Biomechanics of Sports-Induced Axial-Compression Injuries of the Neck

I prepared 5 fresh-frozen human osteoligamentous cervical spine specimens (2 men, 3 women; age = 86.8 ± 4.9 years). The specimens included occiput through C5 (specimens 2 and 3) or C6 (specimens 1, 4, and 5). Apart from typical age-related degenerative changes, the specimens did not have any disease that could have affected the osteoligamentous structures. The specimens were mounted in resin at the occiput and most inferior vertebra (C5 or C6) with the occiput flexed 30° relative to the occipital mount and the inferior endplate of C5 or C6 parallel with its mount. This cervical spine posture with lordosis removed represents a straightened neck rather than normal lordotic posture.¹⁷  The occiput and C5 or C6 vertebra were secured rigidly to custom aluminum fixtures within their mounts using screws. The aluminum fixture within the occipital mount had 4 rigidly attached bolts protruding superiorly for subsequent rigid attachment of the surrogate head. The mount for C5 or C6 and its aluminum fixture were designed for subsequent rigid attachment to a mass on a custom-built sled using a laterally aligned, double-ended threaded stud. I did not observe loosening at the mounting interfaces or mount failure during preliminary studies. Lightweight motion-tracking markers (model ASIN B000FMYIP6; SmallParts, Seattle, WA) were attached rigidly to each cervical vertebra and to both mounts.

A Hybrid III surrogate head (Humanetics Innovative Solutions, Plymouth, MI) was modified to include a 6-component load cell (MC3A force/torque sensor; Advanced Mechanical Technology, Inc, Watertown, MA) with an ultimate shear force of 3300 N, ultimate axial force of 6600 N, and ultimate sagittal moment of 165 Nm. The head (mass = 4.6 kg, sagittal moment of inertia = 0.0214 kg m²) was attached rigidly to the occipital mount in anatomic location. The head had a motion-tracking marker rigidly fixed to its right side at its sagittal center-of-mass location. To provide postural head and neck stability and passive resistance to motion during head-first impact, the head and neck were stabilized using muscle-force replication, which was modified from a previous study.¹⁸  The muscle-force replication was symmetric about the midsagittal plane and consisted of 2 anterior and 2 posterior cables originating from the head and anchored to preloaded springs with stiffness coefficients of 8.0 N/mm. The preload in each cable was 50 N, resulting in 200 N total compressive spinal preload immediately before impact. This preload was within the range of the estimated in vivo spinal compression computed using mathematical models: between 100 and 122 N in relaxed posture and 1164 and 1700 N with maximally active neck muscles.^19,20 

Each specimen was subjected to a single head-first impact with velocity at impact of 4.1 ± 0.3 m/s. The impact apparatus consisted of a mass rigidly fixed to the sled with the C5 or C6 mount of the specimen rigidly fixed at the front of the mass (Figure 1). The mass (55.5 kg) represented the equivalent torso mass of a heavier athlete or an obese male.²¹  I assumed that the entire torso mass acted on the neck due to its forward momentum during head-first impact. An upward force was used to counterbalance the head weight and to maintain the head and neck posture before impact.

The sled was mounted on low-friction linear bearings, which translated along 2 precision-ground, stainless-steel shafts. Motion-tracking markers were fixed to the sled and padded impact barrier. A high-speed potentiometer (model 3540S-1-203; Bourns, Riverside, CA) mounted to the sled was used to confirm sled translation computed using the flag marker data. The sled was accelerated using a custom-built acceleration generation system consisting of a piston, high-energy compression springs, and computer-controlled electromagnetic release. At the time of electromagnet release, a trigger signal initiated high-speed camera (MotionPRO; Redlake MSAD, San Diego, CA) recording. In preliminary studies, nonphysiologic neck motions and injuries were documented due to head impact and not due to the initial sled-acceleration phase. The padded barrier consisted of a pneumatic system (Original Line; Bimba, University Park, IL) that permitted adjustment of its energy absorption from rigid, no-energy absorption to gradual deformation. The padded barrier consisted of a 5-cm-thick, open-cell, highly deformable, polyurethane foam pad (density = 0.0288 g/cm³, compressive strength = 3.8 kPa, tensile strength = 172 to 207 kPa) mounted to a rigid aluminum plate (model 325-4197; Enco, Fernley, NV). This foam pad provided head constraint and pocketing during head-first impact.

The impact apparatus was controlled fully by a personal computer and custom LabVIEW software (version 8.5; National Instruments, Austin, TX). Uniaxial load cells were used to measure impact barrier loads with an ultimate load of 10 kN (model LCCA-750; Omega Engineering, Inc, Stamford, CT) and loads at the base of the neck with an ultimate load of 40 kN (model LCCA-3K; Omega Engineering, Inc). Biaxial accelerations were measured at the head center of mass and C5 or C6 vertebra and had a capacity of 50g (model ADXL250JQC; Analog Devices, Norwood, MA). Load cell, accelerometer, and potentiometer data were sampled continuously at 1 kHz using an analog-to-digital converter and the LabVIEW program. The high-speed digital camera recorded the sagittal kinematics at 500 frames per second. Macroscopic neck injuries were determined by fluoroscopy and visual inspection after the impacts.

Representative time-history responses were determined for intervertebral rotations and cervical spine curvature during head-first impact. Neck buckling, which was defined as rotation in opposing directions throughout the cervical spine, has been observed in previous head-first impact simulations of cadavers.⁷  A vertebral inflection point was defined as the point above and below which the spinal levels rotated in opposing directions. The intervertebral rotations for the spinal levels above and below the inflection point were summed separately and normalized to the corresponding physiologic rotations obtained from intact flexibility tests. To visualize the cervical spine curvature, a phase diagram was created by plotting the normalized upper and middle cervical spine rotations.²² 

Specimen-specific kinetic and kinematic time-history responses during head-first impact were determined using the load cell and accelerometer data and analyses of the high-speed videos. Impact velocity was determined by numerical differentiation of the sled potentiometer data. The global coordinate system was fixed to the ground and had its positive z axis oriented anteriorly, positive y axis oriented superiorly, and positive x axis oriented to the left relative to the specimen (Figure 1). Loads included forces in lateral shear (Fx), tension and compression (Fy), and anteroposterior shear (Fz) and moments in flexion and extension (Mx), axial torque (My), and lateral torque (Mz). Loads were computed at the impact barrier (Fy), occipital condyles (6 components), and center of mass of the C5 or C6 vertebra (Fy). Mean ± standard deviation errors in the computed loads were 0.5 ± 1.8 N for shear force, −0.1 ± 2.2 N for axial force, and 0.7 ± 0.7 Nm for moment.²³  Accelerations were determined at the centers of mass of the head (Az and Ay) and C5 or C6 vertebra (Ay). Occipital condyle loads and head accelerations were expressed in the head coordinate system, which was fixed to and moved with the head. The global and head coordinate systems were aligned immediately before impact.

Custom MATLAB (The MathWorks, Inc, Natick, MA) programs automatically tracked the coordinates of all flag markers during the impacts and were used to compute head and torso motions. These motions consisted of translations, including axial separation and compression (Ty) and anteroposterior shear (Tz) and rotation (Rx), expressed in the global coordinate system. The mean ± standard deviation errors were −0.06° ± 0.17° for rotation²⁴  and 0.3 ± 0.2 mm for translation.²⁵ 

All data were filtered digitally using a third-order, dual-pass, low-pass Butterworth filter with a cutoff frequency of 150 Hz. Average occurrence times of the first local and global compression force peaks at the impact barrier, occipital condyles, and C5 or C6 vertebra were compared using analysis of variance and post hoc Bonferonni pairwise tests. The α level was set at .05. I used Minitab 16 (Minitab, Inc, State College, PA) to analyze the data.

A representative example was chosen to illustrate the original high-speed movie images during head-first impact (Figure 2A). The high-speed movie images demonstrated dramatic neck compression, which was exemplified at 50 and 60 milliseconds. Head flexion remained within physiologic limits that was less than 10°. Excessive axial compression caused the cervical spine to buckle and fail. During buckling, nonphysiologic motions were observed in flexion at the upper cervical spine, head–C1, and C1–2 and extension at the subaxial spinal levels (Figure 2B). The phase diagram of normalized upper and middle cervical spine rotations included the physiologic rotation bands for the upper cervical spine and middle cervical spine (Figure 2C). The intersection of the 2 bands represents the physiologic rotation region. The areas outside this region represent potential injury zones due to nonphysiologic rotation. The C3 vertebra was identified as the inflection point for reversal of motion direction. The neck buckled in an S-shaped curvature, with the onset of nonphysiologic motion occurring first in extension at the middle cervical spine (32 milliseconds) and subsequently in flexion at the upper cervical spine (34 milliseconds). The time courses for head and C5 translations indicated that the onset of head and torso compression occurred at 10 milliseconds before the onset of intervertebral rotation (Figure 2D). The onset of posterior head translation commenced at 20 milliseconds. Head and torso rebound occurred simultaneously beginning at 56 milliseconds.

The specimen-specific time-history responses for loads, accelerations, and motions appear in Figure 3. No apparent lag time was observed among the onset of compression forces at the impact barrier, occipital condyles, and C5 or C6 vertebra, but the barrier and occipital condyle loads increased more rapidly than the C5 or C6 loads (Figure 3A and B). Predominant loads at the occipital condyles included compression coupled with anterior shear and flexion moment (Figure 3B and C). Average peak deceleration at the head was 27.9g at 17.6 milliseconds and at the C5 or C6 vertebra was 16.7g at 45.6 milliseconds (Figure 3D). The head accelerated posteriorly and then anteriorly (Figure 3E). Average peak motions of the head relative to the torso were 6.0 cm in compression at 56 milliseconds, 2.4 cm in posterior shear at 56 milliseconds, and 6.4° in flexion at 86 milliseconds (Figure 3F). The average deformation of the impact barrier pistons was 0.9 cm as measured after impact.

The first local compression force peaks at the impact barrier, occipital condyles, and C5 or C6 vertebra occurred earlier than all global compression peaks (F range = 9.1–157.5, P < .001; Table). The first local compression peaks reached a mean of 3070.0 ± 168.0 N at the impact barrier, 2868.1 ± 732.4 N at the occipital condyles, and 2884.6 ± 910.7 N at the C5 or C6 vertebra.

The fluoroscopic and visual inspections revealed severe macroscopic injuries, including occipital condyle fractures (specimens 1 and 5), atlas fractures (all specimens), type II (specimen 4) and type III (specimen 2) odontoid fractures, and subaxial comminuted burst and facet fractures (all specimens).

The present cadaveric head-neck model simulated horizontally aligned head-first impacts with a straightened neck, replicating traumatic neck compression during football, rugby, hockey, diving, wrestling, gymnastics, trampolining, and equestrian sports. Average peak neck compression force (7531.6 N) and velocity at impact (4.1 m/s) were consistent with the ranges computed for real-life head-first sports impacts that have caused cervical quadriplegia in athletes: 3611 to 8138 N for neck compression and 3.4 to 6.6 m/s for impact velocity (Table).^1,2  The impacts in this study caused clinically relevant compression fractures, including occipital condyle, atlas, odontoid, and subaxial comminuted burst and facet fractures. These results are consistent with those reported in previous clinical studies in which researchers^26–28  described multiple noncontiguous neck fractures due to traumatic head impacts.

The limitations of this model should be considered before interpreting these results. Given a lack of young cadaveric material, the sample size was limited to 5 specimens with an average age of 86.8 years, which was outside the age range of previous cadaveric studies (age range, 35–82 years).^7–10  The present specimens and the older specimens of the previous studies likely had decreased bone mass, density, and strength compared with the younger athletic population that sustains neck injuries due to head-first impacts. To determine occipital condyle loads, the model included a Hybrid III surrogate head modified to include a 6-component load cell rigidly attached to the occipital mount. I used a torso-equivalent rigid mass of 55.5 kg, which represented the mass of a heavier athlete or obese male.²¹  I assumed that the entire torso mass acted on the neck due to its forward momentum during head-first impact. This represented the worst-case scenario for causation of neck injuries. In real-life head-first impacts, neck loading due to head impact and forward torso momentum is decoupled. The present cadaveric neck model did not include an active neuromuscular response; however, axial preload, postural head and neck stability, and passive resistance to motion were provided by muscle-force replication. The first local neck compression force peak occurred within 25 milliseconds on average, which was earlier than the active neuromuscular reflex (Table).²⁹  The importance of neck preload has been exemplified in previous studies, in which researchers¹³  have shown that preload alters the dynamic head-neck responses during head-first impact. Although limited by factors associated with all cadaveric models, the present biomechanical model of head-first impact can be used to understand the mechanisms of neck injury and failure.

Previous vertical drop tests of human head-neck specimens in inverted postures have demonstrated a lag time between the onset of forces at the ground impact barrier and T1 vertebra.^7,9,10,30  The average lag time was 1.7 and 6.9 milliseconds for impacts into rigid and padded barriers, respectively, indicating a decoupled response of the head and neck. The apparent lack of lag time in the present horizontally aligned model can be attributed to several factors. The inverted posture of head-neck specimens with no axial preload in the previous vertical drop tests placed the neck in tension immediately before head impact, which might have accentuated the decoupled response.^7,9,10,30  This is in contrast to neck compression induced by muscle bracing in real-life head-first impacts. The compressive spinal preload of 200 N was nearly double that used in a previous study of a surrogate head-neck¹³  but less than that produced with maximally active neck muscles. The present specimens were from older donors, and their cervical discs were degenerated, which might have contributed to the lack of decoupled head and neck response. Whereas I did not observe lag time, I did observe a more rapid initial increase in occipital condyle loads compared with C5 or C6 loads (Figure 3A and B), with the first local compression peak occurring initially at the occipital condyles (19.6 milliseconds) and subsequently at the C5 or C6 vertebra (25.0 milliseconds; Table). In a recent study of vertical drop tests of a surrogate head-neck, researchers¹³  demonstrated a reduction in lag time from 6.2 milliseconds with no preload to 2.2 milliseconds with 104 N of preload. This decrease in lag time with increasing axial preload indicates that muscle bracing by athletes in anticipation of head-first impact, which can exceed 1000 N,¹⁹  might minimize the lag time.

In the previous cadaveric studies, investigators have defined the onset of bony injury as the first decrease in axial neck load with increasing spinal deformation. In their series, Nightingale et al⁷  observed an average compressive neck load of 1727 N at the onset of bony injury, occurring from 14 to 19 milliseconds for impacts into a padded ground barrier. The present model demonstrated that the first local peak in compression force occurred at the occipital condyles at 19.6 milliseconds and at the C5 or C6 vertebra at 25.0 milliseconds (Table). These transient decreases in compression force occurred with increasing axial compressive deformation, ranging from 0.9 to 2.4 cm. These results suggest that the onset of the observed fractures of the occipital condyles and upper cervical vertebrae might have occurred before the onset of fractures at the inferior vertebrae due to abrupt deceleration of the head as loads were transferred inferiorly through the neck. The onset of the observed fractures of the inferior vertebrae were likely due to continued forward torso momentum that transferred loads superiorly through the neck, which likely exacerbated the injuries to the occipital condyles and upper cervical vertebrae. The magnitudes of the first local peaks in compression force were 2868.1 ± 732.4 N at the occipital condyles and 2884.6 ± 910.7 N at the C5 or C6 vertebra (Table), both of which were higher than the 1727 N local compression peak reported by Nightingale et al.⁷  This difference might be due in part to the higher impact velocity used in the present study (4.1 m/s) than in the study by Nightingale et al⁷  (3.2 m/s).

In the previous biomechanical studies, neck injuries due to head-first impact were attributed to abrupt stoppage and rebound of the head combined with continued torso momentum.^7,9,10,30  Between 38% and 54% of total neck compression force was attributed to head rebound.⁷  In these previous studies, head-neck specimens were dropped vertically onto a nondeformable steel plate with and without a padded surface. Head-first sport impacts most often involve the head striking a padded, deformable surface, which might reduce or negate head rebound. These surfaces include the padding and soft tissues of an opponent, compressive playing surfaces, hockey rink boards, and goal posts.¹⁶  Helmeted impacts increase the energy absorption due to padding within the helmet. I replicated these injury scenarios by incorporating a padded, pneumatic impact barrier, which deformed 0.9 cm on average during the impacts. The barrier also allowed angular deformation, as demonstrated in the representative high-speed movie images (Figure 2A, times 30–80 ms). This caused posterior head translation (Figures 2D and 3F) and might have accentuated the S-shape neck curvature (Figure 2C). I observed continued forward motion of the head after the onset of injury. The onset of head rebound occurred simultaneously at 56 milliseconds at the time of peak head and torso motions in axial compression and posterior shear (Figure 2D). Neck injuries due to excessive axial compression loads in this model were caused by abrupt deceleration of the head and continued forward torso momentum before simultaneous rebound of the head and torso.

The present results are consistent with those reported in previous studies in which investigators noted neck compression injuries within 20 milliseconds of impact before substantial head rotation.^7,9  Nightingale et al¹⁰  observed that a large amount of head rotation might protect the neck from injury by allowing it to escape the path of the oncoming torso. The present results indicated that head rotation consistently remained within physiologic limits (<10°) and average peak head posterior translation was 2.4 cm (Figure 3F). These data showed that head motions were of insufficient magnitude to allow the head and neck to escape the path of the oncoming torso. In contrast, average peak axial compression reached 6.0 cm with an average onset at 10 milliseconds. This compression is greater than the 2.5- to 3.5-cm compression observed with vertical impacts of unrestrained cadavers in inverted postures, during which the head flexed abruptly, causing chin-to-chest contact and allowing the head and neck to partially escape the path of the oncoming torso.⁸  These cumulative findings suggest that axial compression load and neck compressive deformation are the predominant mechanisms that cause early injuries during head-first impact, whereas subsequent injurious neck buckling causes nonphysiologic intervertebral motions with minimal head rotation.

The present study determined the head and neck biomechanical responses and neck failure mechanisms using a cadaveric cervical spine model in simulated head-first impacts. The first local compression force peaks at the impact barrier, occipital condyles, and neck occurred earlier than all global compression peaks. Head rotation remained within physiologic limits. Neck injuries due to excessive axial compression occurred within 20 milliseconds of impact and were caused by abrupt deceleration of the head and continued forward torso momentum before simultaneous rebound of the head and torso. Continued biomechanical and epidemiologic studies of neck injuries and injury mechanisms due to head-first impact ultimately will lead to reduced frequency and severity of sports-induced neck trauma.

This research was supported by grant 5R01CE001257 from the Centers for Disease Control and Prevention, Atlanta, Georgia.