|Home | About | Journals | Submit | Contact Us | Français|
Background.It is unknown whether adverse birth outcomes are associated with maternal highly active antiretroviral therapy (HAART) in pregnancy, particularly in resource-limited settings.
Methods.We abstracted obstetrical records at 6 sites in Botswana for 24 months. Outcomes included stillbirths (SBs), preterm delivery (PTD), small for gestational age (SGA), and neonatal death (NND). Among human immunodeficiency virus (HIV)–infected women, comparisons were limited to HAART exposure status at conception, and those with similar opportunities for outcomes. Comparisons were adjusted for CD4+ lymphocyte cell count.
Results.Of 33 148 women, 32 113 (97%) were tested for HIV, of whom 9504 (30%) were HIV infected. Maternal HIV was significantly associated with SB, PTD, SGA, and NND. Compared with all other HIV-infected women, those continuing HAART from before pregnancy had higher odds of PTD (adjusted odds ratio [AOR], 1.2; 95% confidence interval [CI], 1.1, 1.4), SGA (AOR, 1.8; 95% CI, 1.6, 2.1) and SB (AOR, 1.5; 95% CI, 1.2, 1.8). Among women initiating antiretroviral therapy in pregnancy, HAART use (vs zidovudine) was associated with higher odds of PTD (AOR, 1.4; 95% CI, 1.2, 1.8), SGA (AOR, 1.5; 95% CI, 1.2, 1.9), and SB (AOR, 2.5; 95% CI, 1.6, 3.9). Low CD4+ was independently associated with SB and SGA, and maternal hypertension during pregnancy with PTD, SGA, and SB.
Conclusions.HAART receipt during pregnancy was associated with increased PTD, SGA, and SB.
(See the Editorial Commentary by Watts and Mofenson, on pages 1639–41.)
Maternal human immunodeficiency virus (HIV) infection has been associated with adverse birth outcomes [1–10], but it remains unclear whether the use of highly active antiretroviral therapy (HAART) contributes to this risk. Observational studies in both developed and developing countries have reported conflicting evidence on the association between HAART exposure and preterm delivery (PTD), low birth weight, and stillbirth (SB) [10–22] and on the association between protease inhibitors and PTD [11, 13, 14, 16, 17, 23, 24].
Conflicting data from past studies may be related to differential handling of noncomparable exposure groups and lack of power to detect modest effect sizes. In an attempt to overcome these limitations, we performed the largest surveillance study to date among HIV-infected women in the HAART era and restricted analyses to comparable exposure groups.
Approval was granted by human subjects committees in Botswana and at the Harvard School of Public Health. We included all women who delivered live births or stillbirths at a gestational age ≥20 weeks at 6 government facilities in Botswana. Surveillance occurred at Princess Marina Hospital (PMH) in Gaborone, Scottish Livingstone Hospital (SLH) in Molepolole, Deborah Retief Memorial Hospital (DRM) in Mochudi, Nyangabgwe Hospital (NH) in Francistown, Letsholathebe Memorial Hospital (LMH) in Maun, and Gantsi Primary Hospital (GPH) in Gantsi. These public hospitals were selected for geographic diversity and inclusion of primary and tertiary levels of obstetrical care (PMH and NH are referral hospitals for obstetric complications in southern and northern Botswana, respectively). Surveillance began on 1 May 2009 at PMH in Gaborone and between 1 October and 1 November 2009 at the remaining 5 sites, and continued through 30 April 2011. We reviewed obstetric records upon discharge from each maternity ward. In the event of multiple gestations, the outcome of the first-born infant was recorded.
The Botswana Ministry of Health recommends routine (“opt-out”) HIV testing in pregnancy and prioritization of CD4+ lymphocyte cell count testing for HIV-infected pregnant women not yet on HAART . During the study period, HIV-infected women with CD4+ counts<250 cells/μL or World Health Organization clinical stage 3 or 4 conditions were eligible to initiate HAART during pregnancy for maternal treatment. Women already receiving HAART at conception were advised to continue during pregnancy. In most cases, HAART consisted of nevirapine (NVP) plus zidovudine (ZDV) and lamivudine (3TC). Women not receiving HAART were eligible for ZDV monotherapy for the prevention of mother-to-child HIV transmission (PMTCT). Beginning in October 2009, a limited number of HIV-infected women with CD4+ counts >250 cells/μL had access to HAART for PMTCT through a pilot Botswana government program. During the study period, this program was initiated at SLH, with 89 (24.8%) of 359 eligible women enrolled. The program also had a minor impact at other sites, with <5% of eligible women receiving HAART for PMTCT. The regimen provided through this program was usually lopinavir/ritonavir (LPV/r) plus ZDV/3TC.
The information abstracted from obstetric records included maternal demographics, medical history, laboratory values measured in pregnancy (hemoglobin and rapid plasma reagin, a test to detect syphilis infection), vital signs, HIV status and antiretroviral (ARV) history, and birth outcomes. We defined anemia during pregnancy as a recorded hemoglobin ≤10 g/dL. Maternal hypertension in pregnancy was defined as a systolic blood pressure measurement >140 mmHg or a diastolic measurement >90 mmHg at any visit before labor, admission to hospital for hypertension, delivery complicated by hypertension, or induction for preeclampsia. We extracted estimated gestational age at delivery from obstetric records. For most pregnancies in Botswana, estimated date of delivery is based on date of last menstrual period alone. However, if the date is uncertain or discordant with fundal measurements, gestational age is estimated using ultrasonography.
For HIV-infected women, the date of HIV diagnosis, up to 2 CD4+ cell counts in pregnancy, and details of antiretroviral therapy were extracted from the records. A hospital database was searched to locate unrecorded CD4+ results. We used the most recent CD4+ cell count in all analyses. Antiretroviral use during pregnancy was classified as HAART continued from before the current pregnancy, HAART initiated during pregnancy, ZDV monotherapy, or no antiretroviral drugs received prior to delivery. HAART was defined as 3 or more antiretroviral drugs. In the event that an antiretroviral regimen changed during pregnancy, the last regimen was utilized in the analyses. We excluded interventions given during labor, such as zidovudine or single-dose nevirapine, in analyses.
The primary birth outcomes were SB, PTD, and small for gestational age (SGA). SB was defined as fetal death with an Apgar score of 0; preterm delivery as delivery <37 weeks gestation; and SGA as below the 10th percentile of birth weight by gestational age using norms created from infants born to HIV-uninfected women in Botswana . Because accurate norms were not available for deliveries before 26 weeks and after 40 weeks gestation, these deliveries were excluded from the outcome SGA. We also evaluated neonatal death (NND), defined as death during the same hospitalization and within 28 days of a live delivery.
Statistical analyses were performed using SAS, version 9.3 software (SAS Institute, Cary, NC). All reported P values are based on 2-sided tests. Unadjusted and adjusted logistic regression analyses of HIV-infected versus HIV-uninfected women were performed for SB, PTD, SGA, and NND. Adjusted analyses were performed using stepwise selection for logistic regression analyses, and covariates with a significance level ≤0.05 were included in the model.
Among HIV-infected women, we also used stepwise selection for logistic regression modeling. We included CD4+ cell count (categorized as presence of CD4+ count in pregnancy ≤200 cells/μL, >200 cells/μL, or unknown) in all models, in an attempt to account for confounding by indication. We first compared women who continued HAART from before pregnancy with all other HIV-infected women, to allow for evaluation of early events (beginning at first antenatal visit) for all women. We next compared women who initiated HAART or ZDV monotherapy during pregnancy by 34 weeks gestation, restricting comparisons to events ≥34 weeks gestation to allow for similar ARV exposure times for each group. Finally, we compared rates of SGA infants among women who continued HAART from before pregnancy with those who initiated HAART during pregnancy. We directly compared rates of SGA between these groups (but not other birth outcomes) because women in each group had similar opportunity to experience the outcome SGA. Because of the limited number of events and the potential for multiple interactions between SGA, PTD, and NND, logistic regression modeling was not performed for the outcome NND.
From 1 May 2009 through 30 April 2011, 33 148 birth outcomes occurred at the 6 study sites: 13 181 (40%) at PMH, 4103 (12%) at SLH, 2967 (9%) at DRM, 7293 (22%) at NH, 4221 (13%) at LMH, and 1383 (4%) at GPH. These deliveries comprised at least one-third of all births in Botswana during the study period .
Data on HIV status and birth outcomes are shown in Table 1. Of the 33 148 women included, 32 113 women (96.9%) had a known HIV status, of whom 9504 (29.6%) were HIV infected. Some differences were noted by maternity site, with HIV prevalence ranging from a high of 34% at NH to a low of 23% at GPH. HIV-infected women experienced significantly higher rates of all adverse birth outcomes, including SB, PTD, SGA, and NND. In adjusted analysis, maternal HIV infection remained significantly associated with an increased risk for SB (adjusted odds ratio [AOR], 1.5; 95% confidence interval [CI], 1.3, 1.7); PTD (AOR, 1.3; 95% CI, 1.3, 1.4), SGA infants (AOR, 1.8; 95% CI, 1.7, 1.9), and NND (AOR, 1.4; 95% CI, 1.2, 1.7) among HIV-infected women compared with HIV-uninfected women. We did not detect an association between maternal HIV infection and congenital anomalies.
Of 9504 HIV-infected women, 9149 (96%) had a known initiation date for antiretroviral drugs received during pregnancy; 2189 (24%) continued HAART from before pregnancy, 1101 (12%) initiated HAART during pregnancy, 4625 (51%) initiated ZDV monotherapy during pregnancy, and 1234 (13%) received no antiretroviral drugs. Maternal characteristics are shown according to antiretroviral drugs received in pregnancy in Table 2.
CD4+ cell count in pregnancy was available for 4492 (49%) women, and rates of CD4+ cell count testing varied significantly according to antiretroviral drugs received in pregnancy (24% among those who continued HAART, 70% among those who initiated HAART, 62% among those starting ZDV, and 20% among those who received no antiretroviral drugs). The overall median CD4+ cell count was 388 cells/μL, and differed by antiretroviral drugs received in pregnancy (Table 2). Among women receiving HAART during pregnancy, 2851 (87%) were noted to have received NVP/ZDV/3TC or did not have a regimen specified (and considered likely to have received NVP/ZDV/3TC), and 312 (9%) were noted to have received LPV/r/ZDV/3TC (34 from conception, and 278 started in pregnancy). Median CD4+ cell count for those receiving LPV/r/ZDV/3TC was 458 cells/μL. Among the women initiating antiretroviral drugs during pregnancy, the median gestational age of HAART initiation was 25 weeks; first–third quartile (Q1–Q3) 20–29 weeks. The median age of ZDV initiation was 29 weeks (Q1–Q3, 28–31).
Among HIV-infected women, the overall rate of PTD was 24% and the median gestational age for PTD was 34 weeks (Q1–Q3, 32–36). Compared with all other HIV-infected pregnant women, HAART exposure from before pregnancy was significantly associated with PTD (AOR, 1.2; 95% CI, 1.1, 1.4) (Table 3). Compared with women initiating ZDV in pregnancy, initiating HAART in pregnancy was also significantly associated with increased odds of PTD (AOR, 1.4; 95% CI, 1.2, 1.8). Additional risk factors for PTD among HIV-infected women in multivariate analysis are shown in Table 3, and include maternal hypertension during pregnancy and anemia.
The rate of SGA among HIV-infected pregnant women was 18%, and the median gestational age of women with an SGA infant was 39 weeks (Q1–Q3, 36–40). Compared with all other HIV-infected women, continuing HAART from before pregnancy was associated with increased odds of having an SGA infant (AOR, 1.8; 95% CI, 1.6, 2.1) (Table 4). Compared with ZDV initiation, HAART initiation during pregnancy was also associated with an increased rate of delivering SGA infants (AOR, 1.5; 95% CI, 1.2, 1.9). When we compared women who continued HAART with those who initiated HAART, the former experienced a higher percentage of SGA infants (AOR, 1.3; 95% CI, 1.0, 1.5). Among all HIV-infected women, CD4+ count ≤200 cells/μL and maternal hypertension were significantly associated with SGA infants.
We observed an SB rate of 5% among HIV-infected pregnant women at a median gestational age of 32 weeks (Q1–Q3, 28–37). Compared with all other HIV-infected pregnant women, continuing HAART from before pregnancy was associated with higher odds of SB (AOR, 1.5; 95% CI, 1.2, 1.8) (Table 5). Compared with women who initiated ZDV, initiating HAART was also associated with an increased risk of SB (AOR, 2.5; 95% CI, 1.6, 3.9). Additional risk factors were CD4+ count ≤200 cells/μL and maternal hypertension.
Neonatal death occurred among 2.3% of infants born to HIV-infected women, and the median gestational age was 30 weeks (Q1–Q3, 26–37). Preterm babies experienced a significantly higher rate of NND compared with those born at term (7% vs 0.8%; P < .0001), as did SGA infants compared with those appropriate for gestational age (3.5% vs 1.5%; P < .0001). However, we did not observe significant differences in NND when we compared women who continued HAART from before pregnancy with all other HIV-infected women (2.0% vs 2.3%; P = .41) in univariate analysis. Women who initiated HAART experienced a higher rate of NND compared with those who initiated ZDV during pregnancy (1.9% vs. 0.8%; P = .002). In univariate analysis, additional risk factors for NND included maternal hypertension (3.2% vs 1.7%; P = .0002), but not maternal CD4+ count ≤200 cells/μL (2.0% vs 1.8%; P = .71).
Sensitivity analyses were performed for multiple gestation and for CD4+ cell count. There were 489 (1.5%) women with multiple gestation, and they were significantly more likely to experience PTD, SGA, and SB. When they were excluded from the analyses, we did not observe changes in the study findings. We also performed sensitivity analyses among only HIV-infected women with a recorded CD4+ cell count (included as a continuous variable) in all logistic regression models; this limited the number included in the models but did not significantly change the magnitude or direction of associations (data not shown). In addition, we stratified analyses by women with CD4+ cell count above and below 200 cells/μL. There were no differences in the direction of each association in these analyses, and we found the greatest magnitude of association between adverse outcomes and either continuing HAART or starting HAART in pregnancy among women with CD4+ count >200 cells/μL (data not shown).
We also evaluated associations between PI-based HAART and PTD, and the duration of HAART exposure and all outcomes. Of 48 women who continued PI-based HAART from before pregnancy, 20 (42%) experienced PTD compared with 522 (26%) of 1998 women who continued a non-PI-based HAART regimen (OR, 2.0; 95% CI, 1.1, 3.6). Of 178 women initiating PI-based HAART, 44 (25%) experienced PTD compared with 131 (20%) of 654 initiating a non-PI-based HAART regimen (OR, 1.3; 95% CI, .9, 1.9). We did not observe significant differences in rates of PTD, SGA infants, or SB according to timing of HAART initiation during pregnancy (before or after 32 weeks gestation; data not shown).
We performed the largest study of birth outcomes to date among HIV-infected women with access to HAART in pregnancy. In this study from Botswana, adverse birth outcomes such as preterm delivery, SGA infants, and SB were more common among HIV-infected women compared with HIV-uninfected women. In addition, HAART exposure during pregnancy (continued from before pregnancy and initiated during pregnancy) was associated with adverse birth outcomes when compared with all other HIV-infected women and with ZDV monotherapy during pregnancy, respectively. We also identified maternal hypertension during pregnancy as an independent risk factor for each adverse outcome.
Previous data for adverse birth outcomes related to HAART use are mixed, and largely from cohorts in the developed world [10–19]. We believe the conflicting results may be related to limited power, differences among the populations studied and the availability of obstetrical care, and confounding by indication for HAART use in pregnancy. The selection of exposure categories has varied widely in previous studies, in particular with respect to the inclusion of HIV-infected women who did not receive antiretroviral drugs in pregnancy. However, the importance of selecting appropriate exposure categories must be underscored, since bias is introduced by many factors, including differences in opportunity to start treatment prior to delivery . We were cautious in our interpretation of results among women who were untreated in pregnancy, since these women had less access to antenatal care and may have delivered before having had the opportunity to receive antiretroviral drugs. We limited most of our comparisons to either those categorized as HAART exposed versus unexposed at conception, or to a time-limited comparison of HAART versus ZDV initiation in pregnancy, and attempted to control for CD4+ cell count differences (including sensitivity analyses with CD4+ cell count as a continuous variable and additional analyses stratified by CD4 categories) to reduce confounding by indication. Through these comparisons, we found in all analyses that HAART exposure was significantly associated with adverse birth outcomes. Our findings agree with several previous associations between HAART exposure and preterm delivery [10, 11, 19, 23, 27, 28], SBs , and SGA infants , including a study comparing historical cohorts from West Africa  and a previous study in Botswana .
NND is an important birth outcome that in part reflects the severity of PTD and SGA, and may also differ by the level of obstetrical and neonatal care. Our evaluation of NND was limited by the smaller number of outcomes, and only captured deaths that occurred before leaving the hospital. We were reassured by the lack of association between NND and women who continued HAART from conception, when compared with all other women. This was not the case for women starting HAART in pregnancy, which may reflect the more targeted comparison with those receiving ZDV in pregnancy that excluded women not receiving any ARVs (who may have had higher risks for other reasons). Additional studies are needed to determine the relationship between HAART exposure in pregnancy and neonatal mortality in different settings, particularly those lacking skilled obstetrical and neonatal care.
Several mechanisms have been proposed to account for associations between HAART and adverse birth outcomes. These include modulation of the immune system by a cytokine-mediated effect from HAART [30, 31], and increased risk for hypertension and preeclampsia in pregnancy . Studies in Spain and the United Kingdom identified significantly higher rates of preeclampsia among HIV-infected women in the HAART era compared with the pre-HAART era [32, 33]. A study of placental pathology among HIV-infected women delivering stillbirths in Botswana identified chronic placental hypertensive damage to be significantly more common in HAART-treated women compared with those not receiving HAART . We found maternal hypertension to be significantly associated with an increased risk for PTD, SGA, and SB (which is consistent with prior studies [35–40]) and believe further clarification of the relationship between HAART and hypertension is needed. Overall, these findings highlight the need for additional research to study potential causal mechanisms between HAART use during pregnancy and adverse birth outcomes. Because hypertension itself was a strong predictor for all adverse outcomes, we believe the management of hypertension in pregnancy should be prioritized for all high-risk women, including those receiving HAART.
Our study was limited by missing CD4+ cell count data, with a recorded CD4+ cell count available in pregnancy for only 49% of all women. However, women initiating either HAART or ZDV in pregnancy had higher rates available (70% initiating HAART and 62% initiating ZDV). We believe a sufficient number of CD4+ results were available from the various treatment groups to reduce bias by indication (in addition to the fact that many of those receiving HAART from conception were already immune reconstituted), and that associations with HAART were not merely residual confounding by maternal disease status. We evaluated all birth outcomes for confounding by maternal CD4+ cell count, and found that the associations between HAART use and adverse birth outcomes were independent of maternal CD4+ cell count. When we restricted our analyses to HIV-infected women with a recorded CD4+ cell count, the associations between HAART and adverse birth outcomes remained significant. Furthermore, when we stratified our data and analyzed only women with CD4+ counts >200 cells/μL, the strength of the associations increased; this suggests the associations with HAART were unlikely to be a marker for low CD4+ cell count or maternal illness. A randomized study comparing HAART with ZDV monotherapy at higher CD4+ cell counts would reduce the possibility of confounding and provide a clearer understanding of this association. However, at lower CD4+ cell counts, the established benefits of HAART make such a trial unethical to perform.
Our study had several additional limitations. Several variables in the obstetric records were self-reported, and some information was missing. The calculation of the gestational age by the use of the last-normal-menstrual-period method (with fundal height and ultrasound confirmation) is less precise than the use of ultrasound, but we would not expect significant differences in gestational age calculations by antiretroviral exposure category. Although we utilized logistic regression modeling, there may have been unmeasured confounding by maternal clinical factors, and women receiving HAART were sicker than other women in the cohort. We did not have access to HIV RNA level data during pregnancy or nadir CD4+ cell counts for women who were receiving HAART from conception, and we acknowledge that the reconstituted CD4+ cell count may be an imperfect marker for a woman's immune status or her risk for adverse events in pregnancy. Although 62% of births occurred at PMH and NH (the 2 referral hospitals), and there were minor differences in rates of adverse outcomes according to site (with higher rates of PTD observed at PMH, Francistown, and Maun, and a higher rate of SB at PMH), the magnitude and direction of the findings were consistent at all sites.
This study highlights the complexity of issues regarding the use of HAART and adverse birth outcomes. The benefits of HAART for the health of immunocompromised HIV-infected persons [41–44] and for PMTCT [45–47] have been well documented. In this study, we observed an increased risk for preterm delivery, SGA infants, SBs, and neonatal death among pregnant women exposed to HAART. Although these data are observational, they serve to underscore the need for further research into potential mechanisms by which HAART may affect birth outcomes as well as investigation of the safest antiretroviral regimens for use during pregnancy. These data also demonstrate the high number of adverse birth outcomes in this population. As more women gain access to HAART during pregnancy, additional efforts are needed to identify those at high risk for adverse outcomes, and to provide intensified support systems that address modifiable risk factors such as hypertension in pregnancy. Public health efforts should focus on supporting high-risk obstetrical and neonatal care services in countries like Botswana to maximize the benefits of HAART use during pregnancy for mothers and their infants.
Acknowledgments.We gratefully acknowledge the contribution of our research assistants, Kelebogile Binda, Rehanna Matsebe, Koziba Malibala, and Aboleleng Ditsheko, as well as the nurses, counselors, obstetricians, pediatricians, and superintendents who worked at the individual study sites (Princess Marina Hospital, Scottish Livingstone Hospital, Deborah Retief Memorial Hospital, Nyangabgwe Hospital, Letsholathebe Memorial Hospital, and Gantsi Primary Hospital). We are also grateful for the administrative support provided by the Botswana-Harvard AIDS Institute Partnership (BHP). We are grateful to the institutional review boards at the Health Research Unit in the Botswana Ministry of Health, Harvard Medical School, and Harvard School of Public Health for reviewing this study.
Financial support.This work was supported by a grant from the Centers for Disease Control and Prevention (U2GPS000941), a grant from the National Institutes of Health (P30 AI060354, Harvard University CFAR, H. J. R), and by research fellowships from the Doris Duke Charitable Research Foundation (J. Y. C., N. P.)
Potential conflicts of interest.All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.