| |
HCV Accelerated Death in HIV+: Danish Study
|
| |
| |
"Impact of Hepatitis C Virus Coinfection on Response to Highly Active Antiretroviral Therapy and Outcome in HIV-Infected Individuals: A Nationwide Cohort Study"
Clinical Infectious Diseases May 2006;42:1481-1487
Nina Weis,1 Bjarne O. Lindhardt,2 Gitte Kronborg,1 Ann-Brit E. Hansen,2 Alex L. Laursen,6 Peer B. Christensen,3 Henrik Nielsen,4 Axel Moller,5 Henrik T. Sorensen,7,8 and Niels Obel3
1Department of Infectious Diseases, Hvidovre Hospital, and 2Department of Infectious Diseases, Rigshospitalet, Copenhagen, 3Department of Internal Medicine, Odense University Hospital, Odense, 4Department of Infectious Diseases, Aalborg Hospital, Aalborg, 5Department of Internal Medicine, Kolding Hospital, Kolding, 6Department of Infectious Diseases, Skejby Hospital, and 7Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; and 8Department of Epidemiology, Boston University, Boston, Massachusetts
note from Jules Levin: for those of you new to the NATAP email list, I have reported numerous studies finding that HCV worsens mortality in HIV & is a leading cause of death in HIV & likely the leading cause of death & hospitalization. It has been well accepted that HIV accelerates HCV disease progression. In addition, in the past year several studies were published that found HAART in coinfected patients slowed HCV disease progression to a rate equal to that in HCV monoinfected patients. One study found that only in coinfected patients achieving undetectable HIV did this benefit accrue. I maintained after these studies were released & continue to maintain these studies did NOT establish HAART slows HCV progression equal to that in monoinfected patients, and it is unlikely that HAART can slow HCV disease progression to that of monoinfected HCV+; perhaps in some, but not in general
Infection with hepatitis C virus (HCV) and HIV are often found in the same individual [1]. After the introduction of HAART and the subsequent decrease in mortality caused by HIV-related diseases, management of coinfection among this population has become increasingly important. Studies of the clinical prognosis of patients coinfected with HCV and HIV conducted prior to the introduction of HAART have shown conflicting results [2, 3]. After HAART became available, several studies examined the impact of HCV coinfection on HIV disease progression, again with discordant results and conclusions [4-11]. These studies often relied on small or select samples or had short follow-up periods.
The present study used a nationwide, population-based design to examine the impact of HCV on response to HAART and outcomes of HCV-HIV coinfection.
ABSTRACT
Background. Coinfection with hepatitis C virus (HCV) in human immunodeficiency virus (HIV) type 1-infected patients may decrease the effectiveness of highly active antiretroviral therapy. We determined the impact of HCV infection on response to highly active antiretroviral therapy and outcome among Danish patients with HIV-1 infection.
Methods. This prospective cohort study included all adult Danish HIV-1-infected patients who started highly active antiretroviral therapy from 1 January 1995 to 1 January 2004. Patients were classified as HCV positive (positive HCV serological test and/or HCV PCR results [443 patients {16%}]), HCV negative (consistent negative HCV serological test results [2183 patients {80%}]) and HCV-U (never tested for HCV [108 patients {4%}]). The study end points were viral load, CD4+ cell count, and mortality.
Results.
Compared with the HCV-negative group, overall mortality was significantly higher in the HCV-positive group (mortality rate ratio, 2.4; 95% confidence interval [CI], 1.9-3.0), as was liver disease-related mortality (mortality rate ratio, 16; 95% CI, 7.2-33).
Furthermore, patients in the HCV-positive group had a higher risk of dying with a prothrombin time <0.3, from acquired immunodeficiency syndrome-related disease, and if they had a history of alcohol abuse. Although we observed no difference in viral load between the HCV-positive and HCV-negative groups, the HCV-positive group had a marginally lower absolute CD4+ cell count.
Conclusions. HIV-HCV-coinfected patients are compromised in their response to highly active antiretroviral therapy. Overall mortality, as well as mortality from liver-related and acquired immunodeficiency syndrome-related causes, is significantly increased in this patient group.
Causes of Death
Risk of liver-related death was significantly increased in the HCV-positive group, compared with the HCV-negative group (table 2). The mortality rates for liver-related deaths were 14 deaths per 1000 PYRs (95% CI, 9.7-21 deaths per 1000 PYRs) in the HCV-positive group and 1.0 deaths per 1000 PYRs (95% CI, 0.5-1.9 per 1000 PYRs) in the HCV-negative group. Patients in the HCV-positive group also had a significantly higher risk of dying with a last measured prothrombin time of <0.3 (table 2). More patients in the HCV-positive group than in the HCV-negative group received diagnoses of ascites, esophageal varices, uncompensated liver disease, and cirrhosis prior to death (data not shown). Also, patients in the HCV-positive group had an increased risk of dying from AIDS-related disease, compared with patients in the HCV-negative group (mortality rate, 17 deaths per 1000 PYRs [95% CI, 12-25 deaths per 1000 PYRs] vs. 11 deaths per 1000 PYRs [95% CI, 9.3-14 deaths per 1000 PYRs]), and appeared to have a higher risk of dying with a CD4+ count >100 cells/L (mortality rate, 33 deaths per 1000 PYRs [95% CI, 25-42 deaths per 1000 PYRs] vs. 13 deaths per 1000 PYRs [95% CI, 11-15 deaths per 1000 PYRs]). Among the patients with fatal cases, 17% of HCV-positive patients were registered with alcohol abuse prior to death, versus 9% of the HCV-negative patients. Additionally, HCV-positive patients had an increased risk of dying with known alcohol abuse (mortality rate, 11 deaths per 1000 PYRs [95% CI, 7.0-17] vs. 2.5 deaths per 1000 PYRs [95% CI, 1.7-3.8]).
RESULTS
Descriptive Data
Baseline clinical and demographic data are provided in table 1. In total, 2734 HIV-1-infected patients initiated HAART in Denmark from 1 January 1995 to 1 January 2004. Of these, 443 patients (16%) were infected with HCV (the HCV-positive group, defined as having a positive anti-HCV serological test result or presence of HCV RNA); 2183 patients (80%) were not infected with HCV (the HCV-negative group, defined as having a negative anti-HCV test result); and 108 patients (4%) were never tested for HCV antibodies or the presence of HCV RNA (the HCV-U group). The study included 12,356 person-years of follow-up, with a median follow-up of 4.7 years. During the study, 3.4% of patients were lost to follow-up because of emigration. The 3 HCV groups were similar with respect to median CD4+ cell count and HIV RNA load at HAART initiation. The proportion who experienced an AIDS-defining event or exposure to antiretroviral drugs prior to the initiation of HAART also did not differ among the groups (table 1). A longer median time from HIV diagnosis to HAART initiation was observed in the HCV-positive group. Patients in the HCV-positive group were significantly less likely to initiate HAART than were patients in the other groups (adjusted incident rate ratio, 0.64; 95% CI, 0.58-0.72). This effect remained after excluding patients with a history of injection drug use (adjusted incident rate ratio, 0.82; 95% CI, 0.69-0.97). In the HCV-positive group, 134 patients (30%) had documented treatment interruption >90 days in duration, compared with 325 (14%) in the HCV-negative group. In 20 patients (4.5%) in the HCV-positive group and 36 patients (1.5%) in the HCV-negative group, gastrointestinal toxicities (including liver-related toxicity) were registered as the reason for treatment interruption.
Viral Suppression and Immune Recovery After Start of HAART
Fewer patients in the HCV-positive group than in the HCV-negative group attained a viral load <500 copies/mL (figure 1), and this difference held after exclusion of injection drug users (data not shown). The prevalence of patients with a viral load >500 copies/mL was higher at week 144 and week 288 in the HCV-positive group, compared with the HCV-negative group (week 144 crude OR vs. week 288 crude OR, 1.6 [95% CI, 1.2-2.0] vs. 1.8 [95% CI, 1.3-2.6]), but this disparity decreased after adjustment (week 144 adjusted OR vs. week 288 adjusted OR, 1.1 [95% CI, 0.80-1.5] vs. 1.3 [95% CI, 0.88-2.0]). After the exclusion of injection drug users, the week 144 crude OR was 1.3 (95% CI, 0.87-1.9) with an adjusted OR of 0.93 (95% CI, 0.61-1.4), and the week 288 crude OR was 1.3 (95% CI, 0.80-2.2) with an adjusted OR of 1.0 (95% CI, 0.60-1.8). It is of note that 1 variable-the presence of a documented treatment interruption >90 days in duration-produced the main change in the estimated effect.
The HCV-positive group attained lower absolute CD4+ cell counts after start of HAART than did the HCV-negative group (figure 2), and this effect was still observed after the exclusion of injection drug users (data not shown) (regression coefficient for week 144 vs. regression coefficient for week 288, 1.9 [95% CI, 1.3-2.6] vs. 1.8 [95% CI, 0.78-2.8]). In an analysis that excluded patients infected via injection drug use, this difference was less pronounced (regression coefficient for week 144 vs. regression coefficient for week 288, 1.9 [95% CI, 0.91-2.8] vs. 0.76 [95% CI, -0.60 to 2.1]).
Figure 2. Median absolute CD4+ cell count, by hepatitis C virus (HCV) infection status and number of weeks after start of HAART. Solid line, HCV-negative patients; dotted line, HCV-positive patients.
DISCUSSION
In this prospective, nationwide study, we found that HCV-HIV coinfected patients had a weaker response to HAART, in terms of viral load suppression and improvement in CD4+ cell counts, compared with patients infected with only HIV. The viral load findings are explained by differences in covariates, mainly injection drug use and compliance with therapy. However, our finding that coinfected patients had increased overall mortality was only partly explained by higher rates of liver- and AIDS-related mortality.
To our knowledge, this is the first study of HCV-HIV coinfection that includes all patients initiating HAART on a nationwide basis. The complete inclusion of patients, the long and nearly complete follow-up, and completeness of data on HCV status are major study strengths. The importance of data completeness is clear from a recent study that demonstrated that missing data on HCV serological status may introduce considerable bias in HCV cohort studies [18].
It must be noted that patients were grouped according to HCV serological status rather than by active HCV replication, as measured by HCV PCR. Thus, some HCV-positive patients may have cleared the virus subsequent to HCV infection, but this proportion seems small [19]. Another possible shortcoming is that HCV status was not handled as a time-dependent variable. However, the bias that this introduces is probably minor and would lead to the underestimation of the impact of HCV infection. The vast majority of patients presented to the clinics with HCV-HIV coinfection, and nearly none were infected with HCV after the start of HAART.
After HAART initiation, HCV-HIV-coinfected patients in our study had lower absolute CD4+ cell counts than did HCV-negative patients. This is in line with the findings of the Swiss Cohort Study, which found that HCV-HIV-coinfected individuals had a blunted response to HAART with less improvement in CD4+ cell count [5]. A recent study also found that immunological parameters were compromised in HCV-infected patients receiving HAART [20].
We observed increased liver-related and total mortality among HCV-HIV-coinfected patients. To reduce potential information bias, we also estimated the risk of dying with a prothrombin time <0.3 and found that this risk increased in HCV-positive patients as well. This observation is consistent with those of the Swiss Cohort Study, which showed that, among 1157 HIV-infected patients receiving HAART, those coinfected with HCV had a higher risk of liver-related death (7.7 deaths per 1000 PYRs) [5]. In contrast, the Johns Hopkins Cohort Study found no increase in mortality in HCV-HIV-coinfected patients receiving HAART [6]. However, the latter study included patients from only 1 treatment center, which makes it difficult to apply its findings to other instances. Both studies differed from ours in that they had shorter follow-up periods, were not completely population based, and included a higher proportion of injection drug users. On the basis of adjusted analyses, a recent study from the EuroSIDA group found no impact of HCV infection on the effectiveness of HAART in terms of changes in viral load or CD4+ cell counts [7]. However, the HCV-infected patients in the EuroSIDA study had a higher overall mortality rate and, in the instance of liver-related death, this group observed an incidence rate ratio of 11.7, which is comparable to that seen in our study.
Although liver-related causes of death explain some of the excess mortality observed among the HCV-positive group, other non-HIV-associated risk factors also seem to increase mortality. In accordance with an article from the EuroSIDA group [21], we found that patients with HCV infection were more likely to discontinue their HAART regimens. In our study, we used HAART discontinuation as a surrogate marker for compliance, because structured treatment interruptions are infrequently included in treatment regimens in Denmark.
Lack of HAART has been shown to be a strong predictor for faster progression to cirrhosis among HIV-HCV-coinfected patients [10, 22-24]. The delayed start of HAART in our HCV-positive group, combined with lower compliance, may therefore have had an impact on the differences observed in HAART response among HCV-HIV-coinfected individuals, compared with the HCV-negative group. It is also likely that patients with more-advanced liver disease were less likely to be prescribed HAART because of their poorer physical condition and the known hepatotoxicity of antiretroviral drugs [11].
Our observed prevalence of HIV-HCV-coinfection was generally low, compared with other cohort studies [4, 6], and not surprisingly, the majority of the HCV-positive patients were injection drug users. The impact of HCV infection on the effectiveness of HAART is, therefore, difficult to distinguish from the impact of injection drug use. However, in analyses excluding patients who use injection drugs, we still observed increases in total and liver-related mortality.
In the present study, the HCV-coinfected group differed markedly from the HCV-negative group in time from HIV diagnosis to HAART initiation. This finding may be partly explained by the significantly higher proportion of HCV-coinfected patients who were diagnosed with HIV during the pre-HAART era. However, when HIV diagnosis prior to 1995 was introduced into the multivariate analysis, it had only a minor influence on outcome estimates.
Our data clearly show an increased risk of liver-related and AIDS-related mortality among HCV-coinfected patients. In addition, our observations of lower compliance, higher risk of not initiating HAART, and increased frequency of injection drug use, as well as the magnitude of alcohol abuse among patients with fatal cases, suggest that HCV coinfection is a marker for psycho-social risk factors that could lead indirectly to increased mortality. Increased attention to these factors is required to improve the prognosis in HCV-HIV-coinfected patients.
MATERIALS AND METHODS
Setting
As of 1 January 2005, the population of Denmark was 5,411,405 persons [12], with an estimated HIV prevalence of about 0.1% in the adult population [13]. Antiretroviral treatment and other medical care are provided free of charge to all HIV-positive residents of Denmark. Patients with HIV infection are treated in 1 of the country's 8 specialized medical centers, where they are seen on an outpatient basis at intended intervals of 12 weeks. During our study's follow-up period (1 January 1995-31 December 2003), the national criteria for initiation of the HAART regimen were HIV-related disease, acute HIV infection, pregnancy, a CD4+ cell count <300 cells/uL, and-until 2003-a plasma HIV RNA load >100,000 copies/mL. Very few HIV-infected patients (<15) have been treated for hepatitis C with IFN during the study period.
Study Population and Data Collection
The Danish HIV Cohort Study is a population-based, prospective, nationwide cohort study of all HIV-infected individuals >15 years old at diagnosis who were treated at Danish HIV centers after 1 January 1995 [14]. The cohort is ongoing, and patients are consecutively enrolled. A unique 10-digit civil registration number assigned to all individuals in Denmark enables the centers to avoid multiple registrations. The cause of death is specified by the treating center and is specifically coded as being AIDS-related, when applicable.
The current analysis focuses on the 2734 patients in the cohort who started HAART (defined as a regimen consisting of at least 3 antiretroviral drugs, including either abacavir, a nonnucleoside reverse-transcriptase inhibitor, or a protease inhibitor) before 1 January 2004.
Medical records of those patients who died from 1 January 1995 to 1 January 2004 were examined by physicians at the treating centers. Deaths were categorized as: (1) definitely non-liver related, (2) probably non-liver related, (3) probably liver related, or (4) definitely liver related. Information on alcohol abuse and most recent prothrombin time were obtained from patient files, as well.
Predictor Variables
The patients were grouped into 3 categories: (1) those having at least 1 serological test result positive for HCV and/or detection of HCV RNA (HCV-positive group), (2) those having at least 1 serological test result negative for HCV and no history of a positive test result (HCV-negative group), and (3) those never tested for HCV antibodies or the presence of HCV RNA (HCV-U group). Patients with 1 positive test result for HCV were included in the HCV-positive group for the entire follow-up period. The following covariates were considered for inclusion in the multivariate models: sex, race, age at HIV diagnosis (>40 years old vs. <40 years old), year of HIV diagnosis (before vs. after 1 January 1995), previous AIDS-defining event, CD4+ cell count at the start of HAART (<100 cells/uL vs. >100 cells/uL), viral load at start of HAART (<100,000 copies/mL vs. >100,000 copies/mL), antiretroviral treatment status at start of HAART (naive vs. experienced), year of HAART initiation (before vs. after 1 January 1998), and documented treatment interruption of >90 days in duration. Injection drug users were defined as those who acquired HIV infection through injection drug use.
Outcomes
Response to HAART was assessed by the proportion of patients with an HIV RNA load <500 copies/mL and by absolute CD4+ cell counts. Values for 12-week intervals were computed as described elsewhere [15].
The following clinical outcomes were used in the mortality analyses: time from initiation of HAART to (1) death from any cause, (2) liver-related death (definite or probable liver-related death, as recorded by the treatment centers), (3) death when the last recorded prothrombin time was <0.3, (4) death when the last recorded CD4+ cell count was >100 cells/L, (5) death with a known history of alcohol abuse, and (6) an AIDS-related death (defined as death attributed by the treatment center to an AIDS-defining disease).
Statistical Analysis
Intergroup baseline characteristics were compared using the x2 test and the Kruskal-Wallis test. P < .05 was considered to be statistically significant.
Viral load. In the analysis of patients whose HIV RNA load was <500 copies/mL at week 144 and at week 288 after HAART initiation, we used logistic regression to estimate prevalence ORs and associated 95% CIs. The "change in estimate" principle was used, so that only factors changing the estimate by at least 10% were included in the final model [16]. The following covariates were forced into the model: AIDS diagnosis prior to HAART initiation, antiretroviral-naive status at HAART initiation, CD4+ cell count at HAART initiation, total treatment interruption >90 days in duration, and age at the time of HIV diagnosis.
CD4+ cell count. Absolute CD4+ cell count was compared between groups at weeks 144 and 288 by means of multivariate linear regression, using the variables listed above. In these analyses, CD4+ cell counts were square root-transformed to obtain a normal distribution of the residuals.
Mortality. Person years at risk (PYRs) were counted from the start of HAART to the date of death or date of last follow-up visit (censored at 384 weeks). Kaplan-Meier analysis was used to construct survival curves. In the analysis of time to liver- or AIDS-related death, or to death with a last recorded prothrombin time <0.3, with a last recorded CD4+ cell count >100 cells/uL, or with known alcohol abuse, patients lacking a given characteristic were censored at death. As previously described, Cox regression analysis was used to estimate mortality rate ratios (MRRs) adjusted for the variables cited above [17]. The proportional hazard assumptions were assessed graphically and found appropriate. Selection of other potential confounders was performed using the "change in estimate" method [16]. In all mortality analyses, the following variables were forced into the models: AIDS diagnosis prior to start of HAART, antiretroviral-naive status at start of HAART, CD4+ cell count at start of HAART, treatment interruption of >90 days in duration, and age at HIV diagnosis. Patients were included in the analysis until death or last follow-up visit, even when their antiretroviral treatment was stopped or changed to <3 drugs. Mortality rates were computed using 1000 PYRs.
HCV infection diagnosis and likelihood of HAART. To examine whether patients who received a diagnosis of hepatitis C initiated HAART at the same frequency as other patients, we computed the observation time as follows: date of fulfilling 1 criterion for HAART initiation (censored at 1 January 1997) until date of last observation, with HAART initiation as the outcome measure. Incidence rate ratios were estimated by Cox regression analysis that included the following variables as possible covariates: sex, race, year of HIV diagnosis, injection drug use, and previous AIDS defining event.
We estimated that <20 patients in the HCV-U group had HCV infection, and therefore, the HCV-U and the HCV-negative groups were pooled in statistical analyses. The study was approved by the Danish Data Protection Agency. SPSS statistical software, version 12.0 (Norusis, SPSS) was used for data analysis.
|
|
| |
| |
|
|
|
