Hepatitis B and long-term HIV outcomes in coinfected HAART recipients
10 September 2009
Hoffmann, Christopher J; Seaberg, Eric C; Young, Stephen; Witt, Mallory D; D'Acunto, Kristin; Phair, John; Thio, Chloe L
aDivision of Infectious Diseases, Department of Medicine, Johns Hopkins School of Medicine, USA
bDivision of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
cUniversity of New Mexico School of Medicine, Albuquerque, New Mexico, USA
dDavid Geffen School of Medicine at UCLA, Los Angeles Biomedical Research Institute at Harbor-UCLA, Torrance, California, USA
eDepartment of Infectious Diseases and Microbiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pennsylvania, US
fDivision of Infectious Diseases, Northwestern University Feinberg School of Medicine, Chicago, Illinois, USA.
"In summary, this study demonstrates that HBV status at HAART initiation does not affect the long-term ability of HIV-infected patients to respond to HAART in terms of HIV RNA suppression and immunological recovery. Thus, CH-B should not diminish enthusiasm for prescribing HAART nor should long-term virologic or immunologic failure on HAART be attributed to CH-B. However, despite effective HAART that includes agents active against HBV, individuals with CH-B did not have an improvement in rate of liver-related mortality and thus remain at considerably higher risk for progressive liver disease, possibly because of incomplete HBV suppression with lamivudine as HBV suppression is an important component of slowing disease progression. Further work is needed to assess the impact of long-term suppressive HBV therapy with agents that are more durable than lamivudine (such as tenofovir disoproxil fumarate), earlier HAART initiation in coinfected patients, and screening for liver disease during HAART to improve liver-related outcomes among the large population of individuals coinfected with HIV and CH-B."
Background: Chronic hepatitis B (CH-B) is common among HIV-infected individuals and increases liver-related mortality in the absence of HAART. The impact of CH-B on long-term HAART outcomes has not been fully characterized.
Methods: To address this question, HAART initiators enrolled in the Multicenter AIDS Cohort Study were retrospectively analyzed. Patients were classified by hepatitis B category based on serology at the time of HAART initiation. The association of CH-B with mortality, AIDS-defining illnesses, CD4 cell rise, and HIV suppression was assessed using regression analysis.
Results: Of 816 men followed for a median of 7 years on HAART, 350 were never hepatitis B virus (HBV) infected, 357 had past infection, 45 had CH-B, and 64 were only core-antibody positive. Despite HAART, AIDS-related mortality was the most common cause of death [8.3/1000 person-years (PYs)]. It was highest in those with CH-B (17/1000 PYs, 95% confidence interval 7.3, 42) and lowest among never HBV infected (2.9/1000 PYs, 95% confidence interval 1.4, 6.4). In a multivariable model, patients with CH-B had a 2.7-fold higher incidence of AIDS-related mortality compared with those never infected (P = 0.08). Non-AIDS-related mortality was also highest among those with CH-B (22/1000 PYs), primarily due to liver disease (compared to never infected, adjusted hazard ratio 4.1, P = 0.04). There was no significant difference in AIDS-defining events, HIV RNA suppression, and CD4 cell increase.
Conclusion: In HIV-infected patients receiving long-term HAART, HBV status did not influence HIV suppression or CD4 cell increase. However, mortality was highest among those with CH-B and was mostly due to liver disease despite HBV-active HAART.
HAART has increased the life expectancy of HIV-infected individuals who maintain long-term suppression of HIV replication and restore their CD4 cell counts [1-3]. Factors such as the initial HAART regimen, baseline HIV RNA, adherence, and side effects influence the success of achieving long-term HIV RNA suppression; however, it is unclear whether hepatitis B virus (HBV) coinfection affects long-term response to HAART. Chronic hepatitis B (CH-B) occurs in 5-10% of HIV-infected individuals and its long-term influences on HIV RNA suppression, CD4 cell recovery, and mortality while on HAART are not fully characterized. Several studies of HIV-HBV-coinfected individuals have examined the short-term response to HAART and are inconsistent. An Italian study found no difference in CD4 cell response between HIV-HBV-coinfected and HIV-monoinfected individuals . A Thai study found a smaller increase in CD4 cell count during the first 4-8 weeks of HAART among hepatitis B surface antigen (HBsAg) positive compared with HBsAg-negative patients . In two Taiwanese studies, HIV-HBV-coinfected patients were less likely to achieve HIV RNA suppression than HIV monoinfected individuals receiving 24 months of HAART [6,7]. Two other studies have found similar rates of HIV suppression and CD4 cell rise when comparing patients with and without HBsAg [8,9].
As HAART use expands in areas with high rates of CH-B, it is important to determine if past or current HBV infection influences the long-term HAART response. To address this question, we studied HIV-infected patients enrolled in the prospective Multicenter AIDS Cohort Study (MACS), assessing the effect of HBV status on mortality and HIV outcomes for up to 10 years following HAART initiation.
In this long-term, prospective analysis of 816 men with a median of 7 years of follow-up on HAART, past or present hepatitis B infection did not impact long-term response to HAART as measured by HIV RNA suppression, CD4 cell rise, or ADI incidence. However, even after receiving HAART that included drugs active against HBV, individuals with CH-B had an increased risk for non-ADI mortality, primarily due to liver disease. There was also a trend for increased AIDS-related death among those with CH-B. Importantly, isolated core status did not impact the response to HAART or liver-related mortality. These findings are particularly timely as HAART is being made available in countries where hepatitis B is highly endemic.
Another study examined clinical outcomes among HBV-HIV-coinfected individuals who were all receiving HAART for greater than 12 months, but, our study differs in important ways . Omland et al. evaluated similar clinical end points of mortality and liver-related mortality; however, they classified the study patients based on a single HBsAg test obtained either pre-HAART or on-HAART. This study design can result in misclassification as the serologies can shift over time especially with the initiation of HAART , which can affect the validity of the results. In our analysis, we classified hepatitis B status for each participant by requiring a minimum of two consistent serologic tests separated in time by at least 6 months. In all cases, at least one of these tests was obtained prior to HAART initiation. Second, our study separated patients without CH-B into isolated core, past infection, and never infected, which other studies, including that of Omland et al., were not able to do [4-6,9]. Heterogeneity in groups can be problematic because even those with past HBV infection may have active HBV replication . Our categorization allowed a clear demonstration that HBV infection, past or current, does not impact the immunologic or virologic response to HAART compared with those who were never infected with HBV.
Other studies have evaluated the impact of CH-B in patients not on HAART  and with heterogenous HAART experience . The latter includes a study by Konopnicki et al. evaluating the impact of CH-B in the Euro-SIDA cohort for short-term HAART response (to 12 months). The longer-term outcomes, such as overall and liver-related mortality, were examined among a mixed population of HAART-naive and HAART initiator patients.
Our study is also unique in that we separated CD4 cell recovery into the first and second phases, which has not been previously done in assessing CH-B. Consistent with some reports [8,9] and in contrast to other previous reports of attenuated initial CD4 cell recovery, we found no difference in CD4 cell response between HBV groups  early during HAART or later during HAART . Assessing the two phases of CD4 cell recovery separately enhances the potential of finding a difference in CD4 cell rise because the different mechanisms believed to lead to the early CD4 cell rise versus the second phase of rise.
Although HAART response, as measured by CD4 cell rise and HIV RNA suppression, was not compromised by CH-B in our study, it is notable that AIDS-related mortality rate was higher in this group compared with the HBV uninfected group. After adjustment for baseline CD4 cells, proportion of visits with HIV RNA suppression, and time of HAART initiation, a trend remained for increased AIDS-related death, but it was no longer statistically significant. Thus, it is difficult to rule out an effect of CH-B also on AIDS-related death. Men with CH-B started HAART at lower CD4 cell counts that could partially explain an increase in AIDS-related mortality. However, if this explained the entire association, one would expect the association to disappear in the adjusted model.
Unfortunately, liver-related mortality remained significantly increased in patients coinfected with CH-B despite a suppressive HAART regimen that generally included an agent active against HBV, lamivudine or tenofovir disoproxil fumarate. A study by Puoti et al.  compared risk of liver-related death by antiretroviral therapy (ART) agent among HIV-HBV-coinfected individuals receiving less than 4 years of HAART. That study reported a decreased risk of liver-related death with use of lamivudine (adjusted relative risk of 0.73 per year). In our study, the liver-attributable mortality rate on HAART was similar to that previously reported from the MACS before and on-HAART (14 per 1000 PYs in an earlier study of the MACS  and 17 per 1000 PYs in this study). There are several possible explanations for this finding that are not inconsistent with the partially protective effect of lamivudine reported by Puoti et al. First, liver-related death was a major cause of death in the Puoti et al. study, accounting for 26% of all deaths. Although there was a decreased risk among patients receiving lamivudine, the incidence remained high. Second, men who died of liver-related disease in our study, and were taking lamivudine, may have developed more lamivudine-resistant hepatitis B due to a longer duration of follow-up than in the Puoti et al. study. Studies of HBV-monoinfected people demonstrate that liver disease advances in the setting of lamivudine-resistant virus . This is a plausible contributing factor as two of the four men with liver-related death in this study had known lamivudine-resistant CH-B. Only one of the men who died of liver disease received tenofovir disoproxil fumarate, so it is plausible that with more effective anti-HBV HAART, liver mortality would decrease. In HBV monoinfection, potent anti-HBV therapy improves liver outcomes, but whether this occurs in HIV coinfection needs to be evaluated once sufficient data are available on patients with CH-B who are treated with tenofovir disoproxil fumarate-containing HAART. Although treatment of HIV-HBV co-infection with tenofovir disoproxil fumarate is recommended in the United States, it is not the current practice in many of the regions, such as Africa, with high prevalence of CH-B. Among individuals coinfected with HIV-HBV living in low-income countries, consideration should also be given to including tenofovir disoproxil fumarate in the first-line regimens. A second plausible explanation for similar liver-related mortality during HAART compared to the pre-HAART era is that these men had more severe liver disease before initiation of HAART and that HAART was not able to reverse their disease; however, the long time on HAART before liver-related mortality, a median of 5 years, among the men who died of liver-related disease makes this less likely. A third plausible explanation is that immune reconstitution syndrome-associated inflammation, which is increased among CH-B-coinfected individuals and usually occurs soon after HAART initiation , led to an accelerated long-term progression to liver disease.
Interestingly, there was an increase in mortality among patients in the isolated core group, which was not attributable to liver disease. We do not anticipate that this was a result of HBV reactivation because reactivation occurs with waning immunity not with the increasing CD4 cell counts that we observed  and liver disease was not a cause of death for any of the isolated core group. Surprisingly, most of the mortality in this group was from cardiovascular causes, leading to the possibility of differences in behavior rather than an intrinsic property of isolated core status. Differing patterns of drug use, including the possibility of more noninjection cocaine use, could potentially explain increased cardiovascular mortality . Chronic hepatitis C has also been associated with increased cardiovascular mortality ; however, only one of the four patients who died of cardiovascular causes had detectable HCV antibody. Significantly increased mortality among isolated core individuals was not reported in the previously described study from Taiwan supporting the possibility that a behavioral or environmental factor may explain the difference.
In adjusted analysis, we did not observe an increased risk for mortality among the past infection versus the never infected groups. Crude mortality was higher among those with past infection; however, this is explained by the lower CD4 cell count and older age at HAART initiation among those with past infection compared with the never infected group. The absence of a difference in mortality is reassuring, given the high fraction of the population with past HBV infection.
Several limitations to this study are worth noting. An assessment of the potential impact of the HBV-suppressive effect of lamivudine, tenofovir disoproxil fumarate, or both on liver-related outcomes was not possible due to the small numbers of patients and the later introduction of tenofovir disoproxil fumarate. Second, we did not analyze hepatotoxicity risks because it is usually a transient event that may be missed with the infrequent (every 6 months) transaminase testing in this cohort. Third, underlying liver disease may have been underreported on death certificates. This would have lead to an underestimate of the total burden of liver disease in this cohort. However, we do not believe that a systematic bias lead to differential underreporting by HBV group.
In summary, this study demonstrates that HBV status at HAART initiation does not affect the long-term ability of HIV-infected patients to respond to HAART in terms of HIV RNA suppression and immunological recovery. Thus, CH-B should not diminish enthusiasm for prescribing HAART nor should long-term virologic or immunologic failure on HAART be attributed to CH-B. However, despite effective HAART that includes agents active against HBV, individuals with CH-B did not have an improvement in rate of liver-related mortality and thus remain at considerably higher risk for progressive liver disease, possibly because of incomplete HBV suppression with lamivudine as HBV suppression is an important component of slowing disease progression. Further work is needed to assess the impact of long-term suppressive HBV therapy with agents that are more durable than lamivudine (such as tenofovir disoproxil fumarate), earlier HAART initiation in coinfected patients, and screening for liver disease during HAART to improve liver-related outcomes among the large population of individuals coinfected with HIV and CH-B.
Of 1371 HAART initiators in the MACS, 816 met our inclusion criteria. Four hundred and forty-three men were excluded because of unknown HBV status at HAART initiation, 106 were excluded because the date of HAART initiation occurred during a window greater than 1 year, and six were excluded because HAART initiation was before 1996. The HBV status of the 816 men included 350 never infected, 357 with past infection, 45 with CH-B, and 64 with isolated core. Of the 64 patients with isolated core, 62 had successful HBV DNA testing of which 11 (17%) had detectable HBV DNA with a median of 74 IU/ml (range 28-323). The patients who were never infected initiated HAART at the youngest age, the latest calendar years, and had the highest median CD4 cell count at HAART initiation (Table 1). Isolated core patients had the highest proportion of ever IDU and HCV. The CH-B group had the lowest median CD4 cells at HAART initiation and highest baseline alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels.
Median follow-up was 7 years [interquartile range (IQR) 3.5-9.0]. Of all patients, 95% received an HBV active agent with either lamivudine, emtricitabine, tenofovir disoproxil fumarate or both tenofovir disoproxil fumarate and lamuvidine or emtricitabine, as part of HAART during the observation period. Ninety percent received lamivudine at any point during the observation period and 48% received tenofovir disoproxil fumarate, with or without concomitant lamivudine or emtricitabine, during the observation period.
Deaths and AIDS-defining illnesses
Overall, 87 patients died after HAART initiation [rate: 17/1000 person-years (PYs), 95% confidence interval (CI) 13, 21]. Despite HAART, AIDS-related mortality was the most common cause of death with 43 deaths due to AIDS (8.4 per 1000 PYs, 95% CI 6.2, 11.2). There were 30 non-AIDS-related deaths (5.9 per 1000 PYs, 95% CI 4.1, 8.5), and 14 patients died from an unknown cause (2.8 per 1000 PYs, 95% CI 1.6, 4.7).
In univariate analysis, the AIDS-related mortality rate was highest among patients with CH-B (17 per 1000 PYs, 95% CI 7.3, 42), followed by isolated core (14 per 1000 PYs, 95% CI 5.9, 34), past infection (11 per 1000 PYs, 95% CI 7.7, 16), and never infected (2.9 per 1000 PYs, 95% CI 1.4, 6.4). After adjustment for baseline CD4 cells, proportion of visits with HIV RNA suppression, IDU, and HAART initiation after 1996, CH-B remained associated with a higher but not statistically significant risk of AIDS-related death (IRR = 2.7, P = 0.08) compared with those who were never infected with HBV (Table 2).
The non-AIDS death rate was highest among those with CH-B (22 per 1000 PYs, 95% CI 7.9, 47) compared with never infected (2.4 per 1000 PYs, 95% CI 0.77, 5.5), past infection (5.8 per 1000 PYs, 95% CI 3.2, 9.7), and isolated core (14 per 1000 PYs, 95% CI 5.7, 34) (Fig. 1). After adjustment for age and baseline CD4 cells, both CH-B (hazard ratio 4.1, P = 0.04) and isolated core (hazard ratio 3.6. P = 0.06) remained associated with a higher risk of non-AIDS-related death compared with the HBV never-infected group (Table 2).
Among patients with CH-B, four of six non-AIDS-related deaths (66.7%) were liver-related (liver specific mortality rate 17 per 1000 PYs, 95% CI 7.1, 41), significantly higher than in the other groups (P = 0.05) (Table 3). The time on HAART among the four individuals with liver-related death ranged from 2.3 to 7.3 years (median 5.0 years), and all received lamivudine. One of these four patients received tenofovir disoproxil fumarate starting approximately 1 year prior to death. Three of the four patients with liver-related death had HBV DNA level assayed within 1 year of death. Of these three, one had a high HBV DNA (3.0 x 107 IU/ml), one had low but detectable HBV DNA (296 IU/ml), and one had HBV DNA less than 20 IU/ml. The two with detectable HBV DNA had documented lamivudine-resistant HBV; however, the patient with the HBV DNA of 296 IU/ml was receiving tenofovir disoproxil fumarate.
Surprisingly, among the isolated core group, four of the five deaths (80%) were cardiovascular-related. Of the four men who died, one was HCV antibody positive and one (a different patient) reported IDU. Only one of the men whose death was cardiovascular-related had detectable HBV DNA (occult HBV).
ADI events occurred at a rate of 28 per 1000 person-years (95% CI 24, 33; 116 events). In a multivariate analysis, CH-B did not increase the risk of ADI events (Table 2).
HIV RNA suppression
HIV RNA suppression was achieved among 60% of patients by the second HAART visit and was achieved by 53% of patients at more than 80% of HAART visits. In the univariate analysis, the isolated core patients were least likely to achieve HIV RNA suppression (OR 0.54, 95% CI 0.32, 0.90, P = 0.02) when compared with the never infected group. When adjusted for significant covariates - baseline HIV RNA, ever IDU, and HAART initiation after 1996 - isolated core status was no longer associated with decreased likelihood for HIV RNA suppression. The adjusted odds ratio for HIV RNA suppression when compared to never infected was 0.68 (95% CI 0.39, 1.2) for isolated core, 0.87 (95% CI 0.64, 1.2) for past infection, and 0.88 (95% CI 0.46, 1.6) for CH-B (P = 0.17).
CD4 cell recovery
HBV status was not associated with CD4 cell rise over the first 3 months of HAART (P = 0.58) or in the second phase of HAART (P = 0.9) (Fig. 2). In both phases, the covariates associated with larger CD4 cell rise were lower baseline CD4 cell and proportion of visits with HIV RNA suppression (P = 0.01 and <0.001, respectively). There were no interactions between any of the covariates and HBV status (data not shown). In a multivariate model, HBV status remained unassociated with CD4 cell recovery in either phase of HAART (P > 0.05).
The MACS prospectively follows men who report having had sex with men from four metropolitan areas in the United States (Baltimore, Maryland; Chicago, Illinois; Pittsburgh, Pennsylvania; and Los Angeles, California) and has had three recruitment periods: April 1984-March 1985, 1987-1991, and 2001-2003 . This current study included men who initiated HAART while enrolled in the MACS and had sufficient HBV serologic data to be classified into a specific HBV infection category, as described below.
HAART was defined according to guidelines from the Department of Health and Human Services (DHHS)/Kaiser Panel . The following regimens were considered to be HAART: two or more nucleoside reverse transcriptase inhibitors (NRTIs) in combination with at least one protease inhibitor or one nonnucleoside reverse transcriptase inhibitor (NNRTI); one NRTI in combination with at least one protease inhibitor and at least one NNRTI; ritonavir and saquinavir in combination with one NRTI and no NNRTIs; and abacavir or tenofovir with three NRTIs in the absence of both protease inhibitors and NNRTIs.
The date of HAART initiation was estimated as the midpoint between the last visit at which the patient was not on HAART and the first visit at which a patient reported having started HAART. Only patients whose HAART initiation date could be estimated within 1 year were included in this study.
Men were prospectively followed every 6 months from HAART initiation (earliest 1996) until death, date of last visit, or 31 March 2006. Alcohol consumption and injection drug use were assessed by self-report at each follow-up visit and categorized as never, former, or current use. Causes of death were obtained from death certificates, autopsy report, or personal contacts.
Hepatitis B virus infection status
We classified each participant's HBV status using the HBsAg, hepatitis B surface antibody (anti-HBs), and hepatitis B core-antibody (anti-HBc) results obtained before HAART initiation using the following categories: 'never infected' defined as either negative for all HBV serology or only anti-HBs positive; 'past infection' defined as anti-HBs and anti-HBc positive; 'CH-B' defined as HBsAg positive on two occasions more than 6 months apart; and 'isolated core' defined as positive for anti-HBc and negative for the other serologies. To select the specific results on which to base each participant's HBV status, we first looked for the latest serology results obtained prior to HAART initiation. If these results were not obtained within 18 months prior to HAART initiation, we then looked for serology results obtained at the first visit less than 6 months following HAART initiation. All participants whose serology results were available during this window, ranging from 18 months before HAART initiation to 6 months after, were classified according to the HBV algorithm defined above. For all other HAART initiations, we looked for the latest results obtained prior to HAART initiation and the earliest results obtained following it. If these two results yielded the same HBV classification as defined above, then HBV status was based on these results. All participants who only had serology results obtained more than 6 months following HAART initiation were reviewed, and those who were never infected with HBV (category 1 above) were also included. Finally, all persons classified as having CH-B were reviewed to confirm that they had a second HBsAg positive result obtained more than 6 months before or following the index result. The study was approved by the Institutional Review Boards at each MACS site, and all participants provided written informed consent.
HBV serologic testing was performed with EIA assays (Abbott Laboratories, Abbott Park, Illinois, USA). HBV DNA levels were determined on patients with isolated core serology using the Artus HBV LC PCR kit (Qiagen, Valencia, California, U.SA) with a detection limit of 40 IU/ml. Hepatitis C virus (HCV) antibody status was assayed by approved third-generation EIA. HCV positive was defined as a single positive assay without HCV RNA testing. Hepatitis B resistance mutations were identified by sequencing a nested 1122 base-pair HBV polymerase product (primers available upon request). Substitutions were identified by comparison to a reference sequence using SeqScape version 2.5 (Applied Biosystems, Foster City, California, USA). CD4 cell count was assayed by flow cytometry (Becton Dickinson, Mountain View, California, USA). HIV RNA levels were determined using Roche Ultrasensitive RNA PCR assay (Hoffmann-LaRoche, Nutley, New Jersey, USA) with a detection limit of 50 copies/ml.
The primary outcome was death, which was divided into AIDS-related and non-AIDS-related causes based on the Centers for Disease Control and Prevention (CDC) AIDS case definitions . Secondary outcomes included the incidence of an AIDS-defining illness (ADI), HIV RNA suppression (<400 copies/ml), and CD4 cell count rise.
Continuous data were compared using the Kruskal-Wallis test, and proportions were compared using chi-squared tests. We calculated incidence rate ratios (IRRs) to assess the association of selected covariates with survival and ADI events using Poisson regression.
For the HIV RNA suppression analysis, we used logistic regression with robust variance estimates using generalized estimating equations (GEEs) with unstructured correlation matrices to estimate the crude and adjusted odds ratios (ORs). The GEE approach was used to account for the within-patient correlation generated by the inclusion of all visits following HAART initiation for each patient.
The analysis of the CD4 cell count rise was divided into the initial 3 months of HAART and subsequent time on HAART because of the biphasic characteristic of CD4 cell rise . We used the time period from baseline to the first on-HAART visit (assuming an average time of visit to be 3 months from HAART initiation) to model the first phase of CD4 cell rise and the second to the 10th on-HAART visits to model the second phase of CD4 cell rise during the subsequent 5 years. We also calculated overall CD4 cell rise for patients with suppressed HIV RNA to compare with previous reports [14,15]. For these analyses, we used linear regression with GEE methods to account for within-patient repeated measures.
For all analyses, we assessed associations between the outcome of interest and year of HAART initiation (1996 versus 1997-2006), age at HAART initiation (continuous variable), baseline absolute CD4 cell count (continuous variable), baseline log10 HIV RNA (continuous variable), history of IDU, history of alcohol abuse, HCV status, proportion of HAART visits with HIV RNA suppression prior to and including each visit (time-dependent covariate), HIV RNA suppression at each visit, and HBV status. In building multivariate models, we initially included HBV status indicator variables and all other independent variables with a P value of less than 0.1 in univariate analysis. We then obtained final models by using a modified backward stepwise algorithm based on significance levels in the full model (P < 0.1). We used a P value of less than 0.05 to indicate statistical significance. STATA 10.0 (StataCorp, College Station, Texas, USA) was used for all analyses.