Multifaceted neural and vascular pathologies after pediatric mild traumatic brain injury

Abstract

Dynamic changes in neurodevelopment and cognitive functioning occur during adolescence, including a switch from reactive to more proactive forms of cognitive control, including response inhibition. Pediatric mild traumatic brain injury (pmTBI) affects these cognitions immediately post-injury, but the role of vascular versus neural injury in cognitive dysfunction remains debated. This study consecutively recruited 214 sub-acute pmTBI (8–18 years) and age/sex-matched healthy controls (HC; N = 186), with high retention rates (>80%) at four months post-injury. Multimodal imaging (functional MRI during response inhibition, cerebral blood flow and cerebrovascular reactivity) assessed for pathologies within the neurovascular unit. Patients exhibited increased errors of commission and hypoactivation of motor circuitry during processing of probes. Evidence of increased/delayed cerebrovascular reactivity within motor circuitry during hypercapnia was present along with normal perfusion. Neither age-at-injury nor post-concussive symptom load were strongly associated with imaging abnormalities. Collectively, mild cognitive impairments and clinical symptoms may continue up to four months post-injury. Prolonged dysfunction within the neurovascular unit was observed during proactive response inhibition, with preliminary evidence that neural and pure vascular trauma are statistically independent. These findings suggest pmTBI is characterized by multifaceted pathologies during the sub-acute injury stage that persist several months post-injury.

Introduction

There are approximately 750,000 new cases of pediatric mild traumatic brain injury (pmTBI; used synonymously with concussion) each year in the United States alone. ¹ Two primary concerns for the field ² are the timeframe for pathophysiological rather than symptomatic recovery from injury and the putative biological basis of persistent post-concussive symptoms (PCS) in approximately 15–35% of pmTBI. Preclinical data suggest a constellation of pathophysiologies that individually and collectively affect major components of the neurovascular unit (neurons, perivascular astrocytes, pericytes, microglia, etc.) post-trauma.^{3 –6} Not surprisingly, mild-to-moderate cognitive dysfunction in the domains of inhibition, attention, working memory and processing speed are broadly implicated in the acute to sub-acute injury phase, with long-term deficits (i.e., multiple months post-injury) on objective cognitive tests more actively debated.^7,8

Functional magnetic resonance imaging (fMRI) has been used to determine the physiological basis of cognitive dysfunction post-pmTBI.^{9 –13} The fMRI signal primarily arises from the neurovascular unit, ⁵ and therefore can be affected by cognitive impairment, non-specific cognitive factors (e.g., effort, pain), neuronal dysfunction, baseline perfusion of cerebral blood flow (CBF), or microvascular injury (pericyte loss, damage to tight junctions) that compromises cerebrovascular reserve post-injury. To date, few cross-sectional or longitudinal studies have applied multimodal imaging to disambiguate the different aspects of neurovascular trauma post pmTBI. ¹⁴

Age-at-injury represents a known risk factor for persistent PCS (adolescence > middle-to-late childhood) ¹⁵ and could moderate potential effects of pmTBI due to either age-related differences in plasticity or by adversely impacting critical periods of development.^{16 –18} Executive dysfunction is common following pediatric TBI. ¹⁹ Cognitive development continues throughout middle childhood and adolescence, and occurs later for more proactive (i.e., the active maintenance of goal-relevant information) relative to reactive aspects of executive control,^{20 –22} including response inhibition. The current study therefore extended previous findings ¹² of functional deficits in proactive response inhibition (i.e., inhibiting responses in a planned and purposeful manner) within motor circuitry post-pmTBI, and examined putative associations between age-at-injury/injury severity with neurovascular unit pathophysiology.

It was hypothesized that pmTBI would exhibit hyperactivity within motor circuits during the proactive phase of response inhibition (post-cue) in conjunction with hypoactivity during the actual processing of stimuli. ¹² Multimodal imaging assessed more global metrics of neurovascular coupling (fMRI) from resting perfusion and vascular trauma (cerebrovascular reactivity [CVR] measured during hypercapnia). ¹⁴ Objective injury characteristics (e.g., loss of consciousness [LOC]/post-traumatic amnesia [PTA]) were expected to associate more with task-related abnormalities and vascular dysfunction, whereas self-reported PCS and clinical risk scores ¹⁵ were not expected to account for significant variance.

Materials and methods

Participants

A total of 214 pmTBI patients (8–18 years old) were consecutively recruited from local Emergency Department and Urgent Care settings in this prospective cohort design (Figure 1; Supplemental Materials). Results from a subset (i.e., less than one-third) of participants have been previously reported. ¹² Inclusion criteria were based on the American Congress of Rehabilitation Medicine (upper injury limit) and the Zurich Concussion in Sport Group (lower injury limit). All pmTBI experienced a closed head injury with Glasgow Coma Score ≥13, loss of consciousness less than 30 minutes, post-traumatic amnesia less than 24 hours, an alteration in mental status, or at least two new symptoms. The sub-acute (SA) visit for pmTBI occurred 7.4 ± 2.2 days post-injury (maximum = 11 days), with the early chronic (EC) visit occurring approximately 4 months later (131.4 ± 14.7 post-injury). Age- and sex-matched healthy controls (HC; N = 186) were recruited from the community (word of mouth or fliers) and scanned at equivalent times (124.4 ± 15.6 days between visits) to control for neurodevelopment and/or repeat assessment. Sample size calculations were performed as part of the grant submission that funded the collection of the presented data.

Figure 1.

Flowchart of enrollment, inclusion and data quality assurance from the sub-acute (SA) and early chronic (EC) phases of injury for patients with a pediatric mild traumatic brain injury (pmTBI) as well as matched healthy controls (HC). The asterisk denotes the total number of participants who were eligible to return, which is the sum of participants with usable functional magnetic resonance imaging data from the proactive response inhibition task and those with quality assurance issues that may not be present at the EC visit.

All participants were excluded if reporting 1) major neurological diagnosis, 2) previous TBI with loss of consciousness >30 minutes, 3) developmental disorder (autism spectrum disorder or intellectual disability), 4) psychiatric disorders other than adjustment disorder, 5) contraindications for MRI (including pregnancy), 6) non-English speaking or 7) substance abuse/dependence. Patients with pmTBI were also excluded if they received general anesthesia as part of routine care. HC were excluded if diagnosed with Attention Deficit Hyperactive Disorder or a learning disability, whereas pmTBI with these conditions were included in the study. Urine-based drug screens were conducted for all participants at both visits, with a positive test resulting in exclusion for all participants. The study was approved by the University of New Mexico Institutional Review Board, and informed consent/assent was obtained based on the Declaration of Helsinki.

Clinical assessments

A Common Data Elements battery of tests were administered at both visits along with retrospective ratings (Supplemental Methods; Table S1). The battery included a modified version of the 5 P risk score ¹⁵ and a semi-structured pediatric interview to ascertain history of previous TBIs. The Post-Concussion Symptom Inventory, Conflict and Behavioral Questionnaire, and Pediatric Quality of Life Inventory served as primary clinical outcome measures. A full list of all primary and secondary clinical assessments is included in Supplemental Materials. Recovery from post-concussive symptoms (i.e., symptomatic versus asymptomatic) was determined using a normative rather than simple change approach to reduce false positives. ²³ A battery of neuropsychological tests investigated cognitive deficits, reading ability and effort (Table S1; Supplemental Materials). The entire clinical and neuropsychological batteries typically required between 70–90 minutes to administer, and individual tests within each battery were always administered in a fixed order.

Task paradigm

A full description of the proactive response inhibition task has been previously reported. ¹² Briefly, a multisensory cue (audiovisual; 300 ms duration) indicated (Figure 2(a)) the inhibition of upcoming motor responses (“NONE” = proactive response inhibition) or focused attention on a specific sensory modality (“HEAR” = attend-auditory; “LOOK” = attend-visual; active trial data will be presented in separate manuscript). Multisensory numeric probes (words = “ONE”, “TWO”, or “THREE”; 300 ms duration) occurred 2460 to 3380 ms post-cue (6 trials per each 8 second block). Stimuli were presented foveally and binaurally via headphones (head-centered). Participants were asked to withhold motor responses following “NONE” trials. Auditory and visual probes were always incongruent (i.e., different auditory/visual number) during NONE trials. Inter-block intervals were jittered (3700 to 5540 ms) to decrease temporal expectations between the cue and probe and minimize non-linear summing of the hemodynamic response function (HRF).

Figure 2.

A schematic (Panel a) depicting the proactive response inhibition trials with “NONE” cues and Arabic numeral probes. Variable inter-block intervals (IBI _v ) were used between trials, and variable inter-stimulus intervals (ISI_v) were used between the cue (corresponding boxes denoted with light grey outline) and the first simultaneously presented multisensory probe (dark grey outline) to generate independent hemodynamic response function for cues and probes. In contrast, the inter-stimulus interval between the remaining multisensory probes was fixed within each block (ISI_f). Auditory stimuli are denoted by speaker icon and quotes. Histograms (Panel b; bin = 1) with density plots (solid black line) representing the percentage (X-axis) of healthy controls (HC; blue shading) and pediatric mild traumatic brain injury patients (pmTBI; red shading) with errors of commission during sub-acute (SA) and early chronic (EC) assessments (Y-axis = percent of trails with erroneous responses). Although the overall error rate was low, pmTBI exhibited a significantly higher number of errors relative to HC across both visits (denoted by centralized asterisk).

MR imaging parameters

Imaging data were collected on a 3 T Siemens TrioTim scanner with a 32-channel head coil at a single site (Supplemental Methods). High resolution T₁–weighted (1.00 mm³), T₂–weighted (1.10 × 1.10 × 1.50 mm), susceptibility weighted images (1.00 × 1.00 × 1.50 mm) and fluid-attenuated inversion recovery images (0.80 × 0.80 × 3.00 mm) were reviewed by a blinded, board-certified neuroradiologist. ²⁴

Task (2 runs) and CVR data were acquired utilizing a single-shot, gradient-echo echoplanar pulse sequence with 56 interleaved slices acquired for whole-brain coverage (3.02 × 3.02 × 3.00 mm). Multiband imaging (slice factor = 8) achieved high temporal sampling (TR = 460 ms) of the HRF. Initial images from task (1208 final images) and CVR (598 final images) runs were discarded to account for T₁ equilibrium effects. Pseudo-Continuous Arterial Spin Labeling (pCASL; 45 tagged/untagged images) and proton density sequences were acquired with 20 interleaved slices for whole brain coverage (3.44 × 3.44 × 6.00 mm). The post-labeling delay (1800 ms) was selected based on recommended values for adolescents ²⁵ and predictions of trauma-related hypoperfusion.^{4,26 –28} All imaging data were collected in a single session in a fixed order (see Supplemental Materials).

MR processing and analyses

Task, CVR and CBF data were assessed for anomalous values and replaced using AFNI’s despiking tool. Task and CVR images were temporally interpolated (first slice), and spatially registered to the reference image in two- and three-dimensional space. Susceptibility distortion correction was accomplished using FSL’s Topup algorithm, non-linearly warped (AFNI 3dQwarp) to stereotaxic space, and spatially blurred (6-mm Gaussian filter). A single HRF was generated using finite impulse response deconvolution for both the cue (14.26 s post-cue onset) and target (20.70 s post-target onset) phases for task-based data. This was feasible secondary to relatively high temporal resolution and the cue-target jitter. Nuisance regressors modelled variance associated with 12 motion parameters, error trials, and a second-order polynomial to reduce hardware-related artifacts. ²⁹ Summed beta coefficients were divided by the average model intercept to calculate percent signal change (PSC) values for peak (2.30–5.98 s) and inhibitory (6.90–11.04 s) activity during the cue phase, and peak (2.30–5.98 s) and late peak (5.98–9.66 s) activity following targets.

Hypercapnia data were processed in MATLAB (version R2014) to develop an end-tidal (ETCO₂) regressor (Supplemental Materials). ³⁰ A spline function produced an ETCO₂ vector which was subsequently smoothed with a median filter (10 s window) and temporally resampled to match image acquisition. This ETCO₂ regressor was fit against global grey matter signal (subject-specific segmentation of T₁–weighted image) using a family of offsets (7.82 to 18.4 s post-image collection). Participants with a global grey matter fit of less than 0.7 were excluded from further analyses based on established criteria. ³⁰ Additional processing steps for BOLD CVR data included detrending with a 2^nd order polynomial function, followed by fitting the family of time-lagged, individualized ETCO₂ regressors on a voxel-wise basis. Resultant variables included maximum fit (highest Pearson correlation coefficient) and associated latency for maximal fit,^31,32 which grossly measure vascular tension and autoregulatory functioning. ³³ Fit maps were transformed to Fisher’s Z scores and the index of maximal fit was converted to real time (7.36 s + index*0.46 s) for maximal fit latency maps.

CBF images were registered to a proton density scan using 2 and 3-dimension motion correction algorithms (AFNI). The images were spatially blurred using a 6-mm Gaussian kernel, with pre-processed labeled image then subtracted from the paired control image. CBF was quantified based on established parameters [blood/tissue water partition coefficient = 0.9 mL/g; longitudinal relaxation time of blood = 1664 ms; labeling efficiency = 0.85; label duration = 1665 ms] and algorithms. ³⁴ T₁ magnetization correction and scaling of CBF was accomplished on a voxel-wise basis with the PD image and spatially normalized.

Statistical analyses

Clinical measures were assessed using age-at-injury (unit=month) and retrospective ratings (when available) as covariates. Clinical analyses were performed using either generalized linear models (GLM; Group effect only) or generalized estimating equations (GEE; Group and Visit) to account for repeated measures with Gaussian, gamma or negative binomial distributions.

A priori regions of interest (Figure 3(a)) were identical to our previous publication ¹² and included the left premotor cortex, bilateral supplemental motor area (SMA) and the left sensorimotor cortex (SMC). Individual 2 × 2 [Cue period: Group (pmTBI vs. HC)× Visit (Sub-Acute vs. Early Chronic)] or 2 × 2 × 2 [Probe period: Group (pmTBI vs. HC) × Visit (Sub-Acute vs. Early Chronic) × Phase (Peak Activation vs. Late Peak)] Linear Mixed Effects models (LME) were respectively performed with age and framewise displacement as covariates within motor ROIs. Similar LMEs (Group×Visit) examined group differences in resting CBF or CVR metrics.

Figure 3.

Panel a depicts a priori regions of interest (ROI) within motor circuitry derived from a previous publication using the same task, including the left sensorimotor cortex (SMC), bilateral supplementary motor area (SMA) and left premotor area (PrMot). ¹² ROI demonstrating increased activation during active trials relative to baseline (Base) across both groups are denoted in warm colors (red: p < 1 × 10⁻⁷; yellow: p < 1 × 10⁻⁹) for the lateral and medial portions of the left (L) hemisphere. Panels b (cue phase) and c (probe phase) depict differences in percent signal change (PSC) values for the entire hemodynamic response function (HRF) during the proactive response inhibition task (“NONE” trials) for pmTBI (red lines) and HC (blue lines) collapsed across visits (error bars =standard error of the mean). Shaded bars indicate the peak (light grey; 2.30–5.98 s) and inhibitory (dark grey; 6.90–11.04 s) phases of the HRF within motor circuitry for the cues, as well as peak (light grey; 2.30–5.98 s) and late peak (dark grey; 5.98–9.66 s) phases for the probes. Asterisks denote main effects of Group whereas guillemets (»«) denote interactions.

Voxel-wise imaging results are presented in Supplemental Results. Voxel-wise, whole-brain LMEs for the task were corrected for family-wise error using statistical (two-sided p < 0.001) and minimum volume thresholds (task = 604 µl) based on 10,000 Monte Carlo simulations and spherical autocorrelation estimates. ³⁵

Results

Demographics

The final sample included 183 pmTBI (see Table S2 for mechanism of injury) and 166 HC at the SA visit, with 137 pmTBI and 153 HC returning for the EC visit. Groups did not differ on biological sex, age, self-reported Tanner stage of development or handedness (all p’s ≥ 0.05; Table S2). Significant group differences were observed for self-reported history of previous head injuries (χ² = 7.89, p = 0.005; pmTBI = 19.7%, HC = 9.0%), parental self-reported psychopathology (Wald-χ² = 16.09; p ≤ 0.001; pmTBI > HC), premorbid reading ability (Wald-χ² = 22.61; p ≤ 0.001; pmTBI <HC), and effort (Wald-χ² = 23.39; p < 0.001; pmTBI < HC). Reading ability and effort were therefore used as covariates during neuropsychological analyses.

Clinical outcomes

Table 1 presents central tendency data for clinical and neuropsychological results. Self-reported PCS severity results indicated a significant Group × Visit interaction (Wald-χ² = 21.03; p < 0.001). Follow-up tests indicated that PCS severity (pmTBI > HC) was greater at the SA (Wald-χ² = 126.44; p < 0.001; 36.6% symptomatic) relative to EC (Wald-χ² = 9.78; p = 0.002; 15.3% symptomatic) visit, suggesting partial recovery for a subset of the sample. The primary quality of life measure and self-reported behavioral disturbances were not significant (all p’s > 0.0167) following Bonferroni correction.

Table 1.

Clinical and neuropsychological data for both groups at both visits.

Metric	Outcome	SA pmTBI(N = 183)	SA HC(N = 166)	EC pmTBI(N = 137)	EC HC(N = 153)
Symptom Measures
PCSI (% Max) b ,c	P	14.3 (4.8–34.5)	2.9 (0–8.8)	4.8 (0–14.7)	4.0 (0.8–8.7)
PROMIS Sleep b	S	19 (13–24)	13 (11–16)	18 (12–23)	14 (11–17)
PROMIS Anxiety b	S	3 (0–8)	1 (0–4.5)	2 (0–6)	1 (0–5)
PROMIS Depression a	S	2 (0–8)	0 (0–4)	1 (0–6)	1 (0–3)
Pain Scale b	S	3 (1–5)	0 (0–1)	0 (0–2)	0 (0–1)
HIT-6 b	S	52 (44–60)	40 (36–47)	44 (40–55)	42 (38–47)
Behavioral & Outcome Measures
CBQ b ,c	P	1 (0–3)	1 (0–2)	1 (0–3)	1 (0–2)
PedsQL	P	–	–	85.9 (76.1–93.5)	90.2 (82.1–95.7)
SDQ a	S	–	–	7 (4–10)	4 (2–8)
GOS-E a	S	1 (1–4)	1 (1–1)	1 (1–2)	1 (1–1)
Cognitive Measures
TOMMe10 a	S	10 (9–10)	10 (10–10)	10 (10–10)	10 (10–10)
WRAT4 a	S	49.3 (44.0–54.7)	54.4 (48.0–62.7)	50.0 (44.7–58.0)	56.0 (49.3–62.7)
PS b	P	47.4 ± 7.4	50.6 ± 8.1	50.7 ± 8.2	52.5 ± 8.8
AT b	P	47.8 (42.2–53.3)	52.2 (46.7–55.6)	50.0 (45.0–55.5)	53.3 (47.8–56.7)
WM	S	47.1 ± 7.9	49.9 ± 10.5	48.7 ± 8.8	50.8 ± 10.9
EF a	S	47.6 ± 7.0	50.5 ± 6.5	50.7 ± 6.7	53.1 ± 6.4
HVLT Delay a	S	8 (6–10)	9 (8–10)	7 (6–9)	9 (7–10)

Notes: Outcomes are classified as primary (P) or secondary (S). SA: sub-acute; EC: early chronic; HC: healthy control; pmTBI: pediatric mild traumatic brain injury; PCSI: Post-Concussion Symptom Inventory (presented as percent of maximum score to account for age-related scale differences); PROMIS: Patient Reported Outcomes Measurement Information System; HIT-6 : headache impact test; CBQ: Conflict Behavior Questionnaire; PedsQL: Pediatric Quality of Life Inventory; SDQ: Strengths and Difficulties Questionnaire; GOS-E: Glasgow Outcome Scale Extended; TOMMe10: Test of Memory Malingering – 10-item short version; WRAT4: Wide Range Achievement Test 4; PS: processing speed; AT: attention; WM: working memory; EF: executive function; HVLT Delay: Delayed recall on Hopkins Verbal Learning Task (measure of long-term memory). Data are either formatted at mean ± standard deviation or median (interquartile range).

^aGroup main effect.

^bGroup×Visit interaction.

^c=Group×Age or Group×Visit×Age interaction.

Secondary clinical measures (Bonferroni 0.05/7 = 0.007) indicated significant Group × Visit interactions for anxiety, sleep, pain and headaches (all p’s ≤ 0.007). Increased symptoms were observed for pmTBI relative to HC during the SA visit (all p’s ≤ 0.001), but group differences were no longer significant for anxiety, pain and headaches at the EC visit (all p’s > 0.05). Sleep complaints remained significantly (p = 0.017) elevated for pmTBI at 4 months, but were of lower magnitude relative to SA visit. Significant main effects (all p’s ≤ 0.001) for self-reported depressive symptoms, a parent-rated multi-dimensional measure of behavioral functioning, and trauma-related functional outcome indicated worse outcomes for pmTBI.

Primary (Bonferroni-corrected p’s < 0.025) cognitive analyses indicated a significant main effect for group (HC > pmTBI) for attention (Wald-χ² = 5.74; p = 0.017) and a significant Group × Visit interaction for processing speed (Wald-χ² = 5.30; p = 0.021). Follow-up tests suggested processing speed deficits at the SA (p = 0.011; HC > pmTBI) but not EC (p = 0.325) visit. A main effect of Group (HC > pmTBI) was observed for secondary cognitive measures of executive dysfunction (p = 0.009) and long-term memory (p < 0.001). The measure of working memory was not significant for group differences (all p’s > 0.0167).

Anatomical imaging and motion results

Positive day-of-injury CT scans (intracranial pathology or skull fracture) were observed in 7.6% of pmTBI during routine care. Four additional pmTBI were diagnosed with probable trauma-related findings on MRI scans. ²⁴ Significant group difference in head motion existed during the task (pmTBI > HC; Wald-χ² = 8.80; p = 0.003), as well as the expected negative relationship between head motion and age (p < 0.001). Motion was therefore used as a covariate for all analyses.

Proactive response inhibition: Behavioral and motor circuitry results

A 2 × 2 (Group×Visit) GEE indicated both increased errors of commission (Figure 2(b)) for pmTBI relative to HC during the proactive response inhibition task (Wald-χ² = 5.50; p = 0.019), and increased errors during the SA relative to EC visit (Wald-χ² = 6.88; p = 0.009). There was also a significant negative association between error rate and age (p = 0.005).

The a priori prediction (Figure 3(b)) for hyperactivity post-cue was observed within the bilateral supplemental motor area (SMA; Wald-χ² = 7.31; p = 0.007) for pmTBI, with trends in the left sensorimotor cortex (SMC; p = 0.088) and premotor cortex (p = 0.093). A main effect of Visit was also observed in left SMC (EC > SA; Wald-χ² = 5.71; p = 0.017). There were no significant main effects or interactions observed during the inhibitory phase, and age-at-injury did not interact with the group effect.

Significant Group×Phase interactions for the left SMC (Wald-χ² = 4.89; p = 0.028) and bilateral SMA (Wald-χ² = 7.06; p = 0.008) supported a priori predictions of motor circuitry hypoactivation during probe phase (Figure 3(c)). Follow-up testing indicated significantly decreased activation in the SMA (p = 0.044) and trend decreased activity within the left SMC (p = 0.077) for pmTBI during the peak phase. Age-at-injury did not interact with the group effect for any of the ROI within motor circuitry.

Significant Visit×Phase interactions were present for the SMC (Wald-χ² = 10.41; p = 0.001), SMA (Wald-χ² = 9.20; p = 0.003) and left premotor cortex (Wald-χ² = 5.26; p = 0.022), with follow-up tests indicating significant differences in the HRF shape for the SMA (p = 0.017; peak > late peak) at the SA visit, and for the left SMC (p = 0.019; late peak > peak) and premotor cortex (p = 0.022; late peak > peak) at the EC visit.

Similar findings of hypoactivation were observed for the SMA and bilateral inferior parietal lobule during voxel-wise analyses of probe data, with ventrolateral prefrontal cortex and cerebellar regions demonstrating increased activation for pmTBI (Supplemental Materials).

Multimodal disambiguation of neurovascular coupling

All motor ROI (Figure 4(a) and (b)) exhibited significant main effects of Group (pmTBI > HC) for both maximal fit of the ETCO₂ regressor (SMC: Wald-χ² = 12.38; p = 0.001; SMA: Wald-χ² = 7.50; p = 0.007; premotor cortex: Wald-χ² = 5.64; p = 0.018) as well as the latency of fit (SMC: Wald-χ² = 4.72; p = 0.031; SMA: Wald-χ² = 6.95; p = 0.009; premotor: Wald-χ² = 4.71; p = 0.031). A Group × Age interaction was significant for maximal fit (Wald-χ² = 4.36; p = 0.038), but all follow-up analyses were non-significant (p > 0.10).

Figure 4.

Schematic design (Panel a) for cerebrovascular reactivity (CVR) equipment and airflow (red arrows) during the task. Variable offsets are applied to the resultant end-tidal CO₂ data (ETCO₂; left y-axis; ribbon of colors, offsets of 7.82–18.40 s) to account for machine delays and individual physiological differences in the blood oxygen level dependent (BOLD) signal on a voxel-wise basis. The same procedures are applied to global grey matter signal (Glb GM; right y-axis; red trace in left column of Panel a, demeaned to overlay offsets) as part of the quality assurance protocol. The timing for the delivery of room air (down) and mixed medical gas (up) is represented by the black square wave. Box-and-scatter plots (elements: median, IQR, and 3*IQR or local maxima/minima) for CVR (Panel b) and cerebral blood flow (CBF; Panel c) are presented for pmTBI (red diamonds) and HC (blue diamonds) collapsed across visits and residualized to remove the effects of age and mean framewise displacement. Separate plots are presented for maximal fit (Fisher’s Z) and latency to maximal fit to the ETCO₂ data, with asterisks denoting significant main effects.

During the CVR analyses, significant main effects of Visit (EC > SA) were also present for latency of fit (SMC: Wald-χ² = 4.73; p = 0.031; SMA: Wald-χ² = 6.23; p = 0.013; premotor cortex: Wald-χ² = 4.30; p = 0.039).

Results from CBF analyses (Figure 4(c)) were null for Group main effects or interactions, although negative associations were observed between age and CBF in the left SMC (Wald-χ² = 210.69; p < 0.001), bilateral SMA (Wald-χ² = 260.65; p < 0.001) and left premotor cortex (Wald-χ² = 180.30; p < 0.001).

Neurovascular abnormalities and clinical correlations

A final series of analyses examined the relationship between clinical indices of injury severity (independent variables: LOC, PTA, mechanism of injury, positive structural finding, previous history of pmTBI, symptom burden and 5P risk score) and BOLD/CVR abnormalities (dependent variables) within the pmTBI group. Specifically, task-related activation in the SMA during the Cue phase, the average of SMC and SMA activation during the peak phase, as well as maximal fit and latency to maximal fit during CVR (average across three ROIs) were used for these analyses.

None of the objective markers of injury severity (LOC, PTA, presence of complicated pmTBI, mechanism of injury) were associated with CVR findings. In contrast, both self-report of a previous history of concussion (Cue: Wald-χ² = 4.19; p = 0.041) and the 5P clinical risk score (Cue: Wald-χ² = 15.31; p < 0.001; Probe: Wald-χ² = 4.05; p = 0.044) were negatively associated with activation during the cue phase and positively associated with activation during the probe phase.

Finally, CVR metrics of delay and maximal fit were not linearly associated with either cue or probe related BOLD activation within pmTBI (p > 0.10), suggestive of independent pathologies.

Discussion

The current study used a multimodal imaging approach to independently isolate neurovascular (fMRI during a proactive inhibitory control task) and vascular (CBF and CVR) pathologies within the motor circuit post-pmTBI. Patients exhibited multidimensional symptoms (cognitive, physical and emotional) approximately one week post-injury, with full recovery on several (e.g., anxiety, pain, headaches, processing speed) but not all (e.g., depression, sleep, attention, executive functioning, long-term memory) measures at four months, even after controlling for repeat visit and potential confounders (e.g., practice effects, reading ability, effort). Current and previous^19,36,37 findings therefore suggest ongoing neuropsychiatric disturbances months post-injury for a minority of pmTBI, and highlight the need for multi-dimensional clinical assessments. ³⁸ Importantly, the majority of these clinical deficits were relatively small in nature (Table 1), and significant only at the group level.

To date, only a handful of studies have examined BOLD deficits during cognitive tasks in well-defined, sub-acute pmTBI samples, ¹⁴ with even fewer prospective studies.^9,12 Similar to neuropsychological testing results, pmTBI exhibited increased errors of commission during the task, indicative of a failure in proactive response inhibition relative to HC. These behavioral deficits were associated with distinct neurovascular coupling abnormalities both immediately following the presentation of cues and during the subsequent processing of probes. As predicted,^11,12 pmTBI exhibited evidence of hyperactivation within the motor network following a cue to withhold a motor response. In the context of the current study, this putatively reflects an inability to suppress neural responses in a proactive fashion. ³⁹

In contrast, pmTBI demonstrated evidence of increased inhibitory activity (i.e., hypoactivation relative to baseline) within the SMA and SMC during the probe phase, which occurred approximately 3 seconds post-cue. This pattern replicates our previous findings, ¹² as well as other studies reporting reversals in signal abnormalities during longitudinal imaging post-pmTBI. ⁹ The SMA has been shown to be critical for inhibiting, planning and coordinating inter-hemispheric movements, and therefore plays a prominent role in proactive response inhibition. ⁴⁰ Both BOLD abnormalities were associated with past history of head injury and the clinically derived 5P risk score for prolonged symptoms. Other fMRI studies have also reported hypoactivation (HC > pmTBI) during both attention and working memory tasks. ¹⁴ Group differences in hyper- and hypoactivation were most pronounced within the SMA during both ROI and voxel-wise analyses, as well as across both phases of the task.

From a cognitive neuroscience perspective, current behavioral and functional imaging findings are suggestive of impaired cognitive planning and a more reactive response inhibition strategy (i.e., increased error rate). Specifically, neuronal activity was not adequately suppressed in pmTBI following the 100% predictive cue, in conjunction with a failure to appropriately release inhibition once probes were actually presented. The execution of proactive control occurs at more advanced cognitive neurodevelopmental stages,^{20 –22} and may susceptible to impairment post-pmTBI. The current proactive response inhibition trials required participants to withhold motor responses while actively attending to multisensory stimuli. Thus, there are limited behavioral metrics (i.e., a lack of reaction time data) through which to measure the deployment of proactive cognitive control compared to other continuous performance cognitive tasks. ⁴¹ Moreover, even though the current task was relatively easy, younger children still experienced more difficulty performing the task relative to adolescents. Future studies could consider developing different cognitive tasks to examine age-at-injury effects, but this approach obviates direct comparisons with age-at-injury as a continuous variable.

Mild TBI has been previously associated with both impaired long-interval intracortical inhibition and disruptions to local neocortical interneurons, as well as corticospinal excitatory deficits in preclinical and clinical studies.^42,43 Recent murine studies have also provided evidence of inhibitory and excitatory deficits post-TBI (mild and severe models) that are dependent on injury severity and time post-injury, as well as rapidly fluctuating compensatory mechanisms such as dendritic remodeling. ⁴⁴ Thus, a loss of excitatory-inhibitory balance post-pmTBI could explain findings of both hyper- and hypoactivity from a neural perspective, which may be further mediated by time-post injury, injury severity and the nature of the task. A reduction in proactive response inhibition may also contribute to more general behavioral issues post-TBI such as increased aggression and impulse control problems. Critically, these deficits were not limited to the motor network, but also occurred in ventrolateral prefrontal cortex, inferior parietal lobule and thalamus. Although several of these regions have been implicated in various aspects of cognitive control, previous studies have also reported disruptions in functional connectivity to these and other networks following pmTBI.^45,46 Future studies are therefore needed to examine the association between putative connectivity changes and dysfunction within the neurovascular unit.

Pure microvascular impairment remains a relatively understudied consequence of trauma that likely contributes to neurovascular coupling deficits in a complex fashion. ⁶ Current findings indicated a strong negative association between age and CBF in motor circuitry. This finding replicates previous reports of decreased CBF during neurodevelopment that may be further mediated by sex-related differences in older adolescents. ⁴⁷ However, in contrast to previous work in pmTBI,^{4,26 –28} there were no differences in resting CBF within the motor circuitry across the two groups. Instead, pmTBI exhibited both increased maximal fit and increased latency of fit during hypercapnia within the motor network. These results replicate other studies^{30,48 –50} reporting a similar constellation of CVR abnormalities (i.e., increased magnitude and latency during hypercapnia) in conjunction with normal CBF. These findings suggest that CVR may either be more sensitive to pmTBI than CBF, and/or require a longer recovery trajectory. CVR deficits may result from altered endothelial cell function, pericyte loss, or from changes in capillary distensibility and total capillary recruitment. All of these may affect cerebrovascular reserve post-injury and are indicative of a purer vascular pathology or disruptions to autoregulation.^6,30,50 These CVR and task-elicited BOLD-based deficits were not linearly related within the pmTBI sample. Thus, current findings suggest independent pathologies, or that vascular reserve affects the shape and timing of the hemodynamic response function in a more complex fashion post-TBI. Similarly, a recent study in adults with moderate-to-severe TBI indicated that traumatic vascular and axonal injury are spatially distinct. ⁵¹

There is great interest in identifying the biological bases of PCS and especially prolonged recovery. ² Previous findings have suggested that the amplitude of the BOLD signal during inhibitory control tasks is related to post-concussive symptoms, ¹¹ but these findings were not replicated in the current study. Similarly, our group has also not observed any relationships between persistent PCS and resting state abnormalities, ⁵² cortical and subcortical atrophy, ⁵³ diffusion ⁵³ or CVR abnormalities. Importantly, only ∼15% of pmTBI exhibited persistent PCS at the EC visit when using more stringent methods that adequately control for inflated false positive rates among healthy individuals. ²³ Thus, the ability to detect relationships between imaging abnormalities and more complex neuropsychiatric phenotypes may be challenging given both lower statistical power and lower test-retest reliability associated with PCS report. ⁵⁴

The spectrum of mTBI is characterized by heterogeneity in terms of mechanism of injury, initial injury severity characteristics (e.g., post-concussive symptoms only versus LOC up to 30 minutes) and injury location, all of which complicate prognosis and synthesis of data across studies. Past mTBI history and clinical risk scores ¹⁵ were associated with BOLD abnormalities in the current sample, whereas other traditional markers of injury severity (LOC and PTA) were not significant. Both preclinical and clinical studies suggest that children are more susceptible to diffuse injuries, and younger children performed worse on the task as expected.^{16 –18} However, age-at-injury was not associated with BOLD-based abnormalities post-cue (hyper-activation) or probe (hypo-activation) within the motor circuitry. Several grey and white matter regions exhibited significant Group × Visit × Age interactions, but the majority these findings were not significant during follow-up testing.

Strengths of the current study include a large, homogeneous and well-characterized pmTBI sample, high retention rates, multimodal imaging approach, and equivalent assessment of an equally powered, typically developing cohort to control for neurodevelopment and practice effects. However, the current study did not collect spectroscopy data to directly quantify inhibitory/excitatory neurotransmitter levels or blood-based biomarkers to complement imaging findings. For example, glial fibrillary acidic protein is increased post-TBI in both animal and human studies.^55,56 Astrocytes perform several critical functions during neurovascular coupling and are essential component for tight junctions. ⁵ Similarly, other blood-based biomarkers (e.g., claudin-5 and claudin-15) may further disambiguate different forms of vascular trauma (endothelial dysfunction versus blood-brain barrier breach). Second, a smaller number of individuals agreed to perform the CVR portion of the study due to parental/child concerns with breathing a medical grade gas mixture, as well as associated data loss due to the increased complexity of CVR data acquisition. Finally, although the current effort represents the largest study to date to characterize neurovascular coupling post-pmTBI, ¹⁴ the sample size may still have been too small to characterize more complex interactions between neurovascular trauma and effects associated with age-at-injury. Similarly, the current study did not examine the potential moderating effects of biological sex as a function of injury, with biological sex demonstrating complex developmental relationships with cognition, cerebral blood flow and CVR even among healthy children.^47,57

Conclusion

In summary, pmTBI remains a large public health concern due to the number of children affected each year and growing evidence of physiological abnormalities that may persist beyond typical symptom-based recovery windows. Current findings suggest neurovascular coupling abnormalities that persist up to four months post-injury during proactive response inhibition, as well as evidence of microvascular trauma that affects vascular reactivity and potentially autoregulation. ³³ These results highlight the role of multimodal neuroimaging for disambiguating various aspects of pathology post-pmTBI. Additional studies are needed to examine for direct evidence of neuronal dysfunction (as measured by electroencephalography or magnetoencephalography) during cognitive control tasks in addition to hemodynamic abnormalities.

Data availability

Supplementary material for this paper is available at: http://jcbfm.sagepub.com/content/by/supplemental-data. In addition, the data that support the findings of this study will be openly available in FITBIR at fitbir.nih.gov upon the conclusion of the study, reference number FITBIR-STUDY0000339.

Data citation

[dataset] Mayer, A.R. (Ongoing). The Impact of Diffuse Mild Brain Injury on Clinical Outcomes in Children. FITBIR (https://fitbir.nih.gov): FITBIR-STUDY0000339