| |
Impact of Maternal HIV Coinfection on the Vertical Transmission of Hepatitis C Virus: A Meta-analysis
|
| |
| |
Clinical Infectious Diseases April 15, 2007;44:1123-1131
Chelsea B. Polis,1 Snehal N. Shah,2 Kristine E. Johnson,3 and Amita Gupta3
Departments of 1Population, Family and Reproductive Health and 2Molecular Microbiology and Immunology, Johns Hopkins Bloomberg School of Public Health, and 3Division of Infectious Diseases, Johns Hopkins University School of Medicine, Baltimore, Maryland
ABSTRACT
Background. Observational studies suggest that maternal human immunodeficiency virus (HIV)-hepatitis C virus (HCV) coinfection is associated with increased odds of vertical HCV transmission. We performed a meta-analysis to summarize current evidence.
Methods. We systematically searched for relevant articles published during the period from January 1992 through July 2006 and independently abstracted articles that met our inclusion criteria. Under a random effects model, we calculated the pooled odds ratio for vertical HCV transmission according to maternal HIV-HCV coinfection status and performed sensitivity analyses.
Results. Ten articles met our inclusion criteria. Study quality varied widely, and study estimates displayed high statistical heterogeneity. Restriction of the analysis to studies that included >50 HIV-HCV-coinfected women provided our most reliable estimate: maternal HIV-HCV coinfection increases the odds of vertical HCV transmission by about 90% (odds ratio, 1.9; 95% confidence interval, 1.36-2.67), compared with maternal HCV infection alone. When we restricted analyses to HIV-infected mothers with HCV viremia, the odds of vertical HCV transmission were 2.82-fold (95% confidence interval, 1.17-fold to 6.81-fold) greater than the odds for HIV-infected mothers without HCV viremia.
Conclusions. HIV-HCV-coinfected women have significantly higher odds of transmitting HCV to their infants than do women who are infected with HCV alone.
Globally, an estimated 170 million people (or 3.1% of the world's population) are infected with hepatitis C virus (HCV), and 40 million individuals are infected with HIV [1, 2]. Because these infections share transmission routes, HIV-HCV coinfection is not uncommon; it affects about 4-5 million people [3, 4]. Both infections can be transmitted vertically from mother to child. Precise quantification of the number of HIV-HCV-coinfected women of childbearing age is difficult, but the shared risk groups and overlapping geographical distribution of the 2 infections suggests that HIV-HCV coinfection is an important public health issue. Efforts to prevent vertical HCV transmission will become increasingly important as the epidemic of HIV-HCV coinfection evolves.
The rate of risk of vertical HCV transmission ranges from 4% to 10%, and viremic women appear more likely to transmit HCV [5, 6]. Unlike HIV, the contributions of other potential risk factors for vertical transmission of HCV, such as breast-feeding and mode of infant delivery, remain poorly characterized. Understanding the epidemiology and risk factors for vertical transmission of HCV is essential in guiding future public health efforts. Although evidence suggests that, for the HCV-infected infant, disease progression is slow during the first 10-15 years of life, little is known about disease progression beyond adolescence [6]. In the United States, the prevalence of HCV infection in children ranges from 0.05% to 0.4% [7].
Observational studies suggest that maternal HIV-HCV coinfection is associated with increased vertical HCV transmission. The potential biological mechanisms responsible for this association are not clearly understood, although virologic and host factors likely play a role. HIV infection has been associated with an elevated HCV load [8], which may facilitate HCV vertical transmission. Other factors believed to influence HCV transmission rates include HCV genotype, severity of HCV-induced liver disease, hepatic inflammation (i.e., elevated alanine aminotransferase level), stage of HIV infection, HIV antiretroviral experience, active injection drug use (IDU), infant sex, and infant prematurity.
Pappalardo's [9] 2003 meta-analysis of 10 observational studies reported an OR of HCV vertical transmission of 2.82 (95% CI, 1.78-4.45) for HIV-HCV-coinfected mothers, compared with HCV-monoinfected mothers. Since the time of Pappalardo's publication [9], 2 additional studies have contributed to the knowledge base, including the largest cohort study to date, which was designed specifically to evaluate vertical HCV transmission in HIV-HCV-coinfected mothers [10, 11]. Because of the availability of these new data, we conducted a meta-analysis of published studies to investigate the impact of maternal HIV coinfection on the vertical transmission of HCV.
RESULTS
Our literature search identified 243 articles. We retrieved 36 full-text articles for review, 26 of which did not meet our inclusion criteria (figure 1). Four included only women with HIV-HCV coinfection, and 14 included data for <20 HIV-HCV-coinfected women. Three articles did not meet our diagnostic criteria, 3 had unclear or missing data, and 2 did not use original data. The funnel plot provided no evidence of publication bias (figure 2) [10, 11, 17-24].
Qualitative analysis. Publication dates ranged from 1993 to 2005 (table 1). Nine studies were conducted in Europe [10, 11, 18-24], and of those, 7 were based in Italy [10, 19-24]. Only 1 study was based in the United States [17]. None of the studies were conducted in developing countries. Nine were prospective cohort studies [10, 11, 17, 19-24], and 1 was retrospective [18]. Studies generally recruited pregnant HCV-infected women, and many participants had a history of IDU. The sample sizes and proportion of HIV-HCV-coinfected mothers were highly variable.
The number of mother-infant pairs in each study ranged from 37 to 1479 pairs, and the proportion of HIV-HCV-coinfected women ranged from 13.7% to 85.7%. Although all studies used standard HCV diagnostic methods, tests were not always routinely applied to all women (table 2). In addition, the exact timing of HCV maternal testing (antenatal, intrapartum, or early postpartum testing) was not specified. Only 4 studies performed quantitative PCR for HCV RNA for all women [21-24]. No studies provided information on maternal HIV stage or HIV load, and only 2 reported partial information on HIV antiretroviral treatment, which may affect the HIV load but has an unclear impact on HCV load [10, 11].
A total of 287 HCV-infected infants were identified, but infant HCV diagnosis and follow-up intervals varied (table 1). Six studies conducted HCV EIA antibody testing on infants after they were 18 months of age [10, 17-19, 21, 23]. Three of these studies also performed 2 HCV RNA PCR tests at least 3 months apart [18, 19, 23]. One study did not use antibody testing and instead diagnosed infant HCV infection by performing 2 HCV RNA PCR tests >3 months apart [24]. Of the remaining 2 studies, one diagnosed infant HCV infection using HCV EIA antibody tests without confirmatory RIBA after the subjects were 18 months of age [20], and the other performed 2 PCR tests at unspecified times [22]. Where reported, the duration of infant follow-up ranged from 6 to 90 months. Given the wide variation in diagnostic methods, misclassification of infant HCV infection may have occurred, although we attempted to correct for potential misclassifications associated with blood transfusions.
Studies were of varied quality (table 3). Six studies [17, 18, 21-24] handled misclassification bias well by identifying the means of transmission and/or by using superior HCV diagnostic criteria, whereas 4 studies [10, 11, 19, 20] were less clear about the possibility of misclassification bias. One retrospective study [18] examined stored blood specimens obtained from pregnant women several years after collection, but poor storage conditions may have led to misclassification.
Only 1 study [17] attempted to make groups comparable by matching, so the potential for selection bias was generally high in these studies. Only 1 study [20] provided clear information on whether differential loss to follow-up was likely. Regarding the comparability of HIV-infected and HIV-uninfected women at baseline, 6 studies provided IDU data [10, 20-24], whereas fewer provided information on Cesarean delivery [21-24], breast-feeding [11, 22-24], serum alanine aminotransferase level [23, 24], premature birth [11, 21], or HIV antiretroviral use [10, 11]. Among the 6 studies that reported IDU according to HIV status, 61% of HIV-HCV-coinfected women reported IDU, whereas only 26% of HCV-only infected women reported IDU, resulting in substantial differences between potentially confounding characteristics of HIV-HCV-coinfected women and women with HCV infection alone. Although 6 studies mentioned HCV genotype [10, 17, 18, 22-24], information generally only confirmed that infant HCV genotype matched the mother's. Only 2 studies [17, 19] provided CD4 cell counts, and 1 study [17] provided information on HIV load. If loss to follow-up is associated with HIV status, which is theoretically conceivable, this also could have introduced bias.
Eight studies performed univariate analysis of risk factors associated with HCV transmission [10, 17-19, 21-24], and of these, 3 identified HIV coinfection as a significant risk factor for vertical HCV transmission [21-23]. Two studies performed multivariate analysis [11, 20]. The first [20] controlled for maternal IDU, maternal HIV infection, breast-feeding, and vaginal delivery but performed the analysis only for a subset of mothers with positive HCV PCR results. The authors concluded that maternal IDU alone was significant. The second multivariate analysis [11] included maternal HIV status, mode of delivery, sex of infant, prematurity, and breast-feeding and found that only the sex of infant was a significant factor. Neither analysis suggested that maternal HIV infection was a statistically significant predictor of HCV transmission.
Quantitative analysis. Of 4424 mother-infant pairs, 858 (19.39%) included HIV-HCV-coinfected women. A total of 278 HCV-infected infants were born. Analysis of these 10 observational studies (after excluding potentially misclassified information) in a random effects model revealed that HIV-HCV-coinfected women have a 2.75 greater odds of transmitting HCV, compared with women with HCV infection alone (OR, 2.75; 95% CI, 1.51-4.99). However, the I2 statistic was 60%, indicating that the study estimates were heterogeneous and ideally should not be pooled (table 4).
We performed a subgroup analysis of 7 studies that reported data on mothers with HCV viremia [10, 17, 19, 20, 22, 24]. The pooled odds of vertical transmission were 2.82 (95% CI, 1.17-6.81) among HIV-HCV-coinfected mothers with HCV viremia, compared with HCV-monoinfected mothers with HCV viremia (figure 3). The I2 statistic was 64%, again indicating that the study estimates were heterogeneous and should not be pooled.
The removal of the retrospective study [18] did not change the OR substantially (OR, 3.02; 95% CI, 1.66-5.50), and the I2 statistic indicated significant heterogeneity.
When we restricted studies to the 5 that had sample sizes of 50 subjects [11, 17, 19-21], the risk of HCV vertical transmission was found to be 1.90 (95% CI, 1.36-2.67), and the I2 statistic for heterogeneity was 50% (I2, -15%).
The removal of the study that had no transmission events in the HCV monoinfection group (and, therefore, an unstable zero cell) [22] produced an OR of 2.31 (95% CI, 1.42-3.75). Furthermore, the I2 statistic decreased to 40%, indicating that this study was the major source of statistical heterogeneity.
DISCUSSION
The results of our meta-analysis found that, compared with maternal HCV infection alone, maternal HIV coinfection increases the vertical transmission risk of HCV. We suggest that our most reliable estimate comes from the pooled analysis that included studies with 50 HIV-HCV-coinfected mothers (OR, 1.9; 95% CI, 1.36-2.67). As a group, these larger studies did not differ drastically in quality, showed low heterogeneity, and were of better overall quality. We, therefore, have determined that the odds of vertical HCV transmission are 90% higher for women infected with HIV and HCV than for those infected with HCV alone.
This estimate is similar to Pappalardo's estimate (OR, 1.97) [9]. However, we reached these estimates using different methods and studies. Our article selection criteria excluded 3 studies included in Pappalardo's analysis (1 conference abstract [25], 1 study that included <20 coinfected women [13], and 1 study that did not meet our criteria on antibody testing [26]). In addition, our review includes 2 new studies [10, 11], one of which was the largest study conducted to date on this topic. We also included the 1995 study by Zanetti et al. [22], which was excluded from Pappalardo's graphical analysis and study table. Finally, our analysis also differs from Pappalardo's in that we excluded potentially misclassified information from the 1998 study by Granovsky et al. [17].
Because observational studies have a greater risk of bias, it is more difficult to confidently use statistical meta-analyses to generate a meaningful summary statistic. The lack of comparability between these studies generally resulted from a lack of standardized HCV diagnostic criteria and the inability to control for known confounders. By nature of the exposure, high-risk individuals must be targeted for enrollment, and this can introduce high rates of loss to follow-up and may limit generalizability to areas where IDU is not the main mode of HIV acquisition. Our findings are likely generalizable to populations in which most cases of HCV infection are related to IDU, but they should be applied with caution to areas such as the developing world, where nutritional status, HIV subtype, HCV genotype, and other differences may influence HCV transmission.
Little information exists on how known confounding factors influence the relationship between HIV-HCV coinfection and HCV transmission. Information on Cesarean delivery may be particularly relevant, because most of the studies were conducted in Italy, where the procedure is common [27]. If Cesarean deliveries are protective against HCV transmission but not considered in studies, researchers may underestimate transmission rates for HIV-HCV-coinfected mothers, because most HIV-infected mothers in the United States and Europe deliver via Cesarean section.
Some studies identify maternal HCV viremia as a risk factor for HCV transmission, but few studies have tested for viremia [28, 29]. In a comparison of analyses restricted to mothers with HCV viremia, our OR was higher than that in the study by Pappalardo [9] (2.82 vs. 1.97), probably because our meta-analysis included 2 additional studies in which a number of mothers were tested for HCV RNA [10, 11]. Misclassification may have occurred among HCV antibody-positive mothers, because 15% of women clear HCV. In addition, improvements in HCV testing over time may have led to decreased instances of misclassification in more recent studies. However, this bias should be nondifferential, because evidence suggests that there is no difference in HCV clearance based on HIV serostatus [5, 23]. Because of the persistence of maternal antibodies and HCV RNA positivity in the absence of antibodies, serial measurements of HCV RNA level should also be performed in infants to avoid misclassification [17, 21, 26].
Our analysis indicates that large studies that control for potential known confounders, use clear selection criteria, show minimal loss to follow-up, and employ standardized HCV testing of infants and mothers using qualitative tests are needed. In addition, better understanding of the underlying biological mechanisms of HCV transmission in relation to HIV infection in both mothers and infants is needed to improve the quality of future studies on this topic. Future research in this area is critical to the development of effective public health measures to reduce vertical transmission of HCV to infants from HIV-HCV-coinfected mothers.
|
|
| |
| |
|
|
|
