all three infections).
The majority of patients were male, women being particularly under-represented in the HIV/HBV and the multiple hepatitis groups; they were more likely to be in the HIV/HCV or HIV alone group [odds ratio (OR) = 8.6; P = 0.0004]. Latinos were more likely to be diagnosed with only HIV in contrast to HIV and HBV or HCV.
We analyzed the time-related variation in liver enzymes and function tests. The total bilirubin increased over the follow-up in the 346 patients in the co-infected group, and decreased in the group with 126 patients with HIV alone (P = 0.02). Other parameters including variations in prothrombin time, serum albumin, aspartate aminotransferase, and alanine aminotransferase levels were not statistically different in the two groups (data not shown).
A total of 134 patients (28%) died during a follow-up of 8353 patient-months.
Unadjusted all-cause mortality rates were similar in the four subgroups: 26% in the HIV/HBV group (19 of 72), 25% in the HIV/HCV group (65 of 256), 50% in patients with multiple hepatitides (nine of 18), and 33% in the group with HIV alone (41 of 126). Kaplan-Meier actuarial survival curves adjusted for initial CD4 cell counts confirmed similar overall mortality in the HIV group compared with the co-infected group (data not shown).
Mortality associated with liver disease was noted in 55 (12%) of patients. Thus, more than one-third of all deaths (55 of 134, 41%) was associated with liver disease and nearly two-thirds (87 of 134, 65%) was associated with HIV.
Liver deaths were noted in seven of 126 (6%) HIV patients, 11 of 72 (15%) HIV/HBV co-infected, 32 of 256 (13%) HIV/HCV co-infected patients and in five (28%) of those with both hepatitis viruses. Liver deaths were significantly more common in patients co-infected with either hepatitis B or C compared with patients that had HIV alone.
In the HIV/HCV co-infected group, 35 patients were treated with at least one dose of interferon, while 221 were not treated. Mortality rates were comparable in the interferon-treated (6% liver- and 14% HIV-related) and untreated (14% liver- and 17% HIV-related) groups, respectively (P = 0.39).
The seven liver deaths in patients with HIV alone were due to: drug hepatotoxicity (one isoniazid, one mitochondrial toxicity), two alcoholic liver diseases, one hepatorenal syndrome with hepatocellular cancer (alcohol use 15 g/day; negative HCV RNA), one had non specified hepatitis and another had HCV in the death certificate. The latter patient had negative anti-HCV and non detectable HCV RNA.
When the entire cohort was evaluated, the variables associated with the endpoints of either liver or HIV death were: initial CD4 cell counts < 200 × 106 cells/l, lifetime alcohol use >= 50 g/day, evaluation between 1993-1996 versus 1997-2001, lifetime use of two or less antiretrovirals versus three or more drugs. The initial CD4 cell counts and the number of antiretroviral drugs were predictors of both liver and HIV death. In addition, viral co-infection was associated with an increased risk of liver death, whereas a decrease of CD4 counts over the observation period was associated with an increased risk of HIV death.
When only the co-infected group was evaluated, liver mortality rates among the three subgroups (HIV/HCV, HIV/HBV or HIV with multiple hepatitides) was superimposable (P = 0.36). Thus, a Cox regression model was constructed combining these three subgroups into a single co-infected group (n = 346). The two parameters significantly associated with liver mortality were; initial CD4 cell counts < 200 × 106 cells/l and less than three antiretrovirals, which implies evaluation before the HAART era.
We noted that 13 of 67 (19.4%) patients presenting with CD4 cell counts > 500 × 106 cells/l were on no antiretrovirals. Similarly, 15 of 88 (17%) patients who had albumin levels lower than 3 g/dl, had not received any antiretrovirals. All of the above patients were included in the group that had received two or less lifetime antiretrovirals, namely the presumed 'pre-HAART' group.
Alcohol use was also associated with mortality in the co-infected subgroup, but did not reach statistical significance. Liver mortality was not associated with gender, ethnicity, age, or mode of contamination. Hepatocellular cancer (HCC) was noted in the death certificate of three patients (0.6%); one with HIV alone and two with HIV/HBV co-infection.
We have prospectively followed a cohort of patients with HIV infection with risk profiles that appear to mirror the general HIV population in the United States. We found that in our cohort of HIV-positive individuals, greater than 40% of deaths were associated with liver disease either as a primary or associated cause of death . The assessment of mortality using death certificates is generally well accepted. However, liver disease may have been overshadowed by AIDS-related diagnoses and not been included. For example, hepatocellular cancer was diagnosed in less than 1% of patients, thus underdiagnosis is a distinct possibility. In addition, there may be differences in the assessment of the type of liver disease on death certificates. Considering the potential underestimation using death certificates , the impact of liver disease appears to be important in patients with HIV infection.
Co-infection with HBV alone (i.e. without co-existing HCV) had a significant negative impact on liver mortality, which appears to be at least as significant as hepatitis C. In the past, hepatitis B was thought to be a benign disease in the setting of HIV infection. A number of papers reported that HBV replication assessed by DNA or DNA polymerase levels, is higher but histology is less severe in HIV-positive patients compared to those without HIV [5,14,15]. However, others found no difference in HBV DNA or ALT levels in co-infected versus singly-infected HBV patients . Colin et al. also found more frequent cirrhosis (28%) compared with non-HIV patients with HBV (13%), although the difference was of borderline statistical significance . Moreover, mortality and its causes were not evaluated . In a large Italian cohort of HIV patients, the presence of HBsAg, older age, CD4 > 200 × 106 cells/l, and alcohol abuse were associated with greater liver mortality . However, 90% of the HIV/HBV co-infected group also had HCV co-infection because of a high rate of injecting drug use in Italy . Recently, high hepatic mortality rates in HIV/HBV co-infected patients were reported in the MACS cohort; co-infected patients were eight times more likely to die of liver disease compared with patients with HIV infection alone . The latter study included mainly men who have sex with men, Caucasians (82%) and low rates (14%) of injection drug use. Our cohort appears closer to the general HIV population seen in industrialized nations, and included significant percentages of minorities. Hepatitis B data are even more compelling considering that HBV, unlike HCV, may have been partially treated with the use of many pre-HAART regimens which included lamivudine. In the absence of such effective antiviral medication , one could speculate that in HIV/HBV co-infection, liver prognosis would have proven worse yet. One potential weakness of our study is the absence of data on specific antiviral use (e.g. lamivudine versus other nucleoside analogs without HBV activity) and past HBV infection. Several of our patients with HIV alone may have had a positive HBcAb, IgG type. It is known the organ donors with isolated positive HBcAb are able to transmit HBV infection to recipients of their donated organs. In addition, HBV has been shown to re-activate in patients with HIV, a negative HBsAg and positive HBcAb . In our group with HIV alone, we did find one patient who died with HCC and another who had hepatitis mentioned in the death certificate. We can only speculate that, in addition to overt HBsAg positive infection, occult HBV infection may also be detrimental in patients with HIV.
Our data also indicate that co-infection with multiple hepatitis viruses may further worsen liver prognosis, as our multiply infected group had the highest liver mortality. However, because of small numbers, statistical significance was not achieved.
Although HIV is known to worsen HCV liver disease, the impact of HCV on HIV is less clear. Three cohorts, one Swiss, one Italian and one Canadian have shown a lower CD4 cell recovery in patients with HCV infection that undergo HAART [10,11,21]. However data derived from these cohorts did not suggest that HCV patients had higher HIV-related mortality. On the other hand we found, like Sulkowski et al.  no impact of hepatitis B or C co-infection on HIV and overall mortality. A recent multicenter study found no evidence that HCV co-infection affects HIV-related or cardiovascular morbidity . As hepatic mortality was worse, but overall mortality was similar in co-infected patients, we are confident that our co-infected cohort did not have any worse HIV-related mortality compared with patients with HIV alone.
As expected, initial CD4 cell counts were associated with HIV mortality. Surprisingly, in patients co-infected with any combination of viral hepatitis B or C (with or without delta hepatitis) CD4 counts were also associated with liver mortality. A cutoff for CD4 (> 200 × 106 cells/l) exists which differentiates good versus poor liver outcome (CD4 < 200 × 106 cells/l). This underscores the need for excellent HIV control in most patients with viral hepatitis co-infection. In addition, evaluation in the pre-HAART era (1993-1996) and the use of less than three concomitant antiretroviral medications independently affected liver mortality (Table 3). Similar findings were noted in a large German cohort, which included 81% hemophiliacs . Our results confirm that the beneficial effects of HAART on liver outcomes can be generalized to HCV and HBV co-infected patients with a wide spectrum of risk factors.
It is possible that a minority of patients may have been untreated because of well-controlled HIV disease, yet they would have been classified in the less than three antiretrovirals category. It is also possible that some patients with very advanced liver disease may not have been treated or may not have taken HAART. These two scenarios would underestimate and overestimate, respectively, the difference in liver mortality between the two or less antiretroviral group and the three or more drug group. Since similar number of patients were found to be in each category (13 patients with CD > 500 × 106 cells/l and 15 patients with low albumin levels), we believe that these patients did not significantly influence the results.
Lifetime alcohol use equal or greater than 50 g/day was associated with mortality (overall and liver-related) but did not achieve statistical significance. Patients with HIV alone also experienced some liver mortality, although significantly less than co-infected patients. Alcoholic liver disease and drug toxicity represented the majority of these deaths.
In summary, we have evaluated prospectively collected epidemiological data to assess survival in patients with HIV infection. The main findings of this study are the following:
(1) co-infection with hepatitis viruses did not afford any disadvantage in terms of overall survival compared to non co-infected patients;
(2) HBV co-infection was associated with a likelihood of liver mortality as high as HCV co-infection;
(3) the use of three or more antiretrovirals, as opposed to the use of two or less drugs, was associated with lower liver mortality in HIV patients co-infected with viral hepatitis.
We believe that while liver disease contributed significantly to mortality in HIV patients before 1996, liver mortality has been lower in the HAART era. In the viral hepatitis co-infected group, the variables associated with higher hepatic mortality were: initial CD4 counts < 200 × 106 cells/l, and the use of less than three antiretroviral drugs.
We conclude that control of immunosuppression using HAART and keeping CD4 cell counts > 200 × 106 cells/l are the first priority in patients with HIV and any viral hepatitis co-infection. Control of hepatitis viruses needs to be considered after HIV is suppressed or as a primary endpoint in selected patients with early untreated HIV disease.
Patients and methods
From August 1993 to November 2001, adult patients with HIV infection were referred to the Los Angeles County - University of Southern California (LAC-USC) HIV/Hepatitis Clinic, for evaluation of abnormal liver tests, usually elevated aminotransferase (ALT), a positive serological test (HBsAg, anti-HCV antibody or both) or for the management of advanced liver disease complications. Patients underwent a complete history and physical examination and filled a demographics questionnaire with particular emphasis on: risk factors for HIV and HCV, estimated duration of infection, alcohol use, antiretroviral drug use and time of the first positive HIV test. The main risk factors for HCV and HIV were: history of male sex with males, any history of injection drug use, blood or blood product transfusions, or non parenteral cocaine use. The acquisition of HIV or HCV was assumed to have occurred at the time of the first blood or blood product transfusion, or the first injection drug use. The total duration of disease was estimated in years by subtracting the date of the earliest exposure from the date of evaluation. In patients with sexual or occupational exposure, exposure to intranasal cocaine, tattoos, or no known risk factors the duration of HIV, HBV or HCV disease was not estimated.
Survival time and change in laboratory values were computed based on the date of first evaluation in the Liver Clinic.
Alcohol use was estimated as an average daily intake (estimated in grams). Routine laboratory data including hepatitis serologies (HBsAg, anti-HCV, delta antibodies) were obtained from the USC-LAC Medical Center Central laboratory in most instances. Patients were divided into two main groups: (1) HIV alone, without HBsAg or anti-HCV and (2) HIV with any viral hepatitis marker as defined below. This group was further divided into three subgroups: (i) HIV/HCV with positive anti-HCV antibody (with or without detectable HCV RNA) or HCV RNA alone; (ii) HIV/HBV with positive serum HBsAg; and (iii) HIV with multiple viral markers (anti-HCV, HCV RNA, HBsAg or delta antibody).
Mortality was attributed to liver disease when any liver disease was mentioned in the death certificate as either the primary or contributing cause of death. In addition, patients that had died with known decompensated liver disease (ascites, endoscopically documented variceal bleed, hepatocellular carcinoma, hepatic encephalopathy) or who had histologic cirrhosis at the time of death were also considered to have died of liver disease.
Conversely, death was attributed to HIV when AIDS, HIV, pneumonia, sepsis or an AIDS-defining condition was noted in the death certificate. Patients whose death was related both to HIV and to liver disease were included in both groups in multivariable analysis as previously reported.
Mortality data were obtained by direct patient follow up and, in the majority of cases was confirmed by data provided by the HIV Epidemiology Program of the Los Angeles County Department of Health Services. This database also was used to confirm the cause of death in those instances in which patients had died at LAC-USC or affiliated hospitals.
Use of antiretroviral drugs was noted as the maximum number of antiretroviral agents ever (in their lifetime) taken concurrently by the patient. We considered patients taking three or more drugs concurrently to have received HAART. Patients were encouraged to follow-up in the Liver clinic every 1 to 6 months, depending on the clinical circumstance The Institutional Review Board of the University of Southern California approved the study (IRB no. 95-8007) and patients gave consent to their participation in the study.
The Kaplan-Meier method was used to construct survival curves for all four subgroups of patients. Hazard ratios [with 95% confidence intervals (CI)] were obtained for variables associated with survival (HIV or liver) from a Cox proportional hazards regression model, using the Breslow method for ties. Chi-square statistics were used as appropriate and P-values reported as two-tailed. All statistical analyses were performed using a Stata statistical package (Stata Corp. Release 8; College Station, Texas, USA).