Research ArticlePreterm Birth

Slower Postnatal Growth Is Associated with Delayed Cerebral Cortical Maturation in Preterm Newborns

See allHide authors and affiliations

Science Translational Medicine  16 Jan 2013:
Vol. 5, Issue 168, pp. 168ra8
DOI: 10.1126/scitranslmed.3004666

Abstract

Slower postnatal growth is an important predictor of adverse neurodevelopmental outcomes in infants born preterm. However, the relationship between postnatal growth and cortical development remains largely unknown. Therefore, we examined the association between neonatal growth and diffusion tensor imaging measures of microstructural cortical development in infants born very preterm. Participants were 95 neonates born between 24 and 32 weeks gestational age studied twice with diffusion tensor imaging: scan 1 at a median of 32.1 weeks (interquartile range, 30.4 to 33.6) and scan 2 at a median of 40.3 weeks (interquartile range, 38.7 to 42.7). Fractional anisotropy and eigenvalues were recorded from 15 anatomically defined cortical regions. Weight, head circumference, and length were recorded at birth and at the time of each scan. Growth between scans was examined in relation to diffusion tensor imaging measures at scans 1 and 2, accounting for gestational age, birth weight, sex, postmenstrual age, known brain injury (white matter injury, intraventricular hemorrhage, and cerebellar hemorrhage), and neonatal illness (patent ductus arteriosus, days intubated, infection, and necrotizing enterocolitis). Impaired weight, length, and head growth were associated with delayed microstructural development of the cortical gray matter (fractional anisotropy: P < 0.001), but not white matter (fractional anisotropy: P = 0.529), after accounting for prenatal growth, neonatal illness, and brain injury. Avoiding growth impairment during neonatal care may allow cortical development to proceed optimally and, ultimately, may provide an opportunity to reduce neurological disabilities related to preterm birth.

Introduction

Survival rates of very preterm infants (≤32 weeks gestational age) have risen markedly owing to advances in obstetrical and neonatal intensive care, but these improvements have not been accompanied by a reduction in long-term morbidity in this population (13). Lower gestational age and birth weight increase the risk for neonatal comorbidities (for example, infection and respiratory complications) (4), and these in turn are associated with adverse white matter development (5).

Intrauterine growth restriction (IUGR) refers to infants whose birth weights are <10th percentile because of growth failure in utero. Premature IUGR newborns demonstrate a pattern of discordant gyrification (6) and reduced cortical volumes (7, 8) when compared to preterm infants born an appropriate weight for gestational age (10th to 90th percentile) and/or full-term controls. Abnormal cortical volumes in premature IUGR infants and in experimental models have been associated with poorer neurodevelopmental outcome (9, 10). Although growth in utero appears to be important for brain development, the majority of premature newborns are not born IUGR (1113). Therefore, IUGR cannot fully account for the extent of abnormal neurodevelopment and brain (gray and white matter) injuries observable on magnetic resonance imaging (MRI) within the preterm population (1416). However, many premature newborns develop persistent growth deficits postnatally, such that by discharge from the neonatal intensive care unit (NICU), the majority of preterm infants are considered growth-restricted, that is, <10th percentile for their postmenstrual age (12, 13). Postnatal growth failure in the NICU is associated with increased incidence of cerebral palsy and neurodevelopmental impairment, after accounting for prenatal growth, systemic illness, and brain injury (17). Additionally, the rate of change in cortical surface area between 24 and 44 weeks postmenstrual age predicts cognitive ability at 2 and 6 years corrected age in children born very preterm (18). At age 7 years, children born preterm demonstrate altered cortical connectivity (19) and synchronization during cognitive tasks relative to full-term controls, even in the absence of major disability (20). However, the etiology for their altered cortical development and processing remains unknown. Given this, we set out to examine the extent to which poorer postnatal growth in the NICU relates to diffusion tensor imaging measures of microstructural development of the cerebral cortex in infants born very preterm.

Diffusion tensor imaging, an MRI technique, can be used to characterize the spatial distribution of water diffusion in each voxel (three-dimensional pixel) of the image, providing a measure of regional brain microstructural development (21). In the cerebral cortex, fractional anisotropy (FA), a measure of the directionality of water diffusion, decreases between 25 and 40 weeks postmenstrual age, coincident with the disappearance of the radial glia and increasing complexity of the developing cortex (2228). In contrast, in the white matter, FA increases with maturation, coincident with maturation of the oligodendrocyte lineage and early events of myelination (5, 29).

Systemic illness and medical interventions are important determinants of brain injury and abnormal brain development (5, 30, 31). Moreover, focal brain injuries have been also found to affect overall brain development (3234), leading to moderate to severe neurodevelopmental disability (14, 15, 35). Therefore, it is important to consider multiple medical confounders in evaluating the relationship between neonatal growth and cortical development in infants born very preterm.

This study examines whether neonatal growth is related to microstructural development of the cerebral cortex in infants born very preterm. We hypothesized that poorer growth in the NICU would be associated with delayed cortical maturation, independent of prenatal growth, systemic illness, and brain injury.

Results

As in previous studies, infants from this cohort were excluded if they had a congenital malformation or syndrome, antenatal infection, or evidence on ultrasound of a parenchymal hemorrhagic infarction >2 cm (33, 36, 37). The characteristics of the 98 included infants are provided in Table 1 and detailed in table S1. Ninety-five newborns had diffusion tensor images of sufficient quality for cortical analyses. Diffusion tensor imaging parameters of FA and eigenvalues (λ1, λ2, and λ3) were collected across eight regions of interest identified in the cerebral cortex (precentral gyrus, postcentral gyrus, secondary somatosensory cortex, superior frontal gyrus, dorsolateral prefrontal cortex, ventrolateral prefrontal cortex, anterior insula, and occipital gray matter; Fig. 1). In univariate unadjusted analyses, infants with poor postnatal weight gain (n = 27) appeared to have higher cortical gray matter FA values at scan 2 (MRI at median 40.3 weeks postmenstrual age) compared to infants with appropriate weight for their postmenstrual age and sex in the NICU, but the difference was not statistically significant [ventrolateral cortex: 95% confidence interval (CI), −0.002 to 0.04; P = 0.069; fig. S1]. The magnitude of this difference was more pronounced and reached statistical significance when infants born small for their gestational age and sex (n = 19) were excluded (ventrolateral cortex: 95% CI, 0.002 to 0.05; P = 0.032; across all regions of interest: 95% CI, 0.001 to 0.02; P = 0.036). Infants born small for their gestational age and sex were therefore included in the longitudinal multivariable models to provide a more conservative estimate of the difference in cortical diffusion tensor imaging parameters related to postnatal growth restriction.

Table 1

Infant characteristics. Scan 1, MRI at ~32 weeks postmenstrual age; Scan 2, MRI at ~40 weeks postmenstrual age; IQR, interquartile range; PVHI, periventricular hemorrhagic infarction.

View this table:
Fig. 1

Description of the regions of interest obtained within the cortical gray matter. (A and B) Diffusion tensor image-encoded anisotropy color maps of an infant born at 26.29 weeks gestation and scanned at 30 weeks postmenstrual age. The images demonstrate the relatively high FA of the cerebral cortex typical for this age. The color convention used to display the predominant diffusion direction has red representing right-left, green representing anterior-posterior, and blue representing superior-inferior anatomical directions (56, 58). Eight cerebral cortical regions of interest were examined, and values of each region were averaged bilaterally: (a) precentral gyrus, (b) postcentral gyrus, (c) secondary somatosensory cortex, (d) superior frontal gyrus, (e) dorsolateral prefrontal cortex, (f) ventrolateral prefrontal cortex, (g) anterior insula, and (h) occipital gray matter.

Weight change in relation to diffusion parameters of the cortical gray matter

Longitudinal models revealed that lower gestational age (effect size, −0.038; SE, 0.011; P < 0.001), birth weight (effect size, <−0.001; SE, <0.001; P = 0.016), and slower weight gain [weight at scan 2 (MRI at ~40 weeks postmenstrual age) − weight at scan 1 (MRI at ~32 weeks postmenstrual age)] (effect size, −0.410; SE, 0.089; P < 0.001) were independently associated with higher FA values in the cortical gray matter, after adjusting for sex, brain injury [white matter injury, intraventricular hemorrhage, and cerebellar hemorrhage (brain injury model)], systemic illness [patent ductus arteriosus, days intubated, postnatal infection, and necrotizing enterocolitis (extended model)], and age at scan (Table 2). Therefore, neonatal growth was associated with cortical gray matter maturation in the NICU, independent of birth weight, brain injury, and systemic illness. Change in FA reflected changes in the radial diffusion axes (λ2 and λ3; Table 3), but not the axial diffusion axis (λ1; Table 4), suggesting a delay in neuronal process formation and/or apoptosis in the cerebral cortices of infants who are born very preterm and have impaired growth.

Table 2

Weight change in relation to mean FA values of eight regions of interest in the cortical gray matter. Weight change = weight at scan 2 (MRI at ~40 weeks postmenstrual age) − weight at scan 1 (MRI at ~32 weeks postmenstrual age).

View this table:
Table 3

Weight change in relation to mean λ2 and λ3 values of eight regions of interest in the cortical gray matter. λ2 and λ3, radial diffusion axes.

View this table:
Table 4

Weight change in relation to mean λ1 values of eight cortical regions of interest in the cortical gray matter. λ1, axial diffusion axis.

View this table:

Weight change in relation to FA of the white matter

Weight change was not significantly associated with FA values in the white matter in the basic statistical model (effect size, −0.035; SE, 0.055; P = 0.529; table S2). Therefore, white matter maturation appears to be relatively spared from the effects of postnatal growth restriction. Rather, postnatal infection (effect size, −0.057; SE, 0.020; P = 0.005) was independently associated with lower FA values in the white matter after adjusting for gestational age, birth weight, sex, brain injury, systemic illness, weight change, and age at scan.

Weight change in relation to diffusion parameters of the cortical gray matter excluding infants who received postnatal corticosteroids

Neither dexamethasone (effect size, −167.044; SE, 223.072; P = 0.454) nor hydrocortisone (effect size, −341.346; SE, 245.124; P = 0.164) was associated with weight change after adjusting for gestational age, birth weight, sex, brain injury, systemic illness, and age at scan. Nonetheless, as a sensitivity analysis, we examined weight change in relation to FA of the cortical gray matter excluding infants who received postnatal corticosteroids. In newborns who did not receive corticosteroids postnatally, lower gestational age (effect size, −0.034; SE, 0.012; P = 0.005), birth weight (effect size, <−0.001; SE, <0.001; P = 0.009), and slower weight gain (effect size, −0.512; SE, 0.114; P < 0.001) between scan 1 and scan 2 were independently associated with higher FA values in the cortical gray matter, in longitudinal models adjusting for gestational age, sex, brain injury, systemic illness, and age at scan (table S3). Given that the relationship between weight change and FA values did not change meaningfully after the exclusion of infants who received postnatal corticosteroids (hydrocortisone and/or dexamethasone), exposed infants were included in all other longitudinal models.

Length change in relation to diffusion parameters of the cortical gray matter

Longitudinal models revealed that lower gestational age (effect size, −0.030; SE, 0.010; P = 0.002), confirmed necrotizing enterocolitis (effect size, 0.125; SE, 0.050; P = 0.012), and slower linear growth (effect size, −0.837; SE, 0.177; P < 0.001) between scan 1 and scan 2 were independently associated with higher FA values in the cortical gray matter, after adjusting for birth weight, sex, brain injury, systemic illness, and age at scan (Table 5). Change in FA reflected changes in the radial diffusion axes (λ2 and λ3: effect size, 0.189; SE, 0.073; P = 0.010) and not the axial diffusion axis (λ1: effect size, −0.043; SE, 0.062; P = 0.488).

Table 5

Length change in relation to mean FA values of eight regions of interest in the cortical gray matter. Length change = length at scan 2 (MRI at ~40 weeks postmenstrual age) − length at scan 1 (MRI at ~32 weeks postmenstrual age).

View this table:

Head circumference change in relation to diffusion parameters of the cortical gray matter

Longitudinal models revealed that lower gestational age (effect size, −0.030; SE, 0.010; P = 0.004) and slower head growth (effect size, −1.090; SE, 0.025, P < 0.001) between scan 1 and scan 2 were independently associated with higher FA values in the cortical gray matter, after adjusting for gestational age, birth weight, sex, brain injury, systemic illness, and age at scan (Table 6). Change in FA reflected change in the radial diffusion axes (λ2 and λ3: effect size, 0.265; SE, 0.098; P = 0.007) and not the axial diffusion axis (λ1: effect size, 0.058; SE, 0.086; P = 0.498). Results from these models are consistent with the models above examining the relationship between weight change and length with diffusion parameters, and therefore provide further support for the finding that neonatal growth over and above birth weight, brain injury, and systemic illness predicted cortical gray matter maturation in the NICU.

Table 6

Head circumference change in relation to mean FA values of eight regions of interest in the cortical gray matter. Head circumference change = head circumference at scan 2 (MRI at ~40 weeks postmenstrual age) − head circumference at scan 1 (MRI at ~32 weeks postmenstrual age).

View this table:

Discussion

This study examined whether neonatal growth is related to microstructural development of the cerebral cortex in infants born very preterm. We found that impaired neonatal growth (weight, length, and head circumference) was significantly associated with delayed cortical maturation. This independent association of postnatal growth with early brain development persisted even after controlling for gestational age, birth weight, sex, postmenstrual age, brain injury, and systemic illness. However, it was the cortical gray matter, rather than the white matter, which appeared to be most susceptible to impairments in postnatal growth. Similar to our previous studies (31, 33, 37), abnormal development of the white matter, on the other hand, was more strongly associated with postnatal infection.

Our finding, which demonstrates vulnerability of the cortical gray matter rather than white matter to the effects of neonatal growth, builds on results from previous studies examining the relationship between IUGR and cortical volumes/microstructure early in life (68). Animal models of IUGR have demonstrated a transient delay in oligodendrocyte maturation and myelination (38). Although markers of myelinating oligodendrocytes were reduced in utero, white matter volumes returned to control levels postnatally and persisted into adulthood (38). Thus, it has been suggested that the altered neurodevelopment associated with IUGR is likely not due to long-term deficits in myelination. Rather, it is the reduction of cerebral cortical volumes and microstructure associated with prenatal growth restriction that have been more pronounced, persistent, and associated with functional impairment (810).

Microstructural integrity of the cortical gray matter can be inferred from the diffusion parameters (22, 23, 3941). Between 25 and 40 weeks postmenstrual age, FA decreases as the developing cortex increases in complexity (22, 23) with arborization of the basal dendrites, formation of thalamocortical and cortical-cortical connections, and disappearance of the radial glia (2428). Therefore, given that higher FA was reflective of change in the radial diffusion axes (λ2 and λ3) between ~32 and 40 weeks postmenstrual age, there may be delayed expansion of neuronal process formations, synaptogenesis, and/or apoptosis in the cerebral cortices of infants who are born very preterm and have impaired growth. Similarly, in IUGR fetal sheep, the numerical densities of synapses in layer 1 of the visual cortex were reduced by 17% compared to controls (42). Moreover, fewer cells were found in the cortical plate and marginal zones for growth-restricted fetuses and infants compared to controls (43).

Lower gestational age and birth weight were associated with higher FA; however, the greater association with growth and cortical maturation appeared postnatally. This finding was supported by Smart et al., who used a rat model to demonstrate that nutritional deprivation during both gestation and neonatal periods had a greater influence on cortical development than deprivation during either one of these periods alone; however, the damage to the forebrain was largely determined by nutritional deprivation in the postnatal period (44). By 20 to 24 weeks gestational age, a large proportion of neurons have been produced in the ventricular and subventricular zones before their migration to the cortical plate (27). It has been well established that these precursor cells are vulnerable to nutrient insufficiency, which can lead to focal or widespread white matter injury, as well as reduced cortical gray matter (34). However, it is the nutritional demand of rapid brain growth, synaptogenesis, and sensory-driven activity between 24 to 42 weeks gestation that leaves the neonatal cortex particularly vulnerable to nutritional insult (45). Postnatal injuries have been shown to alter cortical maturation, synaptogenesis, and activity (4648). Moreover, studies of neonatal rats deprived of adequate postnatal nutrition have provided evidence for altered neuronal activity (4951). In support of these data, we also found that slower postnatal growth (which reflects many factors including neonatal nutrition) between 32 and 40 weeks postmenstrual age, independent of prenatal growth, was associated with impaired microstructural development of the cortical gray matter in infants born very preterm.

“Region of interest” based analysis of cortical diffusion tensor imaging can be limited by reproducibility of voxel sampling between scans, and by the risk of partial averaging. To improve accuracy of our measurements and replication between scans, we compared different sizes and positions for our region of interest voxel boxes, and determined the optimal size and placement for them. Future advances in MRI acquisition and analysis that allow for automatic segmentation and quantification of cortical FA from early in life to term-equivalent age may also refine our ability to detect differences in cortical maturation related to growth and illness.

We were able to collect detailed information regarding the infants’ clinical status and neonatal care, in addition to obtaining serial scans. Age at scan varied in cases where infants were unstable or were discharged before term-equivalent age. Our statistical models included terms for gestational age at birth and postmenstrual age at scan, accounting as best as possible for the timing of scans. This variability enabled us to examine changes in cortical FA across the age spectrum from relatively early in preterm life to term-equivalent age. We have found this approach to be robust when examining the impact of clinical conditions on MR metrics of brain development in this cohort and others [for example, (5, 33)].

We did not have data on neonatal nutrition, caloric intake, and feeding because the infants in this study were participants from a larger longitudinal study examining systemic illness in infants born very preterm (33, 37). The nutrition protocols in place during this study included both starting parenteral nutrition upon admission to the NICU and encouraging the use of breast milk and early trophic feeds. The standard fluid intake was 150 ml/kg per day, with a goal of 120 calories/kg per day. Weight was measured daily unless the patient was too unstable. Fluid and caloric intake was also assessed daily and adjusted to optimize nutrition and growth. Postnatal growth is affected by a multitude of factors, which include fluid management, nutritional and caloric intake, catabolic stressors associated with severity of illness, and endocrine, genetic, and environmental factors, including procedural pain (52). The consistency in findings across measures of weight, length, and head circumference supports the hypothesis that the alterations in cortical development reflect growth rather than fluid management. Our study was able to account for several medical confounders, which were likely to affect both growth and brain development, although residual confounding remains possible. These infants are currently being followed through childhood to determine whether neonatal growth–related delays in cortical maturation are associated with abnormal neurodevelopmental outcomes. Future studies are also needed to examine the specific roles of systemic illness and nutrition and to determine the optimal postnatal growth for cortical maturation in the NICU. Furthermore, additional research is needed to identify the mechanisms that may underlie the association between postnatal growth and cortical gray matter maturation in the NICU.

The results of this study have important clinical implications. Neonatal growth over and above birth weight, brain injury, and systemic illness correlated with cortical gray matter maturation in the NICU. Therefore, by diagnosing, treating, and preventing poor postnatal growth, clinicians may have the opportunity to optimize conditions for cortical development to proceed normally in infants born very preterm.

Materials and Methods

This study was approved by the University of British Columbia/Children’s and Women’s Health Centre of British Columbia Research Ethics Board.

Study cohort

Infants born very preterm (between 24 and 32 weeks gestational age) were admitted to the NICU at the British Columbia’s Women’s Hospital between March 2006 and January 2009. As in previous studies, infants from this cohort were excluded if they had a congenital malformation or syndrome, antenatal infection, or evidence on ultrasound of a parenchymal hemorrhagic infarction >2 cm (33, 36, 37). After parental informed consent was obtained, 98 infants were included in the present study. A neonatal research nurse performed medical and nursing chart review from birth to term-equivalent age or discharge (whichever came first). Data included but were not limited to gestational age, sex, birth weight, presence of patent ductus arteriosus, duration of intubation, infection, necrotizing enterocolitis, and corticosteroid (hydrocortisone and/or dexamethasone) exposure. Infants with clinical sepsis (who had negative cultures but were treated with antibiotics for ≥5 days) or with confirmed infections (positive cultures of the blood, urine, or cerebral spinal fluid, or ≥4 white blood cells found in tracheal aspirates associated with clinical pneumonia) were included in this study because these types of infections are associated with abnormal brain maturation (31). This approach is also consistent with the study by Stoll et al., which demonstrated that neonatal infections among extremely low birth weight infants are associated with poor neurodevelopmental outcome, even in the absence of positive cultures (53). Infants were classified as having necrotizing enterocolitis if they met either stage 2 (clinical signs and symptoms, and pneumatosis intestinalis on x-ray) or stage 3 (critically ill, clinical signs and symptoms, and pneumatosis intestinalis on x-ray) of Bell’s criteria (54). Infants were assessed for neonatal growth (weight, length, and head circumference) at the time of each MRI scan: median of 32 weeks (interquartile range, 30.4 to 33.6; total range, 27.3 to 40.7) and 40 weeks (interquartile range, 38.7 to 42.7; total range, 33.4 to 46.4) postmenstrual age.

Magnetic resonance imaging

Infants were scanned without pharmacological sedation when stable at median 32 (scan 1) and 40 weeks (scan 2) postmenstrual age. All newborns were scanned in an MRI-compatible isolette (Lammers Medical Technology) with a specialized neonatal head coil (Advanced Imaging Research). A Siemens 1.5-T Avanto magnet and VB 13A software were used to obtain the following sequences: three-dimensional coronal volumetric T1-weighted images (repetition time, 36; echo time, 9.2; field of view, 200 mm; slice thickness, 1 mm; no gap) and axial fast spin echo T2-weighted images (repetition time, 4610; echo time, 107; field of view, 160 mm; slice thickness, 4 mm; gap, 0.2 mm). Neuroradiologist K.J.P., blinded to infant medical history, assessed the images for cerebellar hemorrhage and the severity of white matter injury and intraventricular hemorrhage (14, 36). Twenty random scans were rescored; intraobserver reliability of κ > 0.9 was comparable with previous reported scores (14). In addition, K.J.P. identified seven subjects with white matter cysts typical of cystic periventricular leukomalacia on at least one imaging study. Three neonates had mild periventricular leukomalacia, with less than four cysts <2 mm in diameter; four neonates demonstrated cysts >1 cm in diameter. A variable identifying infants with cystic periventricular leukomalacia was not included in the statistical models, given the small number in each category. These infants were, however, included in this study and were identified as having moderate to severe white matter injury.

Diffusion tensor imaging

Diffusion tensor imaging was acquired with a multirepetition, single-shot echo planar sequence with 12 gradient directions (repetition time, 4900; echo time, 104; field of view, 160 mm; slice thickness, 3 mm; no gap), three averages of two diffusion weightings of 600 and 700 s/mm2 (b values), and an image without diffusion weighting, resulting in an in-plane resolution of 1.3 mm. Diffusion tensor imaging allows for the robust and noninvasive capture of maturation-dependent changes to water diffusivity in the cortex of premature newborns, reflecting cortical microstructure. Specifically, diffusion tensor imaging describes an ellipsoid space, where the size, shape, and orientation are given by eigenvalues (λ1, λ2, and λ3), and FA values, which reflect the variance of λ1, λ2, and λ3, and thereby describe overall directionality. λ1 corresponds to axial diffusion. This is the preferred diffusion direction because water preferentially diffuses along white matter tracks and radial glia of the developing cortical gray matter. In contrast, λ2 and λ3 correspond to radial diffusion. Between 32 and 40 weeks postmenstrual age, FA increases with cortical white matter maturation, coincident with changes in λ2 and λ3, reflecting maturation of the oligodendrocytes and myelination (55, 56). At the same time, FA decreases in the cortical gray matter, coincident with changes in the axial diffusion axis (λ1), reflecting neuronal maturation, synaptogenesis, and disappearance of the radial glial cells (22, 23).

Diffusion tensor imaging parameters of FA and λ1, λ2, and λ3 were collected bilaterally in 15 regions of interest by two observers. Eight regions of interest in the cortical gray matter were identified by J.V. in 95 neonates (precentral gyrus, postcentral gyrus, secondary somatosensory cortex, superior frontal gyrus, dorsolateral prefrontal cortex, ventrolateral prefrontal cortex, anterior insula, and occipital gray matter; Fig. 1), and seven regions of interest in the white matter were identified by V.C. in 97 neonates (anterior, middle, and posterior subcortical white matter, genu, and splenium of the corpus callosum; posterior limb of the internal capsule; and optic radiations), as described previously (33). Values from regions of interest on a degraded diffusion tensor image were not measured (7% of regions of interest).

Reliability of the cortical gray matter regions of interest

Many considerations were given to the size and placement of the region of interest voxel boxes in the cerebral cortex. First, it was important to determine the size of voxel box that could fit within the thin layer of cortical gray matter, which is about 2 mm thick in the newborn (6). We observed that 2 × 3 voxel boxes could fit within the boundaries of the cerebral cortex and surrounding structures.

Second, replication of the regions of interest over time was complicated by the fact that there is a marked change in the complexity of the cortex, with increasing sulcation and gyration between 32 and 40 weeks (32, 39). However, we found that 2 × 3 voxel boxes could be reliably placed at the height of the gyrus in the cortical gray matter for both the first and second scans. Intra-rater reliability was calculated on 20% of the regions of interest in the cortical gray matter, by Bland Altman analyses (57), and values were compared with those previously published in the literature (37): Scan 1 showed an FA mean difference of 0.001 (limits of agreement, −0.001 to 0.003), and scan 2 showed an FA mean difference of −0.002 (limits of agreement, −0.004 to 0.000).

Finally, reduction of cortical diffusion takes place according to an inside-out laminar gradient (40), thereby introducing the possibility for partial averaging within the measured regions of cortex. To address this issue, we considered whether the values of the top three voxels of the 2 × 3 voxel box compared favorably with the bottom three voxels of the 2 × 3 voxel box. FA mean differences for the top three and bottom three voxel boxes across all regions of interest were minor: At scan 1, the mean difference was 0.007 for all regions of interest, with a mean difference range of −0.010 to 0.020 across individual regions of interest; at scan 2, the mean difference was 0.006 for all regions of interest, with a mean difference range of −0.010 to 0.020 across individual regions of interest. Given that there were no systematic differences between using 1 × 3 versus 2 × 3 voxel boxes, and that use of 2 × 3 voxel boxes improved reliability, we proceeded to use 2 × 3 voxel boxes to extract data from the regions of interest within the cerebral cortex.

Reliability of the white matter regions of interest

On the basis of the repeated analysis of 20% of the regions of interest in the white matter (31) by Bland Altman analyses, intra-rater reliability was considered high: FA mean difference of 0.001 (limits of agreement, −0.018 to 0.017).

Data analyses

Statistical analysis was performed with R version 2.13 (R Development Core Team 2011). Normality plots were examined, and skewed variables [diffusion tensor imaging parameters (FA and λ1, λ2, and λ3)] and growth measures [change in weight (gram), head circumference (centimeter), and length (centimeter) between scan 1 and scan 2] were log-transformed. t tests were also used to examine whether there were differences in the cortical gray matter FA values for infants that were growth-restricted (<10th weight percentile) compared to those with an appropriate weight for their postmenstrual age and sex in the NICU at scan 1 and scan 2, and whether these differences were affected by the exclusion of infants born small for their gestational age and sex (<10th weight percentile at birth). Then, linear mixed effects models (LMEMs) were used to examine longitudinal associations between change in weight and diffusion tensor imaging parameters between scan 1 and scan 2 in the cortical gray matter and white matter. Included in the LMEMs were terms for multiple regions of interest (eight cortical gray matter regions or seven white matter regions) and interaction terms for region of interest and postmenstrual age. Splines (values produced by three smooth polynomial segments) were used to account for the nonlinearities between postmenstrual age and FA values (fig. S2), and growth over time relative to birth weight. The independent variables entered in the initial model were gestational age, birth weight, and sex. If weight change was a significant predictor of FA in the basic model (step 1), we extended the model to include brain injury (step 2: white matter injury, intraventricular hemorrhage, and cerebellar hemorrhage) and systemic illness (step 3: patent ductus arteriosus, days intubated, infection, and necrotizing enterocolitis). This model was then reapplied while excluding infants who had received postnatal corticosteroids. Steps 1, 2, and 3 were repeated to examine the relationship between weight change and radial (λ2 and λ3) and axial (λ1) diffusion axes. Moreover, if weight change was a significant predictor of FA, results were confirmed by repeating steps 1 to 3 for length change and head circumference change.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/5/168/168ra8/DC1

Fig. S1. Delayed cortical gray matter maturation in preterm infants with postnatal growth restriction.

Fig. S2. Nonlinear decrease in fractional anisotropy of the cortical gray matter with increasing postmenstrual age.

Table S1. Clinical and brain imaging data obtained from the study cohort (Excel file).

Table S2. Weight change in relation to mean fractional anisotropy values of seven regions of interest in the white matter.

Table S3. Weight change in relation to mean fractional anisotropy values of eight regions of interest in the cortical gray matter excluding infants who received postnatal corticosteroids.

References and Notes

  1. Acknowledgments: We thank the children and their parents who participated in this study, and P. McQuillen for reviewing the manuscript. Funding: This work was supported by the Canadian Institutes of Health Research (CIHR) (MOP79262 to S.P.M. and MOP86489 to R.E.G.). S.P.M. is currently Bloorview Children’s Hospital Chair in Paediatric Neuroscience and was supported by a Tier 2 Canadian Research Chair in Neonatal Neuroscience and the Michael Smith Foundation for Health Research Scholar Award. R.E.G. was supported by a Senior Scientist award from the Child and Family Research Institute (CFRI). CIHR Frederick Banting and Charles Best Canada Scholarship Masters & Doctoral Awards, Pain in Child Health (CIHR Strategic Training Initiative in Health Research) trainee support, and CFRI Graduate Studentship were awarded to J.V. Author contributions: S.P.M. and R.E.G. conceptualized and designed the study. K.J.P. and V.C. contributed to the acquisition of data. R.B. conducted the statistical analyses. J.V. acquired data and drafted the manuscript for review. All authors contributed to the interpretation of the data and provided critical review of the manuscript for publication. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article