Timing is everything: assessing the impact of maternal HIV infection diagnosis timing on infant outcomes in a ten-year retrospective cohort study in South Carolina

Abstract

The primary aim was to evaluate the impact of maternal HIV infection diagnosis timing on infant outcomes. The outcomes of interest included preterm delivery (PTD) and confirmed infant HIV infection. Data for all HIV-exposed infants from 2004–2014 were obtained from the state. Maternal HIV infection diagnosis timing was categorized as: perinatal (PHIV), adult infection before conception, and after conception. Infant outcomes were compared according to timing of maternal HIV diagnosis. Bivariate outcomes were compared using χ² tests. Continuous variables were compared using Kruskal-Wallis tests. Logistic regression was used to determine predictors of PTD and infant HIV infection. Most women were diagnosed with HIV before conception: PHIV (21, 3%), adult before conception (431, 69%), and post-conception (176, 28%). Women diagnosed with HIV as an adult before conception were more likely to deliver preterm (P = 0.007). Prenatal care was associated with lower risk of PTD (aOR 0.1, 95% CI 0.04–0.5). Six infants contracted HIV. Infant HIV infection was more likely in women who did not take antiretroviral therapy (aOR 13.5, 95% CI 2.5–72.1) or delivered preterm (aOR 5.3, 95% 1.1–25.1). Women with PHIV were more likely to deliver at term, and there were no HIV infections among PHIV-exposed infants. These findings are reassuring to PHIV women who desire pregnancy.

Introduction

According to a recent HIV surveillance report, 5,180 U.S. adult and adolescent women are living with perinatally acquired HIV (PHIV), accounting for 0.5% of all persons living with HIV.(Centers for Disease & Prevention, 2018) Given improved survival with effective combination antiretroviral therapy (cART), PHIV women are reaching reproductive age.(Kapogiannis et al., 2011) Secondary to life-long HIV infection, persons with PHIV are more likely to be exposed to multiple antiretroviral medications (ARVs) and subsequently develop ARV drug resistance.(Badell, Kachikis, Haddad, Nguyen, & Lindsay, 2013; Lazenby et al., 2016) ARV drug resistance may necessitate alternative cART use, which impacts the care of PHIV women during pregnancy.(Cruz et al., 2010; Lazenby et al., 2016; Munjal et al., 2013)

In addition to increased medical complexity due to ARV drug resistance, PHIV women may be at increased risk of pregnancy complications. PHIV pregnant women from a small case series had increased rates of anemia and hypertensive disorders of pregnancy.(Williams, Keane-Tarchichi, Bettica, Dieudonne, & Bardeguez, 2009) PHIV women have reportedly poorer HIV viral load suppression during pregnancy, leading to increased rates of cesarean delivery (Badell et al., 2013; Munjal et al., 2013; Phillips et al., 2011), and lower CD4 cell counts.(Agwu, Jang, Korthuis, Araneta, & Gebo, 2011; Jao et al., 2017; Phillips et al., 2011) Some infants exposed to PHIV were more likely to be born preterm,(Kenny, Williams, Prime, Tookey, & Foster, 2012; Williams et al., 2009) low birth weight,(Jao et al., 2017) and small for gestational age.(Jao et al., 2012) Despite the potential for poorer HIV control and adverse infant outcomes, smaller observational studies have concluded that infants born to women with PHIV are not more likely to acquire HIV infection.(Calitri et al., 2014a; Crane, Sullivan, Feingold, & Kaufman, 1998; Cruz et al., 2010; Jao et al., 2012; Lundberg, Andersson, Machado, Costa, & Hofer, 2018)

South Carolina participates in the Enhanced HIV/AIDS Reporting Surveillance System (eHARS), a web-based application developed by the Centers for Disease Control and Prevention (CDC) which investigates of all mother-baby pairs of women with known HIV infection.(Centers for Disease & Prevention, 2015) In a previous evaluation of perinatal HIV infections in South Carolina, we determined that 1.2% of HIV-exposed infants contracted HIV between 2004–2014.(Lazenby, Powell, Sullivan, & Soper, 2018) This estimate is consistent with predictions of the overall rate of perinatal infections in the U.S.(Nesheim, Wiener, Fitz Harris, Lampe, & Weidle, 2017) Approximately 1% of persons living with HIV in South Carolina have PHIV, a relatively high proportion compared to national estimates.(Centers for Disease & Prevention, 2018; South Carolina & Environmental Control, 2017) Given the frequency of PHIV in South Carolina, we sought to use retrospective data to evaluate the impact of maternal perinatal HIV infection (PHIV) on infant outcomes. The outcomes of interest included preterm delivery (PTD), low birth weight (LBW), small for gestational age (SGA), and confirmed infant HIV infection.

Materials and Methods

This was a retrospective cohort study approved by the South Carolina Department of Health and Environmental Control (SCDHEC) Institutional Review Board (IRB15–013). De-identified data were obtained from SCDHEC eHARS database for all women with HIV giving birth between 2004 to 2014. Data for each mother-baby pair for which the mother is known to be HIV positive are collected from prenatal records, hospitalizations, lab reports, birth and death certificates, and case interviews.(Centers for Disease & Prevention, 2015) Data presented here represent a secondary analysis of the data obtained from the 10-year study period. The primary analysis has been published.(Lazenby et al., 2018)

The timing of maternal HIV diagnosis was categorized as: PHIV, adult HIV diagnosis before conception, or adult HIV diagnosis after conception. A woman was considered to have PHIV if her HIV serostatus was confirmed and determined to be acquired from her biological, serostatus-confirmed HIV-infected mother in the absence of any other risk factors (i.e. blood transfusion). A woman was considered to have non-PHIV infection if she was diagnosed with HIV > 11 years of age in the absence of questionable perinatal infection (HIV-infected mother) or other risk factors for childhood infection (breastfeeding from an HIV-infected mother or blood transfusion).(South Carolina & Environmental Control, 2017) We chose to separate women diagnosed with HIV as adults before and after conception, given differences in previous ARV exposure. Non-PHIV women diagnosed with HIV before the index prgnancy were considered an adult diagnosis of HIV before conception. Women who were diagnosed with HIV during pregnancy were categorized as an adult diagnosis after conception. This analysis excluded women who were diagnosed with HIV during labor or postpartum.

For most women, ARVs prescribed during pregnancy were available. ARVs were reviewed and categorized by the number of drugs and ARV classes represented in each subject’s ARV regimen. Antiretroviral regimens were categorized as no ARV use, mono ARV (mARV), dual ARV (dARV), and combination antiretroviral therapy (cART). An antiretroviral regimen was considered to be mARV if it included a single antiretroviral drug, dARV if it included two drugs, or cART if it included at least three drugs in at least 2 antiretroviral classes. Subjects who did not take ARVs during pregnancy were categorized as no ARV use. Alternative ART was defined as ARV regimens that contain integrase inhibitors and/or fusion inhibitors. ARV regimens that contained protease inhibitors were classified as PI-based ART.

In addition to ARV use, the following maternal characteristics were available: race, ethnicity, number of prenatal visits, mode of delivery, year of HIV diagnosis, suspected source of HIV infection, and the most recent HIV RNA viral load (copies/mL) recorded before delivery. Race classification was limited to white or black, and ethnicity was reported as either Hispanic or non-Hispanic. The potential sources of maternal HIV infection were: sexual contact, any current or historical intravenous drug use, perinatal infection (PHIV), or unknown. For the purpose of this analysis, an HIV viral load of < 40 copies/mL was indicative of viral suppression.

Infant outcomes obtained from the database included: year of birth, gestational age at delivery, birth weight, vital status, and status of HIV infection (positive, negative, or indeterminate). Preterm delivery (PTD) was defined as delivery before 37 weeks of gestation. Low birth weight (LBW) was defined as a birth weight < 2,500 grams. Small for gestational age (SGA) was defined as a Z-score with a critical value < 2 standard deviations below the mean. A Z-score for SGA was calculated using race, birth weight, gestational age, and infant gender. Data from all 1989 US black or white singleton live births between 25 and 42 completed weeks gestation were used to assign each baby a Z-score reflecting the number of standard deviations above or below the mean birth weight for each week of gestation, adjusted for the baby’s race (black vs. white/hispanic) and gender. Data for infants born in weeks 20–24 were stratified by race only. This more limited stratification was employed for extremely early deliveries since the smaller number of births reduced the stability of birth weight means and standard deviations, and since in extremely early deliveries differences by sex and race are likely to be negligible. For infants born in week 25 and later, Z-scores were adjusted for both factors.(Zhang & Bowes, 1995) Periviable delivery was defined as delivery before 24 weeks gestation. An infant’s vital status was reported as living or deceased at the time of data collection, and neonatal death was defined as the death of an infant prior to 6 weeks of life. An infant’s HIV status was categorized according to CDC recommendations for children under the age of 13. An infant’s HIV status was categorized as indeterminate if the infant was lost to follow up or did not meet criteria for HIV negative status according to CDC criteria for perinatal HIV infection.(Centers for Disease & Prevention, 2014)

A sample size calculation was performed to find a 4-fold difference in the proportion of preterm deliveries between PHIV women and women acquiring HIV as an adult. Assuming a power of 0.8 and alpha of 0.05, the necessary sample size would be 200 women, 40 women with PHIV and 160 women diagnosed with non-PHIV.

Maternal and neonatal variables were not reported for all subjects. For analysis where data were not available, the denominator reflects the number of results available. Statistical analysis was performed using SAS® 9.4 (Cary, NC). A Shapiro-Wilk test for normality was performed on all continuous variables. All continuous variables available for analysis were non-normal in distribution and are reported as medians with corresponding interquartile ranges. Continuous variables were compared using Kruskal-Wallis tests. Bivariate outcomes were reported as percentages and compared using χ² tests. Univariate logistic regression was performed to determine potential predictors of PTD and infant HIV infection. All factors with a P <0.2 in the univariate model were included in a multivariate logistic regression model and reported with an adjusted odds ratio and 95% confidence interval. Interactions between variables included in multivariate analysis were evaluated.

Results

During the 10-year study period, 666 women with HIV delivered 885 infants. Seventeen of these infants were excluded, because their mothers were diagnosed with HIV in labor or postpartum (13) or the timing of maternal HIV diagnosis was unknown (4). The timing of HIV maternal diagnosis was known for the remaining 868 infants: PHIV (26), diagnosis before conception (657), and after conception (186). Among these women, there were 203 subsequent pregnancies and deliveries, which were excluded, leaving 665 infants from the first pregnancy within the study period. When evaluating infant outcomes, we excluded infants from multiple gestations due to a strong association with PTD, LBW and SGA. The final analysis included 665 women with 627 singleton infants. (Figure 1)

Figure 1.

Selection of women with HIV and HIV-exposed infants for study cohort

Women were diagnosed with HIV between 1985 and 2014. Maternal HIV diagnosis was known prior to pregnancy in 481 women (72%): 21 women with PHIV and 460 non-PHIV women diagnosed before conception. One hundred and eighty-four non-PHIV adult women (28%) were diagnosed during pregnancy. Potential sources of HIV infection were reported for all women with an adult HIV diagnosis (644): sexual contact (467, 73%), unknown (141, 22%), and past or present intravenous drug use (36, 5%). The majority of subjects self-identified as non-Hispanic black (79%). Women with PHIV were more likely to be non-Hispanic (P <0.0001) and deliver later in the study period (P = 0.003). They were more likely to take cART during pregnancy, and their ARV regimens were more likely to be PI-based (P =0.0002) or contain alterative ARVs (P = 0.0003). Although there was a trend towards lower median HIV RNA viral loads before delivery among non-PHIV women diagnosed after conception, there were no significant differences in HIV viral load suppression or poor HIV virologic control between these groups of women. (Table 1)

Table 1.

Maternal characteristics of women with HIV according to timing of maternal HIV diagnosis

	PHIV (n=20)	Before conception (n=460)	After conception (n=184)	P
Maternal race, black	17 (85)	366 (78)	132 (72)	0.1
Maternal ethnicity, Hispanic	0	15 (3)	21 (12)	<0.0001
Intravenous drug use	0	31 (7)	5 (3)	0.07
Year of delivery	2010 (2008–2013)	2007 (2005–10)	2007 (2005–10)	0.003
Delivery of twin gestation	0	22 (5)	6 (3)	0.4
Prenatal care	16 (80)	377 (95)	159 (97)	0.4
Prenatal visits, number	9 (6–10)	9 (6–12)	10 (7–13)	0.5
No antiretroviral use during pregnancy	1(5)	61 (13)	23 (13)	0.5
Mono antiretroviral therapy	2 (10)	75 (19)	25 (16)	0.4
Dual antiretroviral therapy	0	91 (23)	48 (30)	0.008
Combination antiretroviral therapy	18 (90)	233 (58)	88 (55)	0.01
Protease inhibitor-based combination antiretroviral therapy	19 (91)	213 (46)	78 (42)	0.0002
Alternative combination antiretroviral therapy	4 (19)	15 (3)	4 (2)	0.0003
HIV RNA viral load, copies/mL	67 (31–137)	74 (20–506)	47 (0–303)	0.06
HIV RNA viral load < 40 copies/mL	8 (42)	133 (38)	67 (49)	0.1
HIV RNA viral load > 1,000 copies/mL	2 (11)	78 (22)	23 (17)	0.2

These descriptive data are from the 1^st pregnancy within the study period for women living with HIV. Data were missing for one woman with PHIV. Binomial variables are described as percentages and continuous variables are described as medians with a corresponding interquartile range.

Women with PHIV were less likely to have a PTD (5%) compared to women diagnosed before (23%) and after conception (13%) (P =0.007). Other infant outcomes, LBW and SGA, were similar between PHIV women and non-PHIV women. When accounting for potential confounding variables for PTD (no ARV use, HIV RNA viral load ≥ 1,000 copies/mL, timing of maternal HIV diagnosis), any prenatal care before delivery was associated with a significant reduction in PTD (aOR 0.1, 95% CI 0.04–0.5). Non-PHIV women diagnosed with HIV before conception were more likely to deliver preterm when compared to women diagnosed after conception (aOR 1.9, 95% CI 0.99–3.4). Interactions between any prenatal care and no ARV use in the multivariate model were evaluated and found to be insignificant (p=0.99). Although a model evaluating an interaction between no ARV use and an HIV viral load ≥ 1,000 copies/mL at delivery was not significant (p=0.52), we did see a trend in the effect of viral load on preterm birth. Among women with an HIV viral load ≥ 1, 000 copies/mL, those who were not taking ARVs were more than twice as likely to experience PTB compared to those taking ARVs (uaOR 2.6, p=0.06). In contrast, the corresponding OR between ARV and PTB was weaker and not significant among women with an HIV viral load < 1,000 copies/mL (uaOR 1.7, p=0.3). Because neither subgroup analysis was significant, we did not include this in the final multivariate analysis. (Table 3)

Table 3.

Predictors of preterm delivery in singleton infants exposed to HIV

	Unadjusted Odds Ratio (95% Confidence Interval)	P	Adjusted Odds Ratio (95% Confidence Interval)	P
Prenatal care	0.2 (0.1–0.4)	<0.0001	0.1 (0.04–0.5)	0.004
No maternal ARV use	2.2 (1.4–3.7)	0.001	0.8 (0.2–2.6)	0.7
HIV RNA viral load > 1,000 copies/mL	1.9 (1.1–3.1)	0.02	1.6 (0.9–3.0)	0.1
Timing of maternal HIV diagnosis	Diagnosis after conception=reference category
Diagnosis before conception	1.9 (1.2–3.2)	0.03	1.9 (0.99–3.4)*	0.051
Perinatal HIV	0.35 (0.04–2.7)	0.2
Black, non-Hispanic mother	1.2 (0.8–2.0)	0.4

Adjusted model, n=393.

^*We excluded PHIV women from the multivariate analysis. Due to a small number (n=20), the inclusion of this group caused the adjusted model to fail.

Six infants contracted HIV infections, but there were no HIV infections among infants born to PHIV women. Additionally, periviable delivery or neonatal death did not affect infants of PHIV women. In contrast, non-PHIV women diagnosed before conception had poorer infant outcomes, including increased PTD (23%), confirmed infant HIV infection (1.4%), and neonatal death (1.2%). (Table 2) When accounting for intravenous drug use and prenatal care, HIV infection in infants was associated with no maternal ARV use (aOR 13.5, 95% CI 2.5–72.1) and PTD (aOR 5.3, 95% CI 1.1–25.1). Both of these factors were less common among PHIV women. Details of this regression analyses are reported in Table 4.

Table 2.

Infant outcomes according to timing of maternal HIV diagnosis

	PHIV (n=20)	Before conception (n=431)	After conception (n=176)	P
Cesarean delivery	11 (52)	247 (58)	89 (52)	0.3
Gestational age at delivery	38 (38–39)	38 (37–39)	38 (37–39)	0.1
Preterm delivery	1 (5)	96/423 (23)	23/175 (13)	0.007
Birth weight, grams	3148 (2741–3316)	2911 (2579–3260)	2922 (2519–3311)	0.4
Low birth weight	2 (10)	99 (23)	42 (24)	0.3
Small for gestational age	0	8/423 (1.9)	5/175 (2.9)	0.6
Periviable delivery	0	6 (1.4)	1 (0.6)	0.6
Neonatal death	0	5 (1.2)	2 (1.1)	0.9
Confirmed infant HIV infection	0	4 (0.9)	2 (1.1)	0.9
Indeterminate infant HIV infection	0	28 (7)	17 (10)	0.2

Data included for 1^st pregnancy and all multiple gestation deliveries were excluded. Data for 15 infants: 13 whose mothers were diagnosed in labor or postpartum and 2 for whom maternal diagnosis timing was unknown.

Table 4.

Predictors of HIV infection in singleton infants exposed to HIV

	Unadjusted Odds Ratio (95% Confidence Interval)	P	Adjusted Odds Ratio (95% Confidence Interval)	P
No maternal ARV use	7.9 (2.1–30.2)	0.002	13.5 (2.5–72.1)	0.002
Preterm delivery	5.1 (1.4–19.4)	0.02	5.3 (1.1–25.1)	0.04
Intravenous drug use	5.2 (1.0–26.0)	0.04	4.3 (0.7–28.7)	0.1
Prenatal care	0.2 (0.03–0.9)	0.04	2.5 (0.3–20)*	0.4
Cesarean delivery	2.8 (0.6–13.5)	0.2
HIV RNA viral load > 1,000 copies/mL	2.6 (0.4–15.9)	0.3
Timing of maternal HIV diagnosis	1.4 (0.3–6.7)	0.7

Adjusted model, n= 560.

^*Prenatal care became a positive association in multivariate analysis. We believe this occurred because there were so few women in this category; there were zero women with an HIV infected infant who were not taking ARVs and did not receive PNC, but there were 2 women taking ARVs who did not receive prenatal care that had HIV-infected infant.

Discussion

In this 10-year retrospective study of pregnant women from a statewide database, women with perinatal HIV infection did not experience worse infant outcomes. Women with PHIV were less likely than non-PHIV women who were diagnosed before conception to experience preterm delivery. PHIV women did not experience any periviable deliveries nor neonatal deaths. Some previous studies have reported PHIV women are more likely to have PTD, LBW, and SGA infants.(Badell et al., 2013; Calitri et al., 2014b; Crane et al., 1998; Jao et al., 2017; Jao et al., 2012; Phillips et al., 2011; Whitmore, Zhang, Taylor, & Blair, 2011) In our cohort, preterm delivery was reduced for women receiving any prenatal care. Due to lifelong HIV infection and potential continued engagement in HIV care, we speculate that PHIV women may have been more likely to be engaged in prenatal care.

Infants born to PHIV women in this cohort did not contract HIV infection. Factors associated with infant HIV infection were no ARV use, PTD, and maternal intravenous drug use. PHIV women in this cohort were more likely to use cART, and they were less likely to experience PTD and none reported intravenous drug use. We cannot assess pre-pregnancy use of cART in either PHIV or non-PHIV women diagnosed before conception due to database limitations. Given the potential association with cART use at conception and PTD in other studies,(Fowler et al., 2016) it is interesting that both of these groups would be likely to have cART exposure at conception and had such different rates of PTD. Despite more cART use among PHIV during pregnancy, the proportions of women with virologic suppression and cesarean deliveries were similar. Given this, there were likely other factors that were not appreciable in this study that may have contributed lower infant HIV infections among PHIV women. Given our small sample size, there were potentially not enough observations in PHIV women to draw a strong conclusion. We speculate based on our clinical observations that women with HIV diagnosed before conception may have social and behavioral factors that contribute to lower compliance with ARV use and subsequently poorer HIV virologic suppression.

Our retrospective cohort study has several limitations, including missing data collection within the eHARS database. The database lacked information on the timing of ARV drug initiation or duration of the ARV drug use before or during pregnancy. This study was unable to determine if women diagnosed with HIV before conception were taking ART prior to conception. As a result, this study is unable to address concerns that pre-conception ART use increases risk for PTD.(European Collaborative, Swiss, & Child, 2000) Although these data included more infants exposed to PHIV mothers than most previously published observational studies, there were only 26 PHIV women with 21 first singleton births. Despite these limitations, our study strength is in the number of observations in a state-wide database that includes a large cohort of HIV-exposed infants and women with HIV over a 10-year period.

The data presented here are representative of a state in the Southeastern U.S. with a high prevalence of HIV, including persons living with PHIV. These data are likely generalizable to similar populations in the Southeast. This study uniquely separates non-PHIV women into those diagnosed before and after conception. In contrast to previous studies, we were able to demonstrate worse outcomes for these non-PHIV women diagnosed before conception in comparison to PHIV and non-PHIV women diagnosed after conception. Clinicians can reassure PHIV women who desire pregnancy that cART adherence and prenatal care will likely result in good infant outcomes. From these data, there was no association with cART use or PI-based cART use and PTD. As described in previous studies, PHIV women often have other psychosocial barriers to medical care which need to be addressed in addition to providing excellent HIV care..(Millery et al., 2012) We have previously shown that HIV-centered obstetric care, which includes patient advocacy, social work, and case management, can further potentially reduce new perinatal HIV infections and increase uptake of reliable contraception.(Powell, DeVita, Ogburu-Ogbonnaya, Peterson, & Lazenby, 2017)

Conclusions

This study contributes to existing knowledge of the risks attributed to PHIV and pregnancy. Contrary to some previous reports, our findings suggest that PHIV women are not more likely to experience adverse infant outcomes, including PTD, LBW, or SGA. Additionally, our findings suggest that ARV use does not increase the risk of PTD when compared to no ARV exposure in pregnancy. Future studies should consider prospective data collection of PHIV women during pregnancy, including use of cART at conception and during pregnancy.