[Skip to Navigation]
Sign In
January 4, 2022

Understanding the Effects of the COVID-19 Pandemic on Infant Development—The Preterm Problem

Author Affiliations
  • 1Gillings School of Global Public Health, Department of Epidemiology, University of North Carolina at Chapel Hill
JAMA Pediatr. 2022;176(6):e215570. doi:10.1001/jamapediatrics.2021.5570

As the COVID-19 pandemic nears the end of another year, it makes sense to wonder about the effect on children conceived and born in its shadow. Some of the consequences can be seen in the increasing rates of COVID-19 among infants and young children,1 particularly among those with comorbid medical conditions.2 But less obvious sequelae also should be considered: are infants born during the pandemic at greater risk for behavioral or neurodevelopmental problems, either due to exposure to maternal SARS-CoV-2 infection or to the more global effects of trauma and stress? Pregnant women and their infants are vulnerable to the effects of disasters, and evidence suggests that disasters affect maternal mental health and some perinatal health outcomes, particularly among highly exposed women.3 Prenatal exposure to some viral infections, such as rubella and HIV, increases the risk of neurobehavioral problems in children,4 and some have hypothesized that SARS-CoV-2 infection could have a similar outcome via in utero exposure to maternal fever, hypoxia, or inflammation.4,5

The study by Shuffrey et al6 in JAMA Pediatrics aims to provide evidence regarding these concerns. In this study, the authors evaluated the neurodevelopment of infants born during the pandemic, and compared the outcomes of children born during the pandemic with a small group of historical controls, as well as the outcomes of children with and without prenatal exposure to maternal SARS-CoV-2 infection. Mother-infant dyads born in New York City during the pandemic with evidence of SARS-CoV-2 infection during pregnancy were enrolled in the study during the pregnancy or up to several months post-partum. Mothers without evidence of infection during pregnancy were invited to participate if they matched an exposed pregnancy based on infant sex, gestational age at delivery, date of birth, and mode of delivery. In total, 1706 mother-infant dyads were invited to participate and 255 were ultimately included in the study; no information is available on differences between the included and excluded dyads. The final sample consisted of 317 infants, including 255 infants in the pandemic cohort (n = 114 exposed to SARS-CoV-2 in utero; n = 141 unexposed) and 62 infants in the historical cohort.

The outcomes, defined from a validated questionnaire reported by the mothers when the infants were aged 6 months, cover attainment in 5 key developmental domains, such as communication and gross and fine motor skills at age 6 months. In statistical models that compared the mean neurodevelopmental domain scores of infants with and without prenatal exposure to SARS-CoV-2, the 2 groups had near-identical outcomes, with 1 point or less separating the exposed and unexposed infants; multivariable adjustment for maternal race, ethnicity, age at delivery, educational level, parity, and mode of delivery made very little difference. In addition, analyses that compared the number of infants in each group with scores below a prespecified clinical cutoff indicating delay suggested that more infants in the exposed group scored below this cutoff, except for scores in the communication domain. However, these comparisons were not adjusted for covariates, making them difficult to interpret. Despite this limitation, the authors suggest that there were no meaningful differences in neurodevelopment associated with prenatal SARS-CoV-2 exposure. By contrast, the comparison with the historical cohort revealed significantly lower scores among infants born during the pandemic, suggesting a substantial decrement in neurodevelopment associated with birth during the pandemic regardless of actual exposure to SARS-CoV-2 infection.

Several aspects of the study design and analysis related to gestational age at birth deserve further attention. The historical cohort of infants born before the pandemic were enrolled from the Well Baby Nursery and therefore did not include infants born before 37 completed gestational weeks or infants admitted to the NICU. In an attempt to make comparisons with the historic cohort more ‘fair’, preterm deliveries (n = 17) and term infants who were admitted to the NICU (n = 11) were excluded from the pandemic cohort in this set of analyses.6 Within the pandemic cohort, preterm and NICU-admitted infants were included, and infants with prenatal SARS-CoV-2 infection were matched to those without prenatal infection based on gestational age at delivery; gestational age was further included as a covariate in statistical models.6

Accumulating evidence suggests that infants born to women with SARS-CoV-2 infection are more often preterm, particularly for instances of severe maternal illness.7 Preterm delivery is a leading cause of neonatal morbidity,8 including developmental delay9: it is such an important cause and effect of so many maternal and infant morbidities that many researchers have a strong intuition that analyses of perinatal exposures should control for gestational age at delivery. However, this intuition is almost always wrong, and it is important to understand why.

For many years, perinatal epidemiologists described a seemingly paradoxical finding: prenatal exposure to cigarette smoke is associated with low birth weight and increased infant mortality. However, among very low-birth-weight infants, the effect is reversed: babies born to mothers who smoked were more likely to survive than babies born to nonsmokers. It was later observed that, of all the reasons an infant might have a very low weight at birth, smoking was one of the more benign. In fact, a major cause of both very low birth weight and infant mortality is severe congenital malformations, such as anencephaly. The combination of very low birth weight and nonsmoking is indicative of a serious birth defect, which produces the paradoxical association.10 Analyses that control for the presence of malformations generally recover the expected associations.

The birth weight paradox, an example of collider stratification bias, is the result of a distorted association in the study population that is induced from conditioning on a common effect (eg, birth weight) of 2 or more causes (eg, malformations and tobacco use).11 This bias is present in the study population even when the causes are entirely independent in the source population. Collider stratification bias can occur in any situation in which an analysis is conditioned on an effect of exposure and there are uncontrolled common causes of this effect and the outcome.

In the present study by Shuffrey et al,6 the analyses were conditioned on gestational age at delivery by virtue of excluding preterm infants, matching on gestational age, and controlling for gestational age in statistical models. Gestational age at delivery is strongly associated with SARS-CoV-2 infection.7 As for common causes of lower gestational age at birth and neurodevelopment, some of the most likely suspects are genetic or other heritable factors, such as family environment, but the list is certainly long and includes many variables beyond those measured and controlled for in this study. It is possible to hypothesize, for example, that being born at term, despite prenatal SARS-CoV-2 infection exposure, indicates the presence of protective factors for both lower gestational age at birth and neurodevelopment.

Preterm birth, or more generally gestational age at delivery, will almost always be an important prognostic factor in studies focused on prenatal exposures and infant or child outcomes because it indicates the presence of at least 1 pathological state during pregnancy.12 This means that gestational age at delivery is plausibly an intermediate in many studies. When planning such a study, researchers must understand what question they are trying to ask. In the study by Shuffrey et al,6 the questions the authors apparently intended to ask were, what are the effects of either 1) birth during the pandemic or 2) prenatal exposure to maternal SARS-CoV-2 infection with infant neurodevelopment? These are questions about total effects and therefore the analysis should control for common causes of the exposure and the outcome, but not for intermediate variables.13 Because the uninfected mothers were selected for inclusion based on gestational age and the historical cohort was limited to full term infants, the design of the study does not align with these questions and therefore cannot provide information about the total effect of these exposures.

Modern epidemiologic methods include an array of tools such as the potential outcomes framework, causal graph theory, and target trial emulation, all of which can be combined with robust statistical approaches to address causal questions.14 Future studies on the relationship between prenatal exposure to the pandemic or SARS-CoV-2 and infant neurodevelopment should consider these methodological approaches for answering these important questions.

Back to top
Article Information

Published Online: January 4, 2022. doi:10.1001/jamapediatrics.2021.5570

Correction: This editorial was corrected on February 22, 2022, to adjust content based on a previous version of the related study.

Corresponding Author: Mollie E. Wood, PhD, MPH, Gillings School of Global Public Health, Department of Epidemiology, University of North Carolina at Chapel Hill, 135 Dauer Dr, 2101 McGavran-Greenberg Hall, CB #7435, Chapel Hill, NC 27599 (mwood@unc.edu).

Conflict of Interest Disclosures: None reported.

Siegel  DA, Reses  HE, Cool  AJ,  et al; MAPW1.  Trends in COVID-19 cases, emergency department visits, and hospital admissions among children and adolescents aged 0-17 years—United States, August 2020-August 2021.   MMWR Morb Mortal Wkly Rep. 2021;70(36):1249-1254. doi:10.15585/mmwr.mm7036e1 PubMedGoogle ScholarCrossref
Kompaniyets  L, Agathis  NT, Nelson  JM,  et al.  Underlying medical conditions associated with severe COVID-19 illness among children.   JAMA Netw Open. 2021;4(6):e2111182. doi:10.1001/jamanetworkopen.2021.11182 PubMedGoogle ScholarCrossref
Harville  E, Xiong  X, Buekens  P.  Disasters and perinatal health: a systematic review.   Obstet Gynecol Surv. 2010;65(11):713-728. doi:10.1097/OGX.0b013e31820eddbe PubMedGoogle ScholarCrossref
Zimmer  A, Youngblood  A, Adnane  A, Miller  BJ, Goldsmith  DR.  Prenatal exposure to viral infection and neuropsychiatric disorders in offspring: a review of the literature and recommendations for the COVID-19 pandemic.   Brain Behav Immun. 2021;91(91):756-770. doi:10.1016/j.bbi.2020.10.024 PubMedGoogle ScholarCrossref
López-Díaz  Á, Ayesa-Arriola  R, Crespo-Facorro  B, Ruiz-Veguilla  M.  COVID-19 infection during pregnancy and risk of neurodevelopmental disorders in offspring: time for collaborative research.   Biol Psychiatry. 2021;89(5):e29-e30. doi:10.1016/j.biopsych.2020.09.011 PubMedGoogle ScholarCrossref
Shuffrey  LC, Firestein  MR, Kyle  MH,  et al.  Association of birth during the COVID-19 pandemic with neurodevelopmental status at 6 months in infants with and without in utero exposure to maternal SARS-CoV-2 infection.   JAMA Pediatr. Published online January 4, 2022. doi:10.1001/jamapediatrics.2021.5563PubMedGoogle ScholarCrossref
Villar  J, Ariff  S, Gunier  RB,  et al.  Maternal and neonatal morbidity and mortality among pregnant women with and without COVID-19 infection: the INTERCOVID multinational cohort study.   JAMA Pediatr. 2021;175(8):817-826. doi:10.1001/jamapediatrics.2021.1050 PubMedGoogle ScholarCrossref
Platt  MJ.  Outcomes in preterm infants.   Public Health. 2014;128(5):399-403. doi:10.1016/j.puhe.2014.03.010 PubMedGoogle ScholarCrossref
Morse  SB, Zheng  H, Tang  Y, Roth  J.  Early school-age outcomes of late preterm infants.   Pediatrics. 2009;123(4):e622-e629. doi:10.1542/peds.2008-1405 PubMedGoogle ScholarCrossref
Hernández-Díaz  S, Schisterman  EF, Hernán  MA.  The birth weight “paradox” uncovered?   Am J Epidemiol. 2006;164(11):1115-1120. doi:10.1093/aje/kwj275 PubMedGoogle ScholarCrossref
Hernán  MA, Hernández-Díaz  S, Robins  JM.  A structural approach to selection bias.   Epidemiology. 2004;15(5):615-625. doi:10.1097/01.ede.0000135174.63482.43 PubMedGoogle ScholarCrossref
Goldenberg  RL, Culhane  JF, Iams  JD, Romero  R.  Epidemiology and causes of preterm birth.   Lancet. 2008;371(9606):75-84. doi:10.1016/S0140-6736(08)60074-4 PubMedGoogle ScholarCrossref
VanderWeele  TJ, Mumford  SL, Schisterman  EF.  Conditioning on intermediates in perinatal epidemiology.   Epidemiology. 2012;23(1):1-9. doi:10.1097/EDE.0b013e31823aca5d PubMedGoogle ScholarCrossref
Lesko  CR, Keil  AP, Edwards  JK.  The epidemiologic toolbox: identifying, honing, and using the right tools for the job.   Am J Epidemiol. 2020;189(6):511-517. doi:10.1093/aje/kwaa030 PubMedGoogle ScholarCrossref