Fatty acid signatures in people with HIV: association with adverse pregnancy outcomes and offspring anthropometrics

Abstract

Background

We assessed the association of polyunsaturated fatty acids (PUFAs) in pregnant people with HIV (PWH) with pregnancy outcomes and offspring anthropometrics.

Setting

This is a cohort of 264 pregnant PWH, and their HIV-exposed uninfected children, enrolled in the Pediatric HIV/AIDS Cohort Study Nutrition sub-study from 2009-2011.

Methods

We measured third-trimester plasma omega-6 and omega-3 PUFA content, each as a percentage of total fatty acid content, via esterification and gas chromatography. Omega-6:omega-3 ratios were calculated. Pregnancy outcomes were hypertensive disorders of pregnancy, preterm birth (<37 weeks’ gestation), and small-for-gestational age (birthweight <10^th percentile). Childhood anthropometrics outcomes were Z-scores for age and sex: 1) weight and length/height (birth to 5 years of age), 2) head circumference (1-2 years), and 3) triceps skinfold thickness (2-5 years). Log-binomial regression models estimated pregnancy outcome prevalence ratios by omega-6:omega-3 ratios as a continuous variable. Linear regression models using generalized estimating equations assessed childhood anthropometric outcomes in those with omega-6:omega-3 ratios >25^th versus ≤25^th percentile.

Results

Each 1% increase in the omega-6:omega-3 ratio was associated with a 25% (95% confidence interval [CI] 8-43%) and 10% (95% CI 3-18%) higher prevalence of hypertensive disorders of pregnancy and preterm birth, respectively, and 13% (95% CI 1-23%) lower prevalence of small-for-gestational age. A difference in childhood anthropometric outcomes was not identified at any time point between exposure groups.

Conclusion

Higher omega-6:omega-3 ratios in pregnant PWH were positively associated with hypertensive disorders of pregnancy and preterm birth, inversely associated with small-for-gestational age birth, and not associated with childhood anthropometric trajectories.

Introduction

Suboptimal nutritional status periconceptionally and during pregnancy is a potentially modifiable risk factor for adverse pregnancy outcomes (APOs) and dysregulated offspring growth and development.^1,2,3 Polyunsaturated fatty acids (PUFAs) are critical nutrients affecting metabolic and inflammatory homeostasis during pregnancy, with omega-6 (n-6) and omega-3 (n-3) PUFAs considered “essential fatty acids” because they cannot be synthesized by the body and must be acquired through diet.⁴ The n-6 PUFAs generate pro-inflammatory and pro-oxidant eicosanoids and other oxylipins, while n-3 PUFAs and their biologically active metabolites share anti-inflammatory and antioxidant properties.⁴ The relative balance of n-6 and n-3 PUFAs mitigate oxidative stress and inflammation in utero, thereby influencing placental development and function.^4-6 Higher plasma content of n-6 relative to n-3 PUFAs has been associated with placentally-driven APOs, including hypertensive disorders of pregnancy (HDP), preterm birth (PTB), and fetal growth abnormalities.^6,7 Further, the relative balance of the placental transport of n-6 and n-3 PUFAs affects fetal metabolic imprinting with ramifications for early childhood growth and development.^8,9

Pregnant people with HIV (PWH) may be at risk for higher rates of APOs, due to chronic immune activation and dysregulation, acute and chronic inflammation, oxidative stress, and aberrant placental development and function associated with chronic infection and exposure to antiretroviral therapy (ART), even in the setting of viral suppression.^10-13 Exposure to similar pathogenic processes may contribute to depressed childhood growth among HIV-exposed uninfected (HEU) children.¹⁴ Additionally, pregnant PWH have higher content of n-6 PUFAs than pregnant persons without HIV, and their n-6:n-3 ratios frequently exceed the recommended ratio.^15-19 However, evaluation of PUFAs in pregnant PWH and their association with APOs and early childhood growth in their HEU offspring are limited, particularly in resource-rich settings.

Thus, we sought to evaluate whether a higher n-6:n-3 PUFA ratio was associated with APOs including HDP, PTB, and small-for-gestational age (SGA), and early childhood anthropometric and adiposity measures in a United States (US) cohort of pregnant PWH and their HEU children.

Materials and methods

Cohort selection

Since 2007, the Surveillance Monitoring for ART Toxicities Study (SMARTT) of the Pediatric HIV/AIDS Cohort Study (PHACS) Network has enrolled pregnant PWH and their children at 22 sites across the US, including Puerto Rico. From 2009-2011, the Nutrition sub-study (R01HD060325) of SMARTT enrolled PWH in the late second or third trimester of pregnancy (≥24⁰ weeks^days gestation) at 15 of the 22 sites.²⁰ This analysis included pregnant PWH from the Nutrition sub-study of SMARTT, but excluded those with multifetal gestations, a history of bariatric surgery, or incomplete PUFA data. All participants provided written informed consent for their own and their children’s participation. The Institutional Review Board at each recruitment site, the University of Miami Human Subjects Research Office, and the Harvard T.H. Chan School of Public Health (HSPH) approved the Nutrition sub-study.

Exposures

The primary exposures were maternal plasma 1) n-6 content, 2) n-3 content, and 3) n-6:n-3 PUFA ratios. Maternal n-6 and n-3 PUFAs were measured, each as a percentage of total fatty acid content, in plasma samples collected at time of enrollment during the late second or third trimester (≥24⁰ weeks’ gestation) per the Nutrition sub-study protocol.²⁰ Plasma specimens were stored at −70°C, and sent to the Department of Nutrition at HSPH for analysis. Esterification and gas chromatography techniques were utilized to measure maternal plasma content of individual PUFAs, each as a percentage of total fatty acid content.²¹ Total lipids were extracted from plasma and PUFAs were transmethylated with methanol and sulfuric acid by previously defined processes.^21-23 Briefly, specimens were evaporated after esterification, and the fatty acid methyl esters were redissolved in iso-octane. An Agilent Model GC 6890 FID gas chromatograph with 7673 autosampler injector (Palo Alto, CA), splitless injection port at 240°C, and a constant flow hydrogen carrier gas at 1.3 ml/min separated the PUFAs.

While forty-two fatty acids were identified with this methodology, our analysis focused on six n-6 PUFAs and four n-3 PUFAs (Supplemental Digital Content, Figure). The six n-6 fatty acids included in the overall n-6 content were linoleic, gamma linolenic, eicosadienoic, dihomo-gamma-linolenic, arachidonic, and adrenic acid. The four n-3 fatty acids included in the overall n-3 content were alpha-linolenic, docosapentaenoic, eicosapentaenoic, and docosahexaenoic acid. The n-6:n-3 PUFA ratios were calculated, and represent PUFA inflammatory signatures. We did not analyze individual PUFAs due to risk of introducing false discovery with multiple comparisons.

Outcomes

Obstetric outcomes were HDP, PTB, and SGA. HDP included preeclampsia with or without severe features, eclampsia, and HELLP (hemolysis, elevated liver enzymes, low platelets) syndrome, as defined by the American College of Obstetricians and Gynecologists; gestational hypertension was not available as an outcome reported in this cohort.²⁴ PTB was defined as delivery prior to 37 weeks’ gestation. SGA was defined as neonatal birthweight <10th percentile based on US standards.²⁵ HDP and PTB were ascertained from the obstetric chart, with SGA ascertained from the neonatal chart.

Child anthropometric outcomes were: 1) Z-scores for weight (WTZ) and length (LNZ) at birth for gestational age and sex based on Olsen et al.,²⁵ 2) WTZ (at 1-5-year study visits) and Z-scores for length/height (LNZ/HTZ, at 1-5-year study visits) and head circumference (HCZ, at 1- and 2-year study visits) based on the Centers for Disease Control and Prevention (CDC) 2000 growth charts,^26,27 and 3) Z-scores for triceps skinfold thickness, TSFT (TSFZ, at 2-5-year study visits) as an estimate of adiposity.²⁸ Trained personnel performed child anthropometric measurements in triplicate using standard techniques.²⁹ Birth length was measured generally within 72 hours of birth, but accepted within 14 days of birth. Birthweight was obtained from the obstetric chart. At annual follow-up study visits, measurements were performed for weight (from birth to 5 years of age), stature (length measured from birth through 35 months of age; standing height measured starting at 36 months of age), head circumference (at the 1- and 2-year study visits), and TSFT (annually starting at the 2-year study visit).^26,30,31

Other Covariates and Data Collection

We collected data on sociodemographic characteristics, obstetric history, and other clinical characteristics through interview and medical record chart abstraction. Characteristics included pre-pregnancy maternal body mass index (BMI), trimester of prenatal care initiation, gestational age at blood draw, earliest HIV RNA viral load (VL) and CD4 count in pregnancy, ART drug class in the first/second trimester, timing of ART initiation, and gestational age at delivery. Gestational age was determined by best obstetric practice for establishment of dating (i.e. by last menstrual period [LMP] or earliest prenatal ultrasound if LMP was unknown or inconsistent with menstrual dating per established dating criteria).³² Children’s health histories were obtained at annual study visits.

Statistical analysis

We produced frequency distributions for categorical variables and plotted distributions for continuous variables. Missing values were queried. A missing category was included in analyses for pre-pregnancy BMI.

For the obstetric outcomes, we used Wilcoxon rank-sum tests to compare the distribution of the median (IQR) of n-6 and n-3 PUFA content, and the n-6:n-3 ratio, among those with versus without the APO of interest (HDP, PTB, SGA). Log-binomial regression models were fit for each of the three APOs with each separate exposure (n-6 PUFA content, n-3 PUFA content, and the n-6:n-3 ratio) to estimate the prevalence ratios (95% confidence intervals, CI) of each outcome for each one unit increase in the exposure. Models of obstetric outcomes were adjusted for potential confounders that were selected based on prior literature and construction of directed acyclic graphs. Maternal age at LMP and annual household income were included in the adjusted models of obstetric outcomes. In a sensitivity analysis, we evaluated models of obstetric outcomes that additionally adjusted for HIV VL.

For child outcomes, we generated box plots representing the distribution of each anthropometric Z-score across age by maternal plasma n-6:n-3 ratio >25^th vs. ≤25^th percentile (Appendix), and compared the distributions using the Wilcoxon ranked-sum test. Pearson correlation coefficients were calculated between each anthropometric Z-score and the n-6:n-3 ratio as a continuous variable at each yearly visit. Linear regression models were fit using generalized estimating equations with the robust variance estimator to evaluate the association of maternal plasma n-6:n-3 ratios >25^th versus ≤25^th percentile with each childhood anthropometric Z-score outcome across age. For each outcome, we present the estimated mean difference in the Z-score outcome between those with n-6:n-3 ratios >25^th versus ≤25^th percentile at ages 2 and 5 years for WTZ, LNZ/HTZ, TSFZ, and at ages 1 and 2 years for HCZ.

For analyses of WTZ and LNZ/HTZ, we excluded values with standard deviations (SD) < −5 and >5 as they were unlikely values. We used inverse probability of censoring weights to account for possible selection bias due to loss-to-follow-up or administrative censoring. Based on prior literature and directed acyclic graphs, we considered the following potential confounders for the models of childhood outcomes: ethnicity, regional sites, socioeconomic characteristics, and the following maternal factors: HIV VL and CD4 count, ART during pregnancy, substance use, and pre-pregnancy BMI. However, final regression models included n-6:n-3 ratios >25^th versus ≤25^th percentile, child age, n-6:n-3 ratios >25^th vs. ≤25^th percentile*child age (interaction term), maternal age at LMP, and household income. In a sensitivity analysis evaluating early childhood anthropometrics, we added HIV VL to the adjusted models. All statistical calculations were performed using SAS Version 9.4 (SAS Institute, Cary, NC).

Results

A total of 317 pregnant individuals were initially deemed eligible and enrolled into the Nutrition sub-study. Of these, 21 pregnant individuals dropped out prior to starting the study or were lost to follow-up, with 296 providing plasma samples for PUFA measurement. Of these, 264 maternal-infant pairs were included in this analysis, with reasons for exclusion indicated in Figure 1. Participants predominately had low annual household income <$10,000 (58%), self-identified as Black (69%), had an overweight or obese pre-pregnancy BMI (66%), and received their HIV diagnosis before the index pregnancy (80%) (Table 1). Ninety-four percent received ART by the second trimester with the majority receiving PI-based ART (84%), typically with a backbone of tenofovir disoproxil fumarate plus either emtricitabine or lamivudine. Blood specimens for PUFA assays were collected at a median (IQR) of 32 (30-36) weeks’ gestation, although specimen collection ranged from 24-40 weeks. Three participants had plasma samples collected at time of study enrollment in the second trimester; the remainder had plasma samples collected at time of study enrollment in the third trimester.

Figure 1.

Flowchart showing study population derivation

PUFAs—polyunsaturated fatty acids

Table 1.

Participant characteristics at enrollment

Sociodemographic characteristics		(n, %) or Median (IQR) N=264	Clinical characteristics		(n, %) or Median (IQR) N=264
Maternal age at LMP (years)		27.5 (22.5, 32.6)	Trimester of prenatal care initiation	1st	178 (68%)
				2nd	74 (28%)
Black race		176 (69%)		3rd	9 (4%)
Hispanic ethnicity		90 (34%)	Gestational age at blood draw in weeks		32 (30, 36)
Education	< High school	82 (32%)	HIV diagnosis before current pregnancy		209 (80%)
	≥ High school	178 (68%)
Annual household income	≥ $10,000	99 (42%)	Trimester of ART initiation	Taking at LMP	95 (36%)
	<$10,000	139 (58%)		1st	55 (21%)
Marital status	Living with partner/spouse	131 (50%)		2nd	97 (37%)
				3rd	14 (5%)
	Single	130 (50%)		Not on ART	17 (6%)
Substance use, ever in pregnancy	Illicit drugs	26 (10%)	ART class at either LMP or 1st/2nd trimester initiation	NRTI use (no PI or NNRTI)	10 (4%)
	Alcohol	26 (10%)		NNRTI use (with or without PI)	19 (7%)
	Tobacco	51 (20%)		PI use (without NNRTI)	203 (77%)
				Other	15 (6%)
Maternal pre-pregnancy BMI (kg/m2)	≥ 35.0	44 (21%)	Earliest CD4 count (cells/mm3)	< 200	29 (11%)
	25.0-34.9	96 (45%)		200-499	99 (40%)
	<25.0	73 (34%)		≥ 500	121 (49%)
Daily dietary PUFA intake1 (grams/day)	n-6	14.47 (10.01, 20.95)	Earliest HIV VL (copies/mL) in the 1st/2nd trimester	≥ 10,000	57 (23%)
	n-3	1.61 (1.06, 2.34)		1001-10,000	61 (25%)
				400-1000	19 (8%)
				<400	106 (44%)

Abbreviations: IQR: interquartile range; LMP: last menstrual period; BMI: body mass index; PUFA: polyunsaturated fatty acid; ART: antiretroviral therapy; PI: protease inhibitor; VL: viral load

Missing data: maternal age at LMP (n=3), race (n=9), education (n=4), annual household income (n=26), marital status (n=3), illicit drug use (n=3), alcohol use (n=3), tobacco use (n=3), maternal pre-pregnancy BMI (n=51), trimester of prenatal care initiation (n=3), HIV diagnosis before current pregnancy (n=3), trimester of ART initiation (n=3), Zidovudine at first ART use (n=17), PI use (n=3), earliest CD4 count (n=15), earliest HIV VL (n=21).

¹Average daily dietary PUFA intake was assessed via 24-hour dietary recalls obtained in accordance with the multiple pass dietary recall method asking about all foods and beverages consumed on three separate occasions including two weekdays and one weekend day over a 2-week period.⁷⁴ All recall periods were separated by at least 1 day. Nutritional supplement use was collected at the end of the first 24-hour recall covering the previous 30 days.

APO prevalence was 5.3% for HDP, 17.0% for PTB, and 8.7% for SGA (Table 2). For each APO, the median n-6 and n-3 content were similar for those with versus without the outcome. However, the median n-6:n-3 ratio was higher in pregnant PWH with versus without HDP (14.6 vs. 13.0; p=0.04) and higher in those with versus without PTB (13.8 vs. 12.9; p=0.02). In contrast, n-6:n-3 ratio was lower in those with SGA versus non-SGA (11.7 vs. 13.2; p=0.03). In the adjusted models, for each 1% increase in n-6:n-3 ratio, we identified an estimated 25% (95% CI 8-43%) higher prevalence of HDP, 10% (95% CI 3-18%) higher prevalence of PTB, and 13% (95% CI 1-23%) lower prevalence of SGA (Table 2). In sensitivity analysis, our results for all APOs were unchanged when HIV VL was additionally included as a covariate in adjusted models.

Table 2.

Association of maternal PUFA content and ratios with adverse pregnancy outcomes in pregnant people with HIV

PUFA content (measured as a % of total fatty acid content) or ratio	Adverse pregnancy outcome		p-value6	Prevalence Ratio7 (95% CI)
	Absent Median (IQR)	Present Median (IQR)
Hypertensive Disorders of Pregnancy (HDP)1	N=250 (94.7%)	N=14 (5.3%)
n-6 4	37.3 (33.8, 40.3)	38.1 (35.7,40.7)	0.54	1.06 (0.96, 1.16)
n-3 5	2.8 (2.4, 3.4)	2.6 (2.1, 3.0)	0.13	0.45 (0.17, 1.21)
n-6:n-3	13.0 (10.7, 15.3)	14.6 (12.9, 19.0)	0.04	1.25 (1.08, 1.43)
Preterm Birth (PTB)2	N=219 (83.0%)	N=45 (17.0%)
n-6 4	37.4 (33.5, 40.3)	38.1 (34.7, 41.0)	0.19	1.05 (1.00, 1.11)
n-3 5	2.8 (2.5, 3.4)	2.8 (2.4, 3.1)	0.13	0.74 (0.52, 1.05)
n-6:n-3	12.9 (10.5, 15.2)	13.8 (12.2, 16.5)	0.02	1.10 (1.03, 1.18)
Small-for-Gestational Age (SGA)3	N=241 (91.3%)	N=23 (8.7%)
n-6 4	37.4 (34.1, 40.2)	38.1 (31.2, 42.1)	0.54	0.97 (0.89, 1.05)
n-3 5	2.8 (2.4, 3.4)	3.1 (2.4, 3.8)	0.19	1.34 (0.95, 1.90)
n-6:n-3	13.2 (11.0, 15.4)	11.7 (8.8, 13.8)	0.03	0.87 (0.77, 0.99)

CI: confidence interval; PUFA: polyunsaturated fatty acid; IQR: interquartile range; n-6: omega-6; n-3: omega-3

¹Hypertensive Disorders of Pregnancy included preeclampsia with or without severe features, eclampsia, and HELLP (hemolysis, elevated liver enzymes, low platelets) syndrome, as defined by the American College of Obstetricians and Gynecologists.²⁴

²PTB: Preterm birth was defined as delivery prior to 37 weeks’ gestation.

³Small-for-gestational age was defined as birthweight <10th percentile

⁴n-6 fatty acids included: linoleic, gamma linolenic, eicosadienoic, dihomo-gamma-linolenic, arachidonic, and adrenic acid (n=6).

⁵n-3 fatty acids included: alpha-linolenic, docosapentaenoic, eicosapentaenoic, and docosahexaenoic acid (n=4).

⁶Median (IQR) PUFA content and n-6:n-3 ratios among cases of hypertensive disorders of pregnancy, preterm birth, and small-for-gestational age, were compared to controls that did not have an adverse pregnancy outcome, using the Wilcoxon rank-sum test.

⁷Models of prevalence ratios were adjusted for maternal age at LMP and annual household income

The distribution of most child Z-score outcomes at each age were similar between those with maternal plasma n-6:n-3 ratio >25^th versus ≤25^th percentile (Table 3). Neither WTZ (Rho=0.018) or LTZ (Rho=0.003) were correlated with n-6:n-3 ratio at birth. Child Z-scores across age for WTZ, LNZ/HTZ, and TSFZ were similar by maternal n-6:n-3 ratio >25^th versus ≤25^th percentile (Figure 2). Although HCZ from the 1- to 2-year study visit was similar by maternal plasma n-6:n-3 ratio groups, those with maternal plasma n-6:n-3 ratio >25^th percentile had slightly lower HCZ scores on average across age. Finally, the estimated mean difference in child Z-scores for WTZ, LNZ, TSFZ, and HCZ across age did not differ by maternal plasma n-6:n-3 ratios >25^th versus ≤25^th percentile (Table 3). In sensitivity analysis, our results for all childhood anthropometric measures were unchanged when HIV VL was additionally included as a covariate in adjusted models.

Table 3.

Anthropometric and adiposity Z-scores by n-6:n-3 ratio >25^th vs. ≤25^th percentile

Age at measurement		n-6:n-3 ratio		p-value*	Pearson correlation coefficient (r)	Estimated mean difference in Z-score by n-6:n-3 ratio** >25th vs. ≤25th percentile	p-value
		≤25th percentile Median (IQR)	>25th percentile Median (IQR)
Birth		(n=66)	(n=198)
	Weight	−0.37 (−0.96, 0.50)	−0.18 (−0.74, 0.25)	0.54	0.018	-	-
	Length/Height	−0.07 (−0.75, 0.51)	−0.10 (−0.67, 0.40)	0.92	0.003	-	-
	Triceps	-	-	-	-	-	-
	Head circumference	-	-	-	-	-	-
48 weeks (1 year)		(n=49)	(n=161)
	Weight	−0.06 (−0.57, 0.44)	−0.12 (−0.87, 0.63)	0.66	−0.047	-	-
	Length/Height	0.21 (−0.37, 1.14)	0.18 (−0.52, 0.70)	0.30	−0.097	-	-
	Triceps	-	-	-	-	-	-
	Head circumference	0.63 (−0.19, 1.52)	0.32 (−0.55, 1.18)	0.08	−0.18	−0.18 (−0.74, 0.39)	0.54
96 weeks (2 years)		(n=48)	(n=155)
	Weight	0.40 (−0.20, 1.01)	0.26 (−0.56, 1.14)	0.65	−0.063	0.10 (−0.17, 0.37)	0.47
	Length/Height	0.54 (−0.35, 1.08)	0.06 (−0.45, 0.73)	0.06	−0.095	0.00 (−0.24, 0.25)	0.99
	Triceps	0.12 (−0.34, 0.88)	0.04 (−0.63, 0.85)	0.33	−0.025	−0.41 (−0.89, 0.06)	0.087
	Head circumference	0.57 (0.20, 1.15)	0.41 (−0.39, 1.07)	0.11	−0.13	−0.17 (−0.75, 0.41)	0.57
144 weeks (3 years)		(n=47)	(n=152)
	Weight	0.58 (0.00, 1.39)	0.47 (−0.21, 1.46)	0.53	−0.037	-	-
	Length/Height	0.73 (−0.41, 1.36)	0.35 (−0.22, 0.94)	0.078	−0.095	-	-
	Triceps	0.26 (−0.58, 0.95)	0.22 (−0.34, 1.03)	0.92	0.080	-	-
	Head circumference	-	-	-	-	-	-
192 weeks (4 years)		(n=48)	(n=145)
	Weight	0.60 (−0.19, 1.32)	0.52 (−0.27, 1.40)	0.77	−0.014	-	-
	Length/Height	1.07 (−0.18, 1.51)	0.59 (−0.20, 1.28)	0.43	0.015	-	-
	Triceps	0.37 (−0.33, 1.28)	0.38 (−0.35, 1.24)	0.85	−0.012	-	-
	Head circumference	-	-	-	-	-	-
240 weeks (5 years)		(n=32)	(n=92)
	Weight	0.89 (0.32, 1.79)	0.64 (−0.11, 1.46)	0.21	−0.04	0.16 (−0.12, 0.43)	0.26
	Length/Height	1.22 (0.27, 1.70)	0.62 (−0.11, 1.45)	0.11	0.001	0.05 (−0.20, 0.30)	0.69
	Triceps	0.72 (0.05, 1.66)	0.74 (−0.17, 1.52)	0.95	0.031	−0.39 (−0.88, 0.10)	0.12
	Head circumference	-	-	-	-	-	-

IQR—interquartile range

^*Difference in Z-score distributions were assessed using the Wilcoxon rank-sum test.

^**Regression models include n-6:n-3 ratios >25^th versus ≤25^th percentile, child age, n-6:n-3 ratios >25^th vs. ≤25^th percentile*child age (interaction term), maternal age at LMP and household income.

Figure 2.

Distribution of child anthropometric Z-scores across age by maternal plasma n-6:n-3 ratio >25^th vs. ≤25^th percentile

Comment

Principal findings

In this US cohort of pregnant PWH, higher n-6:n-3 ratios were positively associated with HDP and PTB, whereas this pro-inflammatory eicosanoid ratio was inversely associated with SGA. Higher n-6:n-3 ratios in pregnant PWH were not associated with growth and adiposity trajectories in their offspring from birth through 5 years of age.

Results in context

HDP and PTB

In our cohort of pregnant PWH in the US, we found higher n-6:n-3 ratios in maternal plasma were associated with higher prevalence ratios for HDP and PTB. Considering lack of data on gestational hypertension diagnoses, HDP prevalence is likely underestimated in this cohort and the strength of the association identified between n-6:n-3 PUFA ratios and HDP may also be underestimated. However, this association was not present when n-6 and n-3 PUFA content was assessed separately. In pregnancies without HIV, international data similarly suggest higher n-6:n-3 ratios raise the risk of HDP, but higher n-3 PUFA content may also lower the risk.³³ A US-based prospective cohort study in the general obstetric population identified higher first-trimester dietary intake of n-3 PUFAs was associated with a lower risk of HDP, but n-6 PUFA intake was not associated with HDP risk.⁷ Similarly, a Cochrane review suggested higher n-3 PUFA intake possibly reduced rates of preeclampsia and PTB,⁶ and data from the Danish National Birth Cohort identified low n-3 plasma content to be a risk factor for PTB, excluding preeclampsia cases. Our study complements this literature by expanding the investigation of PUFAs among pregnant PWH.

SGA

Contrary to our hypothesis, we found higher n-6:n-3 ratios to be associated with a lower prevalence of SGA offspring of pregnant PWH. Neither n-6 nor n-3 PUFA content alone were associated with offspring SGA. In contrast, in a US cohort of pregnant people without HIV, the content of several individual n-6 PUFAs had an inverse association with fetal growth, and higher n-3 PUFA content was positively associated with fetal growth trajectories.³⁴ Similarly, a Danish study identified higher SGA risk associated with higher content of most n-6 PUFAs and lower n-3 maternal plasma content in pregnancies unaffected by HIV.³⁵ Findings from the Generation R cohort suggested higher n-3 content relative to n-6 content in maternal plasma was associated with higher fetal growth velocity and higher birthweight.³⁶ Differences in PWH, with alterations in PUFA metabolism related to viremia, may partially explain why our findings are not congruent with those in the general pregnant population.¹⁵ Markedly few studies have evaluated the association of n-6:n-3 PUFA ratios in pregnant PWH with fetal or neonatal growth. One study from Malawi identified neonates of pregnant PWH had lower LNZ compared to those born to pregnant people unaffected by HIV, but identified that this association with impaired offspring growth was no longer observed when pregnant PWH received lipid-based supplementation with the n-6 PUFA linoleic acid and n-3 PUFA alpha-linolenic acid.³⁷

Child anthropometrics

We did not detect a significant association between anthropometric measures in HEU offspring with maternal plasma PUFA ratios. In a Chinese cohort of pregnant people without HIV, higher maternal erythrocyte arachidonic acid (AA, n-6) was inversely associated with offspring WTZ, but not LNZ, within the first two years of life, though maternal erythrocyte n-6:n-3 ratios were not associated with offspring WTZ or LNZ.³⁸ With respect to adiposity, the Generation R cohort found no association between maternal n-6:n-3 PUFA ratios and neonatal subcutaneous fat mass within the first two years of life, similar to our findings.³⁹ Similar to the literature on fetal growth and infant birthweight, the literature on early childhood growth anthropometrics associated with in utero PUFA exposure is limited, heterogeneous, and largely absent in HEU children. A Ugandan cohort that included HEU children ages 6-10 years identified higher serum n-6 PUFA content, but not n-3 PUFA content, was positively associated with linear growth, but correlation with maternal PUFA content was not evaluated.^40,41 Understanding if an optimal antenatal n-6:n-3 PUFA ratio exists to ensure positive growth outcomes following in utero HIV exposure is an important area of investigation.¹⁴

Clinical implications

HDP and PTB

Several potential mechanisms could underlie the relationships identified herein. HDP manifests due to abnormal placentation, as evidenced by impaired vascular remodeling, and the subsequent maternal systemic inflammatory response.⁴² The n-6 PUFA AA and its prostaglandin derivatives mediate vascular permeability at the maternal-fetal interface and the process of cytotrophoblast invasion.⁴² However, an imbalance in the n-6:n-3 ratio, with excessive action of these n-6 prostaglandin derivatives and/or reduced n-3 PUFAs, promotes excessive inflammatory and oxidative stress processes that may disrupt normal embryo implantation and placentation, predisposing to development of HDP.⁴² HIV infection promotes a pro-inflammatory PUFA milieu in pregnancy with high n-6:n-3 ratios.¹⁵ HIV viral particles specifically activate toll-like receptors that induce rapid activation and upregulation of n-6 PUFA metabolism and their prostaglandin derivatives, and lower n-3 PUFA content has also been observed in pregnant PWH who have active viral replication.^15,43,44 Low levels of n-3 PUFAs relative to n-6 PUFAs may further perpetuate placental secretion of soluble fms-like tyrosine kinase-1 (sFlt1) into the maternal bloodstream that contributes to endothelial dysfunction and the clinical manifestations of HDP.^42,45 HIV viral accessory and matrix proteins further contribute to the altered placental angiogenesis, endothelial dysfunction, oxidative stress, and inflammation at the maternal-fetal-placental interface that predispose to HDP and PTB.^10,46,47

Oxidative stress can further induce DNA damage, telomere shortening, and telomere-dependent senescence of fetal membranes, both contributing to iatrogenic PTB due to HDP and spontaneous PTB. In the case of spontaneous PTB, inflammatory activation following telomere shortening leads to rupture of membranes and/or parturition.^48-51 Excessive n-6 PUFA content has been positively associated with telomere shortening, whereas higher n-3 PUFA content has been associated with reduced rates of telomere shortening, in other disease processes.^52-54 Placental telomere shortening in pregnant PWH, as compared to those without HIV, has been observed and may be a potential mechanism linking maternal HIV, inflammation, and PTB.⁵⁵ It is plausible that imbalances of n-6:n-3 PUFA ratios may contribute to placental chorioamniotic membrane telomere shortening and induce early parturition in pregnant PWH. Therefore, upregulation of n-6:n-3 ratios in pregnant PWH may amplify HDP and PTB risk via the effects of a pro-inflammatory PUFA milieu on placental vascular and telomere function.

SGA

It remains unclear why higher n-6:n-3 ratios, but not n-6 or n-3 PUFA content, were associated with a lower prevalence of SGA in pregnant PWH, divergent from that reported in pregnancies without HIV. This finding suggests HIV infection may inherently alter how the content of n-6 PUFAs relative to n-3 PUFAs mitigate abnormal fetal growth compared to in pregnancies without HIV. Prior work has identified pregnant PWH have higher n-6 content, including AA, than HIV-seronegative pregnant people, with higher n-6:n-3 ratios associated with viremia.^15,56 AA and its eicosanoid derivatives regulate osteoclast differentiation and bone growth and development.^57-59 AA also positively correlates with plasma concentrations of insulin-like growth factor I, which promotes hypertrophic cell size and growth velocity in children.^57-59 It is possible the higher transplacental passage of AA and other n-6 PUFA precursors of AA (i.e. linoleic acid) relative to n-3 PUFAs in pregnant PWH may be protective against SGA by promoting linear fetal growth, though this requires further study. Further investigation is required to better understand the optimal relative balance of n-6 to n-3 PUFAs to promote fetal growth, both in pregnancies with and without HIV.

Research implications

Further investigation is required to elucidate the mechanisms underlying our findings with respect to PUFA ratios and SGA in pregnant PWH, in contrast to pregnancies without HIV, and the observed lack of significant association with childhood growth outcomes. Evaluation of placental histopathology and size, as well as inflammatory cytokines correlated with PUFA ratios, may provide helpful mechanistic insight into the associations identified between PUFA ratios and APOs. Future studies should also evaluate the association of individual n-6 and n-3 PUFAs with the obstetric and child anthropometric outcomes assessed, and how the trajectory of PUFA signatures longitudinally from early to late pregnancy may affect obstetric and postnatal outcomes. Evaluation of individual PUFAs may offer insight into the following: 1) the differential immune effects of individual n-6 and n-3 PUFAs with downstream impact on placental function (e.g. two AA-derived oxylipin subtypes are considered pro-inflammatory [prostaglandins, thromboxanes] while a third subtype [lipoxins] are considered anti-inflammatory^60,61) and 2) identification of a specific PUFA(s) associated with APOs and/or childhood growth may better inform dietary supplementation or other intervention. Evaluation of PUFA trajectories longitudinally may help determine if there is a critical window for intervention among pregnant PWH to reduce APOs associated with metabolic alterations in this vulnerable population.

Strengths and limitations

Our study fills a critical knowledge gap in improving our understanding of how imbalances in n-6:n-3 ratios may alter obstetric outcomes associated with PUFA ratios observed in pregnant PWH, a population at high risk of both nutritional deficiencies and APOs. The longitudinal follow up over five years represents an important strength of the study in demonstrating long-term implications, or lack thereof, of imbalances in these PUFA profiles during pregnancy for postnatal childhood growth and adiposity in this HIV-exposed population.

There are some limitations to consider. Our study generates rich hypotheses, but we cannot infer causality of PUFA ratios with obstetric and early childhood outcomes in PWH and their HEU children. We do not have dietary data from early in pregnancy to evaluate n-6 and n-3 dietary intake as a potential intervention to improve pregnancy outcomes, which otherwise could be analyzed using observational data to emulate a clinical trial. There may also be unmeasured confounding from other clinical factors for which we did not have data (e.g., hepatitis B or C co-infection, diabetes) that may influence our findings, but we did account for several relevant confounders including BMI as described above.

Additionally, gestational age at diagnosis of HDP was not available in this dataset; while it is possible some HDP diagnoses were made prior to plasma collection for measurement of PUFAs, all samples were collected prior to birth and preceding the postnatal diagnosis of SGA. Based on the gestational age at enrollment of study participants, we may not have captured all preterm births occurring; however, as we suspect this to be a low number, this would be unlikely to alter the preterm birth incidence in this population to a significant degree or impact our overall findings. In our analysis, we observed a PTB rate of 17%, similar to what is published in other studies in this population.^62-64 Although we did not measure PUFA ratios longitudinally across gestation, prior longitudinal studies in pregnancy have demonstrated relative stability of PUFA concentrations and ratios from early to late gestation.^15,65,66

Lack of a comparison group without HIV limits our ability to evaluate whether PUFA ratios differentially influence obstetric outcomes and early childhood growth in pregnant PWH and their offspring compared to those without HIV. Considering the study period, this analysis also lacks inclusion of pregnant PWH treated with contemporary ART regimens such as integrase strand transfer inhibitors (INSTIs), and the majority (77%) of this cohort was exposed to PI-based therapy. Now preferred agents for treatment of pregnant PWH per US perinatal guidelines, INSTIs have been associated with higher weight gain in pregnancy – a risk marker for APOs and childhood obesity – compared to those receiving PI-based therapy, which may influence PUFA content.^67-69 Ueland et al. identified n-6 PUFA content was similar in non-pregnant adults exposed to PI- and INSTI-based ART. However, future studies evaluating the role of contemporary INSTI-based ART on PUFA ratios in pregnant individuals, and downstream effects on APOs and child growth, are warranted.

Conclusions

In conclusion, we identified that higher n-6:n-3 ratios in pregnant PWH were positively associated with HDP and PTB among pregnant PWH. However, this pro-inflammatory eicosanoid ratio was inversely associated with SGA and not associated with early childhood growth and adiposity trajectories. Larger contemporary, longitudinal studies of individual n-6 and n-3 PUFAs, and their ratios, in pregnant PWH, with comparison to those without HIV, are needed to better understand the role of PUFA signatures within the context of HIV infection in the development of adverse pregnancy and early childhood growth outcomes.

Supplementary Material

1

Supplemental Digital Content Figure. Relationship of n-6 and n-3 PUFAs with inflammatory eicosanoids

PUFAs—polyunsaturated fatty acids

Dotted arrows denote multiple steps in the metabolic pathway.⁷⁵

Ratios denote number of carbon atoms:number of double bonds, and the series number refers to the number of double bonds in each eicosanoid.⁷⁶

Supplemental Digital Content Table. Sensitivity analyses incorporating VL into regression models of obstetric and childhood growth outcomes

Appendix. Rationale for selected statistical methods

Justification for exclusion of HIV VL and CD4 count from adjusted models of obstetric outcomes

In the main analysis for obstetric outcomes, we adjusted for maternal age at LMP and annual household income as potential confounders, but not HIV VL and CD4 count; however, HIV VL was included in regression models in our sensitivity analysis. We previously demonstrated greater HIV viremia is associated with higher n-6:n-3 ratios in pregnant PWH. HIV VL was not included as a covariate in the main model as HIV viremia can be considered on the causal pathway for the obstetric outcomes assessed.¹⁵ However, it is possible HIV viremia alters lipid profiles, which may be associated with the outcomes evaluated. We did not adjust for CD4 count or ART exposure in the main model or sensitivity analysis, as we previously demonstrated these were not associated with n-6:n-3 ratios in this population.¹⁵

Justification for child anthropometric z-score discriminatory threshold at the 25^th percentile

We evaluated child anthropometric Z-score across age by maternal plasma n-6:n-3 ratio >25^th vs. ≤25^th percentile. We chose the 25^th percentile as a discriminatory threshold between a more pro-inflammatory (>25^th percentile) versus anti-inflammatory (≤25^th percentile) PUFA signature, as standardized cut-off values and reference ranges for normal and abnormal n-6:n-3 ratios have not been standardized in pregnancy due to variation based on the specific laboratory assay used and heterogeneous units of measure reported across studies. Although a 6:1 n-6:n-3 ratio is recommended for optimal health, prior work suggests the average (± standard deviation) n-6:n-3 ratio in pregnant people without HIV may range from as low as 6.1 (±1.7) in a cohort from the Netherlands and as high as 29.5 (±11.7) in a cohort from Spain.^9,70-73,65 In this cohort of pregnant PWH, we previously reported the median n-6:n-3 ratio to be 12.99 (interquartile range [IQR] 1.75, 23.15).¹⁵ Use of a discriminatory cut-off, in lieu of continuous measures, is clinically meaningful as it allows for direct clinical application broadly despite heterogeneous assays or units of measure.

Data availability:

Data available upon request to the PHACS Network.