Cerebral Blood Flow Predicts Recovery in Children with Persistent Post-Concussion Symptoms after Mild Traumatic Brain Injury

Abstract

Persistent post-concussion symptoms (PPCS) following pediatric mild traumatic brain injury (mTBI) are associated with differential changes in cerebral blood flow (CBF). Given its potential as a therapeutic target, we examined CBF changes during recovery in children with PPCS. We hypothesized that CBF would decrease and that such decreases would mirror clinical recovery. In a prospective cohort study, 61 children and adolescents (mean age 14 [standard deviation = 2.6] years; 41% male) with PPCS were imaged with three-dimensional (3D) pseudo-continuous arterial spin-labelled (pCASL) magnetic resonance imaging (MRI) at 4–6 and 8–10 weeks post-injury. Exclusion criteria included any significant past medical history and/or previous concussion within the past 3 months. Twenty-three participants had clinically recovered at the time of the second scan. We found that relative and mean absolute CBF were higher in participants with poor recovery, 44.0 (95% confidence interval [CI]: 43.32, 44.67) than in those with good recovery, 42.19 (95% CI: 41.77, 42.60) mL/min/100 g gray tissue and decreased over time (β = -1.75; p < 0.001). The decrease was greater in those with good recovery (β = 2.29; p < 0.001) and predicted outcome in 77% of children with PPCS (odds ratio [OR] 0.54, 95% CI: 0.36, 0.80; p = 0.002). Future studies are warranted to validate the utility of CBF as a useful predictive biomarker of outcome in PPCS.

Introduction

Traumatic brain injury (TBI) is the leading cause of morbidity and mortality worldwide, and its incidence is highest in children.^1,2 Mild TBI (mTBI) accounts for 90–95% of all TBIs.^2,3 The annual incidence of mTBI in children visiting the emergency room is 691 per 100 000 people in the United States and Canada^1,4–6; 20% of youth experience at least one mTBI by 16 years of age.⁷ Approximately 25% of these will have persistent post-concussion symptoms (PPCS) lasting ≥1 month or longer.^8,9 These neuropsychological problems span physical (e.g., headaches, dizziness), cognitive (e.g., problems with memory and concentration), behavioral (e.g., anxiety and mood disturbances), and sleep complaints.^10–12 These symptoms are often increased during activities such as homework and physical exercise.^12,13 There is a considerable burden associated with PPCS and a significant impact on all domains of health-related quality of life, and yet there are few evidence-based treatments.^10,14–16 In order to develop better treatments for children with PPCS, it is essential to determine the neurobiological underpinnings of poor outcome. PPCS have been linked to increased cortical network activation,^17,18 decreased cerebrovascular reactivity,^19,20 and, more recently, changes in cerebral blood flow (CBF). Unfortunately, many neuroimaging parameters correlate poorly with outcome, partly because the absence of symptoms does not always coincide with brain recovery.^21,22

Altered CBF is a well-recognized phenomenon in TBI and is likely to play a role in the secondary brain injury that evolves following the primary insult.^20,23–25 However, there is a significant knowledge gap about the role of CBF in the pathophysiology of PPCS and how it changes over time, especially in children. Recent research suggests it may play a more complex role than previously appreciated.^26,27 CBF alterations during recovery can occur as a result of structural microvascular injury, changes in cortical excitability and network activation,¹⁷ altered physiological coupling of CBF to metabolic demand,²⁷ and/or abnormal cerebrovascular reactivity (CVR).²⁸ These changes in CBF have potentially important implications for the management of mTBI: therapies that increase cortical activation in the setting of decreased CBF could exacerbate secondary injury²⁹; chronic decreases in perfusion may lead to altered inflammatory responses and long-term mood/cognitive problems³⁰; and pragmatically, abnormal cerebral physiology may decrease the reliability of functional magnetic resonance imaging (fMRI), which is a popular research tool in brain injury.³¹ Further, if CBF is indeed related to recovery, it may offer an objective biological marker of injury and/or treatment response, particularly if combined with serological markers of injury.³²

Arterial spin labeling (ASL) is a quantitative method of assessing cerebral tissue perfusion using MRI.³³ It is performed without manipulating the cerebral circulation (i.e., inducing vasodilation by breath holding or carbon dioxide inhalation), and is therefore particularly suitable for research with children.³⁴ It is intrinsically quantitative and reproducible, allowing it to be used for longitudinal evaluation.³⁵ Studies examining CBF in mTBI that have examined cohorts with persistent symptoms report both increased and decreased CBF patterns.^26,27,36,37 We have previously shown that PPCS at 4–6 weeks post-injury are associated with global increases in CBF, whereas children with mTBI who had clinically recovered had decreases in CBF.²⁶ The aim of this study was to examine how CBF changes in children with PPCS over time. We hypothesized that clinical recovery would be associated with decreases in CBF when compared with children who remained symptomatic.

Methods

Participants 8–18 years of age with a medically diagnosed mTBI were recruited from the Alberta Children's Hospital Emergency Department. Concussion/mTBI was defined using the American Academy of Neurology criteria,³⁸ confirmed by physician at medical assessment.^8,39 Concussion was considered to be part of the spectrum of mTBI.⁴⁰ Participants were enrolled using a two-step process: first, by telephone at 2–4 weeks, and then in person 2 weeks later. Eligible children were invited to participate in this neuroimaging study as part of the Play Game trial (NCT01874847).⁴¹ We enrolled children if they had PPCS, and a ≥10-point increase in their total symptom score on the Post-Concussion Symptom Inventory (PCSI) post-injury when compared with their pre-injury scores (assessed at enrollment).⁴² Children were ineligible if they had a significant medical or psychiatric history, a previous concussion within the last 3 months, persistent symptoms following a previous concussion, or a more severe TBI previously. Other exclusions included: use of neuroactive drugs, claustrophobia, and inability to complete questionnaires. Spoken and written consent was obtained from parents/guardians and children, respectively. At the time of enrollment, a standardized interview and medical examination were performed. The PlayGame trial was a clinical trial assessing the efficacy of melatonin (compared with placebo) for the treatment of PPCS. Neuroimaging studies were performed before the study drug was started (Session 1) and after the study treatment had ended (at least 48–72 h; Session 2). Details of the study process are given in Figure 1. Ethical clearance was granted by the University of Calgary Conjoint Health Research Ethics Board (REB13-0372) and the University of Queensland (2017001523).

FIG. 1.

Details of study recruitment.

Outcome measures

Recovery status

Recovery was assessed at 8–10 weeks post-injury and was determined using standardized clinical interview and examination and the Post-Concussion Symptom Inventory, Youth Report (PCSI-Y). This standardized questionnaire of 26 symptoms provides an overall rating of PPCS. Each symptom is rated on a Likert scale between 0 and 6, (with a range of total scores from 0 to 156). The PCSI has four specific domains derived from factor analysis (physical, cognitive, emotional, and sleep) and a high level of internal consistency and reliability, α = 0.92.^43,44 Children were considered to have had a good recovery if their symptoms were at or below pre-injury levels and they had returned to normal activities.^9,41,45 Those whose symptoms were above pre-injury levels were considered to have had a poor recovery during this time period.^41,46

Neuroimaging

Neuroimaging was performed at 4–6 weeks post-injury and repeated 4–6 weeks later. A high-resolution anatomical three-dimensional (3D) T1-weighted Bravo scan (slice thickness = 0.8 mm, 226 slices) and a 3D pseudo-continuous arterial spin-labelled (pCASL) MRI scan (slice thickness = 3.5 mm, labelling duration = 2 sec, 34 slices) was performed on a 3T GE MR750w Discovery scanner with a 32-channel head coil. The 3D ASL scan was automatically processed into quantitative CBF maps using the scanner-integrated pipeline with default settings (partition coefficient of 0.9; blood T1 of 1.6 sec).

One compartment model was applied to convert 3D pCASL scans to quantitative CBF maps using a proton density map. Pre-processing of MRI scans was completed in Statistical Parametric Mapping (SPM12).⁴⁷ Each anatomical scan was manually reoriented to the anterior commissure-posterior commissure line. Anatomical scans were then segmented (gray matter, white matter, cerebrospinal fluid, bone, soft tissue) using default tissue probability maps. A regression algorithm was used to correct for partial volume effects.⁴⁹ All CBF maps were co-registered with gray matter tissue maps for each patient to account for patient head motion between scans. CBF was normalized with mean gray matter CBF in order to reduce data noise.⁴⁹ The co-registered CBF map and anatomical scan were both normalized to Montreal Neurological Institute space in SPM12. Finally, the CBF map was smoothed with a Gaussian kernel (full-width half maximum [FWHM] = 8 × 8 × 8 mm) to increase signal-to-noise ratio and generate a Gaussian distribution of voxels, and minimize inter-individual variability in data noise contributed by global CBF.⁴⁹

Statistical analysis

Second-level analyses were performed using SPM12 and statistical non-parametric mapping (SnPM) toolboxes to evaluate between group differences in relative CBF (rCBF) using cluster level analyses. SnPM utilizes a Monte-Carlo permutations approach to correct for multiple comparisons, and reduce false positive error rates, minimizing parametric assumptions relating to the null distribution hypothesis.⁵⁰ A two-sample t test was first performed to contrast rCBF differences in recovered versus unrecovered groups at Session 1. In order to analyze the effect of time and recovery on rCBF, a repeated measures analysis of variance (ANOVA) examined the effect of time (Session) and group (good recovery at Session 2 vs. poor recovery). Although all imaging was performed off treatment, any potential confounding effect of melatonin treatment on CBF change was assessed using a repeated measures ANOVA comparing treatment groups (placebo, melatonin 3 mg, and melatonin 10 mg). All statistical contrasts included age and gender as covariates. For all rCBF analyses, the height (voxel-level) threshold was established at p < 0.001_uncorrected using a conservative search threshold (cluster size >60), and the spatial extent threshold was set using familywise error (FWE) for multiple comparisons (p_FWE = 0.05, cluster size >100). Multiple comparisons correction was achieved using 10000 Monte-Carlo permutations using SnPM.

Normalcy was established using the Kolmogorov–Smirnov Normal Test. Group differences were analyzed using t tests for parametric data (body mass index, days post-injury) and Mann–Whitney U test (age, income, PCSI-Y) for non-parametric data. Chi-squared or Fishers Test was used to compare proportions between groups. Mean absolute CBF was calculated as the averaged CBF of gray matter. A random-effects linear mixed-model analysis was used to model mean global CBF as a function of recovery groups (i.e., good recovery, poor recovery) and time, with participant entered as a random factor. Fixed factors were group, pre-injury total PCSI-Y score, pre-treatment total PCSI-Y score, age at injury, and gender. Post-hoc comparisons were corrected using the Scheffe test. Logistical regression was used to model recovery as a function of mean global CBF at Session 1. Statistical analyses were performed using SPSS (IBM SPSS Statistics for Mac, Version 25.0. Armonk, NY: IBM Corp.) and Stata Statistical software Release 15.

Results

After removing 14 scans because of motion artefacts, normalization failure, and slice truncation, 71 participants had pCASL imaging at Session 1 and 61 (80%) of these participants (mean age 14 [standard deviation 2.6] years; 41% male) had imaging repeated at Session 2, Figure 1. Participants did not differ significantly in their demographic, clinical or injury characteristics from those who did not have imaging performed. Twenty-three participants had recovered 4 weeks later at the time of the second scan (Good Recovery group). The clinical and demographic details are reported in Table 1. Clinical and demographic characteristics in recovery groups were not statistically different at the first neuroimaging session (Session 1) other than the pre-injury PCSI-Y total score, which was higher in the Good Recovery group, U = 627; p = 0.004. As expected, the post-injury PCSI-Y score was higher in the Poor Recovery group at session 2, U = 138; p < 0.001.

Table 1.

Demographic and Injury Details of Children with PPCS and Different Recovery Trajectories

	Good recovery (n = 23)	Poor recovery (n = 38)	Test stat	p value
Age, mean (SD) years	14.4 (2.9)	14.0 (2.4)	514	0.252
Gender, male n (%)	13 (52)	12 (48)	3.69	0.055
Handedness (R), n (%)	20 (87)	34 (90)	0.089	0.765
Body mass index	21.0 (4.3)	20.0 (3.6)	0.894	0.397
Average household income, mean (SD)	136,369 (56,973)	128,853 (47,925)	456	0.778
Migraine, n (%)	13 (33)	23 (41)	0.087	0.768
ADHD, n (%)	4 (10)	4 (7)	1.15	0.356
Learning support, n (%)	8 (20)	12 (21)	0.0	0.654
Injury details
Loss of consciousness, n (%)	2 (9)	5 (14)	3.98	0.136
Cause of injury, n (%)			2.87	0.574
• Sport-related	19 (83)	26 (68)
• Fall	1 (4)	4 (11%)
• MVA	0	3 (8)
• Other	3 (13)	5 (13)
Days post-injury (mean, SD)
• Session 1	37 (6.5)	37 (5.3)	-0.06	0.951
• Session 2	69 (6.7)	70 (6.5)	-0.66	0.507
PCSI-Y (median, IQR)
• Pre-injury	8 (3, 24)	2.5 (1, 7)	627	0.004
• Session 1 (pre-treatment)	30 (17, 43)	43 (27, 70)	307	0.053
• Session 2 (post-treatment)	1 (0, 10)	20 (11, 43)	138	0.000

ADHD, Attention deficit hyperactivity disorder; IQR, interquartile range; MVA, motor vehicle accident; PCSI-Y, Post-Concussion Symptom Inventory Youth report; PPCS, persistent post-concussion symptoms; SD, standard deviation; Test stat

Relative CBF maps

Those participants with poor recovery had higher rCBF, demonstrated in Figure 2. Using a repeated measures ANOVA and controlling for age and gender, there were no significant differences in rCBF over time or by recovery group. There was no effect of melatonin treatment group on rCBF change over time compared with placebo.

FIG. 2.

Relative cerebral blood flow (CBF) maps demonstrating regions of increased relative CBF (rCBF) (red) in those participants with persistent post-concussion symptoms (PPCS) who failed to recover over the next 4–6 weeks. Color image is available online.

Mean gray matter CBF

Mean absolute gray matter cerebral blood flow (aCBF) at 4–6 weeks was not significantly correlated with total PCSI-Y score (Pearson's = -0.194, p = 0.133). aCBF, however, was higher in those participants with poor recovery, 44.0 (95% confidence intervals [Cis]: 43.32, 44.67) compared with those with good recovery, 42.19 (95% CIs: 41.77, 42.60) mL/min/100 g gray matter tissue as demonstrated in Figure 3. aCBF decreased over time (β = -4.38; p < 0.001) across groups (β = -1.75; p < 0.001), and there was a significant time by recovery interaction (β = 2.29; p < 0.001, Fig. 3).

FIG. 3.

Mixed effect linear model of mean absolute gray matter cerebral blood flow (CBF) changes between 4 and 6 weeks post injury (Session 1) and 8–10 weeks post-injury (Session 2) in children with persistent post-concussion symptoms. CBF significantly differed at both time points in those children with good recovery (blue) and those who did not recover (red). The trajectory of CBF change was steeper in children who went on to recover by 10 weeks post-injury. Color image is available online.

Predicting recovery using CBF

Using logistical regression, mean aCBF at 4–6 weeks post-injury significantly predicted recovery in 77% of children with PPCS, controlling for any effects of age and gender. Participants with lower aCBF were less likely to have poor recovery (odds ratio [OR] 0.54, 95% CI: 0.36, 0.80; p = 0.002). Receiver operating characteristic analysis demonstrated an area under the curve (AUC) of 0.77 (95% CIs: 0.69, 0.89) indicating a good accuracy profile, see Table 2 and Figure 4. Further, mean aCBF remained a significant predictor of recovery (OR: 0.54, 95% CI: 0.36–0.80; p = 0.002) when the effects of total PCSI-Y score were considered (OR: 0.97, 95% CI: 0.94–1.0; p = 0.088).

Table 2.

Logistical Regression Analysis to Predict Recovery in 61 Children with Persistent Post-Concussion Symptoms

Predictor	Beta	SE Beta	Wald's χ2	df	p	Odds ratio	95% CIs
Constant	28.002	9.029	9.617	1	0.002
Mean aCBF	-0.624	0.206	2.909	1	0.002	0.536	0.358, 0.803
Age	-0.060	0.118	0.254	1	0.615	0.942	0.747, 1.188
Gender (female = 1)	0.349	0.630	0.307	1	0.580	1.417	0.413, 4.869
Hosmer and Lemeshow Test			3.650	8	0.887
Model Summary	-2 Log likelihood		Cox & Snell R2		Nagelkerke R2
	62.538		0.259		0.353

aCBF, gray matter absolute cerebral blood flow; CI, confidence interval; SE, standard error.

FIG. 4.

Receiver operating characteristic (ROC) curve demonstrating the discriminative ability of mean gray matter absolute cerebral blood flow at 4–6 weeks post-injury to predict recovery over the next 4–6 weeks in children with persistent post-concussion symptoms (PPCS). Area under ROC curve = 0.77 (95% confidence interval [CI]: 0.65, 0.89). Color image is available online.

Discussion

This is the first study to examine the changes in CBF over the subacute to early chronic recovery period in children with PPCS. We hypothesized that CBF would change over time in children with persistent symptoms and that these changes would be correlated to recovery. Our hypotheses were fully supported. We observed significant differences in both relative and global absolute CBF at 4–6 weeks post-injury in those participants who recovered over time compared with those who failed to recover. Mean aCBF decreased over time in both groups, although the rCBF did not change significantly. Those participants who recovered had lower aCBF at 4–6 weeks post-injury and also had a steeper decline in aCBF over time. Further, aCBF at 4–6 weeks post-injury significantly predicted subsequent recovery at 8–10 weeks post-injury with good accuracy (AUC: 0.77).

This study continues on from our previous study in which we found significant differences in rCBF in children with mTBI according to their recovery pattern at 4–6 weeks post-injury.²⁶ Children with PPCS had higher rCBF than children with a good early recovery at the same time point 4–6 weeks post-injury. The recovery rates observed in this study are similar to those observed in our previous natural history studies.^8,9,45 The current study demonstrates not only that CBF is different in children with PPCS, but also that the rate at which CBF decreases correlates with subsequent recovery. PPCS symptom scores did not correlate with CBF nor were they predictive of remaining symptomatic. Indeed, aCBF remained a significant predictor of outcome even when symptom scores, age, and gender were added to our regression model. Taken together, these results support our hypothesis that CBF is a useful biomarker in pediatric PPCS.

Cerebral perfusion is a process that, under normal conditions, dynamically responds to the continually changing metabolic demands of the brain.⁵¹ This critical function must be tightly coupled to neuronal energy demands to ensure adequate oxygen, glucose, and metabolic substrates to maintain normal brain function.^52,53 Cerebral regional vascular blood flow is regulated by the neurovascular unit (NVU). This is an intricate functional and anatomical complex consisting of cellular components (i.e., neurons, vascular cells, and glial cells) and their metabolic interactions, and is responsible for regulating regional CBF.^54,55 Although the metabolite carbon dioxide tension is the most well-known mediator of cerebral vessel tone (and therefore CBF), other substances (products of neuronal and glial metabolism) also have vasodilator properties. This cerebrovascular reactivity is often impaired after TBI.^31,56 Traumatic injury can disrupt this system is several ways. Nitric oxide synthesizing interneurons are particularly important in the control of CBF. These belong to a family of γ-aminobutyric acid (GABA)-ergic inhibitory interneurons and are particularly susceptible to trauma.^57–59

Traumatic cerebrovascular injury can occur as a result of primary and secondary brain injury mechanisms.^60–62 The primary injury of TBI results from the biomechanical insult. Secondary brain injury, however, occurs as a result of the complex pathophysiological processes and metabolic cascades that ensue after the primary injury. In animal models, acutely there is loss and constriction of cerebral microvasculature.⁶³ By 2 weeks, there is regrowth of vessels, although many of these are abnormal. Although there is less pathological information in acute human mTBI, there is abundant evidence of vascular injury with microscopic hemorrhages (especially in the pericontusional region) in moderate and severe TBI. By 2–4 weeks, similar to in animal models, extensive cortical changes of vascular injury are present, including apoptotic and regenerative changes.⁶⁴ These CBF changes are age dependent however, with immature animals demonstrating increased hyperemia that resolves over time.^65,66

Although it is well recognized that CBF generally decreases after human acute mTBI, the literature describing the longitudinal CBF changes after is scarce, especially in children.^67–70 Most studies involve adult athletes with typical recovery patterns; that is, recovery occurring by 4 weeks.^{36,60,61,67,71–74} Overall, most of these studies report decreases in CBF in acute to chronic TBI.^{26,37,60,62,74–78} There is a paucity of literature, however, examining the relationship between symptoms and CBF.^60,61,73,79 Most studies have found that increased symptoms are associated with increased rCBF,^{26,61,68,76,80} whereas others have found the reverse.^60,67 Few studies, however, have participants whose recovery took >30 days.⁷⁰ Few studies have examined CBF in children, in whom, similar to in animal models, areas of increased and decreased CBF are seen.^26,37,61,81 As a result, there is a significant knowledge gap about CBF, its relationship with recovery, and its temporal change in children. Our study goes someway to address this knowledge gap, as we focused our study specifically on those children with delayed recovery after mTBI in order to gain further insights into pathophysiological processes underpinning pediatric PPCS. Indeed, global aCBF at 4–6 weeks post-injury predicted subsequent outcome. Others have also demonstrated the potential of CBF to predict outcome, albeit in more typical recovery patterns.^67,70

These changes in CBF have important implications for the management of mTBI. Decreased CBF in the setting of increased cortical activation could exacerbate secondary injury,¹⁷ and chronic decreases in CBF may lead to altered inflammatory responses and long-term mood/cognitive problems.⁸² Further, CBF may also be a treatable biological target as it is amenable to manipulation with many pharmacological (e.g., sildafenil) and non-pharmacological (e.g., exercise) therapies that may improve outcome.^83,84

Although our study has several strengths (longitudinal design, well-phenotyped cohort, large sample size, and low attrition), there are some limitations. Recent recommendations suggest that a sample size of between 20 and 37 participants per group is needed to detect a 10% between-group difference in CBF with 80% power.^39,85–87 Although we generated relative CBF maps that have greater efficiency of power,⁴⁹ it is possible that our sample size was inadequate to detect rCBF changes over time. Global CBF is also susceptible to physiological effects that may increase within- and between-subject variability.⁴⁹ Although unlikely, it is possible that these factors affected recovery groups differently. These children were part of a clinical trial examining the effects of melatonin treatment in PPCS, and there was no effect of treatment group on changes in relative CBF over time. A control group would have provided further information about the role of mTBI on the observed CBF changes over time.

In sum, children with PPCS and poor recovery have higher relative and absolute CBF when compared with children with good recovery over the following 4 weeks. Mean global gray matter CBF at 4–6 weeks post-injury is a good predictor of outcome over the next 4–6 weeks, correctly predicting recovery in 77% of children with PPCS. As yet, there are insufficient data to support the widespread clinical use of ASL MRI in PPCS, and future research is necessary to verify these results in a validation cohort study.

Conclusion

There is accumulating evidence to support the important role of CBF in the pathophysiology of pediatric PPCS, and that it may differ from adult mTBI. CBF at 4–6 weeks post-injury is a promising predictive biomarker and therapeutic target in pediatric mTBI. Although the results of this study need to be confirmed in a validation cohort, CBF assessed using ASL fMRI may aide in the assessment of treatment and rehabilitation strategies employed in PPCS.

Authors' Contributions

Prof. Barlow conceived and designed, analyzed, and interpreted the data, and drafted and finalized the article. Dr. Iyer analyzed and interpreted the data, and critically revised the article. Dr. Scurfield, Ms. Yan, and Dr. Carlson were involved in data acquisition, analyzing the data, and critically revising the article. Prof. Wang was involved in data interpretation, and critically revising the article. All authors approved the version to be published.

Funding Information

This work was supported by the Canadian Institutes of Health Research (grant 293375), the Alberta Children's Hospital Research Institute, and the University of Calgary (grant 10006634).

Author Disclosure Statement

No competing financial interests exist.