Exercise Treatment for Postconcussion Syndrome

A Pilot Study of Changes in Functional Magnetic Resonance Imaging Activation, Physiology, and Symptoms

Abstract

Purpose: 

To compare functional magnetic resonance imaging (fMRI) activation patterns during a cognitive task, exercise capacity, and symptoms in postconcussion syndrome (PCS) patients who received exercise treatment (n = 4) with a PCS placebo stretching group (n = 4) and a healthy control group (n = 4).

Methods: 

Subjects completed a math processing task during fMRI and an exercise treadmill test before (time 1) and after approximately 12 weeks (time 2). Exercise subjects performed aerobic exercise at 80% of the heart rate (HR) attained on the treadmill test, 20 minutes per day with an HR monitor at home, 6 days per week. The program was modified as the HR for symptom exacerbation increased.

Results: 

At time 1, there was no difference in fMRI activation between the 2 PCS groups but healthy controls had significantly greater activation in the posterior cingulate gyrus, lingual gyrus, and cerebellum versus all PCS subjects (P < .05, corrected for multiple comparisons). At time 2, exercise PCS did not differ from healthy controls whereas placebo stretching PCS had significantly less activity in the cerebellum (P < .05 corrected) and in the anterior cingulate gyrus and thalamus (P < .001, uncorrected) versus healthy controls. At time 2, exercise PCS achieved a significantly greater exercise HR (P < .001) and had fewer symptoms (P < .0004) than placebo stretching PCS. Cognitive performance did not differ by group or time.

Conclusions: 

Controlled aerobic exercise rehabilitation may help restore normal cerebral blood flow regulation, as indicated by fMRI activation, in PCS patients. The PCS symptoms may be related to abnormal cerebral blood flow regulation.

RECENT developments have supported the use of standardized exercise testing and controlled aerobic exercise treatment as an approach to treating postconcussion syndrome (PCS).¹ This diagnostic and treatment approach proposes that one fundamental cause of refractory PCS is persistent physiologic dysfunction.² Physiologic dysfunction may include altered autonomic function and impaired autoregulation and distribution of cerebral blood flow (CBF).² Recent studies have reported about the efficacy, safety, and reliability of exercise assessment of concussion and for a program of controlled and progressive subsymptom threshold exercise rehabilitation in ameliorating PCS.^1,3 Exercise assessment and aerobic exercise training may reduce concussion-related physiological dysfunction by restoring autonomic balance and by improving autoregulation of cerebral blood flow.^4–8

Functional magnetic resonance imaging (fMRI) measures differences in the magnetic properties of oxygenated versus deoxygenated blood, which is called blood oxygenation level dependent (BOLD) imaging. Functional MRI is thus an indirect indicator of changes in local CBF to areas of increased neural activity. A number of studies have examined patterns of fMRI hemodynamic BOLD response related to PCS using working memory tasks. Several early studies reported abnormal activations during working memory tasks after mild traumatic brain injury.^9,10 Chen et al¹¹ found that 13 of 15 symptomatic athletes had a more widely distributed fMRI activation pattern relative to healthy controls during a working memory task even though their performances were similar. In a separate study, Chen et al¹² divided athletes with PCS into low and moderate symptom groups. The moderate symptom group showed reduced performance on a computerized cognitive measure whereas both PCS groups showed atypical fMRI activations during a working memory task when compared with controls.

Lovell et al¹³ compared fMRI (using the N-back test), a computerized battery of cognitive tests, and symptom reports of 28 concussed athletes with 13 healthy controls. They found that athletes with increased activation on initial fMRI had a delayed recovery versus athletes without increased activation. Initial changes in brain activation were associated with changes in self-reported symptoms and computerized cognitive test performance. Chen et al¹⁴ used fMRI at 2 time points with 4 athletes who recovered from PCS, 5 athletes with ongoing PCS, and 6 healthy controls. At time 1, athletes with PCS showed decreased activation in a region where normal controls showed increased activation during a working memory task. The athletes with PCS also showed increased activations at time 1 in other areas not seen among the normal controls. At time 2, the athletes who had recovered were similar to controls whereas PCS athletes showed activations similar to their time 1 patterns.

To our knowledge, there is no study that has used fMRI to assess activation changes before and after a specific treatment program for PCS. Reduction in symptoms with controlled exercise treatment may be associated with normalization of autoregulation of CBF as reflected by normalization of hemodynamic changes during a working memory task. In this study, consistent with Chen et al,¹¹ we interpret changes in fMRI BOLD signals to reflect changes in cognitive task-related neural activations that serve as an indirect measure of local CBF. The purpose of this study was to compare fMRI brain activation patterns in PCS patients who were given exercise treatment versus a placebo stretching treatment versus healthy controls. Prior to treatment, we expected different patterns during a working memory task for individuals with PCS when compared with healthy controls. After exercise treatment, we expected no difference between the exercise-treated PCS group and the healthy control group.

METHODS

Participants

Study participants included 10 consecutive patients (5 females; age range, 17–52 years) from a University Concussion Clinic with PCS and 5 healthy controls (4 females; age range, 18-34 years), who were recruited from the community. Patients were eligible for the study if (1) a clinical interview confirmed that the patient's symptoms met the World Health Organization International Classification of Diseases, Tenth Revision, criteria for PCS¹⁵; (2) they had an identifiable threshold (heart rate [HR] and systolic blood pressure) for exacerbation of symptoms on a progressive exercise test¹; and (3) they were symptomatic for 6 weeks or more but 12 months or less postinjury. Patients were otherwise “apparently healthy,” which was defined as absence of signs and symptoms of and low risk for cardiopulmonary disease according to the American College of Sports Medicine.¹⁶ Athlete was defined as someone involved in competitive scholastic/collegiate athletics or a competitive recreational athlete training on a regular basis. All participants gave informed consent in accordance with the University's Health Sciences Institutional Review Board.

Of the 10 patients with PCS, 2 were dropped from the study after the initial MRI examination because of concomitant/other diagnoses that could confound the fMRI results (posttraumatic stress disorder and malingering). One healthy control dropped out of the study after the first fMRI examination due to a scheduling conflict. Healthy controls (n = 4) were free from previous head injury. Some subjects had a history of prior concussions but we did not examine the effects of these because the sample was too small. See Table 1 for demographic and clinical summary of the study participants. The stretching group had 2 athletes (1 high school and 1 collegiate), the exercise group had 2 athletes (both college athletes), and the control group had 2 athletes (1 high school and 1 collegiate). Participants in the controlled aerobic exercise treatment group (n = 4) and the stretching (placebo) treatment group (n = 4) were blind to the expected effectiveness of their treatment condition. Participants in the placebo treatment group were subsequently offered exercise treatment.

TABLE 1: Demographics, time intervals, heart rate and symptoms for the PCS exercise, PCS stretching, and healthy control groups at time 1 and time 2

Design

The first 5 participants eligible for treatment were assigned to controlled aerobic exercise treatment. The second 5 participants with PCS were assigned to the placebo stretching group. This choice for assignment was based on the fact that interaction among participants was probable and we wanted participants to share a view of their assigned treatment as likely to be effective. Healthy controls were recruited to match the PCS group on the basis of age, sex, and athletic status.

There were no preenrollment interventions in this study. The PCS aerobic exercise group participants were instructed to perform a controlled and progressive aerobic exercise program at 80% of the HR attained on a treadmill test.¹ The treadmill test was performed in our clinic. Participants exercised for 20 minutes per day with a Polar heart rate monitor (Kempele, Finland), 6 days per week, at home or in a gym. This program was modified as the HR for symptom exacerbation increased.¹ Subjects were considered ready for the second fMRI scan when they were able to exercise up to age-predicted maximum HR without exacerbation of symptoms.¹ The PCS stretching group participants were provided with a booklet that explained (in written and figure format) a standardized and gradually progressive 12-week low-impact breathing and stretching program developed at the University. The PCS controls were instructed not to exceed a low target HR (40%-50% of age predicted max) so as not to affect cardiovascular fitness. Examples of stretches included quadriceps, double or single knee to chest, sitting hamstring, and so forth. Participants stretched with an HR monitor for 20 minutes per day, 6 days per week. All participants had an fMRI examination at baseline and again after approximately 12 weeks (Table 1). The healthy control participants also completed 2 fMRI scans with approximately the same time interval between the scans.

Treadmill test and symptom reports

The PCS subjects exercised on a treadmill at time 1 and time 2 following a Balke treadmill protocol to assess for symptom exacerbation as per previous studies.^1,3 Heart rate, blood pressure, symptoms, and perceived exertion were recorded every 2 minutes until either symptom exacerbation or exhaustion.¹ Resting concussion symptoms were recorded at time 1 and time 2, prior to exercise testing, using the Postconcussion Scale (PCS), a validated assessment instrument that includes 22 symptoms of concussion (headache, dizziness, photophobia, etc) with sound psychometric properties and normative data for men and women.¹⁷

fMRI conditions

Participants completed an fMRI math task that was modeled after the math task from Automated Neuropsychological Assessment Metrics (ANAM). This computerized cognitive test is widely used for baseline and postinjury assessment of concussion and had been used in previous behavioral studies in our group.¹⁸ The construct validity of the ANAM Math Processing subtest as a measure of processing speed and working memory has received support from confirmatory factor analyses with traditional neuropsychological measures.¹⁹ Participants saw simple math problems that were displayed in white 50-point font on one line centered on a black background. The math problem consisted of the addition and subtraction of 3 numbers. Participants were asked to press a button under their right index finger if the answer was less than 5 and to press a button under their right middle finger if the answer was greater than 5. Participants were asked to respond as quickly and as accurately as possible. Seventy-two trials were completed during the 5-minute run. The accuracy (percent correct) and speed (mean reaction time) were recorded for each participant. Subjects heard instructions over headphones just prior to the task. Stimuli were presented with Resonance Technology headphones, goggles, and a button response system (Resonance Technology Inc, Northridge, California; www.mrivideo.com).

fMRI acquisition

Functional and structural MRI data were acquired using a 3T GE Signa LX Excite 12.0 scanner (Buffalo Niagara MRI Center, Buffalo, NY) with an 8-channel head coil. Functional images were acquired using gradient echo T2* Echo Planar Imaging, which generated 33, 5-mm thick slices with no gap between slices (repetition time = 2500 msec, 144 repetitions for math task, echo time = 35 msec, voxel size = 1 × 1 × 5 mm³, matrix size = 64 × 64, field of view = 24 mm², flip angle = 90°). Image slices were aligned with the AC-PC plane. Functional images were overlaid on a high-resolution structural fast spoiled gradient scan (repetition time = 9.2 msec, echo time = 4.1 msec, voxel size = 1 × 1 × 1 mm³, flip angle = 20°).

fMRI analysis

Statistical Parametric Mapping 5 was used to analyze all fMRI experiments (Wellcome Department of Cognitive Neuroscience, London, England). Oblique axial images were realigned, coregistered, and normalized. Images were smoothed using a full-width half maximum 8-mm Gaussian smoothing kernel. A height threshold of P < .001 was selected for initial comparisons. Regions were considered significant that survived false discovery rate threshold of P < .05, which corrects for multiple comparisons. Given the few subjects in each group in our pilot study, we also report potential findings of interest at P < .001, uncorrected. The resting condition was subtracted from the math task condition. At time 1 and time 2, group comparisons included exercise PCS compared with controls, stretching PCS compared with controls, exercise and stretching PCS groups combined compared with controls, and exercise PCS compared with stretching PCS.

RESULTS

Table 1 presents the demographics, maximum HR achieved during the exercise treadmill test, and number of symptoms (as recorded on the Postconcussion Scale) reported at rest at time 1 and time 2 for each subject. A paired Student t test (2-tailed) comparing the increase in HR from time 1 to time 2 for the PCS exercise group was significant at P < .001. For the PCS stretching group, it was nonsignificant (P < .75). A paired Student t test (2-tailed) comparing the number of symptoms at time 1 to time 2 for the PCS exercise group was significant at P < .0004. For the PCS stretching group, it was nonsignificant (P < .16).

Table 2 presents the average accuracy and the average mean reaction times during the math processing task for the exercise, stretching, and healthy control groups. Paired Student t tests (2-tailed) revealed that there were no significant differences among all 3 groups at time 1 or at time 2.

TABLE 2: Average accuracy and mean reaction times for the PCS exercise, PCS stretching, and healthy control groups at time 1 and time 2^a

fMRI time 1 results

At time 1 there were no significant differences between the symptomatic exercise and stretching PCS groups; similarly, there were no differences between the exercise PCS group itself versus healthy controls and the stretching PCS group itself versus healthy controls. Table 3 and Figure 1 show that, at time 1, when compared with the 2 combined PCS groups, the healthy control group showed increased activation in the posterior cingulate gyrus, lingual gyrus, and cerebellum during the ANAM Math Processing subtest.

Figure 1: Math-Rest, time 1: Posterior cingulate and cerebellum activation greater in normal controls than all postconcussion syndrome subjects. Peak voxels are significant at P < .05, false discovery rate; Image threshold = P < .001, uncorrected.

TABLE 3: Math-Rest, time 1: regions more active in control subjects than in concussed patients (exercise and stretching PCS groups combined)^a

fMRI time 2 results

At time 2, there were no significant differences between the PCS exercise treatment group and the healthy control group. Table 4 and Figure 2 show that, at time 2, the PCS stretching group had less activity in the cerebellum, cingulate gyrus, and thalamus versus healthy controls. There were no regions showing increased activity for PCS stretching versus healthy controls at time 2. Although there were no clusters that survived our threshold for significance (P < .05, false discovery rate), the PCS exercise group had a small cluster (7 voxels) of increased activity in the cingulate gyrus (0, 17.52, 39.66 mm, Talairach coordinates) at time 2 when compared with the PCS stretching group (P = .055, cluster-wise, family-wise error). There was no region where the stretching PCS group had significantly increased activation versus the exercise PCS group.

Figure 2: Math-Rest, time 2: (A) Left cerebellum (P < .05, false discovery rate, corrected); (B) Left cingulate; and (C) thalamic (P < .001, uncorrected) activation in normal controls greater than stretching postconcussion syndrome subjects.

TABLE 4: Math-Rest, time 2: regions more active in control subjects than the stretching PCS group

DISCUSSION

This is the first study to provide support for the hypothesis that a specific treatment program of controlled aerobic exercise may restore patterns of hemodynamic response on BOLD fMRI to normal control levels in PCS patients during a cognitive task to a greater degree than a placebo stretching program. For the Math-Rest condition, the combined exercise and stretching groups of 8 PCS participants showed different activation patterns in several brain regions at time 1 when compared with the 4 participants in the healthy control group. At time 2, after treatment, the exercise PCS group had fMRI activation patterns that did not differ from the control group whereas the stretching PCS group differed from controls.

Interestingly, the PCS group performed similarly to the healthy control group for accuracy and speed on the working memory task at the pretreatment time point (time 1). The mean scores for our control group were at or below normative values for the ANAM Math Processing subtest; thus, it is unlikely that a ceiling effect would explain the lack of a difference between the groups. Pretreatment performance in the PCS participants, however, was associated with the “cost” of altered CBF distribution as measured by fMRI hemodynamic response. Prior to treatment, each PCS participant had a unique pattern of increased or decreased activation that differed from other individual participants. Individuals in the healthy control group did not show this somewhat random, individually unique pattern of increased or decreased activation. These findings are consistent with previous studies that used different working memory tasks and found that each individual PCS subject had a unique activation pattern versus controls.^11–14 These findings have been interpreted to reflect a compensatory mechanism whereby the brain allocates additional processing resources to accomplish a task that has become more difficult than before injury. Although performance may remain average for a period of time, it is likely that compensatory mechanisms lead to fatigue over an extended period of cognitive activity. Consistent with this, our PCS participants reported a greater number of symptoms at rest at time 1, including fatigue, when compared with normative data for noninjured individuals.¹⁷ We hypothesize that fatigue, a common complaint of PCS patients,^17,20 and perhaps specifically fatigue after working memory challenges,²¹ as well as other PCS symptoms, may be related to the increased activation seen on fMRI during cognitive tasks in symptomatic patients. We hypothesize that controlled aerobic exercise treatment may help to alleviate PCS symptoms by restoring normal CBF regulation and that this change may be associated with restoration of a normal hemodynamic response to working memory tasks.

The combined PCS groups in this study experienced an exacerbation of symptoms before they reached their maximum HR during a structured exercise assessment at time 1. At time 2, however, the PCS exercise group could reach maximum exercise capacity whereas the PCS stretching group could not. A recent study of pediatric sport concussion reported significant alterations in global CBF within 72 hours of injury in concussed athletes versus controls that persisted beyond 30 days in some subjects.²² One interpretation of our findings is that the controlled exercise treatment helped to restore normal autoregulation of CBF. It is conceivable that problems with control of CBF may be associated with changes in fMRI hemodynamic response during a working memory task, a hypothesis that warrants further study. In contrast to the exercise group, altered hemodynamic fMRI response persisted at time 2 for the PCS stretching placebo group versus healthy controls. This could be consistent with the different exercise capacities observed between the 2 PCS groups at time 2 as well as the difference in symptom reports at rest between the two groups at time 2. We hypothesize that the exercise and rest symptoms reported by PCS patients reflect abnormal patterns of hemodynamic response brought on by physiological and cognitive stressors. Controlled progressive aerobic exercise treatment may help to restore normal CBF regulation by conditioning the brain to gradually adapt to repetitive mild elevations of systolic blood pressure.²³ There might be an association between this treatment effect and normalization of hemodynamic responses seen in this study. This hypothesis could be tested in future studies that measure actual global CBF autoregulation during exercise as well as global CBF and fMRI activation patterns during a cognitive task.

Chen et al¹⁴ found no initial differences between their 3 groups of concussed subjects (recovered, did not recover, and normal control) in behavioral performance using a working memory task different than ours. The groups with PCS showed decreased activation in the left dorsolateral prefrontal cortex (Brodmann area 9/46), a region where normal controls showed increased activation. The groups with PCS also showed activations in other areas not used by normal controls in performing the task. As noted earlier, the authors interpreted this as compensatory recruitment of other brain regions to carry out the task.¹⁴ The participants in our study showed variable changes in activation without a consistent increase or decrease in the left dorsolateral prefrontal cortex. This discrepancy may reflect the different working memory task used by Chen et al¹⁴ versus the ANAM Math Processing task used in this study. Because the ANAM Math Processing task is widely used clinically to determine recovery from concussion,¹⁸ and is used in return to play decisions for athletes, the changes in activation seen during this task and the corresponding changes in self-reported symptoms and exercise capacity are especially interesting.

We studied the response to a specific treatment program rather than the post hoc spontaneous recovery in PCS studied by Chen et al.¹⁴ Nonetheless, there are similarities in the results. Consistent with our results, Chen et al¹⁴ found that athletes who spontaneously recovered showed hemodynamic response patterns that were similar to controls whereas athletes who did not recover had patterns similar to their own time 1 patterns. In both studies, healthy controls showed similar time 1 and time 2 activations, indicating effective CBF regulation as well as good retest reliability for the fMRI hemodynamic response during the working memory tasks. Self-report of symptoms reflected the changes over time in fMRI activations in both studies.

Lovell et al¹³ also found that initial PCS versus healthy control differences in fMRI brain activation patterns were associated with self-reported symptoms. In addition, they found decreased computerized cognitive test performance. In their study, though, there was increased activation in particular regions on initial fMRI associated with greater symptoms, decreased cognitive performance, and prolonged recovery. In our study, both PCS groups showed reduced activation when compared with healthy controls, which is similar to the findings of Chen et al.¹⁴

In interpreting the results of the group comparisons, a number of caveats need to be considered. Previous studies have shown that compensatory activations during working memory tasks in PCS patients vary among individuals. Thus, our findings may represent coincidental individual differences in recruitment of additional processing resources to accomplish a working memory task. An additional consideration is the possibility of a type II error due to the relatively small sample size. For example, there was not a significant difference for the PCS exercise group versus the healthy control group at time 2. Nonetheless, even considering the possibility of a type II error, greater differences were found between the PCS stretching and the healthy control groups at time 2 than between the PCS exercise and the healthy control groups. The time between concussion and first MRI was much greater for the exercise versus the stretching group (171 vs 65 days), which conceivably could have led to findings related to inherent group differences rather than a treatment effect. The exercise-treated group, however, recovered despite greater symptom duration whereas the placebo group did not.

The specific regions of fMRI activation that were different across groups can be interpreted cautiously with reference to previous studies. At time 1, there was decreased activation in the left posterior cingulate gyrus and in the right and left cerebellum for the 2 PCS groups when compared with the healthy control group. The posterior cingulate region has shown decreased activation in previous functional imaging studies among patients with mild memory problems related to Alzheimer disease.^24,25 Support for involvement of the cerebellum in cognition, including verbal working memory and executive functioning, has also been reported.^26,27

At time 2, the PCS stretching group showed decreased activation in the left cerebellum, left anterior cingulate, and the thalamus when compared with the healthy control group. In addition to the support for the cerebellum's role in cognition noted earlier, an fMRI study reported increased activity in the anterior cingulate cortex and prefrontal cortex during complex working memory tasks.²⁸ This result was interpreted to reflect engaging attention control and selection processes that support active maintenance in working memory. There is also support for a role for the thalamus in contributing to the control of visual attention and awareness.²⁹

In summary, a specific treatment program of controlled aerobic exercise treatment restored fMRI hemodynamic response during a cognitive task to normal control levels in PCS patients to a greater degree than a placebo stretching program. Hemodynamic response during the math cognitive task showed the expected differences between the groups. Before exercise treatment, the 2 PCS groups combined were different from healthy controls but not from each other. After treatment, the exercise group was not different from healthy controls whereas the placebo group showed several regional differences from healthy controls and also showed a small regional difference approaching significance when compared with the exercise group. Controlled exercise treatment may therefore help to restore normal local CBF regulation, at least as reflected by fMRI BOLD activation patterns, during a cognitive task. The PCS patients had reduced exercise capacity, more fatigue and other symptoms, and showed activations in other areas not used by healthy controls in performing the cognitive task; therefore, some PCS symptoms, such as fatigue after extended working memory activity, may be related to abnormal CBF regulation. This preliminary study sets the stage for a future randomized study with a larger sample (we estimate requiring 30 subjects per group) and more homogeneity among groups, which would correlate symptoms and cognitive performance with fMRI results. Furthermore, perhaps the same physiologic dysfunctions, problems with CBF regulation, and autonomic imbalance that are associated with exacerbation of symptoms during exercise are responsible also for the symptoms PCS patients experience during prolonged cognitive working memory tasks, such as fatigue and difficulty concentrating.