Low Birth Weight in Human Immunodeficiency Virus–Exposed Uninfected Infants in Bronx, New York

The results of this large study performed at 1 institution, which included adjustment of potential confounding variables and comparison to human immunodeficiency virus (HIV)-uninfected infants from the same population, revealed low birth weight in HIV-exposed but uninfected infants

Abstract

Background

Prevention of mother-to-child transmission of human immunodeficiency virus (HIV) with antiretroviral therapy (ART) has been highly successful. However, HIV-exposed uninfected (HIV-EU) infants might be at increased risk for low birth weight and/or preterm birth. We compared the birth weights and gestational ages of HIV-EU infants to those of HIV-unexposed control infants in Bronx, New York, an epicenter of the HIV epidemic in the United States.

Methods

This study was performed with a retrospective cohort of HIV-EU infants born at Montefiore Medical Center between 2008 and 2012 and HIV-unexposed control infants. Each HIV-EU infant was matched according to year of birth with 5 HIV-unexposed controls from the New York City Department of Health and Mental Hygiene birth certificate database. We used regression models to assess the association between HIV exposure and birth weight while controlling for potential confounders. A secondary analysis was performed to determine the association of maternal protease inhibitor–based ART use and birth weight among HIV-EU infants.

Results

We included 155 HIV-EU infants born between 2008 and 2012 (51% female, 61% black, 32% Hispanic) and 775 HIV-unexposed infants. The mean (± standard deviation) unadjusted birth weights were 2971 ± 616 g (HIV-EU infants) and 3163 ± 644 g (HIV-unexposed infants) (P < .01). Multivariable regression revealed significantly lower birth weight for the HIV-EU infants (difference, −101.5 g [95% confidence interval, −181.4 to −21.6]). We found no difference in mean birth weight or gestational age with maternal protease inhibitor–based ART use when compared to the use of other regimens.

Conclusions

We found significantly lower birth weight among HIV-EU infants. Long-term prospective studies are necessary to determine the implications of this finding on infant growth and development.

Antiretroviral therapy (ART) has been used successfully for the prevention of mother-to-child transmission (PMTCT) of human immunodeficiency virus (HIV) since 1994 in the United States, and it is now the standard of care in all countries around the world [1]. The estimated percentage of HIV-infected pregnant women who underwent ART in 2015 internationally was 77%, and the rate was greater than 95% in the United States [2]. The widespread availability of ART and effective use of PMTCT protocols have led to a decrease in the number of infants diagnosed with perinatally acquired infection. In New York State specifically, no infant was diagnosed with perinatally acquired HIV infection in 2015 [3]. As a result of the dramatic decrease in perinatally acquired infections, there exists a rapidly increasing number of HIV-exposed uninfected (HIV-EU) infants.

In the United States, Europe, Asia, and Africa the association of HIV exposure and/or ART with growth of the fetus and preterm birth have been conflicting. Some results have suggested that HIV-EU infants are at increased risk for low birth weight (LBW) and preterm birth, and others have indicated that HIV and ART exposure has little or no effect [4–18]. Before the widespread use of ART, HIV exposure had been found to be associated with increased rates of premature birth [19]. After the initiation of widespread ART, the effects of HIV on birth weight and preterm delivery became more difficult to assess because of differences in the timing of ART initiation and adherence and the inability to adjust completely for potential confounders such as ethnicity, smoking, alcohol use, illicit drug use, preexisting maternal comorbidities, pregnancy-related maternal comorbidities, maternal viral load, CD4 count, and socioeconomic factors, including access to care. For example, studies in Botswana and Tanzania found an increased risk for prematurely born and small-for-gestational-age infants from mothers who initiated or continued multidrug ART during pregnancy compared to mothers who received intrapartum zidovudine alone or no therapy [20, 21]. In addition, the type of ART prescribed might contribute to differential outcomes among studies. A 2007 meta-analysis of European and US studies found that although the use of ART was not associated with an increased risk of preterm delivery, such risk was increased significantly in the subset of women who were undergoing protease inhibitor (PI)-based therapy [6]. This result was replicated in a 2013 study that included data from 2 large prospective cohorts of HIV-infected pregnant women and their infants in the United States; the study found a significantly increased risk of preterm delivery in mothers who used a PI-based regimen than in those who used a non-PI based regimen during the first trimester of pregnancy but not in the second or third trimester [22]. A limitation of these studies was the absence of an HIV-unexposed control group for comparison.

New York has the highest prevalence of HIV-exposed infants in the United States, and the Bronx has the highest HIV prevalence of any borough in New York City [23]. We performed a retrospective study of HIV-EU infants and HIV-unexposed infants born between 2008 and 2012 at Montefiore Medical Center, one of the largest providers of care to HIV-infected pregnant women in the United States, to determine if perinatal HIV exposure is associated with LBW or preterm birth. We performed a secondary analysis to determine if PI-based maternal ART is associated with LBW or premature birth among HIV-EU infants in the same population.

METHODS

Study Population and Data Collection

We performed a retrospective cohort study to compare perinatal HIV-EU infants to HIV-unexposed infants born at Montefiore Medical Center, a tertiary care referral center for both the Bronx and Westchester counties of New York and northern New Jersey. After approval from the Montefiore Medical Center Institutional Review Board, we identified through a clinical database at Montefiore Medical Center all infants born to a mother with HIV infection between January 2008 and December 2012. We used maternal medical record number, infant date of birth, and maternal date of birth to then identify these infants in the New York City Department of Health and Mental Hygiene (DOHMH) birth certificate database. HIV-unexposed controls born in the same time period at Montefiore Medical Center were selected randomly from the birth certificate database using a 5:1 control/case ratio and matched for year of birth.

Study Outcomes

The primary study outcome was birth weight. We assessed birth weight as a continuous variable and LBW (≤2500 g) as a dichotomous variable. In a secondary analysis, we examined the associations of maternal ART regimen type with birth weight and gestational age.

Exposure of Interest and Other Variables

The primary exposure of interest was perinatal HIV exposure, defined as maternal HIV infection without subsequent perinatal infection. A secondary exposure of interest was maternal ART regimen stratified according to type (ie, PI-based vs other ART regimen). Data on maternal ART use, HIV viral load (third trimester), and CD4 count (third trimester) were collected by medical chart review. From the New York City DOHMH birth certificate database we obtained other variables, including gestational age, infant sex, and the following maternal variables: education (did not complete high school, completed high school or General Equivalency Development, or some school beyond high school), race/ethnicity (Hispanic, non-Hispanic white, non-Hispanic black, Asian/Pacific Islander, or other), number of prenatal care (PNC) visits (≤3 [inadequate], 4 to 9, or ≥10 visits [adequate]), trimester of first PNC visit (no PNC, first trimester, second trimester, or third trimester), diagnosis of prepregnancy or gestational diabetes, diagnosis of prepregnancy or gestational hypertension, documented maternal infection during pregnancy (including gonorrhea, syphilis, genital herpes, chlamydia, hepatitis B virus, and/or hepatitis C virus [HCV]), smoking 3 months before or during pregnancy, alcohol use during pregnancy, drug use during pregnancy, and insurance type (Medicaid, private insurance, or other). Only 8 of the 910 participants were identified in the “other” insurance category; therefore, they were combined with the Medicaid-insured group.

Statistical Analysis

Using Stata 12.0 (College Station, Texas), we created linear and logistic regression models to assess the association between infant HIV exposure and birth weight. Separate multivariable models were created using birth weight as a continuous variable and as a dichotomous variable. The fully adjusted models were created using an a priori approach in which potential confounders considered to have a biologically plausible relationship to the outcome and primary variable of interest were included. A backward-elimination approach was used also to determine the most parsimonious model by individually removing potential confounders that were not associated with the outcome using a P value of >.05 as the cutoff. Regression diagnostics were performed to assess the validity of the models.

We also performed a secondary analysis to determine whether the type of ART (PI versus non-PI based) regimen is associated with LBW or gestational age. In addition to the aforementioned potential confounders, these analyses also controlled for maternal CD4 count in the third trimester. Maternal plasma viral load at delivery and CD4 count were collinear; therefore, only the CD4 count was included in the analysis.

RESULTS

We identified 155 infants born to an HIV-infected mother in the 5-year study period, and all of them remained uninfected. Of these infants, 51% were female, 61% were black, and 32% were Hispanic. The median CD4⁺ lymphocyte count during the third trimester was 450 cells/μL (interquartile range, 311–622 cells/μL), and 82% had an undetectable HIV viral load. We found no significant association between mean birth weight and mode of delivery among the HIV-EU infants (3000 ± 542 g [vaginal delivery] vs 2851 ± 755 g [cesarean delivery]; P = .81). Among those with a detectable HIV viral load (n = 27), the mean log₁₀ viral load was 6.9 ± 2.2 copies/mL. Maternal ART included PI-based (66% [n = 102]) and non–PI-based (15% [n = 24]) regimens; information on the ART regimen was not found during chart review for 19% (n = 29) of the mothers.

The infants were matched according to year of birth to 755 HIV-unexposed control infants (1:5 case/control ratio). In an unadjusted comparison, the mean (± standard deviation [SD]) birth weight of HIV-EU infants was lower (2972 ± 616 g) than that of the HIV-unexposed controls (3167 ± 644 g) (P < .01). However, we found no difference in their mean gestational ages (38.3 ± 2.5 vs 38.4 ± 2.5 weeks, respectively; P = .09).

We found several other maternal demographic and clinical factors that differed among participants and controls that could contribute to LBW (Table 1). A higher proportion of HIV-EU infants were born to a mother who did not complete high school (P < .01) and who identified as non-Hispanic black (61%). Prepregnancy hypertension, gestational hypertension, sexually transmitted infection, hepatitis B virus, HCV virus, smoking, illicit drug use, and Medicaid insurance were documented in a significantly higher proportion of HIV-infected mothers. In contrast, 73% of HIV-infected mothers had adequate PNC (≥10 visits), compared to 61% of the controls (P < .01).

Table 1.

Demographic, Clinical, and Maternal Characteristics of HIV-EU and Matched Control Infants, Montefiore Medical Center, 2008–2012

Characteristic	HIV-EU (n = 155) (% [n] or mean ± SD)	HIV Unexposed (n = 775) (% [n] or mean ±SD)	P
Birth weight (g)	2971.8 ± 616.3	3166.6 ± 644.1	<.01
Low birth weight (≤2500 g)	16	11	.09
Gestational age (wk)	38.3 ± 2.5	38.4 ± 2.5	.09
Female sex	51	51	.93
Maternal age (y)	29.2 ± 7.1	28.1 ± 6.4	<.01
Maternal ART regimen (n = 126)
PI-based regimen	80.9 (102)	NA	—a
Non-PI based regimen	19.2 (24)	NA	—a
Maternal education
Did not complete high school	36	22	<.01
Completed high school or GED	26	44
Some school beyond high school	39	56
Maternal ethnicity			<.01
Hispanic	32	51
Asian/Pacific Islander	1	5
Non-Hispanic white	6	9
Non-Hispanic black	61	35
Other	0.7	1.7
No. of prenatal care visits
≤3 (inadequate)	1.3	4.3	<.01
4–9	26	35
≥10 (adequate)	73	61
Trimester of first prenatal care visit
No prenatal care	0.7	0.3	.24
First	62	57
Second	30	31
Third	7.7	12
Prepregnancy diabetes	3.2	1.7	.20
Gestational diabetes	11	8	.20
Prepregnancy hypertension	11	3.9	<.01
Gestational hypertension	5.8	2.5	.03
Infections
Gonorrhea	2	1	.41
Syphilis	1.3	0.1	.07
Herpes simplex virus	16	4	<.01
Chlamydia	11	5	<.01
Hepatitis B virus	19	6	<.01
Hepatitis C virus	15	6	<.01
Smokingb	11	5.8	.02
Alcohol usec	1.9	0.5	.09
Drug usec	4.5	1.2	<.01
Insurance type
Medicaid	86	67	<.01
Private	13	32
Other	1	0.7

Abbreviations: ART, antiretroviral therapy; GED, General Equivalency Development; HIV-EU, human immunodeficiency virus–exposed but uninfected; NA, not applicable; PI, protease inhibitor; SD, standard deviation.

^aNo P-value could be calculated because there is no comparison to be made between the groups.

^bThree months before or during pregnancy.

^cUse during pregnancy.

In multivariable linear regression analyses, when we adjusted for gestational age, maternal ethnicity, maternal education, number of PNC visits, maternal HCV infection, maternal smoking, prepregnancy hypertension, and insurance type, we found that HIV exposure was statistically significantly associated with a reduction in birth weight (β coefficient, −101.5 g [95% confidence interval (CI), −181.4 to −21.6 g]) (Table 2). When the model was reduced to the most parsimonious model, including only HIV exposure and gestational age, HIV exposure remained significantly associated with lower birth weight (β coefficient, −121.2 g [95% CI, −196.7 to −45.6 g]). In multivariable logistic regression analysis examining LBW (defined as <2500 g), we found that HIV exposure was associated with increased odds of LBW, although this result was not statistically significant (odds ratio, 1.4 [95% CI, 0.6 to 2.9]). When this model was reduced to the most parsimonious model, including HIV exposure and gestational age only, we found similar results (odds ratio, 1.5 [95% CI, 0.8 to 2.8]).

Table 2.

Adjusted Associations of HIV Exposure With Birth Weight, Maternal ART Regimen With Birth Weight, and Maternal ART Regimen With Gestational Age

Variable	β	95% CI	P
Outcome = birth weight (β values are in g)
HIV exposure and birth weight (n = 930)
HIV exposure	−101.5	−181.4 to −21.6	.01
Gestational age (per wk)	183.6	171.9 to 195.2	<.01
Maternal race/ethnicitya
Non-Hispanic black	−113.2	−224.5 to −1.9	.05
Hispanic	−27.7	−137.7 to 82.2	.62
Asian/Pacific Islander	−161.7	−334.1 to 10.7	.07
Other	−279.7	−529.1 to −30.2	.03
Maternal educationb
Completed HS or GED	75.5	7.4 to 158.4	.07
Some school beyond HS	29.3	−45.8 to 104.4	.44
No. of prenatal care visitsc
4–9	52.5	−100.4 to 205.3	.50
≥10	73.2	−78.1 to 224.5	.34
Hepatitis C virus infection	8.7	−100.6 to 117.9	.87
Smoking	−68.5	−183.9 to 47.0	.25
Prepregnancy hypertension	39.6	−91.5 to 170.7	.55
Non-Medicaid insuranced	31.2	−39.3 to 101.8	.39
Maternal ART regimen and birth weight (n = 126)
PI-based (vs non-PI-based) regimen	−27.9	−223.2 to 167.4	.78
Gestational age (per wk)	142.9	112.6 to 173.3	<.01
Maternal ethnicitya
Non-Hispanic black	−59.5	−352.1 to 233.2	.68
Hispanic	203.6	−114.1 to 521.3	.20
Asian/Pacific Islander	−305.3	−926.2 to 315.6	.21
Other	−485.2	−1306.1 to 335.7	.24
Maternal educationb
Completed HS or GED	93.3	−110.9 to 297.6	.37
Some school beyond HS	110.4	−71.3 to 292.1	.23
No. of prenatal care visitsc
4–9	−244.4	−1050.7 to 561.9	.55
≥10	−94.5	−886.6 to 697.6	.81
Hepatitis C virus infection	−143.2	−351.6 to 65.2	.18
Smoking	−176.3	−416.1 to 63.5	.15
Prepregnancy hypertension	−27.6	−260.7 to 205.5	.81
Non-Medicaid insuranced	−17.8	−249.5 to 213.8	.88
Maternal CD4 count (per cells/µL)	0.14	−0.1 to 0.4	.30
Outcome = gestational age (β values are in wk)
Maternal ART regimen and gestational age (n = 126)
PI-based (vs non-PI-based) regimen	−0.3	−1.5 to 1.0	.68
Maternal race/ethnicitya
Non-Hispanic black	−0.8	−2.7 to 1.1	.43
Hispanic	−0.3	−2.3 to 1.7	.76
Asian	−1.8	−5.9 to 2.2	.37
Other	0.6	−4.7 to 5.9	.83
Maternal educationb
Completed HS or GED	−0.7	−2.1 to 0.6	.27
Some school beyond HS	0.6	−0.6 to 1.7	.34
No. of prenatal care visitsc
4–9	−2.5	−7.7 to 2.7	.34
≥10	−1.8	−6.9 to 3.3	.49
Hepatitis C virus infection	−0.3	−1.6 to 1.1	.71
Smoking	−1.0	−2.5 to 0.5	.20
Prepregnancy hypertension	−1.0	−2.5 to 0.5	.19
Non-Medicaid insuranced	−0.1	−1.5 to 1.5	.96
Maternal CD4 count (per cells/µL)	−0.1	−0.1 to −0.1	.08

Abbreviations: ART, antiretroviral therapy; CI, confidence interval; GED, General Equivalency Development; HIV, human immunodeficiency virus; HS, high school.

^aReference group is those whose race is white.

^bReference group is those who did not complete HS.

^cReference group is those with ≤3 visits.

^dReference group is those with Medicaid insurance.

We examined the relationship between maternal ART regimen at delivery and birth weight and gestational age for the 126 mothers for whom these data were available; 102 mothers were on a PI-based regimen, and 24 were on a non–PI-based regimen. In bivariate analysis, we found no statistically significant difference in birth weight (2958 and 2976 g, respectively; P = .91) or gestational age (38.1 and 38.1 weeks, respectively; P = .98) between maternal ART regimens. In multivariable analysis, in which we adjusted for biologically meaningful confounders and maternal CD4 count in the third trimester, we also found no difference in birth weight or gestational age among mothers on a PI-based regimen and those on a non–PI-based regimen.

DISCUSSION

HIV-EU infants in our population had lower birth weight than a matched sample of HIV-unexposed infants, even after we adjusted for multiple confounders, which suggests that maternal HIV infection and/or in utero exposure to ART affects the growth and development of uninfected infants. The majority of women in the study were receiving adequate HIV viral suppression, which provided an opportunity to examine the effect of ART-treated HIV on pregnancy outcomes. Our findings underscore the important effects of ART on the developing fetus that also were found in other cohorts in the United States [17, 18] and abroad [4, 7, 14]. Unlike some of those studies, we did not detect a difference in gestational age, and we did not detect a difference in birth weight among infants of mothers on a PI-based regimen and those on a non–PI-based regimen. However, most of the HIV-infected mothers were on a PI-based regimen (102 [81%] of 126), and larger studies are needed to confirm this finding.

Previous studies that evaluated the effects of HIV exposure and ART on the growth of fetuses and infants have yielded conflicting results [4–18] and were likely affected by lack of a comparison group, maternal variables, or socioeconomic variables such as income, education, race, and number of children in the household. The time at which maternal ART is initiated might have confounded these findings further. All of the HIV-infected mothers in our study underwent ART during their pregnancy. Because birth weight and gestational age at delivery can be affected by in utero ART exposure and/or indirect exposure to HIV infection, we were unable to conclude if LBW in this cohort was a result of either of these factors individually.

The strengths of our study include the large number of HIV-exposed infants in our high-HIV-prevalence population and the inclusion of a control group from the same institution. In addition, we accounted for a large number of sociodemographic and clinical variables that previous studies did not include. Furthermore, we are not aware of any similar studies performed in a developed country where there is almost universal adherence to PMTCT protocols. Although this study being performed in a developed country might limit the generalizability of our findings, it has meaningful implications for the increasing number of HIV-EU infants born in the United States and other developed countries.

One limitation of this study is the potential for unmeasured confounding caused by factors such as other coinfections that might be more prevalent in HIV-infected women (eg, cytomegalovirus) or concomitant use of other maternal medication. A number of limitations in the assessment of the association between ART regimen and prematurity or birth weight also exist. As noted, the majority of the mothers were on a PI-based ART regimen, which therefore limits the power of the study to detect an association between PI-based and non-PI based regimens and outcomes. In addition, the ART regimen was recorded at delivery only; thus, we had incomplete information on the timing of initiation of therapy. Factors such as ART adherence were not assessed. Therefore, we cannot completely exclude the possibility of an effect of PI use on prematurity and birth weight.

CONCLUSIONS

We found significantly lower birth weight among HIV-EU infants than HIV-unexposed infants, which suggests that in utero exposure to HIV or ART might indirectly or directly affect fetal growth and development. The clinical consequences of these changes are yet to be determined. The results of some studies have suggested that HIV or ART is associated with an altered microbiome [24], which could affect infant growth and alter immunologic development [25, 26]. For example, previous studies have found that HIV-EU infants are at increased risk for invasive pneumococcal disease [27] and suggested that HIV-EU infants have a decreased response to bacille Calmette-Guérin and pneumococcal conjugate vaccines [28, 29], but the mechanism of these changes is unclear. Long-term prospective studies of HIV-EU infants are needed to determine the role of maternal ART use on fetal growth restriction, infant growth, immunologic development, infant response to vaccinations, and risk for future infections.

Notes

Acknowledgment. We thank Berton Freedman, MPH (New York City DOHMH), for his assistance with the New York City DOHMH birth certificate database.

Financial support. This work was supported by Einstein-Rockefeller-CUNY Center for AIDS Research (grant P30-AI-124414), which is supported by the following NIH cofunding and participating institutes and centers: National Institutes of Allergy and Infectious Diseases, National Cancer Institute, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Heart, Lung, and Blood Institute, National Institute of Drug Abuse, National Institute of Mental Health, National Institute on Aging, Fogarty International Center, and Office of AIDS Research and National Institutes of Health (grant K01-HL-137557).

Potential conflicts of interest. All authors: No reported conflicts of interest. 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.

Ethics statement. Institutional review board approval for the use of retrospective clinical data from the New York City DOHMH database was received from the Albert Einstein College of Medicine (Bronx, New York).