Foetal growth, birth transition, enteral nutrition and brain light scattering

If the brain structure is assessed at neonatal intensive care units, covert clinical events related with subtle brain injury might be identified. The reduced scattering coefficient of near-infrared light (μS’) obtained using time-resolved near-infrared spectroscopy from the forehead of infants is associated with gestational age, body weight and Apgar scores, presumably reflecting subtle changes of the brain related to foetal growth and birth transition. One hundred twenty-eight preterm and term infants were studied to test whether μS’ obtained from the head at term-equivalent age is associated with foetal growth, birth transition and nutritional status after birth, which are key independent variables of developmental outcomes. As potential independent variables of μS’, birth weight, Apgar scores, age at full enteral feeding and post-conceptional age at the study were assessed to represent foetal growth, birth transition and nutritional status after birth. Subsequently, higher μS’ values were associated with higher Apgar scores (p = 0.003) and earlier establishment of enteral feeding (p < 0.001). The scattering property of near-infrared light within the neonatal brain might reflect changes associated with birth transition and nutritional status thereafter, which might be used as a non-invasive biomarker to identify covert independent variables of brain injury in preterm infants.

www.nature.com/scientificreports/ obtained from the forehead showed a positive linear correlation with gestational age 16 . Our study in preterm and term infants further confirmed that μ S ' values obtained shortly after birth were associated with variables, such as antenatal glucocorticoid, emergency delivery, gestational age, body size, Apgar scores, requirement for mechanical ventilation and blood gas data at birth, suggesting the possibility that μ S ' might reflect subtle structural changes in the brain associated with antenatal growth, peripartum stress and birth transition 17 . However, little is known regarding the relationship between μ S ' values obtained from the head of newborn infants and their downstream clinical outcomes. The aim of this study was to test the association of μ S ' measured at term-equivalent period with intrauterine growth, birth transition and nutrition after birth, which are short-term surrogate markers for neurodevelopmental outcomes of hospitalised newborn infants [18][19][20][21] .

Results
Four infants, who developed grade III/IV intraventricular haemorrhage, and one infant, who developed hypoxicischaemic encephalopathy, were excluded, leaving 128 infants within the final study cohort (Fig. 1). These infants had a gestation period of 32.0 ± 4.2 weeks and weighed 1564 ± 688 g at birth, and were studied on 44.8 ± 28.3 days of age or 38.6 ± 2.1 weeks post-conceptional age (Table 1).
For the left and right temporal regions and the posterior region, data were not obtained for 8, 8 and 21 infants, respectively, because of insufficient signals from the head (n = 15), poor probe contact (n = 4) and the use of a cap device for non-invasive respiratory support (n = 2). No further data were excluded because of their poor quality or reproducibility. The mean μ a and μ S ' values for all wavelengths and head positions were 0.126 ± 0.025 cm −1 and 6.453 ± 1.416 cm −1 , respectively.
Dependence of μ a and μ S ' on wavelengths and head positions. The wavelength of 836 nm was associated with higher μ a values, whereas the wavelength of 791 nm was associated with lower μ a values compared to those of 761 nm (both p < 0.001) ( Table 2). The right temporal and posterior regions of the head were associated with higher μ a values compared to those of the anterior region (both p < 0.001).
The wavelengths of 791 and 836 nm were associated with lower μ S ' values compared to those of 761 nm (both p < 0.001). The left and right temporal and posterior regions of the head were associated with higher μ S ' values compared to the anterior region (all p < 0.001).
Dependence of μ a and μ S ' on clinical variables: univariate analysis. The higher μ a values were positively associated with gestational age (p = 0.001), body weight at birth (p < 0.001), blood haemoglobin level at study (p < 0.001) and μ S ' values (p < 0.001), and negatively associated with antenatal glucocorticoid (p < 0.001), cord blood pH (p = 0.003) and postnatal age at study (p = 0.001); relationships with multiple pregnancy (p = 0.016), head circumference at birth (p = 0.005) and body weight at study (p = 0.036) were lost after correction for multiple comparisons (all adjusted for the wavelengths and head positions; Table 2 and Fig. 2).
The μ S ' level was positively associated with gestational age (p < 0.001) and μ a values (p < 0.001), and negatively associated with indomethacin for patent ductus arteriosus (p < 0.001) and postnatal age to achieve full enteral feeding (p < 0.001); relationships with antenatal glucocorticoid (p = 0.013), body weight and head circumference at birth (both p = 0.012), Apgar scores at 1 and 5 min (p = 0.039 and 0.029, respectively) and postnatal age at study (p = 0.003) were lost after correction for multiple comparisons (all adjusted for the wavelengths and head positions; Table 2). See Online Supplemental Tables S1-S3 for findings from analyses performed for each wavelength.
Dependence of μ S ' and μ a on clinical variables: multivariate analysis. Higher μ a values were associated with greater age to achieve full enteral feeding (p = 0.049), greater post-conceptional age at study www.nature.com/scientificreports/ (p = 0.015), higher blood haemoglobin levels at study (p < 0.001) and higher μ S ' values (p < 0.001) ( Table 3). Higher μ S ' values were associated with higher Apgar scores at 5 min (p = 0.003), smaller age to achieve full enteral feeding (p < 0.001) and higher μ a values (p < 0.001). See Online Supplemental Tables S4-S6 for findings from analyses performed for each wavelength.

Discussion
Building on previous studies of TR-NIRS, which suggested that the light scattering within the brain shortly after birth is dependent on variables related to foetal growth, antenatal stress and birth transition, we have demonstrated that higher μ S ' values obtained at term-equivalent age were associated with higher Apgar scores and earlier establishment of enteral nutrition. μ S ' can be a unique and clinically useful biomarker of subtle changes in the brains of newborn infants with respect to antenatal stress, birth transition and nutritional status after birth. Light scattering within a tissue theoretically increases with relatively more complex microstructures due to increased reflection and path length of near-infrared light 14 . Thus μ S ' has a potential to provide microstructural information of the brain. Ijichi and colleagues first reported that μ S ' values of near-infrared light obtained shortly after birth from the foreheads of newborn infants with a gestation age of 30-41 weeks depended on gestational age 16 . Our previous study confirmed that μ S ' values obtained from the foreheads of preterm and term infants assessed shortly after birth were dependent on body size and Apgar scores, as well as on gestational age. These findings suggest the possible utility of μ S ' values as a non-invasive marker to evaluate subtle differences in the brain subsequent to foetal maturation, antenatal stress and birth transition 17 . Our current study further verified that the μ S ' value obtained at term equivalent period is associated with both clinical variables at birth and those related to the nutritional status of the infant after birth. Intrauterine growth and maturation, intrapartum stress and response and postpartum nutrition and growth constitute key independent variables of the neurodevelopmental outcomes of the infant [18][19][20][21] . If the consequence of the intrinsic maturity, extrinsic stress, birth transition and nutritional status of the infant can be assessed using μ S ' values obtained from the heads of newborn infants, along with other substantiations, μ S ' might serve as a clinically useful biomarker of cerebral maturation and Table 1. Background characteristics of 128 infants within the study cohort. Values are number (%), mean ± standard deviation or median (lower/upper quartiles). µ a absorption coeffieicnt, µ s ' reduced scattering coefficient. a Assessed at 36 weeks post-conceptional age (or on day 28 for those born later than 32 weeks gestation). www.nature.com/scientificreports/ damage. Future studies need to address the contribution of other potential independent variables of light scattering as measured from the scalp, such as the gyration of the brain and developmental changes in the layer of cerebrospinal fluid. With regard to the absorption of near-infrared light, only modest relationships were observed between higher μ a values and longer time to achieve full enteral feeding and greater post-conceptional age at the time of the study; robust correlations were only observed between μ a values and priori covariates of the wavelengths of light, head position and blood haemoglobin concentration at the time of the study. Given that absorption of near-infrared light within the range of 750-850 nm is primarily determined by the tissue haemoglobin concentrations 14,15 , μ a values might reflect the maturation of the cerebral tissue via increased complexity of the cerebral vessels and subsequent blood volume. Progression of anaemia and increase in the cerebral blood flow and volume with increasing postnatal age might also affect the dependence of μ a values on clinical variables 12 .

Strengths and limitations.
We were able to elucidate the clinical variables potentially determining the property of light absorption and scattering within the brain in a relatively large cohort of newborn infants. However, we were unable to present a direct association between μ S ' values and microstructure of the brain. As described in the previous section, the observed relationships between μ S ' , μ a and clinical variables can be affected by a range of clinical biases. For example, extremely preterm infants are relatively anaemic at birth and the anaemia progresses with postnatal age without transfusion, potentially leading to lower blood haemoglobin and μ a Table 2. Dependence of µ a and μ S ' on clinical variables: univariate analysis. B regression coefficient, CI confidence interval, µ a absorption coefficient, µ s ' reduced scattering coefficient. *Assessed at 36 weeks postconceptional age (or on day 28 for those born later than 32 weeks gestation). Findings are adjusted for the wavelengths of near-infrared light** and position of the head † . www.nature.com/scientificreports/  www.nature.com/scientificreports/ levels with greater gestational age at birth and greater postnatal age at the time of TR-NIRS study. Although we carefully selected independent variables and covariates to minimise the bias, the findings might still be affected by the bias derived from the collinearity between the variables. Our study cohort comprised newborn infants, who were hospitalised at a tertiary neonatal intensive care unit. Although the observed μ a and μ S ' values were comparable to those reported in healthy newborn infants 22 , extrapolation of our findings into physiological transition and growth in healthy newborn infants must be done cautiously. Finally, the longitudinal follow-up study of the study population is still underway, resulting in the lack of outcome information in association with the light absorption and scattering properties.

Conclusions
The μ S ' values of the near-infrared light obtained at term-equivalent period from the heads of newborn infants were associated with Apgar scores and postnatal age when full enteral feeding was achieved, suggesting a correlation between the light scattering property and stress-response at birth and nutritional status of the infant thereafter. With further validations, μ S ' might serve as a biomarker to distinguish the variation of the microstructural complexity of the brain tissue subsequent to different maturational stage, antenatal stress, tissue damage and repair, nutritional status and growth. Associations between the μ S ' values and detailed clinical courses, macro-and microstructural MRI findings and neuro-developmental outcomes need to be addressed to assess the clinical utility of this non-invasive cot-side tool.

Materials and methods
This study was conducted in compliance with the Declaration of Helsinki under the approval of the Ethics Committee of Kurume University School of Medicine (reference number: 12128). Informed parental consent was obtained for each participating newborn infant. All methods were carried out in accordance with relevant guidelines and regulations.
Study population. This study was performed as a secondary analysis of a prospective longitudinal study, which was performed between June 2009 and January 2015 to serially collected the TR-NIRS data of preterm and term infants hospitalised at a tertiary neonatal intensive care centre of Kurume University Hospital (Kurume, Fukuoka, Japan). Independent variables of μ S ' values obtained shortly after birth from a part (n = 60) of the current cohort have been reported in a previous study 17 . Of 136 newborn infants within the original study cohort, 132 infants, who had TR-NIRS data obtained between 34 and 42 weeks postconceptional age, were considered. Infants with chromosomal aberration, malformation syndrome, grade III/IV intraventricular haemorrhage, hypoxic-ischaemic encephalopathy, congenital hydrocephalus and other major cerebral anomalies were excluded.
Data collection. The μ a and μ S ' values were obtained from the heads of the infants for three wavelengths, 761, 791 and 836 nm, using a TR-NIRS system (TRS-10, Hamamatsu Photonics K.K., Hamamatsu, Shizuoka, Japan) 17 . This system employs the time-correlated single photon counting method to create time response profiles of pulsed laser light penetrating an object. The time response profiles were then fitted into a photon diffusion equation using the nonlinear least square fitting method to obtain μ a and μ S ' for each wavelength 14 . Although the data acquisition for the original study was repeated with intervals of approximately 1 week from birth to discharge, for the current study, a particular value obtained between 34 and 42 weeks of post-conceptional age (closest to 40 weeks gestation if there were multiple records) was used to represent each infant. Data were acquired when the infant was clinically stable and asleep or calmly awake. The TR-NIRS probes were inserted into a rubber holder, with an inter-optode distance of 3 cm, and was applied to a relatively flat part of the head. Data acquisition (10 s) was repeated five times for each of the frontal, left and right temporo-parietal and occipital regions by repositioning the probe each time. In our previous study, which acquired TR-NIRS data using the same protocol to the current one 17 , standard deviations of μ a and μ S ' values for five successionally obtained data within the same head position and infant were, in average, 2.4% and 2.7%, respectively. Based on these small intra-individual and intra-regional differences in μ a and μ S ' values, five readings each of μ a and μ S ' were averaged for each brain region. We confirmed the degree of fit to the photon diffusion equation using the conversion chi-square value index of between 0.8 and 1.2 23 . Data were not collected for brain regions with poor probe contact (typically due to the lack of flat surfaces or use of cap devices for non-invasive respiratory support), poor fit to the photon diffusion equation or insufficient signal-to-noise ratio with the count rate < 100 K counts/s or relative dark-to peak-count ratio of > 0.1. The data were retrospectively assessed to identify those with poor quality or intra-regional reproducibility before being processed for further analysis.
Clinical information. The clinical background information was obtained from the electronic records of the patients, including (1) maternal and antenatal variables (antenatal glucocorticoid, multiple pregnancies and emergency caesarean delivery), (2) variables at and shortly after birth (sex, cord blood pH, Apgar scores at 1 min and 5 min, gestational age, body weight, head circumference, hypoglycaemia within 48 h of birth, indomethacin for the treatment of the patent ductus arteriosus, grade I/II intraventricular haemorrhage and periventricular leukomalacia, (3) variables associated with clinical variables of infants after the transitional period (body weight on the day of study, age when full enteral feeding of > 100 ml/kg/d was achieved and chronic lung disease assessed 36 weeks post-conceptional age or on day 28, whichever was later). In order to assess the influence of intrauterine growth on μ S ' values, body weight and head circumference at birth were expressed as z-scores in accordance with the New Japanese Neonatal Anthropometric Charts for Gestational Age at Birth 24 . www.nature.com/scientificreports/ Data analysis. To minimise biases owing to missing data, multiple imputation of the missing values of less than 10% (excluding for μ a and μ S ') was performed (n = 5 imputations), based on the correlation between variables with missing values and other characteristics of the participants (SPSS ver. 22.0, IBM, Armonk, NY, U.S.A.). Although the property of μ a was out of our study scope, independent variables of both μ a and μ S ' were assessed to clarify the possible influence of light absorption to the relationship between μ S ' values and clinical variables. The generalised estimating equation with a linear model was used to account for repeated sampling of TR-NIRS data for three near-infrared light wavelengths and four head regions. Although the influence of the wavelength is much greater on μ a than on μ S ' 16,17 , the three wavelengths were incorporated within the model for consistency in the analytical procedure. Crude effects of clinical variables on μ a and μ S ' values were assessed using the univariate model adjusting for the wavelengths and head positions. p values < 0.002 were assumed to be significant, correcting multiple comparisons of 25 variables. The final models to explain μ a and μ S ' values were developed based on our hypothesis, which employed the body weight at birth, Apgar scores at 5 min, age to achieve full enteral feeding and post-conceptional age at study; the model was also adjusted for priori covariates, which were known independent variables of clinical outcomes (antenatal glucocorticoid, multiple pregnancies and sex), μ a (wavelength, position of the head and μ S ') and μ S ' (wavelength, position of the head and μ a ). Data were presented as mean ± standard deviation unless specified otherwise.