Geometric Changes in the Cervical Spinal Canal During Impact

Abstract

Summary of Background Data.

Although the extant of injury after cervical spine fracture can be visualized by imaging, the deformations that occur in the spinal canal during injury are unknown.

Study Design:

This study compared spinal canal occlusion and axial length changes occurring during a simulated compressive burst fracture with the residual deformations after the injury.

Methods.

Canal occlusion was measured from changes in pressure in a flexible tube with fluid flowing through it, placed in the canal space after removal of the cord in cadaver specimens. To measure canal axial length, cables were fixed in C1 and led through the foramen transversarium from C2-T1, then out through the base, where they were connected to the core rods of linearly variable differential transformers (LVDT). Axial compressive burst fractures were created in each of ten cadaveric cervical spine specimens using a dropweight, while force, distraction, and occlusion were monitored throughout the injury event. Pre- and post-injury radiographs and computed tomography scans compared transient and post-injury spinal canal geometry changes.

Results.

In all cases, severe compressive injuries were produced. Three had an extension component in addition to compression of the vertebra and retropulsion of bone into the canal. The mean post-injury axial height loss measured from radiographs was only 35% of that measured transiently (3.1 mm post-injury, compared with 8.9 mm measured transiently), indicating significant recovery of axial height after impact. Post-injury and transient height loss were not significantly correlated (r² = 0.230, P = 0.16) demonstrating that it is not a good measure of the extent of injury. Similarly, mean post injury canal area was 139% of the minimum area measured during impact, indicating recovery of canal space, and post-injury and transient values were not significantly correlated (r² = 0.272, P = 0.12). Mean post-injury midsagittal diameter was 269% of the minimum transient diameter and showed a weak but significant correlation (r² = 0.481, P = 0.03).

Conclusions.

Two potential spinal cord injury-causing mechanisms in axial bursting injuries of the cervical spine are occlusion and shortening of the canal. Post-injury radiographic measurements significantly underestimate the actual transient injury that occurs during impact.

During the initial impact, cervical spinal cord injury consists of at least two distinct mechanical events. The first is the destruction of the ligaments and bone of the cervical spine, which leaves the vertebral column mechanically unstable. The second is mechanical trauma to the spinal cord caused by abnormal dislocation of the unstable spine. The strength, stability, and motion of the cervical spine during impact as well as the biological response of the spinal cord to mechanical loading have been studied. What appears to be undocumented is the direct mechanism that creates the spinal cord injury during impact to the vertebral column.

From biomechanical studies, cervical spine loading tolerances on human volunteers have been determined to be in the range of 190 Nm in flexion and 57 Nm in extension,^7–9,17,20 with 10.2 m/sec a threshold for impact injury.¹⁸ The severity of the injury has been related to impact velocities, the effects of seat belts or airbags, and the direction of impact.^3,4,13,15,16 Peak forces of up to 5.7 kN can be resisted, and the mechanism of fracture is arching of the spine and fracture of the spinous processes.⁶ Hodgson and Thomas¹² demonstrated that impact location, restriction of the atlanto-occipital junction, and alignment of the impact forces all influenced the location of the resulting injury and the maximum strain on the vertebrae. Flexion-type injuries were not necessarily dependent on a flexed position of the head. Instead, the initial orientation of the spine axis relative to the axis of impact, was critical in determining the kinematic response of the spine.^14,22 Peak force itself was not a useful predictor of cervical spine injury, and the use of impulse (integral of force over time) and maximum head velocity has been proposed by Alem et al.¹ In further studies, this group found that upon impact, the spine could buckle, producing both extension type injuries at C3–5 and flexion injuries at Tl–4.²¹

With the advent of computed tomography (CT) scanning, bone retropulsed into the spinal canal with resulting compression of the spinal cord was noted as a possible mechanism of injury.²⁴ Schönström, et al²⁵ used CT scans to document changes in the crossectional area of the lumbar canal in cadaver specimens, intact and with the disc excised, in flexion-extension and axial compression. In a retrospective study of 158 patients, Gusta et al¹⁰ found that irreversible spinal cord injuries occurred most often with hyperflexion, followed by crush fractures, and dislocations. In a further clinical study, the same group found that spinal cord injury in 46.3% of patients was caused by an injury involving a shearing mechanism.¹¹ Pintar et al²³ documented the anatomic alterations due to compressive loading of the cervical spine by high-speed video, as well as deep freezing after fracture. Frozen sections revealed the pathologic positions of the vertebrae. The authors concluded that “the spine configurations secondary to axial loading are substantially different when relaxed” (i.e., unloaded), “in contrast to those shown in the present study” (i.e., post-injury).

Their conclusion leads to the basic hypothesis of this study; understanding the transient motion of the vertebrae and changes in the canal geometry during the injury event are necessary for a complete understanding of the mechanisms of spinal cord injury. In this first part of our study devoted to injuries by direct compression, we tested the following three hypotheses: (1) significant transient changes occur in the geometry of the spinal canal during impact; (2) two potentially injurious geometric changes are occlusion of the crossectional diameter of the spinal canal, which can constrict the spinal cord, and changes in the length of the canal; and (3) post-injury radiographic measurements of spinal deformity (used to determine the extent of the injury and the method of treatment) underestimate the actual geometric changes occurring in the spinal canal during impact.

Methods

Specimen Selection and Preparation.

Sixteen cervical spines (Cl-TlO) were removed from donor cadavers. Each was radiographed and inspected for disc degeneration, osteophyte formation, severe osteoporosis, or other pathologic conditions. From the group, ten were selected for the study (Table 1). Muscle tissue was removed from each specimen, ensuring that the intervertebral discs and all ligamentous interconnections remained intact. The spinal cord, dura, and nerve roots were also removed. Specimens were then wrapped in wet towels, placed in plastic bags, and stored at −20°C until used.

Table 1.

Conditions of Specimens Pre/Post Fracture

No.	Pre-injury Condition	Post-Injury Condition
1	Degen disk disease, osteophytes	Vertical compression, stage 3, C3, C4
2	Excellent	Vertical compression, stage 3, C4
3	Good	Vertical compression, stage 2, C5
4	Fair, moderate degeneration	Compressive flexion, stage 2, T1
5	Fair, moderate degeneration	Pedicle fracture, C5
7	Good	Vertical comp, stage 3, C4, comp ext, C6/7
9	Good, mild degen and osteophytes	Vertical compression, stage 2, CS
10	Fair, marked degen and calcif disk	Vertical compression, stage 3, C3
14	Fair, moderate degeneration	Compressive extension, stage 2, C4
16	Fair	Compressive extension, stage 1, C4

Measurement of Transient Occlusion of the Cervical Spinal Canal During Impact.

Transient occlusion of the spinal canal was measured by recording the changes in fluid pressure, in a flexible tube having water circulating through it, placed into the spinal canal of each specimen. The occlusion transducer consists of a pump located in a reservoir filled with water (Figure 1). After removal of the spinal cord, dura, and nerve roots from the specimen, a flexible plastic tube (Tygon R3603, 0.5 inch nominal OD × 0.375 inch ID) connected to the pump, was placed through the spinal canal and then led back into the reservoir. From a tap in the tube, upstream of where it enters the spinal canal of the specimen, a pressure transducer (model P50, Gould Electronics, Oxnard, CA) was used to record system pressure. Changes in the crossectional area of the tube caused by occluding it, increased the upstream pressure.

Figure 1.

A schematic diagram demonstrating the method of measuring spinal canal occlusion and distraction, delivery of the impact by a dropweight, and measurement of the impact force.

To relate pressure changes to occlusion area, two calibrations were performed, one relating pressure to tube diameter, and the other, tube diameter to crossectional area. To relate measured pressure to tube diameter, the tube was pinched using a dial caliper and the pressure increase from base operating pressure was determined. To correlate diameter to area, the tube was filled with wet methacrylate, and the tube was pinched using the caliper. After the methacrylate had cured, the tube was sectioned, and the crossectional area for a known diameter was determined using a video scan of the section.

Extensive static and dynamic studies were performed to define the performance of the system.²⁶ The effect on pressure-occlusion area characteristics of the system with different shapes and lengths of occlusion were determined. Also, the effects of changing tubing geometry, specifically the outer/inner diameter ratio were examined. Dynamic testing of the system was performed by impacting one end of the core rod of an LVDT, whose other end rested against the system tube, and comparing the area change measured by the occlusion transducer, derived from pressure changes, with that given by the LVDT. This was required because the pressure wave created by a dynamic occlusion is not the same as that created by static occlusion of the system. From these tests, the rise time response of the transducer could also be measured. Further dynamic testing consisted of measuring the system frequency response. For this test a DC motor was used to drive the impactor, through the LVDT, against the tube, at frequencies from 5 to 50 Hz since the time to peak amplitude of the pressure pulse during testing was 20 msec and the first full period was 90 msec.

Measurement of Spinal Canal Length Changes During Impact.

Length changes in the spinal canal were measured in the following manner. A flexible cable was fixed to Cl and passed through the foramen transversarium from C2 to Tl on each side of the specimen. At Tl, the cables entered tubes in the specimen potting material, which acted as guides, and around pulleys. The cables turned 90° in direction, arid then were attached to the core rods of LVDTs (model DCD 1000, Schaevitz Engineering, Pennsauken, NJ). A light spring, placed between the core rod and the LVDT barrel, maintained pretension in each cable but was insignificant in altering the displacement properties of the specimen. Length changes in the spinal canal were computed from the mean of the left and right side measurements.

Experimental Procedure.

Each specimen was thawed, radiographed, and CT scanned before potting. After potting, the axial length measurement cables and the occlusion transducer tube were positioned, and the lower specimen pot was fixed to a load cell (model 662.lOA 04, MTS Systems, Minneapolis, MN) which was used to record the impact history. A fixture mounted to the top of the upper pot contained a threaded hole into which was screwed a 2-meter long, 12.7-mm diameter brass rod (Figure 1). The position of this impact guide rod was adjusted to the tenter of balance of the specimen, i.e., the location at which an applied axial load did not produce flexion or extension.

Because of the requirement for high-rate data collection, two personal computers (model 286, PC Tech, Austin, TX) equipped with analog to digital signal conversion boards (model 2801, Data Translation, Marlboro, MA) and data collection software (Labtech Notebook v 4.0, Laboratory Technologies Corp, Wilmington, MA) running at a sampling rate of 450Hz, were used. One system was used to measure the outputs of the two LVDTs, and one the impact force and the pressure changes in the occlusion transducer. A 5.95-kg dropweight was raised to a height of 1.53 meters above the specimen. The weight was dropped along the guide rod, the measurement system was triggered, and the specimen impacted. The energy of impact was 89.2J (19) for all except three smaller specimens, which were impacted with 74.2J.

Radiographic Measurements.

Pre-injury and post-impact plain anterior-posterior and lateral radiographs were obtained. Care was taken not to disturb the positions of the vertebrae post injury. A marker of known size was used to accommodate for magnification of the specimen. Spinal canal length changes were measured from both pre- and post-injury lateral radiographs in the following way. A vertebral body superior to the injury was chosen. In eight cases, this body was C2, although in two cases, Cl was sufficiently clear, non-rotated and oriented on the radiographs to be used instead. An inferior body was chosen (usually C7, Tl, or T2) to encompass as much of the specimen as possible. On the radiograph at each vertebra selected a line was drawn along the endplate and another line extended along the posterior wall of the vertebral body, to define the posterior inferior corner of the cranial vertebral body and the posterior superior corner of the caudal vertebra. The vertical distance between the two points was measured with a dial caliper.

From inspection of the post-injury CT coronal slices, the area of maximum occlusion was identified. At this point, the midsagittal diameter was measured along with the crossectional area of the canal. Also, the midsagittal diameters of adjacent intact vertebrae were determined.

Analysis of the Data and Testing of the Hypotheses.

From the experimentally generated data, the impact force, the maximum pressure recorded by the occlusion transducer, and the initial and maximum LVDT readings were derived. The occlusion transducer pressure value was corrected for its over-damped dynamic response. Also since fit of the tube in the canal was variable, the diameter of the tube was compared with that of the canal at the injury site using the mean of diameters from adjacent intact vertebrae, derived from the CT scans. Then the diameter change was determined and converted to an area change. The maximum left- and right-side LVDT readings were averaged. From the plain radiographs and CT scans, the fracture was classified according to Allen et at.²

To test the hypotheses, we compared the post-injury and transient axial height losses and canal areas using a Student’s t-test to determine whether significant differences existed between data groups. Also, linear regressions were performed to determine whether relationships existed between post-injury and transient data. Statistical analysis was performed using a statistical analysis package (Statview II, Abacus Concepts, Berkeley, CA) on a personal computer (Mac II cx, Apple Computer Co, Cupertino, CA).

Results

Pedormance of the Occlusion Transducer System

Static Performance Characteristics.

The system exhibits a nonlinear, but repeatable response to occlusion as shown in Figure 2. The range of measurement of the transducer does not cover the complete range of occlusions possible (i.e., 0%−100%) because of the thickness of the tubing wall. Careful selection of the tubing diameters can produce a useful range of measurement. There was no detectable effect due to the length of the occlusion which ranged from 0.7 mm to 25.4 mm long. The shape of the occlusion, whether oval, as would occur by having opposite surfaces compress the tubing, or lunar shaped, as would result from a protrusion acting on one surface, did not affect the response, for the same occluded area. There was no effect of a second occlusion on the pressure drop resulting from a primary occlusion if the secondary occlusion is smaller. The largest occlusion in the system dominates the pressure drop. A variety of tubing materials were investigated with the thought being that the material should be as flexible as possible to minimize its influence on the occlusion measurement. However extremely flexible and thin-walled tubing, such as that used in the Penrose drain, pinched, crimped, and had oscillatory behavior in response to a strike. For these measurements,·Tygon formulation R3606 tubing was chosen in a size of 1/2 inch nominal OD × 3/8 inch ID.

Figure 2.

Pressure/occlusion characteristics of the occlusion transducer demonstrating measurement range as a function of tubing wall thickness for a fixed inside diameter.

Dynamic Characteristics.

The transfer function was defined as the ratio of occluded area from the transducer to input area derived from the diameter change recorded by the LVDT. This transfer function was plotted as a function of the percent occlusion of the tube (Figure 3). The results show that the system response is over-damped but becomes slightly less so at higher occlusions. However, this characteristic is repeatable. The impact event time in calibration tests ranged from 10 to 25 milliseconds, which is similar to the time frame of a dropweight impact. The viscoelastic response of the tube to impacts at different strike velocities was determined by comparing occlusion measurements at the same input displacements and different velocities. No significant differences were found. The frequency response analysis showed no difference in the transfer function within a test range of 5–50 Hz. The rise time of the transducer, determined by comparing the applied impulse and the resulting response is shown in Figure 4. From ten trials, the mean rise time of the impulse was 14.9 msec (SD = 1.1 msec) and of the occlusion transducer was 19.9 msec (SD = 0.8 msec).

Figure 3.

Ratio of transducer output to known input as a function of tube occlusion.

Figure 4.

Rise time (time to peak value) of the impact as measured from transverse impact on the tube wall and the response of the occlusion transducer.

Measurement Error.

The system chosen for this study, based on the size of the cervical spinal canals in our specimens, was Tygon R 3606, 1/2 in nominal OD × 3/8 in ID tubing, connected to rigid polyethelyne tubing outside of the spinal canal, with a base system operating pressure of 35 mm Hg and 215 mm Hg (maximum) and a measurement range of 0–56% occlusion. An analysis of inherent errors during the procedure of converting the pressure to an area change showed an absolute uncertainty of ±5% occlusion.

General Characteristics of the Specimens After Impact

In all cases, the impact produced severe damage to the cervical spine specimen. As shown in Table 1, the majority of the injuries were axial compressive burst type with retropulsion of bone into the canal, significant height loss, and posterior translation of the superior vertebra (Figure 5). In three of ten cases, there was an extension component to the fracture. A major difference with this type of injury was the occlusion of the spinal canal by the posterior structures.

Figure 5.

Top: Coronal section of a specimen demonstrating residual compression of the occlusion transducer tube after impact. Bottom: Sagittal reconstruction shows tube compression both anteriorly and posteriorly.

Transient Versus Post-Injury Vertebral Column Height Loss

Figure 6 shows a time history of the change in axial height during impact, taken from specimen 10. The partial recovery of height, to about 60% of the preinjury value, is evident. All of the injuries were produced by a large compressive force, and all specimens, including those three that resulted in some extension displacement, had significant height loss. None of the specimens displayed distraction.

Figure 6.

Axial height loss as a function of time post impact, from the specimen. specimen 10, as recorded by the LVDTs located on either side of the specimen.

The mean post-injury height loss was 3.1 mm (SD = 4.4 mm) compared with a mean transient height loss of 8.9 mm (SD = 4.5 mm), P = 0.003. The data show that 65% of the maximum vertebral height loss under compressive loading is recovered, with 35% remaining as a permanent post-injury height loss.

The relationship between the maximum transient height loss and that measured from the post-injury radiographs was tested to determine if post-injury height loss could be a predictor of maximum height loss. The correlation between these measures (Figure 7) was not significant (r² = 0.230, P = 0.16).

Figure 7.

Linear regression between the maximum transient vertebral column height loss, measured during impact, and that measured from post-injury radiographs.

Transient Versus Post-Injury Spinal Canal Occlusion

Figure 8 provides an example of a force and occlusion transducer pressure time history as recorded from specimen 10. The pressure is the raw output from the occlusion transducer. The force was recorded below the specimen, so the measurement includes the elasticity of the spine in transmission of forces. The occlusion transducer had become saturated in this example, at a pressure equivalent to 56% occlusion.

Figure 8.

An example of the force and occlusion transducer time histories during impact, from specimen 10.

From the postinjury CT scans, the mean canal area in the injured region was 268.4 mm² (SD = 59.2 mm²), while the mean peak transient canal area was 193.2 mm² (SD = 26.3 mm²), which were significantly different (P = 0.000 1). The correlation between transient and post-injury canal areas (Figure 9) was not significant (r² = 0.272, P = 0.12).

Figure 9.

Linear regression between transient minimum spinal canal area and the. post-injury area derived from CT scans.

The midsagittal diameters (MSD) (from which the transient occlusion areas were derived) were also compared, with post-injury MSDs taken from the CT scans at the same time that the areas were measured, since there was potentially less error in determining CT MSD rather than CT area. The mean post-injury MSD was 13.67 mm (SD = 2.71 mm) compared with a mean transient MSD of 5.12 mm (SD = 0. 1.47 mm) (P = 0.001). The correlation between transient and post-injury midsagittal diameters was significant, though not strong (r² = 0.48 1, P = 0.03).

Discussion

The range of occlusion measurements that we chose to measure was 25%−55%. Different ranges can be chosen by selecting different thresholds although the magnitude of the range is constant (Figure 2). The threshold is governed by the inner diameter of the tube relative to its outside diameter, as well as fit in the canal. Occlusion beyond about 60% was not measurable with the tubing selected for this experiment since the inside bore of the tube was completely closed at this point. Since 30% occlusion is usually taken as a clinical indicator of the onset of spinal cord injury in the cervical spine,⁵ we chose 30% occlusion as the measurement threshold.

The occlusion transducer has inherent inaccuracies, one of which relates to its fit in the spinal canal. The spinal canal dimensions vary both in the individual vertebrae and along the spine, while the transducer tube diameter is constant. If it is a loose fit at the site of the injury, then the actual occlusion will be greater than that recorded. For this reason, we determined the region of maximum occlusion from the post-injury radiographs and CT scans, then determined the mean of canal dimensions at adjacent levels and used that as a correction factor for fit. The occlusion transducer may also alter the biomechanics of the injury. If the tube were excessively stiff, it would inhibit occlusion of the canal. However, in eight of ten experiments, the transducer tube was completely occluded. Also, the transducer tube may force the retropulsed bone back into position after the injury. We observed, however, that the reduction of retropulsed bone required that the axial height lost be regained.

The axial height loss is usually determined from the plain radiographs by averaging the anterior and posterior height dimensions. In our study, we measured only the posterior height between uninjured vertebrae adjacent to the injury site. This height loss was a linear measurement, ignoring the cervical lordosis; however, our specimens had very little lordosis. This dimension corresponds to the length change of the anterior aspect of the spinal canal, which corresponded reasonably well to our dynamic length measurement, considering that the foramen transverserium are slightly anterior to the canal.

Because of its inherent lordosis, the ligamentous cervical spine forms an arc in the sagittal plane. Applying compressive loads to the end vertebral bodies creates a bending moment that bends the spine into extension. However, the rod used to guide the dropweight was screwed into the top of the upper pot (i.e., the free end of the specimen). With the bottom of the specimen fixed, manually limiting the motion of the guide rod controlled the angle of extension and permitted for the most part, axial compressive loading of the specimens.

The large regain in axial height of the specimen after impact was interesting to note. The elasticity of the disc contributes to height regain as do the longitudinal ligaments. If bone is extruded radially from the injured vertebra, the longitudinal ligament tensions increase. On high-speed video of the impact, large deformations in the injured vertebra were recovered after the impact. This may be a result of fat and marrow in the interstices of the trabecular bone, which pressurizes under impact, then pushes the bone fragments apart.

The correlation between transient and post-injury midsagittal diameters was significant, whereas between canal areas, was not. The measurement of canal area from the CT scan is less precise than midsagittal diameter because the area in the lateral regions of the canal is sometimes hard to define. The transducer occlusion area is derived from the diameter, so results in another step and an additional source of error. The correlation between transient and post-injury midsagittal diameters indicates that this measure may be used as a rough predictor of the severity of the transient injury.

In summary, three hypotheses were proposed relating to the transient geometrical changes in the spinal canal during cervical spine injuries. For axial compressive bursting injuries, we answered these hypotheses.

1. In this injury pattern, significant changes occur in the geometry of the spinal canal.

2. Two significant geometric changes are canal occlusion and vertebral column shortening. There was no canal distraction. Cord damage may occur due to spinal canal occlusion or excessive height loss.

3. The post-injury geometry of the spinal column does not correlate to its transient geometry at the time of impact. The cervical spine undergoes much more severe injury than the post-injury radiographs and CT scans demonstrate. Post-injury, it regains 65% of the maximum height loss at impact, 139% of the minimum canal area, and 269% of the minimum midsagittal diameter.