Purpose To identify differences in neuronal tissue from retinal and brain structures in children born small for gestational age (SGA) with no abnormality in neonatal brain ultrasonography and no previous neurological impairment, and to evaluate the relationship between retinal structure and brain changes in school-age children born SGA.
Methods Two cohorts of children were recruited: 25 children born SGA and 25 children born with an appropriate birth weight according to gestational age. All the children underwent an ophthalmic examination, which included retinal imaging using spectral-domain optical coherence tomography, and a brain MRI. MRI images were automatically segmented and global and regional brain volumes were obtained.
Results Although visual function did not differ between both groups, the complex ganglion cell and inner plexiform layers (GCL-IPL) was thinner in SGA children. Total intracranial volume, and global grey and white matter volumes in brain and cerebellum were correlated with birthweight centile, as were certain regional volumes (temporal and parietal lobes, hippocampus and putamen). Abnormal GCL-IPL measurements accurately identified SGA children with the most severe grey and white matter changes in the brain.
Conclusions SGA children, both preterm and term born, showed evidence of structural abnormalities in the retina, which may be an accurate and non-invasive biomarker of neuronal damage in brain tissue.
- Child health (paediatrics)
- Optic Nerve
Statistics from Altmetric.com
During the last decades, optical coherence tomography (OCT) measurements have been proposed as an accurate biomarker of neuronal changes in the central nervous system (CNS) in certain neurodegenerative conditions, such as multiple sclerosis or Parkinson disease.1 ,2 It allows direct visualisation and quantification of neuronal tissue, with a fast, non-invasive and highly reproducible examination, even in paediatric populations.3
Children born small for gestational age (SGA) are at increased risk for adverse perinatal outcomes and suboptimal neurodevelopment compared with children born with an appropriate birth weight for their gestational age (AGA).4 Low birth weight is known to exert lasting effects on brain structure, not only in preterm born children but also in children born at term. Compelling evidence supports that children born SGA have poorer neurobehavioral outcomes, including learning, cognitive and attentional problems, than do AGA infants.5 Increasing research is focused on structural reorganisation processes underlying neurological impairments. SGA infants exhibit lower total intracranial volume, smaller cortical surface area, reduced global white and grey matter, and decreased volumes of certain regions, such as hippocampus or basal ganglia.6–8
Our research group first reported thinning in the retinal nerve fibre layer (RNFL) and impaired performance intelligence scores in school-age children born SGA due to exposition to chronic hypoxia and undernutrition in fetal life.9 ,10 It has recently been reported in early preterm infants that thinner RNFL may relate to brain structure and neurodevelopment.11 However, only global structural abnormalities were assessed in brain MRI, and RNFL was not compared with global or regional volumes of brain structures.
Children born SGA account for around 10% of all deliveries. Early and accurate biomarkers of suboptimal neurodevelopment are required to identify children at increased risk of neurological impairments among all the infants born SGA. The main purpose of this study was to evaluate the relationship between retinal structure and brain changes in children aged from 6 to 16 years born SGA. In addition, we aimed to identify differences in neuronal tissue from retinal and brain structures in SGA children with no abnormality in neonatal brain ultrasonography and no previous neurological diagnosis, such as cognitive or motor impairments.
Materials and methods
The study protocol was approved by the local ethics committee. Written informed consent was obtained from the parents or guardians of each child. Examiners were blinded to perinatal data and the study group. All procedures adhered to the tenets of the Declaration of Helsinki.
The study involved two cohort of children enrolled between January 2012 and April 2016. SGA cohort was defined by a birth weight below the 10th centile according to local standards.12 A cohort of children born with an appropriate birth weight (above the 10th centile) paired with the SGA cohort by age (<12 months of difference), and gestational age (GA) at birth was recruited. Demographic and paediatric information was collected from the medical records of the hospital including GA at birth, birth weight, gender, obstetrics characteristics, maternal smoking during pregnancy, perinatal outcomes and childhood diseases.
According to their GA at birth, children were classified as term born (GA ≥ 37 weeks of postmenstrual age) or preterm (GA<37 weeks). All early preterm children (<32 weeks) or with a birth weight under 1500 g received retinopathy of prematurity (ROP) screening starting at 4 weeks of age, which was repeated every 1–3 weeks until the retina was fully vascularised based on funduscopic results.13 Children with any stage of ROP were excluded from the study.
All the children were assessed by a paediatric neurologist before the inclusion in the study. Only children with a normal examination were proposed to participate. Subjects with a history of ocular diseases other than refractive errors, significant refractive errors (>5 dioptres of spherical equivalent refraction or 3 dioptres of astigmatism), genetic or chromosomal defects, congenital malformations, abnormalities in neonatal brain ultrasonography or any cognitive or motor impairment were excluded from the study.
All children underwent full ophthalmological assessments and retinal imaging with SD-OCT. SD-OCT scans were performed by one of two experienced examiners (IA or GG) using the Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, California, USA). The retina was imaged using the Optic disc cube 200×200 and the Macular cube 200×200 protocols. The Optic disc cube protocol analyses a 6 mm2 grid of 200 horizontal scan lines. The software algorithm automatically detects the optic disc centre from this volume scan, positions a 3.46 mm diameter calculation circle over this point and calculates the thickness. The Macular cube 200×200 protocol performs six consecutive macular radial scans, 6 mm in length, centred on the fovea. The images were analysed using the OCT3 mapping software. The ganglion cell analysis algorithm segmented a 14.13 mm2 elliptic annulus area centred on the fovea. An internal fixation target was used to maximise reproducibility. Images with a quality score lower than seven were rejected and repeated. No patient was excluded due to poor image quality. Only one eye per child was included in the analysis, which was selected by the best OCT scan or randomly chosen when both of them were optimal.
The following OCT measurements were analysed: average RNFL thickness, thickness in the four quadrants (superior, temporal, inferior and nasal), disc area, cup-disc ratio, average ganglion cell and inner plexiform layers (GCL-IPL) (which included the complex GCL-IPL) and minimum GCL-IPL. We classified outcomes as abnormal when they were below the 5th centile of our reference normal paediatric values for every measurement.14
MRI data acquisition and analysis
Three-dimensional (3D) high-resolution whole-brain gradient-echo T1-weighted images were obtained on a 1.5-T Optima 360 Advance clinical scanner (GE Healthcare Diagnostic Imaging, Milwaukee, Wisconsin, USA), using a 16-channel array head coil. For each subject, a 3-D high-resolution (1 mm isotropic voxels) structural MRI was acquired using T1-weighted volumetric spoiled gradient-recall echo sequence (repetition time 9.1 ms, echo time 1.7 ms, 20° flip angle, number of excitations 1, matrix size 256×256, slice thickness 1.5 mm, gap 0, yielding 124 transverse slices and voxel size=0.86×0.86×1.5 mm). After acquisition, all images were reviewed by one of the neuroradiologists (MAM and NF) and a computer engineer, who were blind to clinical subgroups in order to ensure data quality. No images were discarded after this check. Anonymised images were then transferred to a workstation for analysis and postprocessing.
All the MRI scans were reviewed by one of the neuroradiologists (MAM and NF). Children with any structural abnormality found in the visual inspection were removed from the study.
The T1-3D images were used for the assessment of the brain volumes. All images were preprocessed using a combination of two spatial filters for increasing signal-to-noise ratio and allowing for a semiautomatic quantitative volumetry. First, a non-local spatial filter was applied to minimise the random fluctuation of the MR signal due to the thermal noise. Additionally, the filtered images were also corrected for MRI signal inhomogeneities, which are normally present in the images and depend on the magnetic field homogeneity. This non-parametric method is based on the minimisation of an entropy-related function cost to produce a non-biased corrected image.
The Freesurfer software package (http://surfer.nmr.mgh.harvard.edu/) was used to perform the parcellation and the volumetric measurement of the different brain areas. The method included several steps to extract white and grey matter surfaces and to compute the volume of cortical and subcortical areas. The following areas were included in this study: total grey matter, total white matter, lateral ventricles, frontal lobe, parietal lobe, temporal lobe, occipital lobe, cingulum, thalamus, putamen, caudate nucleus, globus pallidus, hippocampus, cerebellum (grey and white matter) and amygdala. The anatomic accuracy of the grey and white matter parcellation was checked qualitatively by two trained engineers (GG-M and RS-R) in order to detect pitfalls or inconsistencies in this process.
Statistical analyses were performed using SPSS V.21.0 (SPSS, Chicago, Illinois, USA) statistical software. The Kolmogorov-Smirnov test was applied to assess sample distribution for all the quantitative outcomes. Demographic characteristics of the study groups were compared by means of Mann-Whitney U test for quantitative data and Pearson's χ2 or Fisher's exact test for qualitative ones, to ensure comparability of both samples.
First, univariate analyses were performed to examine the effect being born SGA on retinal measurements and brain volumes. Second, multiple linear regression modelling discerned the effect of the main perinatal determinants (GA at birth, birth weight and birthweight centile) adjusted for child's gender and age. Third, we evaluated the relationship between average GCL-IPL thickness and damage in brain tissues. For this purpose, children were classified by OCT measurements in two groups, defined by average GCL-IPL ≥5th centile and <5th centile of normal paediatric population. Differences in global and regional brain measurements were assessed between these groups of children.
A total of 50 Caucasian children were recruited, 25 born SGA and 25 AGA paired by age at the inclusion and GA at birth. The MRI visual inspection revealed the presence of abnormalities in only one child from the AGA group (ventriculomegaly). Thus the final sample included in the study was 24 AGA and 25 SGA children.
Perinatal and descriptive characteristics of both study groups are reported in table 1. Age at examination ranged between 6.5 and 15.8 years and did not differ between groups, as did not gender and rate of maternal smoking during pregnancy. Also, 13 children from each group were term born and 12 were preterm (6 early preterm and 6 late preterm). All the preterm children had a normal transfontanelar brain ultrasonography performed during the first months of life. None of the preterm infants included in the study had suffered from any important neonatal adverse event, such as bronchopulmonary dysplasia, necrotising enterocolitis, neonatal infections or hypoxic-ischaemic events.
Visual acuity and refractive status are reported in table 2. Stereoacuity was considered as full if it was at least 120″. Out of the 50 patients, 6 did not have full stereoacuity (3 from each study group) and 3 from the SGA group had intermittent strabismus. Although OCT was performed to all the included children, the first 12 lacked of macular imaging since GCL-IPL segmentation was not available at this moment. GCL-IPL complex measurements showed statistically significant differences between both study groups. All the other retinal measurements were thinner in average in the SGA group, although differences were not statistically significant. We also found no influence of prematurity on any of the retinal measurements in our study sample, including GCL-IPL measurements, with average values in AGA of 86.77 μm in term born and 86.33 μm in preterm (p=0.831), and 79.70 in term born and 76.13 in preterm in SGA group (p=0.515).
Brain MRI measurements
In a bivariate analysis, children born SGA showed lower total white matter (p=0.027), parietal lobe (p=0.029) and thalamus volumes (p=0.044).
GA and birth weight showed a different pattern of influence on brain measurements, as reported in table 2. Although both of them were directly correlated with total grey matter, total white matter, parietal lobe, putamen and hippocampus volumes, prematurity seemed to increase lateral ventricles volume and affect frontal lobe and thalamus, while parietal lobe and cerebellum white and grey matter were only influenced by birthweight centile. Measurements provided in table 3 were adjusted for age and gender since both of them can affect brain size and volumes.
Relationship between retinal and brain structures
No direct correlation was found between thickness of retinal layers and global or regional brain volumes. However, children with abnormal OCT outcomes (defined by average GCL-IPL ≥5th centile and <5th centile of normal paediatric population) showed lower total intracranial volume, decreased total grey and white matter volumes, and decreased volumes in certain brain regions (temporal and parietal lobes, thalamus, putamen caudate nucleus, pallidum and hippocampus) as shown in table 4. All the children with abnormal average GCL-IPL were SGA (three preterm and three term born).
In figure 1, we present a comparison between brain MRI and OCT outcomes from an AGA and a SGA child.
Our findings identify retinal ganglion cell layer as a potential biomarker of neuronal damage in brain grey and white matter of children born SGA. We found structural abnormalities in retinal layers related to low birth weight, not influenced by GA at birth. GCL-IPL seems to be the most affected structure, and its thinning corresponds with the most profound and extensive brain anomalies.
Our research group had previously reported changes in the retinal structure of children born SGA,9 ,10 ,15 both in ganglion cells and in their axons constituting the RNFL. The aim of the present study was to identify anomalies in SGA children with no neonatal adverse events, normal neurological assessment at birth and at inclusion in the study, and normal brain ultrasonography at neonatal period; and to evaluate whether retinal changes correlated with brain anomalies. Even in these otherwise healthy children we found thinner GCL-IPL and decreased volumes in certain brain regions.
The influence of prematurity on retinal structure depends on the presence of ROP, its treatment and associated comorbidity of the infants. In a previous study, we found that retinal structure was only affected in school-age children born preterm if prematurity was associated with low birth weight (<10th centile), perinatal infections, hypoxic-ischaemic events or ROP requiring treatment.15 Our present results agree with previous findings since our sample lacked important perinatal adverse events and only preterm children born SGA showed reduced GCL-IPL thickness. Recent evidence provided by Rothman et al 11 has demonstrated that these retinal abnormalities are not developmental changes since they are found in neonatal period.
Children born SGA present specific neurostructural and neurodevelopmental anomalies, even when born at term.16 ,17 The impact of an impaired fetal growth on preterm newborns has been widely studied. SGA children were reported to have smaller total cerebral brain volume, smaller cortical surface area, lower volume of brain white and grey matter and smaller basal ganglia and thalamus compared with AGA children.18 Tolsa et al found significant reduction in intracranial volume and in cerebral cortical grey matter volume in growth-restricted preterm neonates, as measured during the first two weeks of life. In line with these results, Padilla et al 19 reported differences in brain lobar volumes in preterm infants with growth restriction compared with preterm AGA or with healthy term-born infants at 12 months of age. Preterm infants born SGA seemed to exhibit specific patterns of grey and white matter distribution, with reduced grey matter in the temporal, parietal, frontal and insular regions compared with both preterm AGA and term-born infants, and increased white matter in temporal lobe compared with preterm AGA and in frontal, parietal, occipital and insular regions compared with term born.19
Our findings in brain MRI measurements are in agreement with previous evidence. Definition of SGA patients slightly differs among studies. Although most of them consider 10th centile of birth weight as the cut-off level, some others prefer birth weight and/or birth length below −2 SD.7 Birth weight should not be evaluated as an isolated parameter, separated from GA, because the observed weight may be AGA, but considered low for a full-term child. Thus, it is better to analyse birth weight by centile according to GA and gender as a surrogate marker of impaired fetal growth. Our data indicate that birth weight below 10th centile according to GA at birth is associated with diminished total intracranial volume, global white and grey matter, parietal and temporal lobes, putamen, hippocampus, and cerebellum white and grey matter volumes compared with AGA children, and this damage persists during childhood.
Effect of fetal growth restriction on white matter seems to be more pronounced than in grey matter.6 ,7 Our findings are consistent with it, both in brain and in cerebellum structures. Different reasons have been postulated for this finding, both prenatal and postnatal. Since growth restriction takes place during the second half of the pregnancy, a relative sparing of grey matter during this period may exist. Also, differential compensatory growth during early postnatal life has been proposed.
Prematurity has been associated with dilated lateral ventricles, thinner parietotemporal cortex and certain frontal and occipital areas.6 ,20 ,21 Since low birth weight is more frequently related to preterm births, its detrimental effects on CNS may be difficult to separate from those due to prematurity. Birth weight is mostly reported without taking into account GA and birthweight centiles. It entails study groups with both AGA and SGA patients, thus providing inaccurate conclusions. Studies with very restrictive criteria in the selection of the samples are needed to properly address this issue. We have found a similar pattern of damage in preterm and term-born SGA children. Although prematurely born infants undergoing an early growth restriction process suffer from a higher rate of perinatal complications, long-term neurodevelopmental and cognitive outcomes in school-age children are comparable with term-born infants with growth restriction.22
Previous studies performed in adult populations with neurodegenerative diseases, such as multiple sclerosis, Alzheimer or Parkinson diseases, have shown a correlation between anterior visual pathway changes, brain atrophy and functional disability. Gordon-Lipkin et al 2 found a correlation between brain atrophy on MRI and RNFL thickness in adults with multiple sclerosis. Later studies largely corroborated this association and further confirmed correlation with functional impairment.
With regard to perinatal brain damage, Rothman et al 11 recently reported that thinner RNFL in papillomacular bundle and temporal quadrant correlated with higher brain global MRI lesion burden in neonatal period of very preterm infants. However, their protocol of acquisition and analysis of the MRI images did not allow them to assess volumes in the different regions of interest, but instead, a scoring system based on simple, 1-D or 2-D brain measurements was used. In our sample of SGA and AGA children, we found that abnormal GCL-IPL (thickness below the 5th centile) enabled us to identify SGA children with the most severe brain white and grey matter changes. Our cut-off values are in agreement with GCL-IPL provided by other authors from paediatric populations.23
The mechanism for the relationship found between retinal abnormalities and regional brain volumes remains unclear. Retrograde trans-synaptic degeneration seems to be the main explanation in children with periventricular leukomalacia or acquired brain damage.24 Primary injuries in immature optic radiation are frequently associated with structural changes in the retina. Lennartsson et al 25 performed brain MRI and OCT in seven cases of young adults born preterm with known white matter damage due to immaturity and concluded that RNFL loss was more pronounced in subjects with more extensive lesions, following the pattern of the optic radiation damage. We propose a double mechanism for neuronal changes in the retina of children born SGA. Although a trans-synaptic degeneration may result in retinal changes in an infant with brain damage due to perinatal adverse events, derived from growth restriction or prematurity, a primary damage may also occur in the retina since neurological insult is taking place in an early stage of neurodevelopment. Chronic exposure to an adverse intrauterine environment in critical stages of pregnancy is likely to cause not only suboptimal development of retinal structures, but also adaptive mechanisms.
Among all the neuron types found in the retina, the ganglion cells and their axons seem to be the most vulnerable to perinatal damage. A biomarker has been defined by the International Programme of Chemical Safety, led by the WHO, as ‘any substance, structure, or process that can be measured in the body or its products and influence or predict the incidence and outcome of disease’. Thus, before accepting OCT outcomes as a biomarker of brain damage, further studies are required to address this found association.26
The main limitations of our study are the small sample size and the wide range of ages of our patients. In our study, OCT measurements were not corrected for axial length. However, its influence on GCL-IPL thickness remains unclear based on competing evidence.27 ,28 Although visual acuity was normal in all the patients, stereoacuity was not full in all of them and three children presented intermittent strabismus. We encourage further assessment of our findings with larger and more homogeneous samples. Our sample of SGA children has very restrictive inclusion criteria, which exclude any child with neurological impairments or abnormal findings in brain structure detected either by neonatal ultrasonography or brain MRI in childhood. In order to minimise the bias derived from the wide range of ages and gestational ages at birth, our main outcomes have been adjusted for demographic characteristics or have taken them into consideration.
In conclusion, SGA children, both preterm and term born, showed evidence of structural abnormalities in the retinal and brain neuronal tissues at school age, even with no evidence of neurological major impairments. We propose GCL-IPL as a potential biomarker identifying SGA children with the most profound brain reorganisation. An early and non-invasive biomarker would help to identify children with an increased risk of poor neurodevelopment among all SGA born. It would make it possible to perform further studies or to propose early abilitative interventions. Although retinal measurements should not be considered as an isolated biomarker of brain damage in SGA children, they could be helpful in combination with other biological or cognitive assessments.
Contributors Design and conduct of the study: VP, JL-P and MAM. Collection, management and interpretation of the data: VP, TP, IG, IA, EP, GG, RS-R, CP, DO, NF and GG-M. Preparation and review of the manuscript: VP, JL-P, MAM, TP, IG, IA, EP, GG, CP, DO, NF, GG-M and RS-R.
Funding The Spanish government (PI11/02430). RETICS funded by the PN I+D+I 2008–2011 (Spain), ISCIII- Sub-Directorate General for Research Assessment and Promotion and the European Regional Development Fund (ERDF), ref. RD12/0026.
Competing interests None declared.
Patient consent Written informed consent was obtained from the parents or guardians of each child.
Ethics approval Comité ético de Investigación clínica de Aragón (CEICA).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.