|Home | About | Journals | Submit | Contact Us | Français|
Conflicting results have been reported among studies of protease inhibitor (PI) use during pregnancy and preterm birth. Uncontrolled confounding by indication may explain some of the differences between studies.
777 HIV-infected pregnant women in a prospective cohort who were not on ARV at conception were studied. Births < 37 weeks gestation were reviewed and deliveries due to spontaneous labor and/or rupture of membranes were identified. Risk of preterm birth and low birth weight (< 2,500 grams) were evaluated using multivariable logistic regression.
72% of the study population received combination ARV with PI during pregnancy and a total of 130 preterm births were observed. In adjusted analyses, combination ARV with PI was not significantly associated with spontaneous preterm birth compared to ARV without PI (Odds ratio (OR): 1.22, 95% confidence interval (CI): 0.70, 2.12). Sensitivity analyses including women on ARV prior to pregnancy also did not identify a significant association (OR: 1.34, 95% CI: 0.84, 2.16). Low birth weight results were similar.
No evidence of an association between use of combination ARV with PI during pregnancy and preterm birth was found. Our study supports current guidelines which promote consideration of combination ARV for all HIV-infected pregnant women.
The relationship between protease inhibitor (PI) use during pregnancy and preterm birth has been of interest since initial studies examining this association found an increased risk of preterm birth with use of combination antiretrovirals (ARV) with PIs compared to no ARV (1-2). Since then, numerous studies have evaluated the association between ARV including PIs and preterm birth (3-13) with conflicting results. The majority of studies reporting an increased risk of preterm birth with use of combination ARV with PI versus either no therapy or mono/dual therapy (1-10) were conducted in European cohorts, while most studies in US and Latin American women have reported no increased risk of preterm birth (11-13). It has been suggested that differences in the obstetrical care or characteristics of HIV-infected women in Europe and the US may account for the contrasting results (14). However, two recent US studies have found an association between PI use and increased risk of preterm delivery (5, 6). The US study from a single obstetrical service noted that PI use was reserved for women with more advanced HIV disease and those who had failed other antiretroviral therapy (5). It is possible that the association with preterm birth in this study was due to confounding by indication from maternal HIV disease severity.
ARV during pregnancy is indicated for HIV-infected pregnant women to prevent mother to child transmission but can also be indicated for maternal health (15). Pregnant women with symptomatic HIV disease (16), high viral loads, and low CD4+ cell counts are more likely to be treated with highly active antiretroviral therapy, including ARV with PIs (5). Advanced HIV disease has also been associated with preterm delivery (17-20). Failure to adjust for potential confounders related to maternal disease may result in biased estimates. None of the previous studies were able to completely adjust for factors related to maternal disease stage with the majority unable to adjust for HIV viral load, a potentially important confounder of the association between PI use and preterm birth (21).
Although a randomized controlled trial has been recommended to definitively determine the impact of PI use on preterm birth (21), previous studies of existing observational cohorts of HIV-infected pregnant women have not utilized available methodologies to minimize the impact of confounding by indication. Specifically, previous studies have not restricted the study population to women who do not have an indication for antiretroviral therapy at conception. Our study evaluated the effect of ARV with PI use during pregnancy on the risk of preterm birth among HIV-infected pregnant women enrolled in a large US-based multicenter prospective observational cohort study who were not on antiretroviral therapy at conception, adjusting further for markers of disease severity including CD4+ cell count, viral load, and CDC clinical category. Our study also attempted to reduce outcome misclassification by distinguishing spontaneous preterm births from all preterm births, which may also include births with identifiable causes of preterm delivery.
The study population included women enrolled in the International Maternal Pediatric Adolescent AIDS Clinical Trials Group Protocol P1025, which is a prospective observational study designed to assess the use and safety of ARVs and other interventions for HIV-infected pregnant women and their infants. Beginning in October 2002, HIV-infected pregnant women were enrolled if they were ≥ 13 years of age, ≥ 14 weeks gestation or within 14 days postpartum. Since December 2007, enrollment was allowed as early as 8 weeks gestation. Institutional review boards approved the protocol at all 56 clinical sites located in the US and Puerto Rico and written informed consent was obtained from enrolled women. The population eligible for this study consisted of women with an estimated date of conception at least 10 months before March 5, 2008 (date that data were frozen), a singleton pregnancy, a first enrolled pregnancy in P1025, and at least one CD4+ cell count during pregnancy. The primary study population included only women who were not receiving ARV at conception or within 6 months prior to conception. Women who had no reported ARV use during pregnancy or with unknown treatment history were excluded from the study population. Women with current or recent (within 6 months) ARV use at time of conception were excluded from primary analyses but were included in sensitivity analyses.
Sociodemographic characteristics, diagnoses, markers of HIV diseases severity, ARV use, and traditional risk factors for preterm birth were primarily abstracted from medical records. Substance use during pregnancy was ascertained using self-administered behavioral questionnaires that were mailed to the data management center in sealed envelopes without site review. For statistical analyses, substance use was defined as use of amphetamines, methamphetamines, barbiturates, cocaine, MDMA (Ecstasy), heroin, methadone, marijuana or non-prescribed prescription drugs (codeine, valium, xanax). A diagnosis of sexually transmitted disease was defined to include any diagnosis of syphilis, or infection with C. trachomatis, N. gonorrhea, or herpes simplex virus.
ARV use during pregnancy was categorized as single drug, two drugs, combination ARV without PI, or combination ARV with PI. Combination ARV with PI use was defined as the concomitant use of at least three drugs, one of which was a PI. If more than one regimen was used during pregnancy, women were assigned according to their most complex antiretroviral regimen based on the following hierarchy: combination ARV with PI > combination ARV without PI > two agents > single agent. For statistical analyses, combination ARV without PI, dual agent, and single agent ARV were combined into one category due to the small number of women only on single or dual agents during pregnancy.
Gestational age at delivery was estimated using obstetric estimates. Obstetrical gestational age of the infants at delivery was estimated using the date of last menstrual period, initial ultrasound and maternal physical examination, or conception date by assisted reproduction. If menstrual and ultrasound dating were discordant and the difference was within the limits of ultrasound dating, ultrasound dating was used. If calculated gestational ages by last menstrual period and by ultrasound estimates were not available or were discordant by more than 2 weeks, the obstetrical gestational age recorded in the patient chart was used. Preterm birth was defined as delivery at less than 37 weeks of gestation and low birth weight was defined as birth weight less than 2500 grams. All preterm births were reviewed by one study obstetrician (RET) who was blinded to ARV use, and additionally classified into three groups: 1) spontaneous preterm births (documentation of preterm labor or preterm spontaneous rupture of membranes, or labor without either induction medications or amniotomy), 2) possibly related preterm births (diagnosis leading to delivery possibly associated with ARV use), or 3) unrelated preterm births (delivery with known cause not directly or indirectly related to ARV use) (Table 1). The classifications were circulated to P1025 study obstetricians and a consensus was obtained. Since risk of spontaneous preterm birth was the primary outcome of interest, possibly related and unrelated preterm births were excluded from primary analyses. Additional analyses were conducted for combined spontaneous and possibly related preterm births, and all preterm births to assess the effect of outcome misclassification (Table 1).
The chi-square test was used to compare characteristics of women who received ARV with versus without PI. The Mantel-Haenszel chi-square test was used to compare the crude risk of preterm delivery according to ARV use. Statistical significance was defined using a two-sided Type I error rate of 0.05.
Logistic regression models were used to estimate the association of ARV with PI versus ARV without PI with the risk of 1) preterm birth and 2) low birth weight. To construct parsimonious multivariable models, predictors of preterm birth or low birth weight with a likelihood ratio p-value of 0.10 or less in a univariable model were included in respective multivariable regression analyses. In addition, to ensure appropriate adjustment for confounding by indication, clinical markers of HIV disease severity, including CD4+ cell count, viral load, and CDC clinical category, were added to the final regression models even if they were not univariable predictors of preterm birth or low birth weight, if they changed the odds ratio for ARV use by more than 10%. 95% Wald confidence intervals and p-values were used to assess whether there was a statistically significant association between ARV with PI and preterm birth or low birth weight. In sensitivity analyses intended to replicate previous studies, logistic regression models were used to evaluate the association between ARV with PI and the risk of preterm birth and low birth weight in an augmented study population which included women with recent or current ARV use at conception. In further sensitivity analysis to evaluate differences in clinical practice between the US and Europe with regards to use of tocolytic medications to arrest preterm labor, infants of women who received tocolytic medications prior to 37 weeks gestation were reclassified as ‘preterm.’ The association between ARV with PI and the risk of this expanded ‘preterm’ outcome was then evaluated using logistic regression analyses. Analyses were conducted using SAS version 9 (SAS Institute, Cary, NC).
As of March 5, 2008, 1402 women had been enrolled in the P1025 cohort, of whom 1,196 were eligible for this study. Of these, 22 women were excluded from all analyses due to no reported ARV use during pregnancy or unknown antiretroviral history, and 397 women were excluded from primary analyses due to current or recent ARV use at conception, leaving 777 women in the primary study population.
Demographic and clinical characteristics of the 777 women in the study population are provided in Table 2. More than half of the women were older than 25 years, 58% were black, 55% were diagnosed with HIV prior to conception, and 13% had prior preterm births. While no women were on ARV at conception by study design, half were on ARVs for more than 20 weeks during pregnancy and the majority (87%) initiated ARVs in the first or second trimester. Only 15% of women had a documented CD4+ cell count <200 cells/ml, 16% had a viral load >30,000 copies/ml, and 8% had symptomatic HIV disease (CDC category C excluding CD4+ cell count criteria) during pregnancy. There were also few co-morbid diagnoses during pregnancy. These included gestational diabetes (1%), gestational hypertension (3%), hepatitis C infection (4%), and sexually transmitted infections (19%).
Of the 777 women in the study population, 558 (72%) received combination ARV with PI. The most common regimens in this group were zidovudine plus lamivudine with nelfinavir (47%) and zidovudine plus lamivudine with lopinavir/ritonavir (22%). Two hundred and two women (26%) received combination ARV without PI; the most common regimens were zidovudine plus lamivudine plus abacavir, (61%) and zidovudine plus lamivudine with nevirapine (30%). Eleven women (1%) received two agents during pregnancy; 10 received zidovudine plus lamivudine and 1 received emtricitabine plus tenofovir. Six women (1%) received only zidovudine during pregnancy.
There were significant differences between women who received ARV with and without PI (Table 2). Women who were on ARV with PI during pregnancy were more likely to have CD4+ cell counts < 200 cells/mL (18% vs. 10%), viral load > 30,000 copies/mL (20% vs. 6%), and symptomatic clinical disease (10% vs. 3%). They also were more likely to have enrolled into P1025 in their second trimester (34% vs. 24%) and have delivered in the later years of the study, 2007-2008 (22% vs. 14%).
Seventeen percent (130/777) of all births in the study population were preterm. Of the 558 women who received ARV with PI, 102 (18%) delivered preterm. Of the 219 women who received ARV without PI, 28 (13%) delivered preterm. After outcomes review of the 130 preterm births, 58% were classified as spontaneous preterm births, 21% as possibly related preterm births, and 21% as unrelated preterm births (Table 1). There were several diagnoses associated with the possibly related preterm births (Table 1). The majority of the unrelated preterm births were elective cesarean sections or inductions.
Figure 1 shows the frequency distributions of gestational ages at delivery according to ARV use for the spontaneous preterm birth study population (Figure 1a) and all preterm birth study population (Figure 1b). Comparison of the two frequency distributions reveals that the majority (67%) of the twelve very preterm births in women who took ARV with PI during pregnancy had identifiable causes of preterm delivery that may not be associated with ARV use during pregnancy. Three were associated with non-reassuring fetal heart rate tracing, one in a woman with gestational hypertension, and one in a woman with active HSV infection. Two of the very preterm births were intrauterine fetal demises and two were associated with placental abruption. One very preterm birth was associated with pre-eclampsia. Although, there was a significant trend toward a higher risk of preterm birth, particularly very preterm birth (< 32 weeks gestation) among women using ARV with PI in the all-preterm birth study population (p-value: 0.04), no significant difference in the risk was detected in the spontaneous preterm birth study population (p-value: 0.56) or combined spontaneous and possibly related preterm birth study population (p-value: 0.14).
In a multivariable logistic regression analysis of spontaneous preterm birth, no significant association was observed between risk of spontaneous preterm birth and use of ARV with PI compared to ARV without PI (Odds ratio (OR): 1.22, 95% confidence interval (CI): 0.70, 2.12) (Table 3). Similarly, no significant association was observed between use of ARV with PI and risk of combined spontaneous and possibly-related preterm birth or all preterm birth. The outcomes review and exclusion of possibly-related and unrelated preterm births increased the univariable p-values of several known predictors of preterm birth to > 0.10 so that they were not included in the multivariable model of spontaneous preterm birth. Sensitivity analyses including the 397 women who were initially excluded due to ARV use prior to pregnancy also did not detect a significant association between ARV with PI and risk of spontaneous preterm birth (OR: 1.34, 95% CI: 0.84, 2.16). Finally, in further sensitivity analyses re-classifying women who took tocolytic therapy at less than 37 weeks as delivering ‘preterm,’ no significant association between ARV with PI and all preterm birth was detected (OR: 1.25, 95% CI: 0.80, 1.96).
Fourteen percent (110/760) of infants with birth weight available had low birth weight. No significant association was observed between ARV with PI use and risk of low birth weight in the low birth weight study population (OR: 1.36, 95% CI: 0.76, 2.44) (Table 3). Sensitivity analyses including women who were initially excluded due to prior ARV use also did not detect a significant association between ARV with PI use and low birth weight (OR: 1.38, 95% CI: 0.83, 2.28).
In this large cohort of HIV-infected pregnant women in the US we observed no significant association between use of ARV with PI and the risk of preterm birth or low birth weight compared to ARV without PI. Our result is consistent with previous US and Latin-American based studies of HIV-infected women (11-13), and is inconsistent with the findings of large studies from Europe (2-4, 7-10) and two studies from the US (5, 6).
We found a similar risk of all preterm births in our study population (13% among those who received ARV without PI and 18% among those who received ARV with PI) as previous US cohorts of HIV-infected pregnant women (14% and 18% respectively) (11). Our overall risk of preterm births (17%) was the same as that found in the earlier cohort of the European Collaborative Study (2), but differed from the risk reported in the more recent European Collaborative Study (4). There has been a steady increase in preterm births within the European cohort over calendar time and the proportion of preterm births from 2000-2004 of 24.9% is significantly higher than our estimate of 17% from 2000-2008 (p<0.0001) (4). Even when excluding elective cesarean deliveries, the proportion of preterm births within the European cohort is significantly higher than that found in our study within the similar time period (37.1% vs. 15%, p<0.001). It is unclear why preterm births have increased over time in Europe while there has been no such increase observed between the earlier US study and this study.
Core differences in European and US HIV-infected pregnant women populations have been suggested as a possible reason for the conflicting results between European studies and US studies (14). Comparison of some potential risk factors for preterm birth between the more recent European Collaborative Study (4) and our study does identify some differences between the study populations. There were significantly more black women and women who reported using illicit drugs in our study population compared to the European study (58% vs 22%, p<0.001 and 15% vs. 9%, p<0.0001 respectively). There also was a significant difference in the age distribution, with our study having higher proportion of younger (<25 years, 36% vs. 22%) and older (≥ 35 years, 17% vs. 14%) mothers (p<0.0001). Given these differences, one would expect a higher prematurity rate in the US population. This however was not observed. The differences between our study and the European Collaborative Study can be adjusted for in multiple regression analyses, although some residual confounding may persist. If the contrasting results of the two studies are to be considered population specific there must be some yet unidentified characteristic that differs between European and US HIV-infected pregnant women that is associated with both combination ARV with PI use and risk of preterm birth.
A unique feature of our study is that we attempted to address factors that could explain the discrepancy in findings between previous published studies. Previous studies that have shown associations between PI use and preterm delivery have noted that PI-containing regimens were reserved for women with advanced disease (5). Thorne et al did note a greater association with prematurity in women who were receiving ARV prior to pregnancy, presumably for their own health, but did not remove these women from analysis. By excluding women who were receiving ARV at the time of conception, presumably for their own health, we attempted to minimize confounding by indication, i.e. the possibility that PI-containing ARV was preferentially prescribed for women with the most advanced disease who might have been at increased risk for preterm delivery due to disease. We attempted to remove residual confounding by indication in the study population through addition of CD4+ cell count, viral load, and disease in our multivariate models, but some confounding may still persist. By extensively reviewing all preterm deliveries and analyzing separately those with spontaneous preterm birth and those potentially due to other factors, we reduced the effect of outcome misclassification. Additionally, to address differences in management of preterm labor in the US vs. Europe, we modeled the extreme scenario in which all women who received tocolytic medication prior to 37 weeks gestation would have had an immediate cesarean delivery.
Current Public Health Service Task Force recommendations for use of ARV in pregnant HIV-infected women recommend nevirapine initiation in pregnant women with CD4 counts >250 cells/mm3 only if benefit clearly outweighs potential increased risk life-threatening hepatotoxicity (22). Thirty percent of our study population on combination ARV without PI during pregnancy used nevirapine. The majority (85%) of nevirapine use however occurred before 2005, prior to revisions of the ARV recommendations for HIV-infected pregnant women.
Our study has some limitations. Although it is reasonable to assume that women receiving ARV at the time of conception were receiving ARV because of more advanced disease, we were not able to directly assess whether women were receiving ARV during pregnancy for their own health or solely for fetal protection. Due to the selection criteria for our primary analysis, our numbers are smaller than in some previous published studies. In particular, the number of women on ARV without PI is relatively small, and not many women received either mono- or two drug regimens. Many of the women receiving other regimens without PI were nonetheless receiving highly active regimens. Our results may not be directly comparable to those of previous analyses which found associations between PI use and preterm delivery because the comparison groups are quite different. Also, although our 95% confidence intervals suggest that a greater than 2.1-fold increase in the risk of preterm birth with PI use is unlikely, the size of our study population may have limited our ability to identify smaller increases in risk. Finally, the P1025 study enrolled women at either ≥ 14 weeks gestation (2002-2007) or ≥ 8 weeks gestation (2007-present), which limited our ability to assess the effect of combination therapy with PI use on early fetal losses.
In conclusion, our study found no significant association between combination ARV with PI and the adverse pregnancy outcomes, preterm birth and low birth weight. Our findings support the current recommendations for use of antiretroviral drugs in pregnant HIV-infected women by the US Public Health Task Force, which promote the consideration of combination ARV according to preferred regimens for HIV-infected adults for all HIV-infected pregnant women whether given for indications of maternal health or to prevent mother-to-child transmission (22).
The authors thank the mothers and infants for their participation in IMPAACT protocol P1025, and the dedicated individuals and institutions involved in the conduct of P1025.
Financial support: Overall support for the International Maternal Pediatric Adolescent AIDS Clinical Trials Group (IMPAACT) was provided by the National Institute of Allergy and Infectious Diseases (NIAID) [U01 AI068632], the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), and the National Institute of Mental Health (NIMH) [AI068632]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. This work was supported by the Statistical and Data Analysis Center at Harvard School of Public Health, under the National Institute of Allergy and Infectious Diseases cooperative agreement #5 U01 AI41110 with the Pediatric AIDS Clinical Trials Group (PACTG) and #1 U01 AI068616 with the IMPAACT Group. Support of the sites was provided by the National Institute of Allergy and Infectious Diseases (NIAID) the NICHD International and Domestic Pediatric and Maternal HIV Clinical Trials Network funded by NICHD (contract number N01-DK-9-001/HHSN267200800001C).
Potential conflicts of interest: None