November 2016, VOLUME138 /ISSUE 5

Third Trimester Brain Growth in Preterm Infants Compared With In Utero Healthy Fetuses

  1. Marine Bouyssi-Kobar, MSa,b,
  2. Adré J. du Plessis, MDc,
  3. Robert McCarter, PhDd,
  4. Marie Brossard-Racine, PhDe,
  5. Jonathan Murnick, MD, PhDa,
  6. Laura Tinkleman, BSa,
  7. Richard L. Robertson, MDf, and
  8. Catherine Limperopoulos, PhDa
  1. aThe Developing Brain Research Laboratory, Departments of Diagnostic Imaging and Radiology,
  2. cFetal Medicine Institute, and
  3. dDepartment of Epidemiology and Biostatistics, Children’s National Health System, Washington, District of Columbia;
  4. bInstitute for Biomedical Sciences, George Washington University, Washington, District of Columbia;
  5. eDepartment of Pediatrics Neurology, Montreal Children’s Hospital–McGill University Health Center, Montreal, Quebec, Canada; and
  6. fDepartment of Radiology, Children’s Hospital Boston/Harvard Medical School, Boston, Massachusetts
  1. Ms Bouyssi-Kobar coordinated and supervised data collection at 1 site, conducted the MRI processing of the data, analyzed the results, and drafted the initial manuscript; Dr du Plessis co-conceptualized and designed the study, and critically reviewed the manuscript; Dr McCarter determined and performed the statistical analysis and contributed to the writing of the manuscript; Dr Brossard-Racine coordinated and supervised data collection and MRI studies at 1 site, contributed to the analysis of the data, and critically reviewed the manuscript; Drs Murnick and Robertson reviewed the MRI studies at 1 site and critically reviewed the manuscript; Ms Tinkleman contributed to the analysis of the MRI data and critically reviewed the manuscript; Dr Limperopoulos co-conceptualized and designed the study, coordinated and supervised the progress of the study at the 2 sites, analyzed the results, and contributed to the writing of the manuscript; and all authors approved the final manuscript as submitted.


BACKGROUND AND OBJECTIVES: Compared with term infants, preterm infants have impaired brain development at term-equivalent age, even in the absence of structural brain injury. However, details regarding the onset and progression of impaired preterm brain development over the third trimester are unknown. Our primary objective was to compare third-trimester brain volumes and brain growth trajectories in ex utero preterm infants without structural brain injury and in healthy in utero fetuses. As a secondary objective, we examined risk factors associated with brain volumes in preterm infants over the third-trimester postconception.

METHODS: Preterm infants born before 32 weeks of gestational age (GA) and weighing <1500 g with no evidence of structural brain injury on conventional MRI and healthy pregnant women were prospectively recruited. Anatomic T2-weighted brain images of preterm infants and healthy fetuses were parcellated into the following regions: cerebrum, cerebellum, brainstem, and intracranial cavity.

RESULTS: We studied 205 participants (75 preterm infants and 130 healthy control fetuses) between 27 and 39 weeks’ GA. Third-trimester brain volumes were reduced and brain growth trajectories were slower in the ex utero preterm group compared with the in utero healthy fetuses in the cerebrum, cerebellum, brainstem, and intracranial cavity. Clinical risk factors associated with reduced brain volumes included dexamethasone treatment, the presence of extra-axial blood on brain MRI, confirmed sepsis, and duration of oxygen support.

CONCLUSIONS: These preterm infants exhibited impaired third-trimester global and regional brain growth in the absence of cerebral/cerebellar parenchymal injury detected by using conventional MRI.

  • Abbreviations:
    gestational age
    intracranial cavity volume
    intraventricular hemorrhage
    patent ductus arteriosus
    total brain volume
    term-equivalent age
  • What’s Known on This Subject:

    Third-trimester brain development is characterized by critical and energy-dependent biological processes needed to support optimal brain growth. Available evidence points to disturbed brain development in ex-preterm infants at term and beyond compared with their term-born peers.

    What This Study Adds:

    Third-trimester brain growth is disturbed in preterm infants compared with healthy in utero fetuses in the absence of structural brain injury. Neonatal intensive care likely influences third-trimester brain development in preterm infants.

    Despite marked improvements in neonatal intensive care, life-long neurodevelopmental disabilities remain highly prevalent in survivors of prematurity.1 Brain injury is a common complication of preterm birth and is diagnosed by using conventional neuroimaging techniques in about one-third of preterm infants.2 However, the presence of structural brain injury alone does not account for the frequency and scope of neurodevelopmental impairments in surviving preterm infants.3 Brain development can be altered even without evidence of destructive prematurity-related brain injury.4 Multiple risk factors have also been implicated with aberrant preterm brain maturation, including infection, medication administration, and respiratory complications.5 However, the timing, progression, and functional impact of delayed brain maturation in the absence of structural brain injury remain poorly understood.

    A growing body of evidence has shown that brain development is impaired in preterm infants by term-equivalent age (TEA) even in the absence of structural brain injury: compared with healthy term newborns, preterm infants at TEA exhibit decreased cerebral and cerebellar volume, altered cortical surface area,69 and microstructural organization,10 as well as impaired functional connectivity.11,12 To date, most inferences about third-trimester brain development in preterm infants have been based on comparative quantitative MRI measures versus those of term healthy newborns at a single time point at the end of the third trimester (ie, TEA). However, these cross-sectional studies at the end of the third trimester were not designed to identify the onset and evolution of impaired brain development in the premature infant. Few studies have investigated the third-trimester brain maturation in the absence of brain injury after early exposure to extrauterine life.1317 Moreover, another substantial obstacle to progress has been the scarcity of in utero normative fetal MRI studies from which to establish and compare departures from normal third-trimester ex utero brain development.

    Recently, we and others have successfully applied advanced quantitative MRI techniques to study in utero brain development in healthy fetuses.1822 These studies have provided major insights into the rate and progression of normal second- and third-trimester global, regional, and tissue-specific brain development in utero. However, no study to date has prospectively compared in utero brain volume in healthy fetuses versus that of ex utero preterm infants. The aim of the present study was to compare third-trimester global and regional brain volumes and brain growth trajectories in very preterm infants with no parenchymal brain injury versus healthy in utero fetuses by using 3-dimensional volumetric MRI measures. As a secondary objective, we sought to examine the relationship between clinical risk factors and brain volumes in preterm infants.



    We studied preterm infants and in utero healthy control fetuses recruited prospectively from longitudinal observational studies performed at 2 medical centers: Children’s Boston Hospital (Boston, MA) and Children’s National Health System (Washington, DC). The design of the study was cross-sectional, and it included only a single observation for each participant.

    Preterm Cohort

    Very preterm infants born before 32 weeks’ gestational age (GA) weighing <1500 g were prospectively enrolled. We specifically excluded any preterm infants with known or suspected chromosomal anomalies, congenital malformations, central nervous system infection, and metabolic disorders. Also excluded were preterm infants with any evidence of parenchymal cerebral or cerebellar injury (ie, white matter injury, grade III–IV intraventricular hemorrhage [IVH], cerebellar hemorrhage) according to conventional MRI.

    Fetal Cohort

    Healthy pregnant women with normal fetal ultrasounds were recruited as normal control subjects in a prospective study comparing brain development in fetuses with congenital heart disease.2325 We excluded multiple pregnancies, chromosomal abnormalities, and congenital infection.

    The studies were approved by the 2 institutional review boards, and written informed consent was obtained from each participant.

    MRI Acquisition

    Preterm Cohort

    Preterm newborns were scanned under natural sleep by using either a 1.5-T MRI scanner Signa Excite, a 1.5-T Discovery MR450 scanner, or a 3-T Discovery MR750 (all: GE Healthcare, Milwaukee, MI). Preterm infants requiring temperature monitoring underwent scanning by using a Nomag MRI-compatible incubator (LMT Medical Systems GmbH, Luebeck, Germany).

    Fetal Cohort

    Fetal MRI studies were performed at 1.5 T either on an Achieva scanner (Philips Medical System, Best, Netherlands) or on a Discovery MR450 scanner (GE Healthcare).

    For both cohorts, comparable sagittal, axial, and coronal T2-weighted images were collected by using single-shot fast spin echo sequences (2-mm slice thickness, 0-mm gap). An experienced pediatric neuroradiologist at each site reviewed all the MRI studies and noted the presence or absence of extra-axial blood.

    Volumetric Analyses

    After visual quality inspection of the anatomic images acquired, images without motion artifacts were selected and corrected for intensity nonhomogeneity.26 For both cohorts, we reconstructed 3-dimensional images of the brain from the 2-dimensional images acquired27 and used a parcellation pipeline18 to delineate the brain regions of interest. The cerebrum, cerebellum, and brainstem were then manually corrected when necessary by using ITK-SNAP software.28 Figure 1 provides an example of the brain parcellation performed in preterm infants and in utero fetuses. Volumes were computed by multiplying the number of voxels included in each brain region by the volume of each voxel and are expressed in cubic centimeters (milliliters). Total brain volume (TBV) was defined as the sum of cerebral, cerebellar, and brainstem volumes; the intracranial cavity volume (ICV) was the sum of TBV and cerebrospinal fluid.

    FIGURE 1

    Parcellation of preterm (top row) and fetal (lower row) brain. Anatomic T2-weighted images and corresponding parcellation maps in a preterm infant at 31 4/7 weeks GA (top row) and a fetus at 32 2/7 weeks’ GA (lower row). Dark gray: cerebrum; intermediate gray: cerebellum; light gray: brainstem; and white: cerebrospinal fluid.

    Prenatal and Postnatal Clinical Data Collection

    Prenatal information, mode of delivery, birth weight, GA at birth, Apgar score, and sex were collected for both cohorts. Additional clinical risk factors abstracted in the preterm cohort included length of mechanical ventilation and oxygen support, need for patent ductus arteriosus (PDA) ligation, sepsis (ie, confirmed sepsis by positive blood culture), postnatal dexamethasone treatment, and the presence of extra-axial blood on conventional MRI.

    Statistical Analysis

    Descriptive statistics were summarized by using median, range, means, and SDs for continuous variables and proportions for categorical factors. Linear regression models were developed by using Stata 13 (Stata Corp, College Station, TX) to evaluate the relationship between global and regional measurements of brain volume according to GA at MRI in healthy fetuses and preterm newborns. During model development, we evaluated the need for cross-product terms to allow for differing growth trajectories in the 2 groups and for higher order effects to introduce curvilinearity if necessary to improve model fit.

    Within the preterm cohort, linear regression analysis was first used to evaluate the relationship between the presence/level of each clinical risk factor individually (ie, chorioamnionitis, cesarean delivery, PDA ligation, sepsis, dexamethasone treatment, number of days ventilated /on oxygen support, presence of extra-axial blood) and brain volume controlling for GA at birth, day of life at MRI, and sex. We chose to control for GA at birth rather than birth weight in the models because the 2 factors were highly correlated (r = 0.74) and because GA was judged less likely to interfere with risk factor identification. Results were expressed as regression (slope) coefficients with 95% confidence intervals (CIs). We then performed a multiple regression analysis that evaluated the combined main and interactive effects of each of the clinical risk factors associated with differences in global or regional brain volumes. Only those 2- and 3-way interactions that achieved at least a borderline level of statistical significance (P < .12) were retained in final models; the decision on which of the alternative models to use to represent the combined impact of risk factors on brain growth was aided by the use of Akaike information criterion statistics.29,30 We consider these analyses hypothesis-generating and thus did not correct for multiple comparisons, and we were more lenient in choosing terms to include in the model to avoid missing the identification of potentially important risk factors.


    Clinical Characteristics of the Cohort

    A total of 205 subjects were studied: 75 very preterm infants (birth GA: 22–32 weeks) with a mean corrected GA at MRI of 34 ± 2.5 weeks (range: 27–39 weeks) and 130 healthy fetuses with a mean GA at MRI of 32 ± 3.3 weeks (range: 27–39 weeks). All MRI scans in the preterm and fetal cohorts were structurally normal (ie, had no parenchymal brain injury/malformations). The clinical characteristics of the cohort are summarized in Table 1.

    TABLE 1

    Clinical Characteristics of the Cohort

    Comparisons of Brain Volumes by Site

    Table 2 summarizes the brain volumes of the preterm and fetal cohorts by site; the ratio of preterm infants to fetuses was similar at each site. Notably, the few significant differences that were present in the preterm and fetal cohorts at a young age (<33 weeks’ GA) disappeared after controlling for GA at MRI.

    TABLE 2

    Brain Volumes of the Preterm and Fetal Cohorts According to Site and Gestational Age

    Comparisons Between Fetal and Preterm Brain Volumes

    Controlling for GA at MRI and sex, brain volumes were lower in preterm infants compared with in utero fetuses for all the brain regions examined (P < .05) (Fig 2). Third-trimester brain growth trajectories by GA at MRI were slower in the ex utero preterm cohort compared with the fetal cohort for the cerebrum (P = .001), the cerebellum (P = .05), the brainstem (P = .06), TBV (P = .002), and ICV (P = .03).

    FIGURE 2

    Brain volumes plotted against GA at MRI in the preterm and fetal cohorts (controlling for sex).

    Preterm Infants: Clinical Risk Factors and Impaired Brain Volume

    The clinical risk factors were similar in the 2 preterm cohorts except for length of ventilation (Table 3). The statistically significant relationships between individual clinical risk factors and regional brain volume when controlling for GA at birth, day of life at MRI, and sex are summarized in Table 4. Dexamethasone treatment, the presence of extra-axial blood, sepsis, and length of oxygen support were associated with reduced brain volumes. We then included the aforementioned risk factors in a multiple regression model analysis (Table 5). The 3-way interaction effect among dexamethasone treatment, extra-axial blood, and length of oxygen support explained >75% of the reduction in cerebral/cerebellar volumes, TBV, and ICV. In addition, sepsis was significantly associated with reductions in cerebral volume, TBV, and ICV.

    TABLE 3

    Clinical Risk Factors

    TABLE 4

    Relationship Between Brain Volumes and Individual Clinical Risk Factors in the Preterm Cohort (n = 75)

    TABLE 5

    Relationship Between Brain Volumes and Combined Clinical Risk Factors in the Preterm Cohort (n = 75)


    To our knowledge, our study reports for the first time decreased third-trimester global and regional brain growth in very preterm infants without evidence of cerebral or cerebellar parenchymal injury compared with healthy in utero control fetuses. Using volumetric MRI, we found differences in brain growth trajectories between preterm infants who experience their “third trimester” of development ex utero and in utero healthy fetuses. Our findings suggest that even in the absence of structural brain injury, the developmental trajectory of the preterm brain is altered over the third trimester. Evidence of clinical risk factors associated with reduced brain volumes in preterm infants included individual and synergistic effects of dexamethasone treatment, the presence of extra-axial blood, the length of oxygen support, and confirmed sepsis.

    Previous studies suggest that brain volumes in preterm survivors are altered by TEA compared with healthy term newborns and are characterized by decreased parenchymal volumes.9,3135 Alterations in brain volumes at TEA in the preterm infants have also been correlated with worse neurocognitive outcomes.3638 Moreover, prematurely born children and adults exhibit altered regional volumetric brain growth compared with their term-born peers39,40 suggesting long-lasting consequences of early-life disturbances in brain growth.41 However, to date, few studies have investigated third-trimester volumetric brain development in preterm infants and healthy in utero fetuses. Only 1 recent study (by Lefèvre et al42) used a retrospective cross-sectional design to compare 27 preterm infants (birth GA: 25–35 weeks) and 14 fetuses; the investigators found that brain volumes were similar, whereas cortical folding trajectories were altered in preterm infants compared with healthy control fetuses. Our data showed reductions in brain volumes across the third trimester of ex utero life in preterm infants compared with the in utero healthy control fetuses. Our larger sample size (N = 205) suggest that we were better-powered to detect subtle but important differences in regional brain growth. The present results extend earlier findings of altered brain growth in preterm infants by TEA9,32 and demonstrate a progressive fall-off/dysmaturation of the brain after preterm birth over the third trimester.

    The third trimester of gestation is a critical period of prolific brain development. Multiple biological processes are actively taking place, including the beginning of myelination, neuronal organization, spinogenesis, and synaptogenesis.43 It is well established that there is a fourfold increase in brain size during the third trimester of gestation18 accompanied by a dramatic increase in brain surface area with the formation of tertiary sulci and gyri.44 This well-orchestrated and precisely timed development of the neural circuitry may be vulnerable to insults associated with preterm birth. Longitudinal studies have explored the relationship between clinical risk factors and neurodevelopmental outcomes in preterm born infants, including postnatal infections,4547 medications,48,49 and bronchopulmonary dysplasia.50 At TEA, several clinical risk factors have been associated with decreased brain volume, including white matter brain injury,31,34,35 respiratory illness,31,34 small birth weight for GA,34,35 increasing prematurity,9 and low-grade IVH.9,51 Our study extends these observations and shows that among the clinical risk factors examined, postnatal dexamethasone treatment, the presence of extra-axial blood, respiratory illness, and sepsis were inversely associated with brain growth in preterm infants during the third trimester.

    We report an association between dexamethasone treatment and impaired cerebral, cerebellar, and brainstem growth during the third trimester. Our findings are in keeping with previously published research suggesting a detrimental effect of dexamethasone treatment on preterm brain volumes at TEA52,53 and during adolescence.54 Animal models have shown that dexamethasone administration during the critical developmental period is associated with abnormal apoptosis and alterations in synaptic function.55,56 Respiratory distress in the preterm infants results in alterations in cerebral hemodynamics that can also lead to impaired brain development.57 In our study, indicators of respiratory illness such as the length of oxygen support were inversely associated with third-trimester brain volumes in preterm infants, consistent with previous findings reported at TEA.9,31,34

    Prenatal, intrapartum, and/or postnatal infections result in inflammatory substances that have been linked to both brain injury and altered brain development in the preterm population.58,59 In this study, we also report an association between confirmed sepsis and stunted cerebral growth. This specific finding contrasts with 2 other studies, which found no relationship between brain volumes and sepsis in preterm infants at TEA.9,31 The severity, time, and occurrence of sepsis might play a role in the varied findings. Longitudinal data sets will allow elucidating whether there is a catch-up in brain growth by TEA in preterm infants with confirmed sepsis. A large, longitudinal population-based study has shown a negative independent effect of sepsis on neurodevelopmental impairments at 2 years of age.46 Conversely, we found no relationship between chorioamnionitis and third-trimester brain volume, which is in agreement with previous studies.35,60 Chorioamnionitis has been associated with an increased risk for developing brain injury,61,62 and it has been especially correlated with severe IVH.63,64 The direct influence of chorioamnionitis on neurodevelopmental outcome is unclear, and it is likely mediated by the increased risk of postnatal morbidities associated with prenatal infection, including brain injury and neonatal sepsis.61,6567

    The presence of extra-axial blood in our preterm cohort was linked to decreased brain volumes. Low-grade IVH has previously been related to decreased brain volumes in near-term and preterm infants at TEA.9,51 Potential mechanisms include free radical injury mediated by extra-axial blood and IVH, which adversely affects the proliferating cells and leads to subsequent downstream events in cerebral and cerebellar development.6870 Additional studies are needed to quantify the effects of blood on the developing preterm brain.

    Surprisingly, the need for surgical PDA ligation was not associated with impaired brain volumes over the third trimester in the ex utero preterm infants in our study. This finding conflicts with previous literature in which PDA ligation was associated with reduced brain volumes at term age,8,9 but it is in line with a recently published study. Lemmers et al71 postulated that the duration of reduced cerebral oxygenation associated with PDA is likely the main factor responsible for brain growth impairment associated with PDA ligation rather than the need for surgery itself. Ongoing research is warranted to address this intriguing question.

    Our study provides new insights into the role of prematurity-related clinical risk factors on brain maturation during the third trimester in ex utero preterm infants with no evidence of parenchymal brain injury. Ongoing prospective studies are urgently needed to better ascertain what aspects of neonatal intensive care may be harmful versus protective for the developing preterm brain.72

    Our study has several limitations. First, the design of the study was cross-sectional; thus, the brain growth trajectories examined are population based (ie, the trajectories are computed from a cohort with a single observation rather than from individuals with longitudinal MRI data). Serial MRI scans would allow for better delineation of third-trimester longitudinal brain growth in normal in utero versus ex utero environments, and study of this topic is currently underway. Second, we combined participants from 2 regionally distinct cohorts, introducing differences in the MRI scanners used and variability in clinical care. However, all participants were scanned with comparable T2-weighted anatomic sequences, and we found no significant difference after controlling for GA at MRI in brain volumes in both fetal and preterm cohorts performed at the 2 sites. Furthermore, the clinical risk factors were similar in the 2 preterm cohorts. The third limitation relates to our sample size that restricted the number of clinical risk factors we could investigate. We decided to include in our multivariate analysis only the clinical risk factors that were correlated with brain volumes on our univariate analysis. Other statistical approaches could lead to different models; consequently, our results are only hypothesis-generating and need to be evaluated and confirmed by other studies. Moreover, our findings are not brain tissue specific. A segmentation pipeline applied to this data set could provide additional information about different tissue types (ie, cortical gray versus white matter) that may be preferentially affected by preterm birth. Finally, although we specifically excluded preterm infants with evidence of structural brain injury on conventional MRI, it is possible that smaller brain lesions below the current resolution of clinical MRI field strength were missed.73 Specifically, we cannot exclude the presence of white matter microscopic necrosis, which can only be resolved by using a higher MRI field strength (eg, 12 T).74


    Our data strongly support the notion that preterm delivery is associated with disturbances in third-trimester global and regional brain growth even in the absence of parenchymal brain injury. Clinical risk factors associated with preterm birth and impaired brain development include the use of steroids, greater respiratory illness, sepsis, and the presence of extra-axial hemorrhage. Ongoing studies are needed to clarify the effects of altered third-trimester regional brain growth on neurodevelopmental outcome.


    The authors are grateful to the families who participated in the study.


      • Accepted August 23, 2016.
    • Address correspondence to Catherine Limperopoulos, PhD, Developing Brain Research Laboratory, Departments of Diagnostic Imaging and Radiology, Children’s National Health System, 111 Michigan Ave Northwest, Washington, DC 20010. E-mail: climpero{at}
    • FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

    • FUNDING: This work was supported by the Canadian Institutes of Health Research (MOP-81116), the SickKids Foundation (XG 06-069), and the National Institutes of Health (UL1TR000075 and P30 HD040677). Funded by the National Institutes of Health (NIH).

    • POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.