|Home | About | Journals | Submit | Contact Us | Français|
To estimate whether HIV-infected pregnant women were at increased risk of hepatotoxicity when taking nevirapine (NVP) containing regimens compared to HIV-infected pregnant women taking antiretroviral therapy (ART) not containing NVP.
This analysis included HIV-infected pregnant women on ART from two multicenter, prospective cohorts: The Women and Infants Transmission Study (WITS) and the International Maternal Pediatric Adolescent AIDS Clinical Trials (IMPAACT) protocol P1025. Multivariate Cox proportional hazards regression models were used to investigate the association between NVP use and hepatotoxicity. NVP use was dichotomized as use or no use and further categorized according to ART exposure history. We investigated two outcomes: any liver enzyme elevation (LEE) (grade 1-4) and severe LEE (grade 3-4).
A total of 1229 women with ART use during pregnancy were studied, 218 (17.7%) of whom received NVP. Among the women receiving NVP, 137 (62.8%) were NVP naïve. Twenty-nine women (13.3%) who received NVP developed any LEE and one (0.5%) developed severe LEE. Of the 1011 women on non-NVP regimens, 145 (14.3%) developed any LEE and 14 (1.4%) developed severe LEE. There were no maternal deaths. In univariate models, LEE was not significantly associated with CD4+ count > 250 cells/μL or NVP use. In adjusted multivariate models, no significant increased risk of LEE (any or severe) in women taking NVP was detected as compared to those taking other ART regardless of prior exposure history.
We did not observe an increased risk of hepatotoxicity among HIV-infected pregnant women on NVP versus other ART, including women who were ART naïve.
Nevirapine (NVP) use has been described as a risk factor for developing severe hepatotoxicity, particularly in women with CD4+ cell counts above 250 cells/μL [1-4] including pregnant women [5-7]. Reports of six maternal deaths secondary to fulminant hepatitis in previously antiretroviral (ART) naïve women, four of whom had CD4+ cell counts above 250 cells/μL [5,7-9] led to labeling changes by the manufacturer of NVP and a public health advisory by the Food and Drug Administration followed warning of these pregnancy related deaths [10,11]. However, most previous studies examining the relationship of NVP and hepatotoxicity in pregnancy have come from case reports or have been limited to estimates among women using NVP only without a comparison group, which precludes estimation of relative risk and limits interpretation of their results.
It is not clear if continuous NVP use in pregnancy is associated with an increase risk of hepatotoxicity as compared to other ART. The goal of this study was to estimate whether HIV-infected pregnant women were at increased risk of hepatotoxicity when taking NVP containing regimens compared with ART not containing NVP.
The study population included HIV-infected pregnant women on ART from two cohorts in the United States: The Women and Infants Transmission Study (WITS) and the International Maternal Pediatric Adolescent AIDS Clinical Trials (IMPAACT) protocol P1025. Both WITS and IMPAACT P1025 are multicenter, prospective cohort studies of HIV-infected pregnant women [12,13]. This analysis was limited to women enrolled after July 2002, when aspartate aminotransferase (AST) and alanine aminotransferase (ALT) data were collected prospectively. NVP exposure was dichotomized into use or no use and further categorized into three groups: ART naïve (no prior ART exposure), ART chronic (taking ART at time of conception) and ART restart (history of ART use, not taking at time of conception). Women who received no ART or only intrapartum or postpartum treatment were excluded. Women with no aminotransferase data were also excluded.
The period of observation began at the time of ART initiation if women started ART during pregnancy (ART naïve and ART restart) or at the estimated date of conception in patients on ART at time of conception (ART chronic). Data collection continued through their postpartum visit. Background data and baseline laboratory values collected are listed in Table 1. Chronic infection with hepatitis B virus (HBV) or hepatitis C virus (HCV) was defined as the presence of either HCV antibody or HBV surface antigen. Both the WITS and P1025 protocols recorded AST and ALT data at the initiation of ART, every trimester, at time of delivery and at the postpartum visit.
ALT and AST levels were classified based on changes relative to the upper limit of normal (ULN) according to the Division of AIDS toxicity guidelines for adults: Grade 0 = <1.25× ULN; Grade 1 = 1.25 – 2.5× ULN; Grade 2 = >2.5 – 5.0× ULN; Grade 3 = >5.0 – 10.0× ULN; and Grade 4 = >10.0× ULN. The ULN for AST and ALT were individualized for each woman per the institution's laboratory standard. If AST and ALT grades were discordant, the higher of the two was used for classification. Women with elevated pretreatment AST or ALT levels were graded based on changes relative to their baseline: Grade 0 = <1.25× baseline; Grade 1 = 1.25 – 2.5× baseline; Grade 2 = >2.5 – 3.5× baseline; Grade 3 = >3.5 – 5.0× baseline; and Grade 4 = >5.0× baseline .
We investigated two hepatotoxicity outcomes: any liver enzyme elevation (LEE) (grade 1-4) and severe LEE (grade 3-4).
The baseline characteristics of the study population were compared according to NVP exposure. Differences in continuous variables were assessed using F tests, and differences in categorical variables were assessed using the Pearson chi-square test and Fisher's exact chi-square test.
Cox proportional hazards models were used to estimate the association between NVP use and LEE while controlling for potential confounders. Data were organized in the counting process style of input to model NVP and other covariates as time-dependent variables . The only covariates not modeled as time-dependent variables were ART exposure prior to follow-up, risk group for HIV transmission, and race. Univariate analysis was first performed to screen for covariates to be included in multivariate analysis that were significant at P=0.10 level. These covariates were then evaluated in a multivariate Cox regression model stratified by year of delivery and the study in which the woman was enrolled (WITS or P1025) to assess possible interaction. The SAS PHREG procedure (SAS version 8.2; SAS Institute Inc. Cary, NC) was used to perform these analyses.
A sample size of 1230 pregnant women (220 taking NVP containing regimens and 1010 taking non-NVP regimens) would provide 80% power to detect a minimum odds ratio (OR) of 1.7 for any hepatotoxicity and an OR of 3.8 for severe hepatotoxicity between pregnant patients taking NVP containing regimens versus non-NVP regimens, assuming a 5% type I error.
A total of 1358 women with ART exposure during pregnancy between July 2002 and June 2006 were considered for this study. After applying exclusion criteria, 1229 women were included in the study population, 218 (17.7%) of whom received NVP. Among the women receiving NVP, 137 (62.8%) were NVP naïve. 61% of the women were enrolled in IMPAACT P1025 (n=749) and 39% were enrolled in WITS (n=480). The baseline demographic and clinical characteristics are shown in Table 1.
Twenty-nine women (13.3%, CI 9.1 – 18.5%) who received NVP developed any LEE (grade 1 or higher) and one (0.5%, CI 0.01 – 2.5%) developed severe LEE (grade 3 or 4). Of the 1011 women on non-NVP regimens, 145 (14.3%, CI 12.2 – 16.7%) developed any LEE and 14 (1.4%, CI 0.8 – 2.3%) developed severe LEE. The only patient exposed to NVP who developed severe LEE was in the CD4+ > 250 cells/μL category. Of the 29 women exposed to NVP who developed any LEE, 21 (72.4%) had CD4+ > 250 cells/μL and 7 (24.1%) had CD4+ < 250 cells/μL. One woman did not have CD4+ data available. These results are similar to the CD4+ cell counts of women who developed LEE among the non-NVP group. Specifically, 9 of the 14 women who developed severe LEE were in the CD4+ > 250 cells/μL category as compared to 3 of those in the CD4+ < 250 cells/μL category (two did not have CD4+ data available). Of the 145 women not exposed to NVP containing regimens who developed any LEE, 111 (77%) had CD4+ > 250 cells/μL and 22 (15%) had CD4+ < 250 cells/μL. Twelve did not have CD4+ data available. There were no maternal deaths. The median time from initiation of therapy to any LEE in patients treated with NVP or other ART during pregnancy was 163 days and 157 days, respectively.
Univariate analysis showed LEE was significantly associated with chronic hepatitis, alcohol use, IV drug use, thrombocytopenia (< 150,000 mm3), immunological stage 3, baseline elevated liver enzyme and history of CDC class C events at P = 0.10 level. CD4+ count > 250 cells/μL and both ART and NVP naïveté were not significantly associated with LEE.
Multivariate Cox regression analysis estimated associations between various categories of NVP and ART exposures and any LEE (Table 2). Each model provides a different reference group against which the other exposures were compared. For example, Model 4 uses women who were treated with NVP for the first time in pregnancy (NVP +, NVP naïve) as the reference group. The relative risk for developing LEE in women who were also treated with NVP in pregnancy, but had previously been exposed to NVP (NVP +, NVP exposed) was 0.97 relative to the reference group. Although CD4+ count was not found to be significantly associated with LEE in the univariate analysis, we included a separate model examining LEE in women with CD4+ > 250 cells/μL stratified by NVP exposure due to unique concerns of hepatotoxicity in this population [5,7,16] (Model 5). When CD4+ > 250 cells/μL was included in the multivariate analysis for the other models, the results continued to demonstrate no significant difference in LEE between the reference group and the comparison groups. In all the models, NVP use during pregnancy was not significantly associated with either any or severe LEE.
The frequency of both grade 1 or higher (13.3%, CI 9.1 – 18.5%) and grade 3 or higher hepatotoxicity (0.5%, CI 0.01 – 2.5%) in pregnant women using NVP in our study is less than that reported in other studies. Hitti et al reported the phase I results of the PACTG 1022 trial which found a 29.4% rate of treatment discontinuation among the 17 pregnant patients taking NVP for suspected hepatotoxicity . However, one of the five patients who discontinued treatment developed only right upper quadrant pain with no LEE. This may reflect provider ascertainment bias in this non-blinded intervention study leading to premature discontinuation of therapy in the NVP group secondary to a preexisting concern that they were at increased risk for developing hepatotoxicity. The authors acknowledge in this study that the small sample size may have resulted in an overestimation of the actual incidence of hepatotoxicity. In a retrospective study, Lyons et al described grade 1 or higher LEE occurring in 34.1% of 85 women using NVP . These two studies and others  have reported a 6.5-11.8% rate of grade 3 or higher hepatotoxicity, as compared to only 0.5% (CI 0.01 – 2.5%) in our study. Our lower rates of both grade 1+ and grade 3+ LEE are more similar to results reported in studies from Thailand , Brazil , Mozambique  and London . While ethnic differences have been suggested as a possible factor in these discordant findings , our study population was very similar to the ethnic composition of the PACTG 1022 study.
We observed no maternal deaths in our study of 1229 pregnant women on ART. This included 218 women taking NVP-containing regimens, 137 (62.8%) of whom were NVP naive. This is somewhat in contrast to reports of six maternal deaths associated with continuous NVP use in pregnancy [5,8,9]. Reporting bias may play a role in this apparent difference in mortality as only one of these six deaths occurred in the setting of a prospective trial, while the remainders were case reports or personal communications. We also did not find an association between CD4+ count > 250 cells/μL and risk of hepatotoxicity in patients taking NVP in both univariate and multivariate analysis. The lack of evidence of an effect of CD4+ count on LEE is consistent with the largest review on NVP and hepatotoxicity which also was unable to demonstrate a CD4+ count cutoff in women that was associated with an increased risk of LEE . Individual genetic susceptibility may play an important role in the development of hepatotoxicity in relation to CD4+ cell count .
Our study has potential limitations. First, confounding by indication may have occurred if clinicians avoided prescribing NVP to women at greatest risk for developing hepatotoxicity, leading to underestimation of risk in these patients. In our study, women using NVP during pregnancy were significantly less likely to be NVP naïve and thus may have differentially included women with tolerance to the hepatic effects of NVP. Still, the control group did generally have similar background characteristics to the NVP group, including baseline CD4+ cell counts. In addition, the majority of the data in this study were obtained prior to the FDA public advisory and NVP manufacturer's labeling changes (2005) warning against potential increased risks of hepatotoxicity making it less likely that these concerns biased the conclusions of this study. Finally, we attempted to control for confounding by indication through consideration of several potential confounding variables.
Another potential limitation was that we were unable to differentiate possible obstetrical causes of LEE from drug-induced hepatotoxicity. As such, it is reasonable to assume a portion of the patients in our study had LEE independent of their ART exposure. A 15 month prospective study of 4,377 deliveries found a 3% incidence of liver enzyme elevation during pregnancy . Pregnancy specific causes such as preeclampsia comprised the majority of these cases (67%), whereas drug toxicity accounted for less than 2% of the pregnancies with elevated liver enzymes. Other studies have shown that HIV infected pregnant women are not at significantly greater risk of developing these pregnancy conditions most often associated with liver dysfunction . Considering the incidence of LEE in our study (14%) was significantly higher than in the Ch'ng study (3%), it is reasonable to assume that the majority of the additional LEE was associated with ART use, although a causal association could not be proven.
In regards to the disproportionate contribution of women from P1025, a sensitivity analysis was performed that stratified the multivariate analysis by study (WITS and P1025) and no significant differences were seen from the combined analysis. This likely reflects the fact that the WITS and P1025 trials were both prospective, observational studies that are very similar on multiple levels such as inclusion and exclusion criteria, population characteristics, and frequency of study visits [12,13]. In addition, by limiting the WITS population to those enrolled after 2002, we ensured that the schedule for LFT monitoring would be the same between the two trials and that the studies would cover the same calendar time frame.
Strengths of this study include the large number of women enrolled which provided statistical power to detect clinically important increases in risk, the inclusion of a reference group of HIV-infected pregnant women not on ARV, and the prospective and standardized nature of the data collection. Even if pregnant women are uniquely at increased risk for hepatotoxicity, our data suggest this risk is not substantially higher with NVP exposure as compared to other ART. The risk of hepatotoxicity associated with NVP use may be less than other studies have estimated and clinicians have feared. Specifically, the 95% confidence interval for the adjusted relative risk of any LEE among NVP unexposed versus NVP exposed women ranged from 0.64 to 1.57 which suggests that a substantial increase in risk of LEE with NVP use in pregnancy is unlikely (Table 2).
We would still exercise caution when administering any antiretroviral medication during pregnancy, with close clinical surveillance for symptoms of hepatotoxicity and laboratory monitoring for LEE, especially when initiating or altering a regimen. As with the decision to prescribe any medication, health care providers must carefully consider the risks and benefits of specific ART regimens and individualize therapy to each patient.
Principal investigators, study coordinators, program officers and funding for the Women and Infants Transmission Study (WITS) include: Clemente Diaz, Edna Pacheco-Acosta (University of Puerto Rico, San Juan, PR; U01-AI-34858); Ruth Tuomala, Ellen Cooper, Donna Mesthene (Boston/Worcester Site, Boston, MA; U01-DA-15054); Phil La Russa, Alice Higgins (Columbia Presbyterian Hospital, New York, NY; U01-DA-15053); Sheldon Landesman, Edward Handelsman, Ava Dennie (State University of New York, Brooklyn, NY; U01-HD-36117); Kenneth Rich, Delmyra Turpin (University of Illinois at Chicago, Chicago, IL; U01-AI-34841); William Shearer, Susan Pacheco, Norma Cooper (Baylor College of Medicine, Houston, TX; U01-HD- 41983); Joana Rosario (National Institute of Allergy and Infectious Diseases, Bethesda, MD); Kevin Ryan, (National Institute of Child Health and Human Development, Bethesda, MD); Vincent Smeriglio, Katherine Davenny (National Institute on Drug Abuse, Bethesda, MD); and Bruce Thompson (Clinical Trials & Surveys Corp., Baltimore, MD, N01-AI-85339). Scientific Leadership Core: Kenneth Rich (PI), Delmyra Turpin (Study Coordinator) (1-U01-AI-50274-01). Additional support has been provided by local Clinical Research Centers as follows: Baylor College of Medicine, Houston, TX; NIH GCRC RR00188; Columbia University, New York, NY; NIH GCRC RR00645; Children's Hospital Boston, MA; NIH GCRC RR 00.
The IMPAACT P1025 study was supported by grants U01AI068632 and 1 U01AI068616 from the National Institute of Allergy and Infectious Diseases, and contracts number N01-HD-3-3365 and HHSN267200800001C (control # N01-DK-8-0001) from the International Domestic Pediatric and Maternal HIV Clinical Trials Network of the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
We gratefully acknowledge the women who have participated in WITS and IMPAACT P1025 and the efforts of the dedicated study personnel at all sites throughout the study who have made this analysis possible.
The content is solely the responsibility of the authors, and does not necessarily represent the official views of the National Institutes of Health.
Author Contributions: Study concept and design: D.W.O., S.B.B., M.L., D.E.S., R.C.H., A.L.F., B.T., R.E.T.Analysis and interpretation of data: D.W.O., S.B.B., M.L., D.E.S., R.C.H., A.L.F., B.T., R.E.T.
Drafting of manuscript and coordination of revisions: D.W.O.
Critical revision of manuscript for important intellectual content: D.W.O., S.B.B., M.L., D.E.S., R.C.H., A.L.F., B.T., R.E.T.
Statistical analysis: M.L., R.M.L., B.T.
Study supervision: R.E.T.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.