Perinatal inflammation, fetal growth restriction, and long-term neurodevelopmental impairment in Bangladesh

Abstract

BACKGROUND:

There are limited data on the impact of perinatal inflammation on child neurodevelopment in low-middle income countries and among growth-restricted infants.

METHODS:

Population-based, prospective birth cohort study of 288 infants from July 2016–March 2017 in Sylhet, Bangladesh. Umbilical cord blood was analyzed for interleukin(IL)-1α, IL-1β, IL-6, IL-8, and C-reactive protein(CRP). Child neurodevelopment was assessed at 24 months with Bayley-III Scales of Infant Development. We determined associations between cord blood inflammation and neurodevelopmental outcomes, controlling for potential confounders.

RESULTS:

248/288 (86%) live born infants were followed until 24 months, among whom 8.9% were preterm and 45.0% small-for-gestational-age(SGA) at birth. Among all infants, elevated concentrations (>75%) of CRP and IL-6 at birth were associated with increased odds of fine motor delay at 24 months; elevated CRP was also associated with lower receptive communication z-scores. Among SGA infants, elevated IL-1α was associated with cognitive delay, IL-8 with language delay, CRP with lower receptive communication z-scores, and IL-1β with lower expressive communication and motor z-scores.

CONCLUSIONS:

In rural Bangladesh, perinatal inflammation was associated with impaired neurodevelopment at 24 months. The associations were strongest among SGA infants and noted across several biomarkers and domains, supporting the neurobiological role of inflammation in adverse fetal development, particularly in the setting of fetal growth restriction.

BACKGROUND

The 2030 United Nations Sustainable Development Goals have brought increasing attention to the prevention of childhood disability in low- and middle-income countries (LMICs).¹ Globally, an estimated 250 million children under the age of 5 are at risk of not achieving their full developmental potential, with the majority (>90%) in LMICs. Perinatal conditions, including preterm birth, fetal growth restriction, and intrapartum-related hypoxic events, increase the risk of long-term neurodevelop-mental impairment.^2,3 Recent attention has focused on small, vulnerable newborns (preterm, small for gestational age [SGA] or low birthweight [LBW]), who comprise 25% of births globally and carry higher risk of short- and long-term adverse outcomes and loss of human capital.⁴ In the 2019 Global Burden of Disease Study, neonatal conditions were the leading cause of health loss globally, accounting for 7.3% of disability adjusted life years (DALYS).⁵

Inflammation is a major risk factor for adverse neurodevelopmental outcomes. The in-utero and neonatal periods are critical for early human brain development, and there is limited research on the effects of inflammation during this period on neurodevelopment in LMICs, where disease burden is high and risk factors for inflammation are common. Maternal and neonatal infections are prevalent in LMICs, where prenatal care coverage is low and 60 million births (> 50%) occur at home without a skilled birth attendant.^6–8 Up to 60% of mothers may experience a urinary or reproductive tract infection at least once during pregnancy,^6,7 and in LMICs, screening and treatment of these infections is inadequate. Limited access to skilled obstetric care increases risk for intrapartum-related hypoxic events and is another major risk factor for perinatal inflammation.⁹ In South Asia, 21% of women of reproductive age are underweight (body mass index [BMI] <18.5 kg/m²),¹⁰ 11% are stunted (height < 145 cm),¹¹ and micro-nutrient deficiencies are common.^10–12 Undernutrition has been linked to chronic inflammation, with complex interactions existing between nutrition and inflammation.¹³ Finally, adversity and psychosocial stress are important and prevalent exposures that place pregnant women and their fetuses at risk for exposure to inflammation in utero.¹⁴

In high-income settings, the adverse effects of inflammation on early brain development have been established in certain populations. Infants exposed to hypoxic birth events, maternal chorioamnionitis, or extreme prematurity carry increased risk of early life inflammation, which may increase risk of brain injury.^15–18 In the fetal and early neonatal period, the rapidly developing brain is highly susceptible to the effects of inflammation and oxidative stress. Peripheral inflammation may activate microglia, leading to the central release of pro-inflammatory cytokines (including interleukin [IL]-1β, IL-6, tumor-necrosis factor α [TNF- α]) and reactive oxygen species, inducing the formation of neurotoxic reactive astrocytes.^19–23 This cascade of events leads to impaired maturation and cell death of developing oligodendrocytes, hypomyelination,^22,24 and focal and diffuse neuronal and axonal white matter injury (WMI).^25,26 In preterm infants, the pre-myelinating cell, or developing oligodendrocyte, is particularly vulnerable,²² and the role of inflammation in WMI is well established.^24,27–29 In the Extremely Low Gestational Age Newborn (ELGAN) study of infants born at < 28 weeks, perinatal systemic inflammation (elevated IL-6, IL-8, IL-1β, C-reactive protein [CRP]) was associated with WMI, microcephaly, impaired cognitive function, cerebral palsy, and attention deficits at 2 years of age.^30–34 Chorioamnionitis and resultant inflammation has been implicated in the pathogenesis of cerebral palsy in term infants.¹⁸ In infants with term neonatal encephalopathy, neonatal cytokine elevations have also been associated with WMI, cerebral palsy, and abnormal neurodevelopmental outcomes.^15,16,35 However, while the role of inflammation has been established in these groups of high-risk infants, there are limited data on the effects of inflammation in the general neonatal population and in other groups of small, vulnerable infants, such as SGA and moderate-late preterm infants (32 to < 37 weeks), who comprise up to one-quarter of births in LMICs.^4,36

Reducing exposure to inflammation is a critical intervention target to optimize early child development in LMICs. Cord blood inflammation at birth reflects exposure to inflammation in utero in the late prenatal period or at birth. The objective of this study was to examine the associations between biomarkers of cord blood inflammation at birth and neurodevelopmental outcomes at 24 months in a population-based, rural birth cohort in Bangladesh. Our secondary objective was to examine associations among small, vulnerable newborns (preterm and SGA infants).

METHODS

Study site

The Projahnmo research site is in the Kanaighat and Zakiganj sub-districts (population: ~500,000) of Sylhet, in northeast Bangladesh, about 350 km from Dhaka. Projahnmo has conducted routine demographic surveillance for all pregnancies and births in the study areas since 2001. The study population is rural, agrarian, low-income (average daily wage <$2/day), and with high burden of disease and mortality rates (stillbirth rate 30/1000 births, neonatal mortality rate 37/1000 live births).³⁷

Participants

As part of an ongoing pregnancy-birth cohort established at the site,³⁸ pregnant women were identified and recruited by community health workers (CHWs) during routine monthly community-based pregnancy surveillance and enrolled at <20 weeks gestation. Ultrasonography was performed to confirm the date of all pregnancies. For this sub-study, we prospectively recruited and enrolled 299 consecutive pregnancies, resulting in 288 live births, within the pregnancy-birth cohort between July 2016 and March 2017.

Data collection

Research CHWs followed pregnant women through pregnancy and delivery, until 24 months postpartum. Data were collected on household sociodemographic, environmental, and pregnancy risks, medical and obstetric histories, pregnancy outcomes, and maternal-infant anthropo-metrics, morbidity, and mortality. Women were incentivized to deliver at a health facility, and research staff attended deliveries to collect biospecimens. Infant weight was measured using a digital infant scale (model: TANITA BD 585 Pediatric Scale; precision: 10 g). SGA was defined as < 10% birthweight for gestational age and sex based on the Intergrowth-21st newborn size standard.³⁹

Umbilical cord blood collection and inflammation protein analysis

Research staff collected drops of umbilical cord blood (0.25–0.20 mL) on Whatman filter paper (FTA Abs) that were dried at room temperature and stored in sealed ziplock bags with desiccant at −80 °C until analysis. Samples were processed and analyzed at the Fichorova Laboratory of Genital Tract Biology, Brigham and Women’s Hospital (Boston, MA), that previously developed and validated laboratory methods for the Extremely Low Gestational Age Newborn (ELGAN) study.⁴⁰ Proteins were eluted using standard validated procedures and total protein determined by Pierce bicinchoninic acid assay. We selected specific inflammation biomarkers based on prior analytic validation and their association with neurodevelopmental outcomes in the prior literature and our experience from the ELGAN study.^{31,33,41–43} Electrochemiluminescence multiplex assays (Meso Scale Discovery Sector Imager S600 [MSD, Gaithersburg, Maryland]) were used to determine cord blood levels of IL-1α, IL-1β, IL-6, IL-8, and CRP. The platform has been analytically and clinically validated compared to traditional enzyme-linked immunosorbent assay (ELISA).^{17,40,44–49} We used protein concentrations to calculate relative luminescent units based on interpolation from log calibrator curves. Reliability was assessed via reanalysis of a quality control split sample on each assay plate (inter-assay variation: 5.54% for IL-1α, 5.96% for IL-1β, 8.44% for IL-6, 10.21% for IL-8 and 5.10% for CRP). Each analyte was normalized to mg total protein. These methods have been used and reported in the ELGAN studies.^{17,30–33,40,41} We classified inflammation as having a biomarker elevation of > 75th percentile (versus the ≤ 75th percentile) of each cord blood inflammatory biomarker distribution. Prior studies from members of our group in the ELGAN studies have used similar laboratory methods and categorization to demonstrate the value of this approach^32,50 in the prediction of abnormal neurodevelopmental outcomes.

Neurodevelopmental assessment

Cognitive, language, and motor function were assessed using three sub-scales of the Bayley Scales of Infant/Toddler Development, Third Edition (Bayley-III) at 24 months (cognitive, language [receptive and expressive communication], and motor [gross and fine motor]).⁵¹ Our child development research team adapted and validated this version of the Bayley-III, which has been used in studies in rural and urban settings and is sensitive to detect group differences and intervention effects.^52–55 As the test is mostly non-verbal, the cultural adaptation mainly focused on replacement of specific images while maintaining the original question intent. Composite scores were calculated from raw scores using age-standardized conversion tables based on Western population norms. Field testing showed positive correlation between composite scores and child nutritional status (p < 0.05 for all three sub-scales) and parental education (p < 0.05 for language and motor component). Short term test-retest reliability (within ~7 days gap) was also high (r > 0.80 for all subscales).^56–58 Two Bayley-III administrators were master’s degree level staff and received rigorous training (3 weeks) by a senior psychologist (supervisor). Data collection was initiated after attaining > 95% agreement between testers and trainer (r = 0.95, p ≤ 0.001 for all three components). Five-to-ten percent of all tests were directly observed by the supervisor throughout the study period for ongoing reliability, along with quarterly refresher training. Based upon our prior experience and studies^59,60 in Bangladesh, we standardized raw Bayley-III scales to generate z-scores (mean = 0, standard deviation = 1). Developmental delay was defined as z-scores <−1.

Statistical methods

Descriptive statistics of key demographic and clinical characteristics, exposure, and outcome variables were reported as mean +/− SD (or median and interquartile range [IQR]) and frequency (n) and percent, for continuous and categorical variables, respectively. The association between inflammatory biomarker elevation and 24-month Bayley performance was modeled using linear and logistic regression, as appropriate. For models of binary outcome, we used Firth’s penalized maximum likelihood logistic regression⁶¹ to minimize bias due to small sample size. Analyses were performed for the total sample, as well as stratified by preterm and SGA status. The conceptual framework for our analysis is shown in Fig. 1. Model 1 examined the unadjusted association between cord blood inflammation and neurodevelopment outcomes; Model 2 included adjustment for a priori-determined confounders (maternal parity, gestational age, sex) as well as postnatal age of Bayley-III assessment. In Model 3, we included additional proximal factors preceding the exposure: socioeconomic status (maternal education), nutrition (maternal BMI), and maternal depression (Patient Health Questionnaire [PHQ-9] score). Given the importance of postnatal factors in shaping child development, we tested an additional model (Model 4) including home environment (Home Observation for Measurement of the Environment [HOME] scores) to control for potential postnatal influences, even though these would not be considered confounders. Results were not adjusted for multiple comparisons as analyses were exploratory. Analyses were performed using SAS 9.4 statistical software (SAS Institute Inc., Cary, NC)

Fig. 1. Conceptual diagram.

Diagram showing the four models tested in our analysis of the associations between cord blood inflammation (exposure) and neurodevelopment at age 24 months (outcome). BMI Body mass index, CRP C-reactive protein, IL Interleukin, SES Socioeconomic status, SGA small for gestational age.

Ethics approvals

This study was approved by the Ethics Review Committee of Johns Hopkins University (IRB No. 00004508), International Centre for Diarrhoeal Disease Research Bangladesh (ICDDR,R) in Bangladesh (PR 12073) and Mass General Brigham (Protocol 2014P001741).

RESULTS

299 pregnancies were enrolled and resulted in 288 liveborn infants, among whom 248 children (86%) were followed up until 24 months of age. Among the 40 infants who were not assessed at 24 months, 14 (35%) infants died and the remaining were lost to follow up. Infants not assessed at 24 months were more likely to have been born preterm (11 [27.5%] v. 22 [8.9%], p < 0.01) and have lower median birthweight (median [IQR]: 2.66 kg [2.15–2.78] v. 2.70 kg [2.46, 3.00], p < 0.05) than infants with 24-month assessments (Supplemental Table 1).

The baseline demographics of the study population are shown in Table 1. Among liveborn infants who completed the 24 month follow up, 8.9% were preterm, 26.6% low birthweight, and 45.0% SGA. Among preterm infants (n = 22), the majority (n = 21; 96%) were moderate-late preterm (gestational age [GA] 34 to < 37 weeks, n = 17; 32 to < 34 weeks, n = 4). Among SGA infants, 104 (96.3%) were full term (term-SGA) and 4 (3.7%) were preterm SGA, and among the 132 non-SGA infants, the majority (n = 128, 97%) were appropriate for gestational age (AGA), while 4 were large for gestational age. The mean maternal age was 23.8 years (standard deviation [SD] 4.5), 29.4% were primiparous with the index child, and 31.3% had BMI < 18.5 kg/m² at enrollment. Women had completed an average 6.1 (SD 2.9) years of schooling, and most (98%) did not work outside the home. The average number of family members living in the household was 6.4 (SD 3.4). Most deliveries (93.2%) occurred in the hospital or a health facility. Signs of potential intrapartum infection (either self-reported or reported by the birth attendant) were present in 18.1% (n = 45) of deliveries, indicated by having fever during delivery (n = 5) and/or receipt of antibiotics for presumed infection (n = 41).

Table 1.

Demographic and clinical characteristics (n = 248).

Variable	Statistic
Maternal age, years
Mean ± SD	23.8 ± 4.5
Parity, n (%)
0	73 (29.4)
1	73 (29.4)
2	49 (19.8)
3	28 (11.3)
4+	25 (10.1)
Maternal educational attainmenta (years)
Mean ± SD	6.1 ± 2.9
Maternal occupation, n (%)
Government job	1 (0.4)
Private job	4 (1.6)
Does not work outside home	243 (98.0)
Maternal BMI at enrollmentb (kg/m2)
< 18.5, n (%)	77 (31.3)
18.5– < 25, n (%)	151 (61.4)
≥ 25, n (%)	18 (7.3)
Maternal PHQ-9 Score at First Postnatal Visit a
Mean ± SD	10.1 ± 4.8
Paternal occupation, n (%)
Unemployed	13 (5.2)
Daily wage earner	114 (46.0)
Farming	37 (14.9)
Private service/self-employed	73 (29.5)
Other	11 (4.4)
Household size (# of people)
Mean ± SD	6.4 ± 3.4
Place of Birth, n (%)
Hospital/facility	222 (93.2)
Home	16 (6.7)
Intrapartum Fever/Infection, n (%)	45 (18.1)
Infant sex, n (%)
Male	112 (45.2)
Female	136 (54.8)
Gestational age at birth (weeks)
Mean ± SD	39.1 ± 1.7
Min–Max	31.7–43.1
Median (IQR)	39.4 (38.2–40.1)
Birth weightc (kg)
Mean ± SD	2.71 ± 0.43
Birthweight Z-score d
Mean ± SD	−1.2 ± 1.1
SGAc, n (%)	108 (45.0)
Preterm, n (%)	22 (8.9)
Total HOME score, mean ± SD	43.8 ± 1.9

BMI Body mass index, IQR Inter-quartile range, HOME Home Observation for Measurement of the Environment, PHQ-9 Patient Health Questionnaire, SD Standard deviation, SGA Small for gestational age.

^aMissing, n = 1.

^bMissing, n = 2.

^cMissing, n = 7.

^dMissing, n = 8.

Cord blood cytokine distribution

We have previously published upon the cytokine distributions of CRP, IL-1α, IL-1β, IL-6 and IL-8 in this cohort, as well as maternal and newborn risk factors for cord blood inflammation.⁶² The distributions of cytokines in the analyzed cohort were right skewed and are shown in Supplemental Fig. 1a–e. There were no significant differences in median cytokine values by preterm or SGA status at birth (Supplemental Table 2a, b; Supplemental Fig. 2a–e).

Perinatal inflammation and 24-month neurodevelopmental outcomes

The mean composite Bayley-III scores at 24 months for cognitive, language, and motor domains were 87.1 (SD 6.9), 88.8 (SD 8.8), and 90.4 (SD 6.9), respectively. Histograms of the distribution of Bayley-III raw z-scores are shown in Supplemental Fig. 4a–c. In this cohort, Bayley-III raw z-scores at 24 months did not significantly differ among infants who were born preterm versus term or SGA versus non-SGA (Table 2).

Table 2.

Bayley-III Scores at 24 months by vulnerable newborn type (SGA and preterm).

	SGA		Non-SGA		Mean difference (SGA-non SGA)a		Preterm		Term		Mean Difference (Preterm-Term)a
	Mean	SD	Mean	SD	Difference	95% CI	Mean	SD	Mean	SD	Difference	95% CI
Total Cognitive Raw Z-score	0.026	1.09	−0.020	0.93	0.046	(−0.21, 0.31)	0.009	1.24	−0.001	0.98	0.010	(−0.43, 0.45)
Receptive Communication Raw z-score	0.034	1.04	−0.016	0.98	0.050	(−0.21, 0.31)	0.11	1.38	−0.011	0.96	0.12	(−0.32, 0.56)
Expressive Communication Raw z-score	0.11	1.02	−0.077	0.98	0.19	(−0.070, 0.45)	−0.20	1.10	0.020	0.99	−0.220	(−0.66, 0.22)
Total Language Raw z-score	0.081	1.02	−0.056	0.98	0.14	(−0.12, 0.40)	−0.077	1.24	0.008	0.98	−0.084	(−0.53, 0.36)
Fine Motor Raw z-score	0.066	0.86	−0.099	0.97	0.17	(−0.080, 0.41)	0.004	1.01	−0.028	0.92	0.031	(−0.38, 0.45)
Gross Motor Raw z-score	0.063	0.79	0.044	0.83	0.019	(−0.19, 0.23)	0.088	0.72	0.050	0.81	0.038	(−0.33, 0.41)
Total Motor Raw z-score	0.11	0.89	−0.051	0.98	0.16	(−0.080, 0.41)	0.062	1.50	0.019	0.87	0.043	(−0.38, 0.46)

Differences in Bayley-III scores by group status were tested using t-tests. SGA, n = 108; non-SGA, n = 132; preterm, n = 22; term, n = 226. The category of non-SGA encompasses both appropriate for gestational age (n = 128) and large for gestational age infants (n = 4). These groups were combined for analysis due to the small number of large for gestational age infants. Bayley-III Bayley Scales of Infant and Toddler Development, Third Edition; CI Confidence interval, SD Standard deviation, SGA Small for gestational age.

^aAll p > 0.05.

In the full cohort, results from linear regression analyses did not suggest mean differences in 24-month Bayley-III performance by inflammatory biomarker status for most biomarkers and domains (Supplemental Table 3). Elevated CRP at birth was associated with significantly lower receptive communication z-scores at 24 months in Model 3 (adjusted mean difference −0.25, 95% CI −0.5, −0.01). While the direction of association was also lower in Models 1, 2, and 4, they were not statistically significant (p < 0.05). Mean Bayley-III scores did not significantly differ by CRP status for other domains in the full cohort, nor by IL-1α, IL-1β, IL-6 and IL-8 status.

Table 3 shows the association of inflammation and developmental delay (z-score < −1). In all models (Models 1–4), CRP and IL-6 elevation were associated with significantly higher odds of fine motor delay (Model 3 [adjusting for prenatal confounders]: CRP, adjusted odds ratio [aOR] 2.75, 95% CI 1.28, 5.87; IL-6, aOR 2.29, 95% CI 1.08, 4.88). IL-8 elevation was associated with higher odds of language delay (OR 2.34, 95% CI 1.09, 5.05) in the unadjusted analysis; however, this association was not significant after adjustment for potential confounders. In Model 2, elevated IL-6 was associated with significantly higher odds of cognitive delay (aOR 2.22, 95% CI 1.04, 4.76) and elevated IL-1β with higher odds of gross motor delay (aOR 4.11, 95% CI 1.07, 15.86), but these associations were not significant in other models. The elevation of the remaining biomarkers at birth were not associated with developmental delays at 24 months.

Table 3.

Associations between elevated cord blood inflammation and odds of developmental delay at 24 months for the full cohort (n = 248).

		Exposure: Inflammatory Biomarker Concentration		Statistical Models
		≤ 75th percentilea	> 75th percentileb	Model 1 (Unadjusted)	Model 2 (adjustedc)	Model 3 (adjustedd)	Model 4 (adjustede)
Bayley-III Domain	Biomarker	n with delay (%a)	n with delay (%b)	OR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
Cognitive Total	IL-1α	30 (16.1)	10 (16.1)	1.03 (0.47, 2.22)	1.00 (0.45, 2.2)	0.96 (0.43, 2.12)	0.92 (0.41, 2.07)
	IL-1β	30 (16.1)	10 (16.1)	1.03 (0.47, 2.22)	1.01 (0.45, 2.23)	1.01 (0.45, 2.25)	1.06 (0.47, 2.38)
	IL-6	26 (14.0)	14 (22.6)	1.81 (0.88, 3.72)	2.22 (1.04, 4.76)*	2.15 (1, 4.63)	2.10 (0.97, 4.52)
	IL-8	28 (15.1)	12 (19.4)	1.38 (0.66, 2.89)	1.29 (0.6, 2.77)	1.32 (0.61, 2.87)	1.35 (0.62, 2.92)
	CRP	27 (14.5)	13 (21.0)	1.58 (0.76, 3.28)	1.94 (0.9, 4.15)	1.98 (0.91, 4.29)	1.91 (0.88, 4.15)
Receptive Communication	IL-1α	18 (9.7)	5 (8.1)	0.87 (0.32, 2.38)	0.86 (0.32, 2.32)	0.82 (0.31, 2.2)	0.76 (0.28, 2.06)
	IL-1β	18 (9.7)	5 (8.1)	0.87 (0.32, 2.38)	0.86 (0.32, 2.3)	0.78 (0.29, 2.11)	0.86 (0.32, 2.33)
	IL-6	16 (8.6)	7 (11.3)	1.40 (0.56, 3.51)	1.22 (0.49, 3.05)	1.18 (0.47, 2.98)	1.18 (0.46, 3)
	IL-8	15 (8.1)	8 (12.9)	1.73 (0.71, 4.23)	1.59 (0.65, 3.88)	1.44 (0.58, 3.57)	1.49 (0.6, 3.71)
	CRP	17 (9.1)	6 (9.7)	1.11 (0.43, 2.9)	1.17 (0.46, 3.01)	1.30 (0.5, 3.37)	1.18 (0.45, 3.11)
Expressive Communication	IL-1α	39 (21)	9 (14.5)	0.66 (0.3, 1.45)	0.57 (0.26, 1.27)	0.55 (0.25, 1.23)	0.54 (0.24, 1.19)
	IL-1β	34 (18.3)	14 (22.6)	1.32 (0.66, 2.66)	1.35 (0.66, 2.79)	1.38 (0.67, 2.88)	1.48 (0.7, 3.1)
	IL-6	32 (17.2)	16 (25.8)	1.69 (0.85, 3.33)	1.63 (0.8, 3.34)	1.59 (0.77, 3.27)	1.59 (0.77, 3.27)
	IL-8	31 (16.7)	17 (27.4)	1.9 (0.97, 3.73)	1.78 (0.88, 3.61)	1.81 (0.89, 3.7)	1.88 (0.91, 3.85)
	CRP	32 (17.2)	16 (25.8)	1.69 (0.85, 3.33)	1.77 (0.87, 3.59)	1.82 (0.88, 3.74)	1.76 (0.85, 3.63)
Language Total	IL-1α	26 (14.0)	6 (9.7)	0.70 (0.28, 1.74)	0.62 (0.25, 1.53)	0.56 (0.22, 1.41)	0.54 (0.21, 1.35)
	IL-1β	24 (12.9)	8 (12.9)	1.03 (0.45, 2.41)	1.06 (0.46, 2.46)	0.92 (0.39, 2.18)	0.96 (0.41, 2.28)
	IL-6	21 (11.3)	11 (17.7)	1.72 (0.78, 3.77)	1.66 (0.75, 3.67)	1.50 (0.67, 3.35)	1.49 (0.67, 3.33)
	IL-8	19 (10.2)	13 (21.0)	2.34 (1.09, 5.05)*	2.10 (0.97, 4.55)	1.82 (0.83, 4.02)	1.86 (0.84, 4.1)
	CRP	22 (11.8)	10 (16.1)	1.46 (0.66, 3.26)	1.50 (0.68, 3.34)	1.59 (0.7, 3.61)	1.54 (0.67, 3.5)
Fine Motor	IL-1α	29 (15.8)	10 (16.1)	1.05 (0.49, 2.29)	1.06 (0.48, 2.31)	0.98 (0.45, 2.16)	0.97 (0.44, 2.15)
	IL-1β	28 (15.2)	11 (17.7)	1.23 (0.57, 2.62)	1.20 (0.56, 2.59)	1.10 (0.5, 2.4)	1.12 (0.51, 2.44)
	IL-6	24 (13.0)	15 (24.6)	2.20 (1.07, 4.51)*	2.41 (1.14, 5.09)*	2.29 (1.08, 4.88)*	2.27 (1.07, 4.83)*
	IL-8	27 (14.6)	12 (19.7)	1.46 (0.69, 3.07)	1.45 (0.68, 3.1)	1.30 (0.6, 2.83)	1.31 (0.61, 2.84)
	CRP	24 (13.0)	15 (24.6)	2.20 (1.07, 4.51)*	2.47 (1.18, 5.18)*	2.75 (1.28, 5.87)**	2.71 (1.27, 5.8)*
Gross Motor	IL-1α	4 (2.2)	2 (3.3)	1.70 (0.35, 8.26)	1.30 (0.33, 5.22)	1.32 (0.35, 4.89)	1.14 (0.3, 4.31)
	IL-1β	3 (1.6)	3 (4.9)	3.12 (0.68, 14.25)	4.11 (1.07, 15.86)*	3.72 (0.97, 14.32)	3.84 (0.99, 14.9)
	IL-6	4 (2.2)	2 (3.2)	1.66 (0.34, 8.07)	1.74 (0.44, 6.83)	1.56 (0.41, 5.84)	1.28 (0.33, 4.92)
	IL-8	4 (2.2)	2 (3.3)	1.70 (0.35, 8.26)	2.10 (0.52, 8.52)	1.84 (0.46, 7.32)	1.75 (0.43, 7.09)
	CRP	5 (2.7)	1 (1.6)	0.81 (0.13, 5.13)	0.75 (0.16, 3.62)	0.58 (0.12, 2.81)	0.48 (0.09, 2.46)
Motor Total	IL-1α	15 (8.1)	1 (1.6)	0.27 (0.05, 1.52)	0.25 (0.05, 1.24)	0.21 (0.04, 1.04)	0.22 (0.05, 1.05)
	IL-1β	14 (7.6)	2 (3.3)	0.50 (0.12, 1.98)	0.49 (0.13, 1.86)	0.43 (0.12, 1.64)	0.41 (0.11, 1.57)
	IL-6	11 (5.9)	5 (8.2)	1.48 (0.51, 4.29)	1.30 (0.44, 3.81)	1.14 (0.38, 3.38)	1.15 (0.39, 3.38)
	IL-8	13 (7.0)	3 (5.0)	0.78 (0.23, 2.65)	0.72 (0.22, 2.41)	0.60 (0.17, 2.06)	0.58 (0.17, 2)
	CRP	12 (6.5)	4 (6.7)	1.11 (0.36, 3.43)	1.16 (0.39, 3.49)	1.31 (0.43, 4.02)	1.35 (0.44, 4.11)

The exposure was biomarker elevation of > 75^th percentile of each cord blood inflammatory biomarker distribution, compared to ≤ 75^th percentile of the inflammatory biomarker distribution. The outcome was developmental delay (defined as a Bayley-III z-score <−1). aOR Adjusted odds ratio, Bayley-III, Bayley Scales of Infant and Toddler Development, Third Edition; CI Confidence interval, CRP C-reactive protein, IL Interleukin, OR Odds ratio.

^*p < 0.05.

^**p<0.01.

^aThe total number of infants with low inflammation for each biomarker (denominator for % calculation) is n = 186.

^bThe number of infants with high inflammation for each biomarker (denominator for % calculation) is, n = 62.

^cModel 2: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment.

^dModel 3: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment, maternal body mass index (BMI), maternal education, maternal depression.

^eModel 4: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment, maternal BMI, maternal education, maternal depression, Home Observation forMeasurement of the Environment (HOME) score.

Associations between cord blood inflammation and composite Bayley-III scores for the full cohort and sub-groups are reported in Supplemental Table 4a–e.

Vulnerable newborns (Stratified analysis by SGA and preterm status)

SGA and non-SGA infants.

In stratified analyses, among SGA infants (n = 108), elevated IL-8 was associated with lower receptive communication (Model 3: −0.40, 95% CI: −0.75 to −0.06) and total language z-scores (Model 3: −0.38, 95% CI: −0.73 to −0.04) in all adjusted models (Supplemental Table 5). Elevated CRP was associated with lower receptive communication z-scores (Model 3: −0.48, 95% CI −0.84, −0.13), and IL-1β elevation was associated with lower expressive communication, gross motor, and total motor z-scores (Model 3: expressive, −0.45, 95% CI −0.85, −0.05; gross motor, −0.45, 95% CI −0.76, −0.14; total motor, −0.35, 95% CI −0.69, −0.0001) in the adjusted models. In Model 4, IL-1β elevation was also associated with lower total language z-scores (−0.40, 95% CI −0.77, −0.03).

Among SGA infants, elevated cord IL-1α was associated with higher odds of cognitive delay (Table 4; Model 3: aOR 3.55, 95% CI 1.08, 11.65). IL-8 elevation was associated with language delay in all models (Model 3: aOR 4.27, 95% CI 1.29, 14.16), particularly in expressive communication (Model 3: aOR 3.49, 95% CI 1.2, 10.15). Elevated IL-1β was associated with expressive communication delay (Model 3: aOR 3.40, 95% CI 1.07, 10.79) and gross motor delay (Model 3: aOR 7.27, 95% CI 1.17, 45.01) in all adjusted models, but it was not associated with delays in the other domains. Elevated CRP was associated with significantly higher odds of delay in receptive communication (Model 3: aOR 3.69, 95% CI 1.04, 13.06).

Table 4.

Associations between elevated cord blood inflammation and odds of developmental delay at 24 months among SGA infants (n = 108).

		Exposure: Inflammatory Biomarker Concentration		Statistical Models
		≤ 75th percentilea	> 75th percentileb	Model 1 (Unadjusted)	Model 2 (adjustedc)	Model 3 (adjustedd)	Model 4 (adjustede)
Bayley-III Domain	Biomarker	n with delay (%a)	n with delay (%b)	OR (95% CI)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
Cognitive Total	IL-1α	11 (13.4)	8 (30.8)	2.86 (1.01, 8.07)*	3.21 (0.99, 10.46)	3.55 (1.08, 11.65)*	3.43 (1.04, 11.32)*
	IL-β	13 (15.3)	6 (26.1)	2.00 (0.67, 5.93)	1.82 (0.55, 6.09)	1.66 (0.5, 5.55)	1.71 (0.51, 5.74)
	IL-6	11 (13.9)	8 (27.6)	2.36 (0.85, 6.55)	3.17 (0.98, 10.21)	3.03 (0.94, 9.77)	2.87 (0.89, 9.22)
	IL-8	10 (13.0)	9 (29.0)	2.71 (0.99, 7.46)	2.95 (0.98, 8.94)	2.81 (0.92, 8.66)	2.75 (0.9, 8.46)
	CRP	13 (16.3)	6 (21.4)	1.44 (0.5, 4.19)	1.93 (0.6, 6.22)	2.00 (0.62, 6.48)	2.07 (0.64, 6.65)
Receptive Communication	IL-α	5 (6.1)	4 (15.4)	2.82 (0.73, 10.87)	2.79 (0.77, 10.14)	2.55 (0.72, 8.97)	1.92 (0.51, 7.2)
	IL-β	6 (7.1)	3 (13.0)	2.09 (0.51, 8.55)	1.99 (0.5, 8)	1.93 (0.5, 7.47)	1.95 (0.49, 7.73)
	IL-6	6 (7.6)	3 (10.3)	1.49 (0.37, 6.01)	1.38 (0.38, 5.05)	1.28 (0.35, 4.62)	1.31 (0.35, 4.9)
	IL-8	4 (5.2)	5 (16.1)	3.39 (0.89, 12.91)	2.94 (0.85, 10.2)	2.64 (0.78, 8.92)	2.61 (0.75, 9.04)
	CRP	4 (5.0)	5 (17.9)	3.98 (1.04, 15.27)*	3.96 (1.12, 14.07)*	3.69 (1.04, 13.06)*	3.83 (1.05, 14)*
Expressive Communication	IL-α	15 (18.3)	3 (11.5)	0.65 (0.18, 2.31)	0.61 (0.17, 2.17)	0.69 (0.2, 2.42)	0.56 (0.15, 2.1)
	IL-β	11 (12.9)	7 (30.4)	2.95 (1, 8.68)	3.42 (1.08, 10.84)*	3.40 (1.07, 10.79)*	4.04 (1.19, 13.74)*
	IL-6	12 (15.2)	6 (20.7)	1.49 (0.51, 4.36)	1.59 (0.54, 4.66)	1.72 (0.58, 5.1)	1.62 (0.52, 4.99)
	IL-8	9 (11.7)	9 (29.0)	3.04 (1.09, 8.52)*	3.17 (1.11, 9.05)*	3.49 (1.2, 10.15)*	3.70 (1.22, 11.19)*
	CRP	11 (13.8)	7 (25.0)	2.11 (0.74, 6.05)	2.11 (0.73, 6.12)	2.27 (0.76, 6.76)	2.22 (0.72, 6.86)
Language Total	IL-α	8 (9.8)	3 (11.5)	1.31 (0.34, 5.03)	1.22 (0.34, 4.4)	1.2 (0.34, 4.28)	0.73 (0.17, 3.13)
	IL-β	8 (9.4)	3 (13.0)	1.56 (0.4, 6.06)	1.49 (0.39, 5.73)	1.46 (0.39, 5.54)	1.48 (0.36, 6.06)
	IL-6	6 (7.6)	5 (17.2)	2.54 (0.74, 8.77)	2.38 (0.73, 7.75)	2.42 (0.73, 8)	2.25 (0.62, 8.16)
	IL-8	4 (5.2)	7 (22.6)	5.00 (1.41, 17.77)*	4.53 (1.36, 15.13)*	4.27 (1.29, 14.16)*	4.06 (1.15, 14.36)*
	CRP	6 (7.5)	5 (17.9)	2.68 (0.77, 9.29)	3.15 (0.94, 10.61)	3.21 (0.94, 10.91)	3.59 (0.95, 13.58)
Fine Motor	IL-α	10 (12.5)	3 (11.5)	1.00 (0.27, 3.73)	1.13 (0.32, 4)	0.96 (0.26, 3.53)	0.85 (0.22, 3.27)
	IL-β	11 (13.3)	2 (8.7)	0.73 (0.17, 3.21)	0.61 (0.15, 2.52)	0.61 (0.15, 2.52)	0.63 (0.15, 2.59)
	IL-6	7 (9.0)	6 (21.4)	2.75 (0.86, 8.85)	3.04 (0.95, 9.76)	3.07 (0.93, 10.11)	2.94 (0.9, 9.63)
	IL-8	9 (11.8)	4 (13.3)	1.21 (0.36, 4.11)	1.25 (0.38, 4.05)	1.04 (0.31, 3.49)	1.04 (0.31, 3.48)
	CRP	8 (10.1)	5 (18.5)	2.06 (0.63, 6.75)	2.64 (0.81, 8.59)	2.12 (0.63, 7.11)	2.13 (0.63, 7.15)
Gross Motor	IL-α	1 (1.2)	1 (4.0)	3.33 (0.32, 34.47)	2.65 (0.41, 17.01)	2.35 (0.41, 13.35)	1.80 (0.31, 10.44)
	IL-β	0 (0)	2 (9.1)	20.85 (0.94, 463.77)	22.46 (2.02, 250.34)*	7.27 (1.17, 45.01)*	6.64 (1.43, 30.88)*
	IL-6	1 (1.3)	1 (3.4)	2.72 (0.26, 27.97)	1.84 (0.29, 11.46)	1.58 (0.28, 8.95)	1.39 (0.25, 7.83)
	IL-8	1 (1.3)	1 (3.3)	2.59 (0.25, 26.64)	2.52 (0.4, 15.77)	2.16 (0.39, 12.03)	1.77 (0.33, 9.53)
	CRP	2 (2.5)	0 (0)	0.57 (0.03, 12.9)	0.51 (0.05, 5.37)	0.55 (0.07, 4.64)	0.70 (0.1, 4.9)
Motor Total	IL-α	4 (4.9)	0 (0)	0.34 (0.02, 6.85)	0.37 (0.04, 3.69)	0.38 (0.05, 3.13)	0.47 (0.06, 3.43)
	IL-β	3 (3.6)	1 (4.5)	1.63 (0.22, 12.07)	1.51 (0.27, 8.53)	1.39 (0.27, 7.31)	1.37 (0.26, 7.2)
	IL-6	2 (2.6)	2 (7.1)	2.89 (0.46, 18)	2.33 (0.52, 10.53)	2.26 (0.52, 9.83)	2.46 (0.57, 10.62)
	IL-8	3 (3.9)	1 (3.4)	1.12 (0.15, 8.18)	1.13 (0.22, 5.76)	1.05 (0.22, 5.05)	1.11 (0.24, 5.22)
	CRP	3 (3.8)	1 (3.8)	1.30 (0.18, 9.57)	1.47 (0.29, 7.5)	1.46 (0.3, 7.13)	1.48 (0.31, 7.16)

The exposure was biomarker elevation of > 75th percentile of each cord blood inflammatory biomarker distribution, compared to ≤ 75th percentile of the inflammatory biomarker distribution. The outcome was developmental delay (defined as a Bayley-III z-score < −1). aOR Adjusted odds ratio, Bayley-III Bayley Scales of Infant and Toddler Development, Third Edition, CI Confidence interval, CRP C-reactive protein, IL Interleukin, OR Odds ratio, SGA Small for gestational age.

^*p < 0.05.

^aThe total number of SGA infants with low inflammation for each biomarker (denominator for % calculation) are as follows: IL-1α, n = 82; IL-1β, n = 85; IL-6, n = 79; IL-8, n = 77; CRP, n = 80.

^bThe number of SGA infants with high inflammation for each biomarker (denominator for % calculation) are as follows: IL-1α, n = 26; IL-1β, n = 23; IL-6, n = 29; IL-8, n = 31; CRP, n = 28.

^cModel 2: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment.

^dModel 3: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment, maternal body mass index (BMI), maternal education, maternal depression.

^eModel 4: Adjusted for gestational age, parity, infant sex, age in years at time of Bayley-III assessment, maternal BMI, maternal education, maternal depression, Home Observation for Measurement of the Environment (HOME) score.

We also explored these relationships among non-SGA infants (n = 132), and there were few significant associations (Supplemental Table 6a, b). CRP elevation at birth was associated with higher odds of fine motor delay among non-SGA infants in all models (Model 3: aOR 3.70, 95% CI 1.33, 10.34).

Preterm vs term infants.

Analyses among preterm infants were limited by small sample size (n = 22), with most infants moderate-late preterm (> 32 to < 37 weeks; 96%). We found that in preterm infants, elevated cord IL-1β and IL-8 (> 75^th percentile) were associated with lower total expressive communication z-scores in the unadjusted analysis (Model 1) (IL-1β: −1.12, 95% CI −2.12, −0.12 and IL-8: −1.31, 95% CI −2.46, −0.16, respectively), and elevated IL-8 was also associated with lower expressive communication z-scores in Model 2 (−1.23, 95% CI −2.43, −0.04) (Supplemental Table 7a). We found no significant associations between elevated cytokines and developmental delay in preterm infants (Supplemental Table 7b).

We found few significant associations among term infants (Supplemental Table 8a, b). In term infants, elevated cord CRP was associated with higher odds of fine motor delay in all models (Model 3: aOR 2.98, 95% CI 1.32, 6.75), and IL-1β was associated with gross motor delay in all adjusted models (Model 3: aOR 4.01, 95% CI 1.02, 15.86) (Supplemental Table 8b).

DISCUSSION

The negative impacts of perinatal inflammation on neurodevelopment have been well established in high income settings in certain high-risk groups. However, there are limited studies that have evaluated the role of perinatal inflammation in population-based LMIC cohorts, particularly among SGA infants. The prevalence of SGA approaches 40% in South Asia, and an estimated 23.4 million infants are born SGA globally each year.⁴ In our birth cohort in rural Sylhet, Bangladesh, we found that elevated CRP and IL-6 at birth were associated with fine motor delay at 24 months. Notably, the associations between perinatal inflammation and neurodevelopmental impairment were most prominent among SGA infants, for whom multiple biomarkers were associated with delays across several domains. These data extend the evidence on the role of inflammation in adverse neurodevelopment in a Bangladeshi cohort where there is a high burden of infectious disease and fetal growth restriction.

The developing infant brain is particularly sensitive to inflammation due to the vulnerability of the developing oligodendrocyte and the activation of microglia that leads to excitotoxicity, neuronal death, and impaired neuronal differentiation.^22,63 Preterm infants comprised 8.9% of our cohort, with most (96%) being moderate-late preterm (32 to < 37 wks), given the low survival rate of extremely preterm infants in Sylhet. In our stratified analysis, preterm infants exposed to elevated IL-8 had lower expressive communication at 24 months. The majority of prior evidence on the impacts of perinatal inflammation has been among extremely preterm infants, among whom inflammation has been associated with white matter brain injury, cerebral palsy, severe IVH, as well as long-term neurodevelopmental outcomes. In the multicenter Extremely Low Gestational Age Newborn (ELGAN) Studies (infants born at gestational age < 28 weeks), inflammation at birth in the infant’s blood (IL-8, SAA [serum amyloid A], MIP [macrophage inflammatory protein], VEGF [vascular endothelial growth factor]) was associated with ultrasonographic evidence of white matter injury, including ventriculomegaly and echo lucent lesions, and attributed to in-utero inflammation effects.¹⁷ Persistent elevation of CRP, IL-6, IL-8 and IL-1β at 3–4 weeks postnatal age was associated with microcephaly and ventriculomegaly and lower Bayley-III scores at 2 years.⁴¹ In a Swedish cohort of infants born before 32 weeks gestation, elevation of IL-8 in cord blood was associated with cerebral palsy at 2 years of age.⁶⁴ The association of perinatal inflammation and neurodevelopmental outcomes has not been as well characterized in moderate-late preterm infants. In our literature review, we identified one study that examined the role of perinatal inflammation in a subgroup of late preterm (≥ 34–< 37 wks) infants in the US Collaborative Perinatal Project.⁶⁵ This study did not find an association between funisitis and later childhood IQ in the late preterm subgroup.

SGA infants are more vulnerable to long-term neurodevelopmental impairment due to placental insufficiency, resulting in chronic fetal hypoperfusion, undernutrition, and hypoxia. Growth restricted infants carry increased risk of neonatal encephalopathy⁶⁶ and cerebral palsy.⁶⁷ In a recent systematic review, SGA infants were more likely than AGA infants to have cognitive delay persisting into teenage years.^68,69 In our cohort, although we did not find lower 24-month Bayley-III scores among SGA infants, we found a consistent pattern demonstrating vulnerability of SGA infants to perinatal inflammation. While animal studies support the role of neuroinflammation in brain injury in fetal growth restriction models, in our literature review (Supplemental Table 9), we did not locate prior reports on the effects of perinatal inflammation in human cohorts of term-SGA infants. Inflammation has been found to characterize a subset of the etiologies of SGA through mechanisms of placental malperfusion, as opposed to mechanisms of aberrant spiral arteriole formation.⁷⁰ Thus, elevated inflammation proteins in the cord blood may be markers of persistent malperfusion of the fetus, including the fetal brain, in a subgroup of intrauterine growth-restricted (IUGR) infants. Furthermore, neuro-inflammation may also mediate neurovascular unit disruption and blood-brain barrier dysfunction in the brains of IUGR infants.⁷¹ In the SGA sub-group, the elevation of several biomarkers at birth was associated with neurodevelopmental impairment across several domains, which was not observed in the non-SGA group. This may represent cord blood inflammatory markers as a biomarker of the nature and severity of perfusion deficits, or alternatively, a “two hit” model of brain damage. Such a “two hit” model has been proposed in rat models, as well as in preterm infants. Intrauterine growth restriction is hypothesized to “sensitize” neonatal rats to later inflammatory stimuli and lead to greater brain injury than in non-growth restricted rats.⁷² Leviton et al. demonstrated in one of the first human studies a similar relationship among extremely preterm SGA infants.⁴⁶ In the ELGAN study, growth-restricted infants were more likely to have increased inflammation biomarkers (including CRP, IL-1β, IL-6, IL-8, TNF-α) on postnatal day 14, and the odds of very low Motor Development Index (MDI) < 55 associated with persistent inflammation of several biomarkers (including IL-1β, IL-6, TNF-α, and IL-8) was considerably higher among SGA versus AGA infants.⁴⁸ We hypothesize that chronic hypoxia resulting from placental insufficiency may disrupt the neurovascular unit, sensitize the brain to inflammatory insults, and place the infant at higher risk of brain injury. IUGR infants have developmental vulnerability in the immature white matter and gray matter, including impaired gyrification and subplate neurons that are more vulnerable to excitotoxic injury.^73,74 These results highlight the therapeutic potential of interventions to reduce exposure to inflammation in-utero to optimize neurodevelopmental outcomes. Such an example is polyphenols, an antioxidant compound that has been shown to lower odds of brain injury when provided to pregnant women with growth-restricted fetuses.⁷⁵

Maternal immune activation (MIA) is believed to adversely affect fetal neurodevelopment via exposure to inflammatory perturbations in utero.⁷⁶ Maternal IL-6 and CRP have been the key inflammation mediators implicated in the MIA hypothesis.⁷⁶ In our study, elevated cord blood IL-6, as a potential indicator of late prenatal maternal concentrations, was associated with cognitive delay at 2 years. Elevated CRP was associated with lower receptive communication scores and elevated CRP and IL-6 with fine motor delays at 2 years. IL-6 activates pathways resulting in expression of acute phase reactants, including CRP and the classical complement pathway. A majority of our knowledge about MIA arises from rodent models, in which maternal inflammation results in placental immune activation,^77,78 alterations in fetal cytokine expression,^79–81 as well as altered postnatal brain cytokine concentrations with regional specific changes.⁸² In animal models, elevated maternal IL-6 alters offspring regional brain development (hippocampus, prefrontal cortex, mid-temporal and parietal lobe)^83–85 and leads to behavioral disturbances (behavioral reactivity, deficits in emotion, attention, and memory) among offspring.^86,87 Observational studies in pregnant women have corroborated the role of maternal inflammation with IL-6 in offspring neurodevelopment.^88–90 Maternal IL-6 has been negatively associated with neonatal functional connectivity⁹⁰ and working memory at 24 months.⁸⁹

IL-8 is a pro-inflammatory chemokine that has been linked with the pathogenesis of cerebral palsy. IL-8 plays a critical role in inciting neutrophil infiltration post-hypoxic injury⁹¹ with subsequent neutrophil-mediated neuronal cell death. IL-8 has also been associated with preterm birth, with elevations of IL-8 noted in amniotic fluid⁹² and has been proposed as a mediator of injury leading to cerebral palsy and motor impairments. In the National Institute of Child Health and Development (NICHD) Neonatal Research Network study of 1067 extremely low birthweight infants, IL-8 was the only one of 5 cytokines to have consistently elevated levels among infants who developed cerebral palsy.⁹³ Infants with elevated IL-8 levels on day 0–4 after birth were more likely to have a Psychomotor Development Index (PDI) score < 70 or moderate-severe cerebral palsy at 18–22 months. Similarly, in the ELGAN study, elevation of IL-8 on day-of-life 1 was associated with 1.7 odds (95% CI: 1.03, 2.9) of motor deficits at 2 years.³¹ In our study, IL-8 was associated with lower receptive communication and language scores, and we did not find significant associations with motor outcomes.

A majority of the studies linking inflammation and neurodevelopment have been conducted in high-income settings. Our study is one of few that has evaluated the impacts of perinatal inflammation on neurodevelopment in a LMIC, where the prevalence of risk factors for inflammation is high. In another cohort of children in Dhaka, Bangladesh enrolled in the Bangladesh Early Adversity Neuroimaging (BEAN) study, fever and elevation in IL-1β and IL-6 in infancy (6 months of age) were associated with a decrease in Bayley-III motor scores from 12 to 24 months of age.⁴³ Furthermore, recurrent inflammation (specifically, CRP elevation) in the first year of life was associated with increased parieto-occipital functional connectivity on EEG and lower cognitive scores as measured by the Mullen Scales of Early Learning at 3–5 years. Increased EEG functional connectivity was postulated to be a maladaptive response due to inflammation disrupting the myelination process.⁹⁴

Strengths and limitations

Strengths of the study include that it is one of the first well-characterized cohorts from an LMIC to assess the relationships of perinatal inflammation and long-term neurodevelopmental outcomes, in a setting with high prevalence of SGA. We used rigorous methods to test cord blood inflammation biomarkers that were developed and validated for newborn blood in the sentinel ELGAN study. Our cohort’s gestational ages were dated by early pregnancy ultrasound, and we had high rates of follow up. Major limitations of this study are its observational nature, the multiple comparisons that were made, and the chance of Type 1 error. We did not adjust for multiple comparisons given the exploratory nature of the study, and, as recommended by Rothman,⁹⁵ we were not addressing a universal null hypothesis. However, the conclusions are supported by similar associations/findings across several biomarkers and domains. A limitation of this study from a mechanistic standpoint is that we only analyzed cord blood at a single time point at birth, and do not have data on placental, maternal prenatal, or infant postnatal inflammation status. We have access to bio-banked maternal and placental samples that may be tested in the future. The ELGAN group has shown that the highest risk of adverse developmental outcomes is among infants with persistently elevated inflammatory proteins at multiple time points.⁹⁶ Postnatal inflammation has also been shown by the ELGAN group and others to be an important predictor of long-term outcomes.^31,41,96,97 We also had a relatively small sample size and were not able to do mediation analysis to examine whether inflammation biomarkers were mediators of brain injury and if biomarker elevations resulted from another predisposing condition. Finally, we acknowledge that a myriad of postnatal factors play a critical role in shaping a child’s development. Our primary interest was in the in-utero exposure to inflammation, and our main models included confounders temporally preceding birth. We examined an additional model (Model 4) adjusting for the postnatal home environment and did find any notable change in our effect estimates. Our study indicates that a one-time non-invasive assessment of inflammation may identify vulnerability at birth that could be targeted by future mechanistic research and preventive interventions.

CONCLUSIONS

In an infant cohort in rural Bangladesh, we found that inflammation at birth was associated with long-term developmental risk, and lower cognitive and motor scores at two years of age. Among vulnerable SGA infants, we noted that exposures to various elevated inflammation biomarkers at birth were associated with neurodevelopmental delays in several domains, supporting the “two hit” hypothesis and the underlying vulnerability of the SGA infant brain. Increased efforts should target interventions to reduce risk factors resulting in inflammation and the prevention of SGA in LMICs to optimize developmental potential.

IMPACT:

• Cord blood inflammation was associated with fine motor and language delays at 24 months of age in a community-based cohort in rural Bangladesh.

• 23.4 million infants are born small-for-gestational-age (SGA) globally each year. Among SGA infants, the associations between cord blood inflammation and adverse outcomes were strong and consistent across several biomarkers and neurodevelopmental domains (cognitive, motor, language), supporting the neurobiological impact of inflammation prominent in growth-restricted infants.

• Prenatal interventions to prevent intrauterine growth restriction are needed in low- and middle-income countries and may also result in long-term benefits on child development.

DATA AVAILABILITY

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.