Clinical Science

Original Paper

Impaired cerebral haemodynamic function associated with chronic traumatic brain injury in professional boxers

Damian M. Bailey, Daniel W. Jones, Andrew Sinnott, Julien V. Brugniaux, Karl J. New, Danielle Hodson, Christopher J. Marley, Jonathan D. Smirl, Shigehiko Ogoh, Philip N. Ainslie


The present study examined to what extent professional boxing compromises cerebral haemodynamic function and its association with CTBI (chronic traumatic brain injury). A total of 12 male professional boxers were compared with 12 age-, gender- and physical fitness-matched non-boxing controls. We assessed dCA (dynamic cerebral autoregulation; thigh-cuff technique and transfer function analysis), CVRCO2 (cerebrovascular reactivity to changes in CO2: 5% CO2 and controlled hyperventilation), orthostatic tolerance (supine to standing) and neurocognitive function (psychometric tests). Blood flow velocity in the middle cerebral artery (transcranial Doppler ultrasound), mean arterial blood pressure (finger photoplethysmography), end-tidal CO2 (capnography) and cortical oxyhaemoglobin concentration (near-IR spectroscopy) were continuously measured. Boxers were characterized by fronto-temporal neurocognitive dysfunction and impaired dCA as indicated by a lower rate of regulation and autoregulatory index (P<0.05 compared with controls). Likewise, CVRCO2 was also reduced resulting in a lower CVRCO2 range (P<0.05 compared with controls). The latter was most marked in boxers with the highest CTBI scores and correlated against the volume and intensity of sparring during training (r=−0.84, P<0.05). These impairments coincided with more marked orthostatic hypotension, cerebral hypoperfusion and corresponding cortical de-oxygenation during orthostatic stress (P<0.05 compared with controls). In conclusion, these findings provide the first comprehensive evidence for chronically impaired cerebral haemodynamic function in active boxers due to the mechanical trauma incurred by repetitive, sub-concussive head impact incurred during sparring training. This may help explain why CTBI is a progressive disease that manifests beyond the active boxing career.

  • boxing
  • cerebrovascular reactivity
  • chronic traumatic brain injury
  • cortical oxygenation
  • dynamic cerebral autoregulation


  • The underlying pathophysiology associated with sports-related traumatic brain injury is not understood. Professional boxing has the potential to provide unique insight into related mechanisms though traditional interest has focused on morphological and neuropsychological correlates to the exclusion of potential abnormalities in the functional integrity of the cerebral circulation.

  • To examine this, we compared male professional boxers against age- and fitness-matched non-boxing controls. Our findings identified that dCA, CO2 vasoreactivity and cortical oxygenation during orthostatic stress were selectively impaired in boxers. Functional impairment was more related to the volume and intensity of sparring during training than the frequency of knockouts inflicted during competition.

  • These findings provide the first clear evidence for chronically impaired cerebral haemodynamic function due to the mechanical trauma incurred by repetitive, mostly sub-concussive, head impact. This may help explain why CTBI is a progressive disease that manifests beyond the active boxing career.


The underlying pathophysiology associated with sports-related traumatic brain injury is not understood despite an estimated 3.8 million cases reported each year [1] at a cost of $56 billion to the U.S.A. alone [2]. Professional boxing has the potential to provide unique insight into related mechanisms given that its intrinsic objective is to render an opponent temporarily unconscious through blunt head injury. Indeed, prolonged exposure to repetitive head trauma can result in CTBI (chronic traumatic brain injury), often referred to as chronic traumatic encephalopathy, a neurological syndrome characterized by progressive impairments in cognitive, behavioural and motor function [3].

Thus early detection of neurological injury in professional boxers is important given its potential to prevent CTBI [4]. Much interest has focused on morphological and neuropsychological correlates to the exclusion of potential abnormalities in the functional integrity of the cerebral circulation [3,5]. In the few studies that exist, regional cerebral perfusion deficits and subsequent hypometabolism have been observed [69] though the underlying mechanisms remain to be established.

View this table:
Table 1 Subject characteristics

Values are means±S.D.

CBF (cerebral blood flow) is regulated through the complex interaction of multiple control systems. Cerebral autoregulation, a homoeostatic mechanism that serves to maintain CBF relatively constant over a wide range of perfusion pressures, and CVRCO2 (cerebrovascular reactivity to changes in CO2) are considered primary inputs that ultimately serve to maintain adequate tissue oxygenation and substrate delivery [1012]. Although both components have been shown to be impaired in the clinical setting [1316], the long-term effects of professional boxing remain to be examined.

In light of these findings, the present study was designed to test the following original hypotheses. First, critical aspects of cerebral haemodynamic function, namely dCA (dynamic cerebral autoregulation), CVRCO2 and cortical oxygenation during orthostatic stress would be compromised in professional boxers. Secondly, the magnitude of functional impairment would be related to the extent of sparring performed during training since this is where the greatest exposure to repetitive head trauma typically occurs [17]. Finally, these haemodynamic tests would serve as a more sensitive correlate of CTBI over more traditional subjective measures of neuropsychological dysfunction. Physical fitness-matched non-boxers served as control comparators to more accurately quantify the magnitude of potential dysfunction given the cerebrovascular adaptations known to occur in response to exercise training [18]. These hypotheses were tested in one of the most elite groups of professional boxers assembled to date.



The study was approved by the University of Glamorgan Human Research Ethics Committee (02/10). All procedures were carried out in accordance with the Declaration of Helsinki [19] of the World Medical Association and informed consent was obtained from all participants.


Participants were asked to attend the laboratory on two separate occasions at least 72 h following completion of their last bout of sparring and outside of their competitive boxing period. On their first visit, they were asked to bring their training diaries and were interviewed extensively to determine their combat history (see below). We also took the opportunity to familiarize them with the laboratory testing environment. All formal procedures were completed on their second visit with the order of testing fixed to minimize potentially confounding carry-over effects. The investigators were not blinded during experimental data collection but were during data analysis. Figure 1 provides a schematic overview illustrating the temporal sequence of the tests conducted.



We recruited 12 male professional boxers that included current World, European, Commonwealth and British champions from the Super Featherweight to Light Heavyweight Divisions. All boxers were free of drug, alcohol or cigarette abuse.


Twelve age- and physical-fitness-matched male non-boxers were recruited as controls. They performed a comparable volume, duration, intensity [HR (heart rate) assessed via telemetry] and mode (with the exception of sparring and skipping) of exercise. They did not engage in any form of contact sport that may have resulted in head trauma and had not sustained a head injury. They were equally free of drug, alcohol or cigarette abuse. Table 1 summarizes the physical characteristics of the subjects.

Combat history

The boxers’ combat histories were assessed following extensive interviews and consultation of their training diaries (Table 1). Their professional careers were preceded by successful amateur careers lasting 6±1 years. Seven (58% of the group) had suffered competitive losses through total knockout not less than 6 months before the study. They typically engaged in an average of 10±4 rounds of sparring per week at an intensity (rated from 1: minimal physical contact to 4: competition standard) of 3±0 AU (arbitrary units). A corresponding sparring index was calculated based on the cumulative rounds sparred over the boxer's amateur and professional careers×average sparring intensity [17].

Clinical examination

Any existing neurological symptoms were classified into motor, cognitive and behavioural features using the CBI (chronic brain injury) scale [20] with scores that ranged between 0 (no CBI) and 9 (most severe CBI) points. All boxers presented with a score of 1–2 points consistent with mild CBI and four boxers had a score of 3 points, which was diagnostic of moderate CBI [20]. Three of the controls achieved a score of 1 point.

Neurocognitive function

Participants completed a battery of neuropsychometric tests (fixed order) that were specifically selected to assess sub-domains of Memory: Rey Auditory Verbal Learning [21], Weschler Memory Scale Digit Span [22]; Attention/Concentration/Visual motor co-ordination: Trail Making Parts A/B, Digit Symbol Substitution [23] and Stroop [24] and Executive function: Grooved Pegboard Dominant/Non-Dominant Hand [25]. The within investigator CV (co-efficient of variation) was determined in a separate study employing 12 participants who performed this specific battery of tests on two separate occasions (having taken into account learning effects) and was <10% (C.J. Marley, D. Hodson, K.J. New, J.V. Brugniaux and D.M. Bailey, unpublished work).

Cerebral haemodynamic function

The within-investigator CV (see above) for all haemodynamic measurements was <10% (D.M. Bailey, S. Ogoh and P.N. Ainslie, unpublished work).

Baseline measurements

MCAv [MCA (middle cerebral artery) blood flow velocity]

The proximal segment of the right MCA was insonated using a 2 MHz-pulsed TCD (transcranial Doppler) ultrasound system (Multi-Dop X4; DWL Elektroniche Systeme). Following standardized search techniques [26], the Doppler probe was secured over the trans-temporal window with a headband device (Spencer Technologies, Nicolet Instruments) to achieve optimal insonation position and maintained in this position for the duration of the study to avoid movement artefact.

MAP (mean arterial blood pressure) and HR

Beat-to-beat MAP was monitored using finger photoplethysmography (Finometer® PRO; Finapres Medical Systems) and a lead II electrocardiogram (Dual BioAmp; ADInstruments) used for HR recordings.

Data analysis

Beat-by-beat data were continuously sampled at 1 kHz using an analogue-to-digital converter (Powerlab/16SP ML795; ADInstruments) and stored on a personal computer for off-line analysis (Chart version 7.2.2, ADInstruments). Chart files were given a coded number (not named) by an investigator that was not involved in this study to ensure that data analysis was ‘blinded’. Both the MAP and TCD channels were ‘time-aligned’ given the time delay (1.07 s) associated with MAP signal processing when using the Finometer® PRO. CVR (cerebrovascular resistance) was calculated as MAP/MCAv and CVCi (cerebrovascular conductance index) as MCAv/MAP. Pulsatility index was calculated as systolic MCAv–diastolic MCAv/MCAv and normalized relative to the prevailing MAP.


Three complementary methods were employed with each participant supine (head elevated to 30°) to quantify dCA in response to transient hypotension via the thigh-cuff inflation–deflation technique [27] and TFA (transfer function analysis) of spontaneously occurring MAP and MCAv oscillations [28].

RoR (rate of regulation)

Bilateral thigh-cuffs were connected to a Hokanson E20 and inflated to 30 mmHg above the recorded SBP (systolic blood pressure) for 3 min. Lower-limb ischaemia was confirmed using Doppler by a lack of blood flow in the dorsalis pedis artery. Cuffs were subsequently deflated (<0.1 s) and the process repeated three times with an 8 min inter-trial recovery period [29]. Control baseline values of MCAv, MAP and CVCi were determined by calculating the 4 s average immediately prior to cuff-release and the (relative) changes incurred during cuff-release subsequently determined. As the rate of change in CVCi during the 1.0–3.5 s interval following cuff-release is directly related to dCA [27] and is a more appropriate index of vascular tone in the current experimental setting [30], the RoR was determined as: Embedded Image where ΔCVCi/Δt represents the slope of the linear regression and ΔMAP was calculated by subtracting the control (relative) MAP from the averaged (relative) MAP during the 1.0–3.5 s period following cuff-release.

ARI (autoregulatory index)

An ARI score was assigned to each of the cuff inflation–deflation trials and the mean value recorded following computation of a second-order linear differential equation [31] given by: Embedded Image Embedded Image where dPn is the normalized change in MAP relative to the control value (MAPbase) adjusted for the estimated CCP (critical closing pressure); x2n and x1n are state variables (0 at baseline), mVn is modelled mean velocity, MCAvbase is baseline MCAv, f is the sampling frequency (100 Hz) and n is the sample number. The mVn generated from ten pre-defined combinations of parameters T (time constant), D (dampening factor) and k (dynamic autoregulatory gain) that best fit (quadratic error) the recorded MCAv was taken as an index of dCA and ranged between 0 (passive, impaired autoregulation) and 9 (brisk, perfect autoregulation).


Steady-state MAP and MCAv signals were obtained over a 5 min period of supine rest and re-sampled at 2 Hz for spectral analysis [28]. The data within each window were linearly de-trended, passed through a Hanning window and frequency-domain transforms computed with a fast Fourier transformation algorithm. The transfer function [H (f)] between MAP and MCAv was calculated as: Embedded Image where Sxx (f) is the autospectrum of changes in the input signal (MAP) and Sxy (f) is the cross-spectrum between the two signals (MAP and MCAv). The transfer function magnitude |H (f)| and phase spectrum |Φ (f)| were obtained from the real part [HR (f)] and imaginary part [HI (f)] of the complex transfer function given by: Embedded Image

The magnitude-squared coherence function [MSC(f)] reflects the fraction of output power (MCAv) that can be linearly related to the input power (MAP) at each frequency given by Embedded Image where Syy (f) represents the autospectrum of the changes in the output signal (MCAv). The spectral power of MAP, MCAv and mean value of transfer function gain, phase shift and coherence function between MAP and MCAv were evaluated in the VLF (very low frequency: 0.02–0.07 Hz), LF (low frequency: 0.07–0.20 Hz) and HF (high frequency: 0.20–0.30 Hz) ranges to reflect different patterns of the dynamic pressure–flow relationship. To ensure that robust phase and gain estimates within the VLF and LF bands were entered for subsequent analysis, we averaged only those gain and phase values where the corresponding coherence was ≥0.5. Accordingly, an increase in coherence and gain and a decrease in phase would be interpreted as evidence for impaired dCA [28].


CVRCO2 measurement was performed in the supine position with each participant's head elevated to 30°. PETCO2 (end-tidal CO2) was sampled from a leak-free mask and analysed via capnography (ML 206; ADInstruments). Following 10 min breathing room air, the inspirate was rapidly changed to 5% CO2 with 21% O2 and balanced nitrogen for 3 min. Following a 5 min recovery breathing room air, participants hyperventilated at 15 breaths per minute for 3 min. From this, CVRCO2 was calculated as the % increase/decrease in MCAv from baseline per 1 mmHg increase/decrease in PETCO2 recorded during the final 30 s (average taken) of the hypercapneic/hypocapneic challenge when steady-state had been achieved: Embedded Image

From these data, we derived the CVRCO2 range as a useful indication of the cerebral circulation's combined ability to respond to differential changes in CO2. This was calculated as the sum of the fractional vasodilation and vasoconstriction incurred during the respective hypercapnea and hypocapnea challenges as described: Embedded Image

The CVRCO2 data were also expressed relative to the changes observed in the prevailing MAP.

Orthostatic challenge

Each participant remained supine (head elevated to 30°) before 5 min of steady-state baseline recordings were obtained (average of last 5 s used for analysis). They quickly assumed the free-standing position (0–3 s) where they remained motionless for 180 s during which time haemodynamic data were continuously recorded and the corresponding nadirs/peaks (1 s average taken) identified [32]. This allowed us to focus on the two main phases of the orthostatic response: (i) the initial response (0–30 s), and (ii) the steady-state response (120–180 s) [33]. We also examined the prevalence of initial orthostatic hypotension defined as a decrease in SBP ≥40 mmHg and/or a decrease in DBP (diastolic blood pressure) ≥20 mmHg during the first 15 s of standing [34].

Cardiac baroflex function

The increase in HR response for a given reduction in MAP incurred by each cuff inflation–deflation manoeuvre (mean of 3 trials taken) and during the orthostatic challenge was incorporated as an index of cardiac BRS (baroreflex sensitivity) [35] given by: Embedded Image where ΔHR=peak HR (post-cuff-release or during standing)−baseline HR (pre-cuff-release or supine) and ΔMAP=baseline MAP (pre-cuff-release or supine)−MAP at nadir (post cuff-release or during standing).

Cerebral oxygenation

Right frontal cO2Hb (cortical oxyhaemoglobin) concentration was monitored continuously during the orthostatic challenge by pulsed continuous wave (780 and 850 nm) NIRS (near-IR spectroscopy) (Oxymon Mk III; Artinis Medical Systems) as described previously [36]. Relative changes (to a normalized baseline arbitrarily defined as 0 μmol·l−1) in cO2Hb were derived using the modified Beer–Lambert law [37] and differential path length factor of 5.93 [38]. We also derived a custom oxygen sensitivity index during the hypercapneic/hypocapneic challenges to determine the ‘efficiency’ of cerebral oxygenation for a given (prevailing) CVRCO2.

Peak aerobic power test

Each participant was seated on an electronically braked semi-recumbent cycle ergometer (Corival; Lode) with the backrest maintained at 70°. The initial workload was set at 35 W for 5 min (70 rev./min) and increased by 35 W·min−1 until volitional exhaustion as described previously [39]. Expired gas fractions were measured using fast responding paramagnetic oxygen (O2) and IR CO2 analysers (Servomex 1400 Series Analyser). The volume of expired gas was measured using a dry gas meter (Harvard) and V̇O2peak (peak oxygen uptake) calculated via the Haldane equation. The within investigator CV was <10% (D.M. Bailey, S. Ogoh and P.N. Ainslie, unpublished work).

Statistical analysis

Shapiro–Wilk W tests confirmed that all datasets were normally distributed. Between group differences were assessed using an independent samples Student's t test and relationships examined using a Pearson Product Moment correlation. Fisher's exact test was incorporated to determine differences in the frequency of initial orthostatic hypotension between groups. Significance was established at P<0.05, and results are expressed as means±S.D.


Baseline measurements

Both groups were well matched in that there were no differences observed in any anthropometric, V̇O2peak (Table 1), cerebrovascular or cardio-respiratory parameters (see Table 4) measured at baseline.

Neurocognitive function

Boxers underperformed on all neuropsychometric tests highlighting functional deficits in memory, attention, concentration, visuo-motor coordination and executive function (Table 2). However, tests scores failed to correlate against any measure of professional combat history.

View this table:
Table 2 Neurocognitive function

Values are means±S.D. †P<0.05.

View this table:
Table 3 Dynamic cerebral autoregulation

Values are means±S.D. P<0.05.


We observed no differences in the frequency-domain profiles of MCAv or MAP variability, gain, phase or coherence when assessing spontaneous oscillations in MAP via the transfer function technique (Table 3). In contrast, the abrupt hypotension induced by the cuff inflation–deflation technique revealed selectively lower RoR and ARI scores indicative of impaired dCA in the boxers despite similar changes in MAP and PETCO2 (results not shown). Figure 2 provides a visual example of the typical differences observed between an individual control and a boxer [with the highest score (3 points) on the Chronic Brain Injury scale] during the thigh-cuff inflation–deflation technique.

Figure 2 Representative traces observed during the thigh-cuff inflation–deflation technique

Data obtained from an individual control (non-boxer) and a professional boxer with the highest score (3 points) on the CBI scale.


Likewise, CVRCO2 was equally impaired as indicated by a selective reduction in the boxers’ ability to vasodilate/vasoconstrict in response to a given PETCO2 stimulus (Table 4 and Figure 3A). As a consequence, the corresponding CVRCO2 range was lower and was more related to the volume and intensity of sparring during training (sparring index) than the frequency of total knockouts suffered during professional competition (Figure 3B) or number of professional rounds fought (r=0.08, P>0.05). A sub-group of four of the more experienced boxers with the highest sparring index scores was identified, characterized by the lowest CVRCO2 range and highest CBI scores. A sparring index ‘threshold’ of 20000 AU corresponding to a CVRCO2 range of ≤4%·mmHg−1 was also apparent, beyond which a boxer's susceptibility to moderate CBI appeared to increase (Figure 3B). When the CVRCO2 data were adjusted according to changes in the prevailing MAP, the impaired response to hypercapnea was found to persist (0.5±0.9%/mmHg/mmHg in boxers compared with 1.4±2.9%/mmHg/mmHg incontrols; P<0.05), whereas differences during hypocapnea (0.3±1.5%/mmHg/mmHg in boxers compared with −1.1±3.5%/mmHg/mmHg in controls) and cumulatively in the form of the CVRCO2 range (0.8±1.6%/mmHg/mmHg in boxers compared with 0.3±4.8%/mmHg/mmHg in controls) and relationship with the sparring index (r=0.30) were no longer apparent (P>0.05). The hypocapneic challenge was associated with a more marked reduction in cO2Hb for any given CVRCO2 as indicated by the lower oxygen sensitivity index in boxers (Table 4).

View this table:
Table 4 Cerebral haemodynamic function during carbon dioxide and orthostatic stress

Values are means±S.D. *P<0.05 compared with the normocapnoea control baseline (taken as 100%) for given group; †P<0.05 between the groups for a given condition. SMCAv/DMCAv, systolic/diastolic MCAv.

Figure 3 Cerebrovascular reactivity to CO2

Values are means±S.D. Open circles highlight those boxers (n=4) with the highest scores (3 points) on the CBI scale consistent with moderate CBI. Numbers next to circles indicate the number of total knockouts suffered by each boxer during their professional career. †P<0.05.

Orthostatic stress

Eleven of the boxers and 8 of the controls (P<0.05) experienced initial orthostatic hypotension whereas none of the subjects was classified as orthostatically hypotensive during the steady-state phase (120–180 s) of standing. Initial orthostatic stress (0–30 s) induced a more marked increase in the pulsatility index and reduction in MCAv, CVCi and cO2Hb in the boxers (Table 4) though none of the subjects reported symptoms consistent with neurogenic syncope. No between group differences were observed in the time taken to reach the nadir for MCAv (9±2 s in controls compared with 10±3 s in boxers; P>0.05), MAP (12±2 s in controls compared with 12±1 s in boxers; P>0.05) and cO2Hb (9±5 s in controls compared with 10±5 s in boxers; P>0.05) or HR peak (11±1 s in controls compared with 10±2 s in boxers; P>0.05).

Cardiac BRS

No differences in BRS were observed during cuff inflation–deflation (−0.74±0.35 beats·min−1·mmHg−1 in controls compared with −0.79±0.22 beats·min−1·mmHg−1 in boxers; P>0.05) or orthostasis (−1.24±0.66 beats·min−1·mmHg−1 in controls compared with −1.22±0.67 beats·min−1 ·mmHg−1 in boxers; P>0.05).

Group correlational analyses


Inverse relationships were observed between VLF gain and the (non-MAP-adjusted) CVRCO2 response to hypercapnea (r=−0.37, P<0.05) and CVRCO2 range (r=−0.36, P<0.05). Inverse relationships were also observed between gain at VLF (r=−0.47, P<0.05), LF (r=−0.41, P<0.05) and HF (r=−0.45, P<0.05) and CVRCO2 response to hypercapnea following MAP adjustment. In contrast, phase did not correlate against any index of cerebrovascular function.

Thigh-cuff inflation–deflation

A positive association was observed between RoR and ARI scores (r=0.97, P<0.05). Both indices correlated positively against the (non-MAP-adjusted) CVRCO2 response to hypercapnea (r=0.48–0.50, P<0.05), hypocapnea (r=0.47–0.55, P<0.05) and corresponding CVRCO2 range (r=0.62–0.64, P<0.05), whereas no relationships were observed against any transfer function index (P>0.05).


Inverse relationships were observed between the reduction in MCAv during orthostasis and RoR (r=−0.54, P<0.05), ARI (r=−0.56, P<0.05), non-MAP-adjusted CVRCO2 response to hypercapnea (r=−0.41, P<0.05) and CVRCO2 range (r=−0.45, P<0.05). In contrast, no relationships were observed between the reduction in cO2Hb and any index of cerebrovascular function.


Consistent with our original hypotheses, the present study has revealed three novel findings. First, in addition to neurocognitive dysfunction, cerebral haemodynamic function was found to be chronically impaired in active professional boxers compared with physical fitness-matched non-boxing controls as indicated by a higher prevalence of initial orthostatic hypotension and lower values in dCA, CVRCO2 and more marked cortical de-oxygenation during both hypocapnea and orthostatic stress. Secondly, unlike neurocognitive measures, haemodynamic impairment in the form of reduced CVRCO2 range correlated against the volume and intensity of sparring where the greatest exposure to head injury is known to occur. Finally, a sparring ‘threshold’ corresponding to a critical CVRCO2 range was identified beyond which a boxer's susceptibility to CTBI was shown to increase. Collectively, these findings provide the first comprehensive evidence for functionally impaired cerebral haemodynamic function across the spectrum of CTBI in professional boxers likely due to the mechanical trauma inflicted by repetitive, mostly sub-concussive head impact.

Given that the structural abnormalities typically observed in boxers on neuro-imaging are not specific to CTBI, researchers have traditionally relied on neuropsychological testing as a more sensitive means for early detection [4]. Our findings are generally consistent with the fronto-temporal cognitive deficits consistently reported in both the boxing [3,4] and clinical TBI (traumatic brain injury) [40] literature. However, unlike a previous study [17], these impairments failed to correlate against any measure of the boxer's professional combat history. Furthermore, the application and interpretation of what are mostly subjective tests that fail to take into account unavoidable differences in baseline education status are not without complication [41], highlighting the need for complementary, more objective measures.

In contrast, the assessment of cerebral haemodynamic function is not routine and the few studies that exist in professional boxers have been confined almost exclusively to resting, static measures of cerebral perfusion. Tracer studies [6,7], single photon emission computed tomography [8] and positron emission tomography [9] have revealed cortical perfusion deficits confined to the frontal, parietal and temporal lobes. In contrast, our study failed to identify any evidence for reduced cerebral perfusion at rest, arguably due to the known limitation of TCD for the comparison of absolute flows between groups [26]. However, when the boxers were subject to haemodynamic stress through the controlled manipulation of MAP and PETCO2, functional deficits in dCA, CVRCO2 and cortical oxygenation became clearly apparent. This need to ‘challenge’ the cerebrovasculature to dynamically ‘force’ the impaired signal out of what would otherwise be interpreted as a functionally intact baseline is an important technical consideration that can improve the diagnostic potential of haemodynamic testing.

Consistent with previous observations in acute TBI (reviewed in [5]), dCA and CVRCO2 (including the MAP-adjusted response to hypercapnea) were shown to be chronically impaired in the boxers, an important observation given the integrative regulation of CBF [11]. We observed an apparent disassociation between spontaneous fluctuations (TFA) and the closely correlated stimulus-induced (RoR/ARI) manipulations in MAP during the assessment of CA (cerebral autoregulation) since impairment was selectively confined to the latter. The reasons for this discrepancy remain unclear though each approach has its limitations with no universally accepted ‘gold standard’ [26] given the complexity of CA. It is possible that the small magnitude and inconsistency of spontaneous pressure oscillations associated with the TFA approach [42] collectively failed to detect subtle imperfections in the boxers’ dCA response. This only became apparent when the MAP input signal was dynamically ‘forced’ during the thigh-cuff inflation–deflation approach which likely served to more reliably engage CA and improve functional detection sensitivity. This observation is consistent with the literature. For example, in anaesthetized cats, abrupt reductions in MAP achieved with sudden arterial occlusion elicited a cerebral vasodilatory response that occurred within 2–3 s [43], whereas in humans latencies of up to 10 s have been observed in response to slow oscillatory fluctuations in MAP [44]. It was also noted that vasodilatation occurred only when a certain threshold of hypotension had been achieved [44]. Collectively, these findings indicate that the magnitude of MAP perturbation influences the nature of the corresponding cerebrovascular response. Future studies need to consider TFA of augmented sinusoidal MAP oscillations forced through repeated squat-stand manoeuvres [44] performed at the appropriate frequencies (0.05 and 0.01 Hz) as a means of optimizing functional detection sensitivity. Indeed, this approach has been shown to increase MAP and MCAv spectral power by up to as much as 100-fold with corresponding improvements in coherence [26]. Furthermore, it has the potential to determine the brain's differential ability to buffer changes in cerebral perfusion in response to (both) hyper and hypotensive challenges, which is important given that hysteresis is a natural characteristic of CA [45].

From a functional perspective, these impairments coincided (though not always correlated since they reflect different functional aspects of the dynamic cerebral circulation) with more marked orthostatic hypotension, cerebral hypoperfusion and corresponding cortical de-oxygenation. The elevated pulsatility index tentatively suggests that an increased resistance/lowered compliance may have reduced the capacity of the cerebrovascular bed to respond to dynamic changes in blood flow. This could not be explained by hyperventilation-induced hypocapnia and associated cerebral vasoconstriction given that the reduction in PETCO2 during the transition from supine to standing was equivalent across groups. An alternative explanation may relate to the fact that boxers typically aim at the carotid sinus in an attempt to make their opponent faint which may cause chronic mechanical irritation thereby predisposing to the carotid sinus syndrome [46]. Although this is unlikely given that the cardiac baroreflex was shown to be intact, we cannot exclude impairment to the sympathetically mediated vasomotor component of the carotid baroreflex which although slower than its cardiac counterpart, is quantitatively more important to blood pressure control [47]. A more detailed examination of the carotid baroreflex is to be encouraged in future studies.

It may be argued that such a small reduction in cO2Hb (<9 μmol·l−1) may be considered clinically insignificant given an estimated 70–100 μmol·l−1 of total Hb in cerebral tissue [48] combined with the fact that none of the boxers were pre-syncopal, indicating that oxygen delivery to the brainstem was likely adequate. However, it is conceivable that in the combined setting of impaired dCA and CVRCO2, any transient reduction in MAP during initial orthostatic hypotension and time delays associated with the cardiac baroreflex and CA (each estimated at ~10 s [49], which is similar to the nadir observed in cO2Hb) may have stressed the brain's oxygen reserve time which has been estimated to be in the order of 6–7 s [50]. Given that clinical TBI is associated with elevated metabolic-neural activity and associated oxygen demand [51], even a subtle reduction in cerebral perfusion and oxygenation has the potential to cause intermittent cerebral hypoxia if not indeed ischaemia, resulting in secondary neuronal–parenchymal damage [14,52]. Such changes may provide an alternative (haemodynamic) mechanism to account for the chronic elevation recently observed in the blood–borne concentration of neuron specific-enolase, an ischaemia-sensitive biomarker of neuronal–parenchymal injury [53].

In addition to the deleterious consequences associated with cerebral hypoperfusion, an inability to buffer spontaneous surges in MAP such as that encountered during intense (boxing) exercise has the potential to cause cerebral hyperperfusion and BBB (blood–brain barrier) disruption, ultimately resulting in capillary stress failure [39], events that are further compounded by cerebral hypoxaemia [54,55]. This may prove to be the common mechanism underlying the systemic extravasation of the BBB-specific protein S100β [53], punctate micro-haemorrhages [56] and neuropathological lesions [41] that have previously been reported in more advanced cases of CTBI.

Although the precise mechanisms underlying these functional deficits remain to be established, the mechanical trauma incurred through repetitive sub-concussive impact which can impart impact forces in excess of 5000 N and translational acceleration of 50 g to the opponent's brain [57] is a likely determinant given the correlations observed with sparring training. Our haemodynamic findings complement previous neuropsychological data [17] and further emphasize that it is the neuro-trauma incurred during sparring training and not competition that is the primary stimulus underlying CTBI. Furthermore, these young boxers may be more vulnerable to stroke in later life given that impaired dCA and CVRCO2 are established independent risk factors [58]. Thus, from a practical perspective, boxers should be encouraged to limit their exposure to sparring when preparing for competition.


Although we recognize the potential limitations associated with NIRS and TCD ultrasonography, these findings provide the first comprehensive evidence for functionally impaired cerebral haemodynamic function in professional boxers. A more ‘vulnerable’ cerebral circulation incapable of responding to changes in blood pressure or CO2 may render the boxer's brain more susceptible to damage. While the underlying mechanisms remain to be established, repetitive sub-concussive impact to the boxer's brain specifically incurred during sparring training is a likely determinant. These findings may help explain why CTBI is a progressive disease that manifests beyond the active boxing career. Follow-up studies are encouraged to determine whether longitudinal monitoring of cerebral haemodynamic function will indeed improve clinical diagnosis of a boxer's trajectory towards CTBI.


This study was funded in part by J.P.R. Williams Research Fellowships (awarded to D.M.B. to support D.H. and C.J.M.).


Damian Bailey was the Principal Investigator and recipient of funding. He led the study conception/design, acquisition/analysis/interpretation of data and wrote the original paper. Daniel Jones, Andrew Sinnott, Julien Brugniaux, Karl New, Danielle Hodson and Christopher Marley made substantial contributions to data acquisition and gave final approval of the version to be submitted. Jonathan Smirl made substantial contributions to data analysis and gave final approval of the version to be submitted. Shigehiko Ogoh and Philip Ainslie revised the original paper critically for important intellectual content and gave final approval of the version to be submitted.


We appreciate the commitment of the subjects, the boxers’ coaches (notably Mr D. Wilson and Mr V. Cleverly) and specialist input from Professor R.B. Panerai and Professor J.E. Hall. Technical support was provided by Mrs B. Wagenaar (ADInstruments) and Mr L. Fall/D. Whitcombe (University of Glamorgan). Finally, we recognize the informative discussions and academic support provided by Mr J.P.R. Williams and Dr P. Williams (retired).

Abbreviations: ARI, autoregulatory index; AU, arbitrary unit; BBB, blood–brain barrier; BRS, baroreflex sensitivity; CA, cerebral autoregulation; CBF, cerebral blood flow; CBI, chronic brain injury; cO2Hb, cortical oxyhaemoglobin; CTBI, chronic traumatic brain injury; CV, co-efficient of variation; CVCi, cerebrovascular conductance index; CVR, cerebrovascular resistance; CVRCO2, cerebrovascular reactivity to changes in CO2; DBP, diastolic blood pressure; dCA, dynamic cerebral autoregulation; HF, high frequency; HR, heart rate; LF, low frequency; MAP, mean arterial blood pressure; MCAv, middle cerebral artery blood flow velocity; NIRS, near-IR spectroscopy; PETCO2, end-tidal CO2; SBP, systolic blood pressure; RoR, rate of regulation; TBI, traumatic brain injury; TCD, transcranial Doppler; TFA, transfer function analysis; VLF, very low frequency; V̇O2peak, peak oxygen uptake


View Abstract