Objective
We assessed brain metabolite levels by magnetic resonance spectroscopy (MRS) in 1-year-old infants born small at term, as compared with infants born appropriate for gestational age (AGA), and their association with neurodevelopment at 2 years of age.
Study Design
A total of 40 infants born small (birthweight <10th centile for gestational age) and 30 AGA infants underwent brain MRS at age 1 year on a 3-T scanner. Small-born infants were subclassified as late intrauterine growth restriction or as small for gestational age, based on the presence or absence of prenatal Doppler and birthweight predictors of an adverse perinatal outcome, respectively. Single-voxel proton magnetic resonance spectroscopy ( 1 H-MRS) data were acquired from the frontal lobe at short echo time. Neurodevelopment was evaluated at 2 years of age using the Bayley Scales of Infant and Toddler Development, Third Edition, assessing cognitive, language, motor, social-emotional, and adaptive behavior scales.
Results
As compared with AGA controls, infants born small showed significantly higher levels of glutamate and total N-acetylaspartate (NAAt) to creatine (Cr) ratio at age 1 year, and lower Bayley Scales of Infant and Toddler Development, Third Edition scores at 2 years. The subgroup with late intrauterine growth restriction further showed lower estimated glutathione levels at age 1 year. Significant correlations were observed for estimated glutathione levels with adaptive scores, and for myo-inositol with language scores. Significant associations were also noticed for NAA/Cr with cognitive scores, and for glutamate/Cr with motor scores.
Conclusion
Infants born small show brain metabolite differences at 1 year of age, which are correlated with later neurodevelopment. These results support further research on MRS to develop imaging biomarkers of abnormal neurodevelopment.
Small-born infants are those with a birthweight <10th centile for gestational age. This represents a common condition affecting up to 10% of all deliveries at term and is associated with a higher risk for adverse neurological and cardiovascular outcome. During gestation, fetal smallness is identified by ultrasound as an estimated fetal weight <10th centile for gestational age, and a thorough Doppler ultrasound assessment is necessary in such cases to evaluate the severity. Among the late-onset forms of fetal smallness, about two thirds present signs of severity, including prenatal Doppler cerebroplacental ratio (middle cerebral to umbilical artery pulsatility index ratio <5th centile and/or mean uterine artery pulsatility index >95th centile and/or estimated fetal weight <3rd centile), and are associated with poorer perinatal outcome. It has been suggested that these cases represent true forms of placental insufficiency and should be considered as late-onset intrauterine growth restriction (IUGR). The remaining one-third of cases of late-onset fetal smallness, without signs of severity, should be considered as small for gestational age (SGA). While SGA fetuses have been commonly defined as constitutionally small, there is no solid evidence to support this and they may represent another form of pathological smallness. Regardless of the clinical presentation, both clinical groups of small fetuses are associated with poorer neurodevelopment as early as the neonatal period and during later stages, affecting mostly functions associated with frontal networking such as attention, creativity, language, memory performance, and learning abilities. The neurostructural and neurofunctional phenotypes associated with fetal smallness, and their relationships, are still only partially understood.
The study of brain metabolism is an important tool to assess neurodevelopment, given their intrinsic relationship. Previous studies have shown that small fetuses have lower frontal lobe ratios of N-acetylaspartate (NAA, a neuronal marker) to choline compounds (Cho, marker of cell membrane turnover) and lower ratios of NAA to creatine (Cr, a potential glial marker implicated in cellular energetics). These metabolic changes have been associated with disrupted brain maturation but it remains unknown to what extent they are a reflection of the brain tissue exposure to an adverse in utero environment, and how they may underlie true neurostructural changes leading to brain remodeling and future changes in neurodevelopmental function.
In this study we addressed the hypotheses that brain metabolite changes were present in the frontal lobe of 1-year-old small-born infants at term, and could be correlated with neurodevelopmental outcome at 2 years of age.
Materials and Methods
Study cohort and clinical perinatal data
This study is part of a larger prospective research program on growth restriction involving fetal, and short- and long-term postnatal, follow-up. The protocol used was approved by the local institutional ethics committee (review board 2010/5736) and all participants gave their written informed consent. A consecutive sample of 70 neonates were prospectively recruited at birth, all delivered at term from singleton pregnancies and presenting normal umbilical artery pulsatility index (<95th centile) at the time of delivery. Subjects were classified as appropriate for gestational age (AGA) (30 subjects) or small (40 subjects), based on their birthweight above or below the 10th centile, respectively. In all cases, gestational age was corrected from fetal crown-rump length in the first trimester. Additionally, the small-born neonate group was subclassified according to recent criteria proposed, considering their birthweight and prenatal ultrasound Doppler parameters from the scan closest to the time of delivery. Thus, small-born infants were classified as late-onset IUGR (31 subjects) if signs of severity were present (cerebroplacental ratio <5th centile and/or mean uterine artery pulsatility index >95th centile and/or a birthweight <3rd centile ), or as SGA (9 subjects) when none of these factors were present. Middle cerebral, umbilical, and mean uterine artery pulsatility index parameters were measured as reported earlier, using a Siemens Sonoline Antares (Siemens, Erlangen, Germany) system with a 6 to 2MHz linear curved-array transducer.
Neonates with congenital malformations, chromosomal abnormalities, infections, chronic maternal pathology, or noncephalic presentations were not eligible for this study. Maternal and perinatal data were prospectively recorded in most study patients.
Magnetic resonance acquisition
Brain magnetic resonance imaging was carried out at 14 (±1.5) months of age, without sedation, during natural sleep. Data were acquired with a 3.0-T scanner (Tim Trio; Siemens Diagnostics Healthcare, Erlangen, Germany) and a head matrix radiofrequency coil was used. The total length of each magnetic resonance examination did not exceed 45 minutes. Reference T1-weighted anatomical images were acquired with magnetization prepared rapid acquisition gradient echo. Single-voxel proton spectroscopy ( 1 H-MRS) was carried out with point-resolved spectroscopy from the frontal lobe region ( Figure 1 , A), using the following parameters: 40 × 20 × 20 mm 3 voxel size, 2000-millisecond repetition time, 30-millisecond echo time, 98 averages, chemical shift selective water suppression, and 3-minute 24-second acquisition time. A reference spectrum with 16 averages was also acquired, without water suppression. Finally, T2-weighted images (half-Fourier acquisition single-shot turbo spin-echo) were obtained; magnetic resonance spectroscopy (MRS) was repeated if artifacts were detected (eg, due to head movements). Structural magnetic resonance images were reviewed for the presence of anatomical abnormalities by an experienced neuroradiologist blinded to group membership.
MRS postprocessing
MRS data with gross visual artifacts and/or absence of an interpretable metabolic pattern were directly discarded. The remaining MRS data were processed using linear combination model-fitting (LCModel; S. Provencher Inc., http://s-provencher.com/pages/lcmodel.shtml ). The basis sets used included a total of 23 metabolites: alanine, aspartate, Cr and phosphocreatine (total Cr), gamma aminobutyric acid, glucose, glutamine, glutamate, glycine, phosphocholine and glycerophosphocholine (choline compounds, Cho), glutathione, myo-inositol, lactate, N-acetylaspartate, and N-acetylaspartyilglutamate (NAAG) (abbrieviated together as total NAA, NAAt), phosphoethanolamine, scyllo-inositol, taurine, threonine, valine, acetate, and ascorbate. Lipid and macromolecule contributions were also included in the modeling. All MRS data analyzed had an estimated signal-to-noise ratio >10 and only spectra with an estimated full width at half maximum <0.1 ppm were selected for further analysis. Those metabolite fittings with Cramér-Rao lower bounds (estimated SDs of the estimated concentration) >50% were also discarded and only metabolite changes of at least twice their Cramér-Rao lower bounds (95% confidence interval) were selected, unless indicated otherwise. The data were analyzed quantitatively, using the reference water scans (estimated mmol/kg water content in brain tissue), and as ratios to total Cr and to Cho compounds (Cho) ( Figure 1 , B, metabolite assignments according to the literature ).
Neurodevelopmental assessment
Neurodevelopmental outcome was assessed at 23 (±1.5) months of age, using the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III). Five distinct areas were evaluated: cognitive, language, motor, social-emotional, and adaptive behavior. Each area consists of a score with a normalized mean of 100 ± 15 (controls). Abnormal BSID-III was defined as a score <85 (1 SD) in any of the 5 scales. All developmental examinations were performed by a single experienced psychologist, blinded to infant medical history.
Statistical analysis
Quantitative and qualitative parameters from the clinical/demographic data were compared with the Student t test for independent samples (or Mann-Whitney U test) and with the Pearson χ 2 , respectively. The Wilcoxon test was used to compare paired groups without normal distributions. MRS differences between cases and controls, and between neurodevelopmental score groups, were assessed using the Student t test for independent samples and with general linear model, adjusting by potential confounding factors: gestational age at birth, sex, maternal breast-feeding >4 months, age at magnetic resonance, and low socioeconomic status, when comparing MRS differences between small-born infants and controls; and age at BSID-III test, when comparing MRS differences between infants with low and normal BSID-III scores. Pearson and Spearman correlation tests were performed to assess univariate correlations between metabolite levels and BSID-III test scores. All statistical analysis were performed with software (SPSS, version 19.0; IBM Corp, Armonk, NY), considering results significant at a P value < .05.
Results
Infant cohort and perinatal clinical observations
After MRS postprocessing and quality assessment, spectra from 59 infants were selected for further analysis (33 small born, 83%; and 26 AGA, 87%) ( Table 1 ). As expected, infants born small were delivered earlier and had longer stays at the neonatal intensive care unit than AGAs. No signs of intracranial pathology were found in any of the anatomic magnetic resonance images. At 1 year of age, small-born infants were still significantly smaller than AGAs (height and weight centiles calculated as reported by others ). With regard to brain MRS data, no significant differences were noticed between the 2 groups as far as signal-to-noise ratio (small born, 26 ± 6.7; AGA, 28.4 ± 5.9) and full width at half maximum (small born, 0.049 ± 0.012; AGA, 0.047 ± 0.012). In 8 (14%) examinations (6 small born and 2 AGAs) the quantifications were limited to metabolite ratios since the water reference scan was not available. No statistically significant differences in perinatal clinical data were found between cases with interpretable and uninterpretable spectra.
| Demographic | AGA (n = 26) | Small (n = 33) | P value |
|---|---|---|---|
| Maternal characteristics | |||
| Maternal age, y | 31.8 ± 7.2 | 32.6 ± 4.8 | .627 a |
| Low maternal socioeconomic status, % | 45.8 | 60.6 | .269 b |
| Maternal smoking during pregnancy, % | 24.0 | 29.0 | .672 b |
| Perinatal data | |||
| GA at birth, wk | 39.8 ± 1.4 | 38.1 ± 0.9 | .000 a |
| Cesarean delivery, % | 22.7 | 45.2 | .093 b |
| Instrumental vaginal delivery, % | 17.6 | 5.9 | .157 b |
| Gestational diabetes, % | 8.3 | 0.0 | .179 b |
| Preeclampsia, % | 0.0 | 6.3 | .501 b |
| Birthweight, g | 3375.2 ± 403.1 | 2304.9 ± 263.8 | .000 a |
| Birthweight centile | 50.7 ± 31.3 | 2.1 ± 2.6 | .000 c |
| Male sex, % | 57.7 | 69.7 | .339 b |
| Umbilical artery pH <7.15, % | 12.5 | 7.1 | .552 b |
| 5-min Apgar score <7, % | 0.0 | 0.0 | – |
| Length, cm | 49.9 ± 1.5 | 45.5 ± 2.2 | .000 c |
| Length centile | 48.2 ± 24.0 | 8.7 ± 7.3 | .000 c |
| Cephalic perimeter, cm | 34.2 ± 0.8 | 32.3 ± 1.2 | .000 a |
| Cephalic perimeter centile | 34.2 ± 19.7 | 12.7 ± 16.9 | .000 c |
| Admission to NICU, d | 0.1 ± 0.4 | 1.2 ± 2.4 | .05 c |
| Infant parameters at 12 mo | |||
| Age at MRI, wk | 56.4 ± 6.0 | 56.0 ± 6.0 | .800 a |
| Weight, kg | 9.8 ± 1.0 | 8.5 ± 1.2 | .001 a |
| Weight centile | 38.4 ± 29.9 | 13.0 ± 21.2 | .000 c |
| Height, m | 0.75 ± 0.03 | 0.71 ± 0.03 | .000 a |
| Height centile | 51.2 ± 29.0 | 20.8 ± 19.6 | .003 c |
| Cephalic perimeter, cm | 46.2 ± 1.3 | 45.2 ± 1.2 | .016 a |
| BMI | 17.3 ± 1.6 | 16.8 ± 2.0 | .352 a |
| Breast-feeding >4 mo, % | 47.1 | 69.6 | .151 b |
| BSID-III scores at 24 mo | |||
| Cognitive | 101.8 ± 15.0 | 99.0 ± 17.1 | .363 d |
| Language | 108.1 ± 19.2 | 95.4 ± 15.1 | .223 d |
| Motor | 104.9 ± 15.5 | 97.1 ± 13.0 | .091 d |
| Social-emotional | 109.3 ± 25.4 | 94.1 ± 24.8 | .070 d |
| Adaptive | 103.9 ± 13.6 | 97.6 ± 15.9 | .448 d |
| Average score of all areas | 106.0 ± 13.2 | 96.5 ± 12.6 | .038 d |
a Student t test for independent samples
d General linear model adjusting by sex, maternal low economic status, GA at delivery, breast-feeding >4 mo, and age at BSID-III as covariates.
Brain metabolite changes in infants born small at term
Estimated metabolite levels
The most consistent, significant difference in small-born infants compared to controls was an increase in brain glutamate levels ( Figure 2 , A). This change was also depicted in the mixed quantification of glutamate and glutamine (glutamate + glutamine; Figure 2 , B). An increase in total NAA (NAAt) was also noticed in infants born small compared to AGAs ( Figure 2 , C). After adjusting for potential confounding variables, the increase in glutamate and glutamate + glutamine remained significant, while the NAAt increase was borderline significant ( P = .050) ( Table 2 ).
| Metabolite | AGA | Small | Change | P value a | P value b |
|---|---|---|---|---|---|
| Glutamate c | 6.0 ± 0.8 | 6.6 ± 1.1 | 10% | .041 | .026 |
| Glutamate + glutamine c | 6.5 ± 1.0 | 7.2 ± 1.4 | 11% | .049 | .032 |
| NAAt c | 4.4 ± 0.8 | 4.7 ± 0.5 | 8% | .073 | .05 |
| Glutamate + glutamine/Cr | 1.5 ± 0.3 | 1.7 ± 0.4 | 7% | .121 | .035 |
| NAAt/Cr | 1.0 ± 0.1 | 1.1 ± 1.4 | 7% | .095 | .039 |
| NAAt/Cho | 4.1 ± 0.5 | 4.4 ± 0.7 | 7% | .099 | .432 |
b General linear model adjusted by sex, maternal low economic status, gestational age at delivery, breast-feeding >4 mo, and age at magnetic resonance imaging as covariates
Metabolite ratios
When considering metabolite ratios, small-born infants showed significant increases in glutamate + glutamine/Cr and NAAt/Cr ( Figure 2 , D and E). Also, a slight but not significant increase in NAAt/Cho was detected ( Figure 2 , F) ( Table 2 ).
Changes in SGA and late-onset IUGR infants
Within the group of 33 small-born infants with usable MRS data, 26 were late-onset IUGR and 7 were SGA. Increased glutamate and NAAt levels were observed in late-onset IUGR infants when compared to controls ( Table 3 ), reaching significance for the NAAt/Cr ratio ( Figure 3 ). Moreover, estimated glutathione levels were significantly decreased in late-onset IUGR infants ( Table 3 ). Although these levels are close to the glutathione concentrations reported in adult brain by MRS, the estimated change detected (–12%) was slightly below the fitting error ranges (Cramér-Rao lower bounds, 8 ± 2%). Metabolite changes in SGA infants were more variable, likely due to the lower sample sizes, but showed the same tendency as in late-onset IUGRs when compared to controls. No significant metabolite differences were detected between SGA and late-onset IUGR subgroups.
| Metabolite | AGA | SGA | Change | P value a | P value b | Late-onset IUGR | Change | P value a | P value b |
|---|---|---|---|---|---|---|---|---|---|
| Glutamate c | 6.0 ± 0.8 | 7.1 ± 1.4 | 18% | .015 | .115 | 6.4 ± 0.9 | 7% | .125 | .111 |
| Glutamate + glutamine c | 6.5 ± 1.0 | 7.9 ± 1.4 | 21% | .015 | .072 | 7.0 ± 1.2 | 8% | .154 | .143 |
| NAAt c | 4.4 ± 0.8 | 4.7 ± 0.7 | 7% | .389 | .118 | 4.7 ± 0.5 | 8% | .076 | .107 |
| NAAt/Cr | 1.0 ± 0.1 | 1.0 ± 0.2 | 3% | .695 | .126 | 1.1 ± 0.1 | 7% | .055 | .004 |
| Glutathione c | 1.4 ± 0.3 | 1.4 ± 0.6 | 2% | .845 | .005 | 1.2 ± 0.2 | –12% | .045 | .641 |
b General linear model, sex, adjusted by maternal low economic status, gestational age at delivery, breast-feeding >4 mo, and age at magnetic resonance imaging as covariates
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
