Meta-analysis: increased mortality associated with HCV in HIV-infected persons is not related to HIV disease progression

Abstract

Background

It is unclear whether coinfection with HCV increases mortality in HIV patients in the HAART era. Using a meta-analysis, we estimated the effect of HCV on HIV disease progression and overall mortality in the pre- and HAART era.

Method

PubMed and EMBASE were searched for studies published until April 30, 2008. Additional studies were identified from cited references. Studies reporting disease progression or mortality in HCV/HIV coinfected patients were selected. Cross sectional studies, studies without HCV negative controls and studies among children and/or post-liver transplant were excluded. Two authors reviewed articles and extracted data on demographics of study populations and risk estimates. Meta-regression was used to explore heterogeneity.

Results

10 pre-HAART studies and 27 HAART era studies were selected. In the pre-HAART era, the risk ratio for overall mortality among HCV/HIV coinfected patients was 0.68 (95% CI: 0.53~0.87) compared to patients with HIV infection alone. In the HAART era, the risk ratio was 1.12 (95% CI: 0.82~1.51) for AIDS-defining events and 1.35 (95% CI: 1.11~1.63) for overall mortality in coinfected patients compared to HIV monoinfection.

Conclusions

HCV coinfection did not increase mortality in HIV patients before the introduction of HAART. In contrast, HCV coinfection increases the risk of mortality but not AIDS-defining events compared to HIV infection alone in the HAART era. Future studies should determine whether successfully treated HCV could reduce this excess risk of mortality in coinfected patients.

Introduction

Hepatitis C virus (HCV) and human immunodeficiency virus (HIV) share routes of transmission, and thus the prevalence of HCV infection is approximately 15-30% in HIV infected patients. Due to enhanced transmissibility of HCV by percutaneous injection compared to HIV, HCV prevalence can exceed 85% in HIV-infected injection drug users [1, 2]. Before the introduction of highly active antiretroviral therapy (HAART), mortality related to HIV overwhelmed that related to HCV. By contrast, while HIV patients live longer in the HAART era, chronic diseases such as viral hepatitis have emerged as important causes of morbidity and mortality [3, 4]. It has been strongly suggested that HIV infection accelerates HCV-related disease progression and mortality [5-7] but the reciprocal effect of HCV on the rate of progression of HIV remains muddled due to the heterogeneity of study results. Understanding whether coinfection with HCV affects progression and mortality related to HIV may elucidate challenging issues for providers treating HIV patients with HCV coinfection and shed insight into the complex interactions between these two persistent viruses.

Some studies have reported a strong association between HCV/HIV coinfection and increased risk of HIV disease progression [8, 9], while other studies have not confirmed this result after the widespread use of HAART [10, 11]. This systematic review estimates the effect of HCV on both mortality and HIV progression in HCV/HIV coinfected patients in the era of HAART. We compared studies from both the pre-HAART and the HAART era, focusing on the latter as these are most likely to inform future practice.

Methods

Data Sources

A literature search was conducted to identify publications reporting disease progression or survival of HIV among HIV and HCV coinfected cohorts using PubMed and EMBASE without language restriction until April 30, 2008. Titles and/or abstract were screened to determine the relevance of studies. Full texts of selected studies were reviewed. Additional studies were identified from cited references.

Study Selection

Publications that reported on disease progression, mortality and/or survival of HIV among adult HIV and HCV coinfected adult cohorts were selected for inclusion, while studies among nonadolescent children were excluded. Studies were classified in the pre-HAART era if occurring entirely in the period before January 1996 and in the HAART era if half or greater than half of the study period was after January 1996, unless specifically split into two categories by the author. Publications on liver disease-related mortality were not included unless overall mortality was also reported due to potential misclassification. Cross-sectional studies, studies on survival after liver transplant, and studies without HIV-monoinfected control groups were excluded.

Data Extraction

One author (TC) completed the search and extracted data from the studies on two occasions. Study design, period and population, number of subjects, median/mean follow up, treatments for HIV and HCV, outcome measures, percentage of intravenous drug users (IDU) and adjustments for potential confounders were extracted by two authors (TC and AK). For studies reporting only incidence rates or cumulative incidence rates of death or AIDS-defining events, risk ratio and 95% confidence interval were calculated from the available data provided. If studies presented several risk estimates, the final adjusted risk estimate was recorded.

Study Quality Assessment

Study quality was assessed for studies in the era of HAART with primary outcomes of overall mortality and/or AIDS-defining diseases. Quality of these studies was evaluated by the following: 1) study design: prospective studies were considered to have higher quality than retrospective; 2) median follow-up time: studies with a longer duration of follow-up were considered more likely to reflect the true effect of HCV on HIV-related disease progression.

Statistical Analysis

Quantitative analyses were conducted for overall mortality in the pre-HAART era and three different outcome measures: overall mortality, AIDS-defining events and the combination of AIDS-defining events and mortality in the HAART era. Conventional random effect models were used to estimate summary risks ratios and 95% confidence intervals given the variability inherent among observational studies [12]. Heterogeneity across studies was evaluated by I² test [13]. I² statistic greater than 20% suggests heterogeneity and an I² statistic greater than 50% indicates substantial heterogeneity. Publication bias was assessed with funnel plots, Begg's and Egger's tests [14, 15]. Sensitivity analysis was conducted by omitting one study at a time to examine the influence of individual studies for overall mortality. Univariate random-effect meta-regressions using aggregate-level data including percentage of patients on HAART, percentage of IDU patients in the study population, study duration and gender were performed for overall mortality in HAART era to assess potential effect modification [16]. The effect of study quality was assessed by subgroup analysis. STATA v. 9 was used for all analysis.

Results

Study Selection

Figure 1 summarizes the study selection process. The electronic search yielded 537 citations in PubMed and 623 in EMBASE. Among the total of 1160 titles/abstracts screened, 52 papers were selected for in-depth review. Seven papers provided data from eras pre- and post- HAART introduction [11, 17-22]. For pre-HAART analysis, 10 studies were excluded for one or more of the following reasons: 1) AIDS and overall mortality were combined as a single endpoint [17, 23, 24], 2) lack of HCV-negative control [21, 25, 26] or 3) data were not able to be pooled [22, 27-29]. As a result, data were extracted for 10 studies [11, 18-20, 30-35]. For analysis of studies conducted in the HAART era, 12 studies were excluded for one or more of the following: 1) they lacked an HCV-negative control group [5, 21, 36-39], 2) used a cross-sectional design [40-42], 3) data were not extractable [22], or 4) the cohort was duplicated [43, 44]. Thus, data were extracted from 33 studies, 10 pre-HAART and 27 HAART era with 4 contributing to both eras.

Figure 1.

Selection process for study inclusion in the meta-analysis of HCV on overall mortality and/or HIV-related progression

Study Characteristics

Ten studies, involving 4413 HCV/HIV coinfected patients and 10213 solely HIV infected patients were included for the pre-HAART era, and 27 HAART era studies including 25,319 HCV/HIV coinfected patients and 61,697 HIV-monoinfected patients were included (Table 1). For the pre-HAART era, study sample sizes ranged from 95 [11] to 10,896 [18]. In the HAART era, study sample sizes ranged from 330 [10] to 23,441 [4]. Study outcomes included: overall mortality in 20 studies [4, 10, 11, 18-20, 45-58], AIDS-defining events in 7 studies [11, 19, 54-57, 59]; and combined AIDS-defining events and mortality in 7 studies [8, 9, 17, 54, 60-62]. One study reported all 3 outcome measures and 5 studies reported both mortality and AIDS-defining events. HAART use was defined as a regimen including at least three antiretroviral agents. Nine studies recruited only patients on HAART [8, 9, 46, 48-50, 58, 60, 61] and 2 studies did not report percentage of patients on HAART [18, 20]. The percentage of patients on HAART in the remaining studies ranged from 35 to 89 %. Most studies either did not report treatment of HCV or reported only small percentage (range: 0~7 %) of coinfected patients on treatment. Percentage of IDU in study populations varied from 0 to 64%. The study without IDU was selected for a subgroup analysis of non-IDU patients [20]. Mean/median follow-up time ranged from 1.1 to 5.9 years for studies conducted only in the HAART era.

Table 1.

Studies included in meta-analysis of HIV and HCV coinfection

Author (published year)	Country	Study design	Study year	Outcome measures	Study population	% of male	Mean/ median age	No. of HCV+/HCV−	% on HAART	HCV Tx	% of IDUΔ
Pre-HAART era
Diamondstone (1995)	USA	RS	1984~1993	M	Hemophilia	100	25	565 / 16	-	-	NR
El-Serag (2005)	USA	RS	1991~1996	M	Hospitalized Veterans	98	45	3048 / 7848	-	-	NR
Haydon (1998)	UK	PC+RS	1985~1996	M	IDU	68	23*	202 / 38	-	-	100
Hung (2005)	Taiwan	PC	1994~1997	M	Adults	92	36	11 / 84	-	-	3
Klein (2003)	Canada	PC+RS	1989~1996	M	Adults	86	36	42 / 620	-	-	5
Macias (1998)	Spain	RS	1989~1996	M	Adults	77	34	214 / 106	-	-	61
Ockenga (1997)	Germany	PC	1993~1996	M	Adults	78	38	56 / 154	-	-	28
Staples (1999)	USA	RS	1992~1997	M	Veterans	99	38	115 / 235	-	-	20
Voirin (2004)	France	PC	1992~1996	M	Adults	81	34	121 / 980	-	-	0
Wright (1994)	USA	RS	1984~1989	M	Adults	98	NR	39 / 132	-	-	40
HAART era
Anderson (2004)	USA	RS	1997~2001	M	Veterans	99	42	306 / 664	73	NR	22
Backus (2005)	USA	PC	1997~2003	M	Veterans	98	46	4668 / 7548	100	3% on Tx	37‡
Bonacini (2004)	USA	PC	1993~2001	M	Adults	84	41	256 / 126	52	NR	21
Braitstein (2005)	Canada	PC	1996~2003	M	Adults	85	37	606 / 580	100	NR	27
Crane (2007)	USA	RS	1995~2005	M	Adults	87	37	144 / 550	100	NR	NR
De Luca (2002)	Italy	PC	1997~2001	C	Adults	75	33	600 / 720	100	No Tx	39
Dorrucci (2004)	Italy	PC	1996~2001	C	Adults	66	27*	458 / 337	35	NR	56
El-Serag (2005)	USA	RS	1996~2001	M	Hospitalized Veterans	98	45	2272 / 4913	NR	NR	NR
Greub (2000)	Switzerland	PC	1996~2000	C	Age~16	71	36	1157 / 1954	100	NR	36
Hung (2005)	Taiwan	PC	1997~2002	M, A	Adults	92	36	42 / 303	85	No Tx	3
Jaen (2008)	Spain	PC+RS	1998~2004	C	Age~16	75	36	642 / 1135	100	NR	33
Jaggy (2003)§	Switzerland	PC	1997~2001	M	Adults	71	36	1645 / 2318	100	NR	36
Klein (2003)	Canada	PC+RS	1996~1999	M, A	Adults	77	38	83 / 456	58	NR	17
Lincoln (2003)	Australia	PC	1999~2002	C	Adults	94	NR	223 / 1481	61	NR	8
Marins (2005)	Brazil	RS	1995~2002	M	Adults	72	30	279 / 554	50	NR	34
Mayor (2006)	Puerto Rico	PC	1998~2003	M	Adults	70	38	193 / 163	45	NR	58
Monga (2001)	USA	RS	1994~1998	M	Veterans	99#	46	166 / 263	42	NR	30
Riley (2005)	USA	PC	1996~2002	M	Homeless	84	40	212 / 118	75	NR	64
Rockstroh (2005)	Europe	PC	1994~2004	M, A, C	Adults	75	36	1960 / 3997	51	2% on Tx	28
Stebbing (2005)	UK	PC	1996~ NR	A	Adults	87	34	85 / 1382	< 49%	NR	NR
Sulkowski (2002)	USA	PC	1995~2001	M, A	Adults	70	37	873 / 1082	61	<2% on Tx	45
Sullivan (2006)	USA	RS	1998~2004	M, A	Age ≥13	72	35	2024 / 8457	70	NR	21
Tedaldi (2003)	USA	PC	1996~2001	M, A	Adults	NR	37	267 / 556	84	7% on Tx	26
Torti (2007)	Italy	PC	1996~2002	C	Adults	78	38	334 / 417	100	NR	41
Voirin (2004)	France	PC	1996~2002	M	Adults	79	35	107 / 1273	NR	NR	0
Weber (2006)	International	PC	1999~2004	M	Adults	76	39	5274 /18167	89	NR	NR
Weis (2006)	Denmark	PC	1995~2004	M	Age > 15	75	39	443 / 2183	100	<1% on Tx	10

IDU: injection drug use, Tx: therapy, HAART: highly-active antiretroviral therapy, PC: prospective cohort, RS: retrospective, NR: not reported. M: overall mortality, A: AIDS, C: combination of AIDS and mortality.

^ΔCalculated according to the proportion reported in HCV positive and negative groups if not reported directly.

^‡Hard drug use.

^*Age at HIV seroconversion.

^#Assumption made according to the study population.

^§Demographic data derived from Greub et al.

Study Quality

Of 27 study designs in the era of HAART, 4 studies reported overall mortality or AIDS-defining diseases as secondary outcomes and hence were excluded from quality assessment [4, 49, 50, 61]. Of 23 studies with a primary outcome of interest, 5 were retrospective designs [18, 45, 51, 53, 56] and 1 collected data both retrospectively and prospectively [19]. There are 8 studies with mean/median follow-up time between 1 to 3 years [8, 11, 19, 47, 53, 55-57] and 10 studies have mean/median follow-up time over 3 years [9, 10, 17, 18, 46, 48, 54, 58-60]. Five studies did not report mean/median follow-up time [20, 45, 51, 52, 62].

Measured Outcomes

Pre-HAART

Overall mortality

The risk ratio for overall mortality among HCV/HIV coinfected patients was 0.69 (95% CI: 0.54~0.88, Figure 2) in comparison to patients with HIV-monoinfection. The chi-square test for heterogeneity was not significant (p=0.67) with I² statistic of 0 (95% CI: 0~62). Sensitivity tests on the influence of individual studies showed pooled risk ratios ranged from 0.66 to 0.87. The study by El-Serag et al. [18] showed a dominant influence on the pooled risk ratio: when removed, the protective effect of HCV coinfection became statistically insignificant with a risk ratio of 0.87 (95% CI: 0.62~1.23). Only one of nine studies from this era included use of antiretroviral therapy as a covariate, precluding analysis on the impact of treatment [34].

Figure 2.

Forest plots of overall mortality in pre-HAART era.

HAART era

Overall mortality

Compared to patients with HIV-monoinfection, the pooled risk ratio for overall mortality among HCV/HIV coinfected patients was 1.35 (95% CI: 1.11~1.63, Figure 3 A). The chi-square test for heterogeneity was not significant (p=0.13) with an I² statistic of 26 (95% CI: 0~58). A sensitivity test on the influence of individual studies showed pooled risk ratios ranging from 1.30 to 1.43, all statistically significant. Evaluation of publication bias by funnel plot showed a symmetric distribution of studies (data not shown). Neither Begg's test (p=0.72) nor Egger's test (p=0.95) suggested publication bias.

Figure 3.

Forest plots in HAART era. A: Overall mortality. B: AIDS-defining events. C: AIDS plus mortality

Potential sources of between-studies heterogeneity were explored by meta-regression stratified analyses and subgroup analyses. Stratified results are summarized in Table 2. Although stratification for sex, age, follow-up duration and treatment did not reveal significant effect modifications, the result suggested that adverse effects of HCV on overall mortality may be stronger among older patients and among patients on HAART. The adverse effect of HCV/HIV coinfection was also more apparent with longer follow-up. Subgroup analyses according to quality of studies yielded pooled estimates 1.22 (95% CI: 0.82~1.81) for 6 retrospective studies [18, 19, 45, 51, 53, 56] and 1.30 (95% CI: 1.02~1.66) for 11 prospective studies [10, 11, 20, 46-48, 52, 54, 55, 57, 58].

Table 2.

Stratified analysis of HCV coinfection and total mortality among HIV positive patients in HAART era

Variable	RR (95% CI)	P for interaction
Sex (n=19)*
Women	2.49 (0.57~11.00)	0.420
Men	1.19 (0.81~1.75)
Age (n=20), year
<38	1.28 (0.97~1.69)	0.611
≥ 38	1.42 (1.07~1.89)
Follow up duration (n=17)
< 3 years	1.11 (0.79~1.55)	0.120
≥ 3 years	1.54 (1.20~1.98)
IDU (n=17)*
Non-IDU	1.57 (0.98~2.51)	0.533
IDU	1.00 (0.37~2.73)
Concurrent Treatment (n=18)‡
Not on HAART	0.81 (0.42~1.56)	0.074
On HAART	1.70 (1.33~2.17)

IDU: injection drug use, HAART: highly-active antiretroviral therapy

^*Sex-specific and IDU-specific results estimated from meta-regression of sex-and IDU-proportion in each trial.

^‡HAART-specific results estimated from meta-regression of proportion of patients ever on HAART in each trial

AIDS-defining events

Seven studies reported data on the risk of developing AIDS-defining events among HIV patients with and without HCV infection (Table 1) [11, 19, 54-57, 59]. The pooled risk ratio was 1.12 (95% CI 0.82~1.51, Figure 3 B), indicating similar risk of developing AIDS-defining events between HIV infected patients with and without HCV infection. The chi-square for heterogeneity was not significant (p=0.88). The funnel plot did not show asymmetry, and neither Begg's test (p=0.13) nor Egger's test (p=0.11) supported publication bias.

Overall mortality and AIDS-defining events

Seven other post-HAART studies reported on the effect of HCV infection on HIV disease progression, defined as either the development of an AIDS-defining disease or overall mortality (Table 1) [8, 9, 17, 54, 60-62], with a pooled risk ratio of 1.49 (95%CI 1.08~2.05, Figure 3 C). No publication bias was observed.

Discussion

Our meta-analysis summarized 10 studies with 14626 patients in the pre-HAART era and 20 studies with 113073 patients in the HAART era. The study demonstrated a 32% reduction of overall mortality risk in HCV/HIV coinfected patient in the pre-HAART era and a 35% elevation of overall mortality risk in the HAART era.

The potentially protective role of HCV in mortality during the pre-HAART is dominantly influenced by El-Serag's study, which included many patients but also unique in its methodology, including restriction to an inpatient population and a retrospective design compared to other studies [19, 20, 34]. Recruiting patients through hospitalization is prone to survival bias. If patient died of AIDS-events prior to diagnosis of HCV, the study would not identify them especially considering that HCV was not routinely screened for in the pre-HAART era. Excluding this study, the remaining of selected studies did not show a protective effect individually or in aggregate.

By contrast, our meta-analysis demonstrates that HCV coinfection is associated with increased mortality in the HAART era. Subgroup analyses showed that: 1) a longer duration of follow-up is needed to observe a significant difference in mortality between HCV/HIV coinfected patients and HIV-monoinfected patients; and 2) the adverse effect of coinfection was more apparent in patients on HAART compared to those not on HAART. The latter finding parallels the observation that HIV-related mortality dominated the pre-HAART era [63]. Patients not on HAART are likely to die of AIDS-related complications regardless of HCV status and increased risk of mortality attributable to HCV coinfection is less likely to be observed.

The meta-analysis showed no statistically significant increased risk of developing AIDS-defining events among HIV patients with HCV coinfection in the HAART era, despite studies that suggested this possibility [25, 59]. There might be some effect of HCV on HIV disease progression that is masked because death from hepatic complications is a competing risk of AIDS-defining events. However, considering the prolonged time period required to develop HCV-related complications and that death due to HIV-related complications remains the leading cause of mortality in HIV-infected individuals [4, 64], competing risk and insufficient end-point events are unlikely to confound our results. When examining the few studies with combined AIDS-defining events and mortality as their outcome, we found increased (~50% higher) risk of HCV/HIV coinfected patients; however determining the relative contribution of each outcome was not possible. Therefore, potential mechanisms by which HCV may accelerate progression to AIDS, (i.e. a blunted immunologic response to HAART), may not carry significant clinical impact [65].

Given the lack of association of progression to AIDS and HCV coinfection, the major contributor to mortality among coinfected subjects during the HAART era is likely liver disease, based on an expanding body of supporting data [39, 66, 67]. Generally, AIDS-related mortality would be heralded by AIDS-related events. As AIDS-related events were not affected by coinfection status, excess mortality during the HAART era in coinfected patients compared to HIV-monoinfected patients was unmasked, coincident with the acceleration of liver disease present in almost exclusively the coinfected group. Unfortunately we did not observe that greater duration of HAART availability positively impacts coinfected patients, but rather the opposite effect—a trend towards a greater effect of HCV on mortality with length of study. While HAART may have benefits for liver disease progression and related-mortality [36, 68] it does not reverse it; implying that broader application of more effective anti-HCV therapies are needed to reduce this excess mortality. Moreover, if HCV is the major direct contributor to increased mortality, death rates between coinfected patients with HCV viremia versus those with spontaneous clearance (and reduced risk of liver-disease progression) should differ. Unfortunately, this parameter was not controlled for in past studies as ascertainment of HCV status was largely determined by antibody.

Independent of HCV pathogenesis, HCV coinfected patients may have increased mortality if they are less likely to be prescribed HAART and/or have poorer adherence to their regimens. Several studies within this meta-analysis observed shorter duration of exposure to HAART for HCV coinfected patients compared to HIV-monoinfection [17, 45, 46, 55, 57, 58]. In one study, coinfected patients were significantly less likely to initiate HAART than HIV-monoinfected patients even after adjusting for IDU [58]. Moreover, HCV coinfection is also independently associated with decreased adherence to therapy [69].

Since HCV is highly correlated with IDU, the increased risk of coinfection on mortality might arguably be confounded by IDU who have less access to HAART [70]. Despite an expectation that IDU would be an effect modifier, coinfection with HCV increased risk of death in both non-IDU and IDU patients in our subgroup analysis. The studies by Weis et al. and Klein et al. demonstrated increased risk of mortality after excluding/stratifying patients with IDU in their studies [19, 58], suggesting that the negative effect of HCV on mortality is independent of IDU. The non-significant interaction could be partially explained by lack of power due to the small numbers of coinfected patients without IDU (reported number ranging from 29 (15% of coinfected patients) [52] to 1774 (38% of coinfected patients) [46]) and misclassification of IDU to non-IDU. Our findings are nevertheless consistent with a recent study that did not find a significant association between IDU and increased mortality in HIV in those initiating HAART [71]. Mortality attributable to IDU is further confounded by the timing of HIV introduction into various populations and whether IDU is historical or current. HIV may have been acquired later in IDU populations compared to MSM [72] and thus HIV-related mortality may have been relatively delayed, resulting in a form of lead-time bias. Current IDU has additional risks from complications such as infections, drug overdose and homicide; however, these covariates were not reported in most studies. Finally, active IDU with the highest mortality rates may not be fully represented in these cohorts due to poor access to care and/or follow-up.

Our study has potential limitations. Although tests for publication bias were negative, studies presented solely in conferences or local journals may have been overlooked. In addition, each outcome analyses in HAART era may not be entirely comparable as studies differed regarding the data provided. Nevertheless, all analyses included more than 15000 HIV patients and sensitivity analysis revealed no significant change when subtracting any single study from the HAART era, suggesting that these limitations are unlikely to weaken the validity of our results. The observational nature of selected studies limits our ability to overcome residual confounding from individual studies, especially time from seroconversion to the initiation of HAART, alcohol consumption and GBV-C coinfection, factors which were analyzed only in a small minority or none of studies. However, twenty out of twenty-seven studies adjusted for important factors such as CD4 counts and HIV therapy. Finally, we could not examine the role of interferon-based treatments, but the lack of their widespread application in coinfected individuals suggests that HCV treatment would not be a major confounder during the era studied [55, 58, 73].

To conclude, this study synthesizes over thirty studies including over 100,000 patients and indicates increased risk of overall mortality in HIV patients with HCV coinfection in the HAART era. The meta-analysis did not demonstrate increased risk of developing AIDS-defining events in coinfected patients. Future studies examining mortality among coinfected subjects should attempt to dissect out the relative contributions of 1) HCV viremia as a surrogate for liver disease risk, 2) whether IDU is current or active and 3) whether broader application of HCV treatment positively impacts mortality of coinfected individuals.