Study design and procedure
Data were collected as part of a large multisite study of pediatric concussion, Advancing Concussion Assessment in Pediatrics (A-CAP)); for a detailed description, see Yeates et al. 2017 19. Participants were recruited from five pediatric Emergency Departments (ED) within the Pediatric Emergency Research Canada (PERC) network 20 in Canada. Participating hospitals were Alberta Children’s Hospital (Calgary), Stollery Children’s Hospital (Edmonton), British Columbia Children’s Hospital (Vancouver), Children’s Hospital of Eastern Ontario (Ottawa), and Centre Hospitalier Universitaire (CHU) Sainte-Justine (Montreal). Participants were recruited if they presented with concussion or orthopedic injury (OI, the comparison group) within 48 hours of injury. This study was approved by the research ethics board at each site (Conjoint Health Research Ethics Board at the University of Calgary, REB152296; Ste Justine Research Institute, University of Montreal, MP-21-2017-1332; CHEO REB16/23E; University of Alberta HREB, 64780; University of British Columbia Children’s & Women’s Research Ethics Board, H16-00104) where all research was performed in accordance with the relevant guidelines and regulations. Informed consent and assent were obtained from the parents/guardians and the youth participants respectively.
Participants
All participants were between the ages of 8.00 to 16.99 years at the time of recruitment. Concussion participants sustained a blunt head trauma and presented with an observed loss of consciousness, a Glasgow Coma Scale (GCS) score of 13 or 14, and/or at least one acute sign or symptom of concussion (i.e., post-traumatic amnesia, focal neurological deficits, skull fracture, posttraumatic seizure, vomiting, headache, dizziness, or other mental status abnormalities). Children who showed delayed neurological deterioration (GCS < 13) or required neurosurgical intervention were excluded. Participants were also excluded if they had a loss of consciousness of greater than 30 min, post-traumatic amnesia of greater than 24 h, or any associated injuries with scores of greater than 4 on the Abbreviated Injury Scale 21.
OI participants sustained an upper or lower extremity fracture, sprain, or strain due to blunt force/physical trauma resulting in an Abbreviated Injury Scale 21 score of 4 or less. Youth were excluded from the OI group if they had any head trauma or suspicion of concussion and concussion-related symptoms at the time of recruitment, or an injury requiring surgical intervention or procedural sedation. Youth were excluded from both groups if they sustained a concussion within 3 months of recruitment. A thorough description of the inclusion and exclusion criteria and the multi-site study are described in the study protocol [17].
MRI and MRS
Participants completed 3T magnetic resonance imaging (MRI) including MRS. The target for the first visit post-recruitment was about 1–2 weeks post-injury (will now be referred to as 12-days) and then the follow-up visits were randomized to either a 3- or 6-month follow-up MRI. The 5 sites included 3 GE scanners (General Electric MR750w in Calgary; General Electric MR750 in Montreal and Vancouver) and 3 Siemens scanners (Siemens Prisma in Edmonton and Montreal; Siemens Skyra in Ottawa).
T1-weighted images were acquired. The sites that used GE scanners used a 3D T1-weighted fast spoiled gradient echo brain volume (FSPGR BRAVO) sequence with a TR/TE/TI = 8.25/3.16/600 ms with a field of view of 24 cm2. The sites that used Siemens scanners used a 3D T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence with TR/TE/TI = 1880/2.9/948 ms and a field of view of 25.6 cm2. Acquisitions from both vendors used 192 slices, a flip angle of 10° with a voxel size of 0.8 × 0.8 × 0.8 mm3.
Single voxel MRS data was acquired in the left dorsolateral prefrontal cortex (L-DLPFC) using Point RESolved Spectroscopy (PRESS) localisation. GE sites used a 32-channel head coil while Siemens sites used a 64-channel head coil. Acquisition parameters were; TE/TR = 30/2000 ms, 96 water suppressed averages (with 8 step phase cycle for GE and 16 step phase cycle for Siemens), 8 unsuppressed water averages, 2 × 2 × 2 cm3 voxel, spectral width 5000 Hz (GE) or 2000 Hz (Siemens), number of points 4096 (GE) or 2048 (Siemens), water suppression methods were CHESS (GE) and WET (Siemens). Each participating site was provided with several reference images to ensure standardized voxel placement, and an example of voxel placement is shown in Fig. 1.
Data analysis
As GE data had individual averages available, a complete pre-processing pipeline was used based on consensus recommendations 22 and included: combination of receiver channels, removal of bad averages, retrospective shot-by-shot frequency and phase correction, left shifting, and zero order phase correction. These pre-processing steps were automated and completed using FID-A 23. The Siemens site data were fully averaged upon scanner export, so the only pre-processing performed was by the vendor software.
Data from both GE and Siemens were quantified with LCModel version 6.3-1J 24. Basis sets customized for each vendor were generated in FID-A using vendor-specific pulse shapes, spectral width, and number of points. The generated basis sets included the following metabolites: Alanine, Aspartate, 𝛽 -Hydroxybutarate, Choline, Citrate, Creatine, Ethanol, Gamma-aminobutyric acid, Glucose, Glutamine, Glutamate, Glycine, Glycerophosphocholine, Glutathione, myo-Inositol, Lactate, N-acetyl-aspartate, N-acetyl-aspartyl-glutamate, Phosphocholine, Phosphocreatine, Phosphoenolamine, Scyllo-inositol, Taurine. The default LCModel macromolecular and lipid basis sets were also included in the quantification.
Tissue correction was performed following analysis according to recent recommendations and guidelines 22. The co-registration and segmentation functions from Gannet (Version 3.1) 25 were used to register the MRS voxel to the corresponding anatomical image and segment it into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF). Tissue correction using literature values for T1 and T2 relaxation times and proton density was completed using the equations specified in the expert consensus in Near et al. 2020 22.
Symptom Measure
Parents completed the Health and Behavior Inventory (HBI) 26, which consists of 20 Likert style questions to assess concussion symptoms and severity and provides summary scores for somatic and cognitive symptoms. It has been adopted as a core measure in the common data elements for pediatric Traumatic Brain Injury (TBI) 27. Parents rated post-injury symptoms at the first (~ 12-days post injury), 4-week (no imaging), 3-month and the 6-month visits. Pre-injury symptoms were also rated retrospectively by parents at the first assessment. Using the pre-injury symptoms, the reliable change 28,29 of the HBI was determined to classify the children with concussion into symptomatic and asymptomatic groups for somatic and cognitive symptoms at 4-weeks, 3-months, and 6-months. This has been validated and done in previous studies 13,29.
Statistics
All statistical analyses were completed using IBM SPSS 26 (IBM Corp. Released 2019. IBM SPSS Statistics for MacOS, Version 26.0. IBM Corp., Armonk, NY, USA).
The first analysis used linear mixed effects models to examine whether metabolites (tNAA, tCr, tCho, Glx, mI) changed over time in concussion participants. A repeated measures linear mixed model was used to account for data loss between timepoints. A second set of linear mixed effects models then examined whether metabolites were affected by different injury groups (Asymptomatic, Symptomatic, OI) over time; injury group was determined by the reliable change at 4-weeks. A timepoint by group interaction term was included to test for changes in the different groups over time. These models included fixed effects for age and sex, while controlling for site and vendor as random effects.
The second analysis used univariate ANCOVAs to examine cross-sectional group differences in metabolites at the 3- and 6-month timepoints. Concussion participants were classified as asymptomatic or symptomatic according to the reliable change index at each timepoint studied. Therefore, 2 sets of analyses were done for each metabolite for 3- and 6-month metabolite data (3-month symptom reliable change for 3-month metabolites, and 6-month symptom reliable change for 6-month metabolites). These models also controlled for age, sex, site, and vendor as covariates.
The third analysis contained 10 sets of linear regressions for each metabolite. They all examined the association between metabolite levels and the concussion symptom ratings (2-measures of symptoms, cognitive and somatic) at 12-days, 3-months, and 6-months in concussion participants. The ten models determined the relationship between each of the following; metabolites at 12-days and symptoms at 12-days, metabolites at 12-days and symptoms at 3-months, metabolites at 12-days and symptoms at 6-months, metabolites at 3-months and symptoms at 3-months, and lastly metabolites at 6-months and symptoms at 6-months. Each model controlled for age, sex, site, and vendor as covariates.