HIV Articles  
Use of ART in pregnant HIV-infected women and the risk of premature delivery:
a meta-analysis
  AIDS: Volume 21(5) 12 March 2007 p 607-615
Kourtis, Athena Pa,b; Schmid, Christopher Hc; Jamieson, Denise Ja; Lau, Josephc
From the aDivision of Reproductive Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Georgia, USA bDepartment of Obstetrics and Gynecology, Eastern Virginia Medical School, Norfolk, Virginia, USA cInstitute for Clinical Research and Health Policy Studies, Tufts-New England Medical Center, Boston, Massachusetts, USA.
Background: The use of antiretroviral agents in pregnant HIV-infected women has been reported to increase the risk of premature delivery in some studies. We performed a meta-analysis on relevant studies to address this question.
Methods: We searched Medline, Embase and the Cochrane Controlled Clinical Trials Register for English language articles. Studies that reported premature delivery for HIV-infected women treated with antiretroviral regimens during pregnancy were selected. Meta-analyses were performed using a random effects model.
Results: Thirteen prospective cohorts and one retrospective study met the inclusion criteria. Antiretroviral therapy during pregnancy did not increase the risk of premature delivery overall [odds ratio (OR) 1.01, 95% confidence interval (CI) 0.76-1.34]. In subgroup analyses, compared with no therapy, monotherapy (mostly zidovudine) conferred an OR of 0.86 (95% CI 0.73-1.01), whereas combination therapy conferred an OR of 1.13 (95% CI 0.79-1.63). The use of protease inhibitor (PI)-containing combinations resulted in an OR for premature delivery of 1.24 (95% CI 0.76-2.02), compared with combinations without PI. The initiation of combination therapy before pregnancy or in the first trimester resulted in an OR of 1.71 (95% CI 1.09-2.67) compared with therapy initiation in the second trimester and beyond. There was a large degree of heterogeneity between studies.
Conclusion: Evidence indicates that antiretroviral therapy during pregnancy is not associated with an overall increased risk of premature delivery. The use of combination regimens before or early in pregnancy may slightly increase the risk of prematurity. Continued surveillance will be necessary to quantify such a risk accurately.
The widespread use of HAART among HIV-infected women in the industrialized world has led to dramatic decreases in the rate of perinatal transmission of HIV to the infant. Before 1994, perinatal transmission rates of HIV in the United States were 20-25%; now rates of less than 2% are achieved with the use of HAART and elective cesarean section [1]. Practice guidelines in the United States emphasize the importance of HAART during pregnancy for maternal health and for the prevention of transmission of HIV to the infant [2]. Current estimates are that 6000-7000 HIV-infected women give birth, and 144-236 babies acquire HIV infection perinatally each year in the United States [1].
The advent of potent antiretroviral agents including protease inhibitors (PI), however, raised the question of the safety of these new medications for the mother, fetus and infant. Apart from the known adverse effects of such agents, some reports from Europe suggested that use of antiretroviral therapy during pregnancy was associated with substantially increased rates of prematurity [3-8]. This risk was particularly pronounced for PI use when started early in pregnancy or before conception [5]. Other studies from Europe [9] and reports from US cohorts have not shown such an association [10-13]. Of note is the fact that the use of zidovudine only during pregnancy had earlier been linked to a decreased risk of prematurity [14], again, a finding not confirmed by later reports [4]. Many of the studies of pregnancy outcomes among HIV-infected women have had limitations, such as relatively small numbers of subjects, a retrospective design, and a lack of control over other factors that could influence the rates of adverse pregnancy outcomes. HIV-infected women often have other risk factors for premature delivery, including previous preterm birth, smoking, illicit drug use, other sexually transmitted infections and advanced disease with immunosuppression [15,16], making the etiology of any adverse obstetric outcome such as prematurity difficult to determine.
The published evidence on the question of antiretroviral drug use during pregnancy and the risk of prematurity in the offspring remains, to this day, controversial. We conducted a meta-analysis of all the relevant studies published to date in order to derive, given a greatly expanded number of subjects, a more reliable overall estimate of the association of the prenatal use of HAART with prematurity, and to explore reasons for discrepancies among studies.
Search results

The initial search identified 633 citations; 29 of these were thought to be relevant and were retrieved for detailed review. Of these, we excluded 12 because they were either reviews or letters to the editor or did not contain an appropriate control group of pregnant women, two [4,14] because these overlapped with more recent publications from the same cohorts and one [10] because the number of subjects exposed to therapy was not provided and we were unsuccessful in obtaining this information from the investigators. The characteristics of the remaining 14 studies that were included in the meta-analysis are summarized in Table 1 and Table 2.
There was wide variation among the studies in terms of design, sample size, number of covariates adjusted for, and the racial composition of their subjects. Abstracts [18,19] were subjected to sensitivity analyses, as the evidence had not yet been subjected to a peer review process. Studies that did not adjust for any confounding factors when exploring the association of prematurity and antenatal antiretroviral use (such as maternal age, race, CD4 cell count, drug use and others, see Table 1) were also subjected to sensitivity analysis. Excluding these studies from the analyses did not affect any of the estimates reported below. Furthermore, given the potential overlap in the study subjects of some studies (Culnane et al. [12] with Tuomala et al. [11], and the European Collaborative Study [4] with Thorne et al. [5]), analysis after excluding Culnane et al. [12] and the European Collaborative Study [4] was performed. In the latter case an earlier report of the Swiss data by Lorenzi et al. [3] was included. The paper by the European Collaborative Study [6] was largely updated by the report in 2004 [5], so the latter was included in all analyses with the exception of the comparison of PI-containing therapy versus non-PI-containing combinations, for which Thorne et al. [5] did not provide information. In the case of the report by Thorne et al. [5], because an attempt to contact the investigators was unsuccessful, we assumed that the premature births by women on no antiretroviral therapy comprised the remaining premature births in the cohort, whose number was not explicitly given. Because the study by Cotter et al. [21] overlapped partly in study subjects with the study by Tuomala et al. [11], a separate analysis after excluding the study by Cotter et al. [21] was also performed.
Five prospective cohorts reported premature births data according to the use of any antiretroviral therapy during pregnancy. A meta-analysis combining these studies found that antiretroviral therapy was not associated with an increased risk of premature delivery overall (OR 1.01, 95% CI 0.76-1.34; Fig. 1a). It should be noted that there is a significant degree of heterogeneity among the studies, particularly when comparing the results of the European with those of the American studies in subgroup analyses (Table 3).
We next examined particular types of therapy. No significant differences were obtained when monotherapy compared with no therapy was examined (OR 0.86, 95% CI 0.73-1.01), or when any combination therapy versus no therapy was examined (OR 1.13, 95% CI 0.79-1.63); monotherapy did not differ significantly in risk from combination therapy, either (OR 0.87, 95% CI 0.72-1.06; Fig. 1b). We then looked at the use of PI regimens during pregnancy, as they in particular seemed to confer the highest risk of premature delivery in some earlier reports [5]. Although when compared with no antiretroviral therapy, the use of PI-based regimens did not confer a statistically significantly increased risk (OR 1.24, 95% CI 0.76-2.02), when compared with other, non-PI-containing combination regimens, the increase in risk reached statistical significance (OR 1.35, 95% CI 1.08-1.70; Fig. 2a).
Because there was a partial overlap in study subjects between the studies by Cotter et al. [21] and Tuomala et al. [11], analysis excluding Cotter et al. [21] was also performed. This did not affect any of the estimates reported above with the exception of the comparison of PI combinations with other combinations, which changed to non-significant from marginally significant (OR 1.29, 95% CI 0.98-1.71).
Few studies have reported the time of initiation of combination antiretroviral regimens in pregnancy. For those that did, the summary estimate of prematurity risk was 1.71 for early initiation (prepregnancy or during the first trimester), compared with initiation in the second or third trimester, and this difference was statistically significant (95% CI 1.09-2.67; Fig. 2b).
The question of whether the use of antiretroviral drugs during pregnancy confers an increased risk of preterm delivery has met with different answers in different studies. Whether HIV infection itself is associated with such an adverse outcome is not conclusive. Some studies from the zidovudine-only era reported a higher incidence of low birth weight and premature delivery among HIV-infected women, compared with uninfected women [22]. Others, however, found that the premature delivery rates in HIV-infected women were similar to those reported for other HIV-uninfected minority populations [16,23]. A meta-analysis that examined the association of maternal HIV infection and perinatal outcomes before the era of PI reported a summary OR for premature delivery of 1.83 among HIV-infected women (95% CI 1.63-2.06) [24], with significant heterogeneity among the studies, suggesting that this result should be interpreted with caution. Recent findings from a large cohort in the United States suggested that HAART is associated with improved obstetric outcomes, including premature delivery and low birth weight [10]. Pregnancy outcomes were reported to be similar to those of HIV-negative women in a small prospective cohort of six US centers [25]. Morris et al. [20] reported a prematurity rate of 19.1% in HAART-treated HIV-infected women, which is consistent with the overall rate of prematurity reported for the US population [26] and with rates reported before the introduction of HAART. We recently undertook an analysis of adverse obstetric outcomes of all hospitalizations of HIV-infected women in the United States; it did not show an increase in the rate of premature delivery or labor from 1994 (pre-antiretroviral era) to 2003 (widespread use of HAART), arguing against a major effect of HAART on preterm delivery rates [27]. Of interest was the fact that we found that HIV-infected women had a consistently higher rate of preterm labor/delivery than did women who were not infected with HIV, even after adjusting for several covariates. A very recent report by Cotter et al. [21] is the first in the United States to report an increased risk of prematurity with the use of PI; of interest is the fact that the study had contributed data to the multisite US study of 2002 [11], which did not report such an association.
This meta-analysis does not show an increased overall risk of premature delivery associated with the use of HAART. The use of PI-containing regimens seems to be associated with a slightly increased risk of prematurity, whereas the use of monotherapy is associated with a slightly decreased risk. These differences are, however, small and not statistically significant. In addition, this meta-analysis confirms the previous observations that the initiation of combination therapy early during pregnancy seems to confer a slightly increased risk of premature delivery, arguing for the early gestational effects of some antiretroviral agents, particularly PI. The available information is limited and the results of the meta-analysis should be interpreted with caution. Most studies used the gestational age of 37 weeks as the cutoff for a definition of prematurity and do not offer data related to more severe degrees of prematurity; however, births at earlier stages of prematurity would be interesting to examine, because they cause most prematurity-related infant morbidity and mortality.
The results of this meta-analysis need to be interpreted in the context of the study methodologies. Observational studies are prone to bias because the groups compared may be dissimilar in characteristics other than their types and date of treatment. Factors other than treatment might be responsible for prematurity, including maternal race, age, previous obstetric history, inadequate prenatal care, stage of maternal illness, tobacco use and illicit drug use. Notably, the European studies had a mostly white patient population, whereas the opposite was true for the American-based studies. Some of these factors were adjusted for in the studies. The European Collaborative Study [4-6] adjusted for maternal CD4 T-cell count, race, age and illicit drug use; the US studies by Tuomala et al. [11] and Cotter et al. [21] also adjusted for previous premature delivery and the use of alcohol and tobacco. The study by Cotter et al. [21] also adjusted for the year of delivery and the duration of antiretroviral therapy during pregnancy. The studies that reported results that were adjusted for potential confounding factors, however, compared different types of regimens, making it impossible to synthesize adjusted results from different studies in the meta-analysis.
Several studies included in this meta-analysis included control groups of women not on antiretroviral therapy during their pregnancy from earlier time periods, when such therapy was not widespread. It is possible that other changes in care might have contributed to the differences in obstetric outcomes between the two study periods. If, for example, there had been advances in preventing premature delivery, then this would minimize (towards the null) the apparent relationship of antiretroviral therapy with prematurity. The use of PI-containing regimens, particularly before or early in pregnancy might be more frequent in women with a more advanced stage of HIV disease, in itself a risk factor for premature delivery, thus accounting for some of the increased prematurity observed in this group. This was observed in the study by Cotter et al. [21] and was also noted by other studies. This might also be the case, particularly in earlier years, for the use of any antiretroviral therapy during pregnancy, and this confounding as a result of selection by indication could result in augmenting the relationship between antiretroviral therapy and prematurity. It is also possible that different antiretroviral classes of agents, or even agents within each class, introduced and adopted in clinical practice at different timepoints might have different effects on the risk of premature delivery. Differences between providers at different centers in the choice of antiretroviral regimens used could have accounted for some of the discrepant results among different studies. In the European Collaborative Study [6], an association was observed with the time of study and increasing maternal age, known to be a risk factor for premature delivery. Advancing maternal age could confound the increase in prematurity seen in later time periods, when antiretroviral use was more widespread. Similarly, there were changes in the racial composition of the European Collaborative Study over time, with more African-born women in later times. Both these factors would augment the apparent relationship of antiretroviral therapy with prematurity.
This meta-analysis concurs with previously published US findings that antiretroviral regimens currently being used to treat HIV-infected women during pregnancy are not associated with an increased risk of premature delivery on a population basis. To assess the risk of prematurity for specific subpopulations, researchers will need to analyse additional prospectively collected data representing different ethnic and geographical populations of pregnant HIV-infected women treated with different types of antiretroviral regimens. Ideally, this should be done in the context of randomized, controlled clinical trials. As antiretroviral regimens become increasingly complex and as newer agents are introduced, such analyses will be important for ensuring that antiretroviral regimens continue to provide optimal health benefits for both mothers and their infants.
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
  icon paper stack View Older Articles   Back to Top