Experimental Design
Sixteen beagles underwent dynamic spinal cord compression. When somatosensory evoked potentials declined by 50%, dynamic compression loading was no longer increased but was sustained. Sustained compression was continued for thirty minutes in eight dogs and for 180 minutes in the other eight. Spinal cord interface pressures were compared at the end of the dynamic loading (time 0) and throughout the period of sustained compression. Somatosensory evoked potentials were regularly monitored throughout the duration of the compression. After the designated time, the compression was discontinued, and somatosensory evoked potentials were monitored over a sixty-minute recovery period.
Functional motor recovery was evaluated for twenty-six days after the injury. After approximately four weeks (mean and standard error of the mean, 28.5 ± 0.72 days), the dogs were anesthetized with sodium thiopental, a magnetic resonance imaging scan was made, and a final recording of somatosensory evoked potentials was carried out. The animals were then killed, and the spinal cords were removed for histological analysis.
Surgical Preparation and Technique
The experimental design and procedures for this study were approved by the Institutional Animal Care and Use Committee. Sixteen adult beagle dogs weighing 11 to 15 kg were anesthetized with sodium thiopental (15 to 20 mg/kg, intravenous induction), a mixture of 50% nitrous oxide and 50% oxygen, and halothane (0.5% to 0.7%). The dogs were intubated, paralyzed with 0.1 mg/kg/hr of pancuronium, and mechanically ventilated to maintain normal respiration and acid-base balance. Body temperature was monitored with an esophageal temperature probe and was maintained between 37°C and 38°C with a heating pad and warming lights. All animals received normal 0.9% saline solution infused at 10 mL/kg/hr.
With use of sterile technique, a laminectomy at the thirteenth thoracic segment (T13) was performed. A device for weight-loading spinal compression (specially constructed by the Engineering Department of Case Western Reserve University) with a hydraulic piston was suspended over the dura and attached to the transverse processes of T13 with 2.0-mm bone screws (
Fig. 1 )
16 . The diameter of the piston was 7 mm, which was similar to the diameter of the spinal canal. A subminiature pressure transducer (Kyowa Electronic Instruments, Kyowa, Japan) attached to the spinal cord contact end of the loading piston relayed real-time spinal cord interface pressures to an amplifier and a computer equipped with a data-acquisition system and custom-written software
17 .
After surgery, the wound layers were closed and the skin was approximated with subcuticular ligatures. After the dogs recovered from the anesthesia, they were allowed free mobility in their cages. Dicloxacillin (10 mg/kg) was administered orally for ten days.
Biomechanical Loading of the Spinal Cord
Dynamic cord compression was initiated with the loading device precalibrated to indent the spinal cord at a constant 0.17 mm/min
18,19 . Somatosensory evoked potentials were continuously monitored. When the amplitudes of the evoked potentials in the lower extremity were reduced by approximately 50% of the baseline value, the position of the piston was maintained. Sustained cord compression was then maintained for either thirty minutes (in eight animals) or 180 minutes (in eight animals). This was followed by removal of the piston, which allowed decompression of the spinal cord.
Monitoring of Somatosensory Evoked Potentials
Somatosensory evoked potentials were obtained by stimulating the median and tibial nerves and recording cortical evoked potentials on a Viking IV Evoked Potential System (Nicolet Biomedical Instruments, Madison, Wisconsin). To control for changes in the evoked potential signal caused by postsynaptic cortical anesthetic effects, we calculated the ratio of the cortical evoked response to stimulation of the median nerve in the upper extremity to the cortical evoked response to stimulation of the posterior tibial nerve in the lower extremity. It is necessary to use this ratio to measure changes specifically related to cord compression because the cortical evoked responses to median and posterior tibial nerve stimulation would be equally affected by cumulative anesthetic effects, but only the cortical evoked response to the posterior tibial nerve stimulation would be affected by the T13 spinal cord compression. The normalized lower extremity-to-upper extremity ratio (LE/UE) was calculated by dividing the mean amplitude of posterior tibial cortical evoked response by that of the median nerve cortical evoked response. Functional recovery was determined by the return of lower-extremity somatosensory evoked potentials.
Functional Outcomes
Coordinated motor function was assessed with use of an open-field test. Motor scores were determined with a modification of the standard Tarlov system
20 . This assessment is based on a 7-point scale in which 0 indicates no movement of or weight-bearing by the hindlimbs; 1 indicates that the dog has barely perceptible movement of the hindlimbs, does not bear any weight on the hindlimbs, and moves around the cage; 2 means that the dog has frequent and/or vigorous movement of the hindlimbs, does not bear any weight on the hindlimbs, and pushes off purposefully; 3 means that the animal can support weight on the hindlimbs, may take one or two steps, and often bears weight on the top of the feet; 4 indicates that the dog is fully weight-bearing, consistently takes steps using the distal portions of the hindlimbs, has limited hip flexion and poor balance, and occasionally bears weight on the dorsum of the foot and the pelvis falls repetitively; 5 means that the dog walks with only a mild deficit, the hindlimbs follow with minimal deviation from the midline, and the dog can stand on the hindlimbs alone; and 6 indicates normal walking, good balance, and recovery from foot slip.
Balance, which was also tested in the open field, was rated as 0 (no balance), 1 (poor), 2 (fair), or 3 (good). Cadence (coordination and spasticity) was evaluated by mildly striking the foot and assigning a rating of 0 if there was no cadence, 1 (poor) if the foot-strike elicited unstable pelvic rock (spastic), 2 (fair) if the foot-strike elicited minor pelvic shift, and 3 (good) if the response was normal.
Stair-climbing was judged on a standard flight of stairs with smooth rubber surfaces. The dogs were led up the stairs, and their ability to climb them was scored, by an independent caretaker, as 0 (no climbing), 1 if they climbed one stair at a time and supported most of their weight on the forelimbs (poor), 2 if they climbed one stair at a time but had better balance and used the hindlimbs (fair), and 3 if they climbed normally (good).
For the inclined-plane test, the dogs were scored according to whether or not they could climb a custom ramp that was 4 ft (1.2 m) in length and at a 20° angle from the floor.
Recovery of bladder function was determined by the need for manual expression of urine from a distended bladder. In the days immediately following the spinal cord injury, the bladder was expressed every eight hours until manual expression was no longer required.
Magnetic Resonance Imaging
Magnetic resonance images of the thoracolumbar spinal column of each dog were made with use of a Siemens Vision magnetic resonance imaging scanner (Erlangen, Germany) with a 1.5-tesla magnet. Sequences for standard T1-weighted and T2-weighted images with a scan thickness of 3 mm were obtained. To obtain high-quality scans, the animals were anesthetized with 15 to 20 mg/kg of sodium thiopental prior to placement in the scanner. T2-weighted images were analyzed for lesion volume (taking into account slice thickness and spacing) with use of a computerized image analysis system (Image Tool; University of Texas Health Science Center at San Antonio, San Antonio, Texas). The values were compared with those for a similarly sized region of normal cord adjacent to the injury zone to determine the cord lesion volume as a percentage.
Histological and Lesion Volume Analysis
The dogs were anesthetized with 15 to 20 mg/kg of sodium thiopental and given an overdose of sodium pentobarbital. The spinal cord (T12 to L1) was dissected, postfixed, cryoprotected, and blocked, and then serial longitudinal sections (20 µm) were cut on a cryostat. Sections were stained with hematoxylin and eosin to determine the volume of the lesion and with Luxol fast blue for histopathological analysis of gray and white matter. On the sections stained with Luxol fast blue, the area containing necrotic or damaged tissue was circumscribed with a computerized image analysis system (MCID; Imaging Research, St. Catherines, Ontario, Canada). Identification of spared tissue was based on positive staining for myelin with Luxol fast blue or a gray matter cytoarchitecture that approximated that seen in normal animals. Areas of cord damage and residual white matter at each horizontal level were determined, and the total volumes of the lesion and of the residual matter were derived by means of numerical integration of sequential areas. All slides were assessed blindly with respect to treatment.
Statistical Methods
Statistical analyses were performed with SigmaStat software (Jandel, San Rafael, California). Analysis of variance, the Mann-Whitney rank sum test, and the Friedman repeated-measures analysis of variance on ranks were used as appropriate. Differences were considered significant at p < 0.05.
Spinal Cord Loading
Spinal cord interface pressures were highest at the end of dynamic cord compression. The peak loading pressure in the thirty-minute compression group (mean and standard error of the mean, 30.4 ± 3.4 kPa) was not significantly different from that in the 180-minute compression group (29.9 ± 2.8 kPa) (
Fig. 2 ). Within five minutes of sustained cord compression, the interface pressures decreased by more than 50% of the maximum in both groups. After thirty minutes of sustained compression, the interface pressures decreased to approximately 25% of the maximum in both groups.
Recovery of Somatosensory Evoked Potentials
At the end of dynamic loading and before sustained compression (time 0), when the interface pressures were the highest, there was a decline in the amplitude of the somatosensory evoked potentials to 19.6% ± 5.2% in the thirty-minute compression group and to 23.4% ± 10.3% in the 180-minute compression group; both declines were approximately 80% from the baseline (
Fig. 3 ).
After decompression, somatosensory evoked potentials returned in all dogs in the thirty-minute compression group, with recovery to 63% of the baseline value at ninety minutes after decompression. Twenty-eight days after the injury, the mean amplitude of the somatosensory evoked potentials was 38% of the baseline in the thirty-minute compression group. In contrast, in the 180-minute compression group, decompression did not result in recovery of somatosensory evoked potentials either early (
Fig. 3 ) or late (twenty-eight days after the injury).
Functional Assessments
The mean modified Tarlov motor scores were significantly better for the thirty-minute group than for the 180-minute group at all time-points (
Fig. 4 ). The Tarlov scores improved significantly over time in both groups (chi square = 28.42, p < 0.001 for the thirty-minute group; chi square = 17.97, p = 0.003 for the 180-minute group). On the day following the surgery, seven of the eight dogs in the thirty-minute group were able to move the hindlimbs and two were able to bear weight on the hindlimbs (
Table I ). Within seven days, seven of the eight dogs in the thirty-minute group were capable of weight-bearing (a Tarlov score of 3 or better). By twenty-one days, all six animals remaining in the group were walking either normally or with only a mild deficit (a Tarlov score of 5 or better) (
Table I ). In contrast, on the day following the surgery, none of the dogs in the 180-minute group were able to move the hindlimbs. By seven days, only two of the dogs in the 180-minute group were able to move the hindlimbs and none had regained weight-bearing (
Table I ). Twenty-one days after the injury, three of the six dogs remaining in the 180-minute group were capable of weight-bearing, although none were ever capable of walking normally or with only a mild deficit.
There was an increase in balance over time in both groups (chi square = 26.75, p < 0.001 for the thirty-minute group; chi square = 14.86, p = 0.011 for the 180-minute group) (
Fig. 4 ). All animals in the thirty-minute group recovered normal balance within the twenty-six-day test period, whereas none of the animals in the 180-minute group recovered normal balance. Accordingly, the thirty-minute group had significantly better balance scores at almost all time-points (p < 0.01).
Cadence also increased over time in both groups (chi square = 28.30, p < 0.001 for the thirty-minute group; chi square = 14.86, p = 0.011 for the 180-minute group) (
Fig. 4 ). All animals in the thirty-minute group recovered normal cadence within the twenty-six-day test period, whereas none of the animals in the 180-minute group recovered normal cadence. The thirty-minute group had significantly better cadence scores two weeks after the injury (p < 0.001).
There was an increase in stair-climbing ability over time in the thirty-minute group (chi square = 28.10, p < 0.001) (
Fig. 4 ). Five of the six animals in that group recovered a normal ability to climb stairs within the twenty-six-day test period, whereas none of the animals in the 180-minute group recovered normal stair-climbing ability. Only one dog in the 180-minute group improved enough to make an attempt to climb the stairs, but that was not until twenty-six days after the injury. All other dogs in the 180-minute group were unable to climb stairs at any time. Accordingly, the thirty-minute group had significantly better stair-climbing scores (p < 0.01).
Initially none of the dogs could walk up the inclined plane (
Table II ), but by fourteen days after the injury all dogs in the thirty-minute group could do so. It took nearly twice as long (twenty-six days) for half of the dogs in the 180-minute group to be able to walk up the inclined plane. At all time-points, beginning at seven days, more dogs in the thirty-minute group than in the 180-minute group were able to walk up the inclined plane.
Recovery of bladder control or reflex emptying occurred more rapidly in the thirty-minute group (
Table III ). On the first day after the injury, five of the eight dogs in the thirty-minute group had recovered bladder function, whereas none of the dogs in the 180-minute group had. Within seven days, none of the dogs in the thirty-minute group had a distended bladder and none required manual expression of urine. In comparison, in the 180-minute group, all dogs did not recover bladder control until the fourteen-day time-point.
Magnetic Resonance Imaging
An abnormally high signal within the central region of the spinal cord on T2-weighted magnetic resonance images indicated spinal cord injury (
Fig. 5 . Changes in signal intensity indicated increased damage. Lesion volumes were significantly smaller (p = 0.04) in the thirty-minute compression group than in the 180-minute group (
Table IV ).
Histological Assessments
The sustained compression injury produced cavitation of the central region of the spinal cord parenchyma at the T13 epicenter, leaving a rim of intact white matter (
Fig. 5 ). It was evident on gross histological observation that the thirty-minute group had more spared tissue and less cavitation than the 180-minute group did (
Fig. 6 ). Quantitative analysis of tissue-sparing revealed that the thirty-minute group had a dramatically smaller lesion volume (p < 0.001) and a greater percentage of residual white matter (p = 0.005) than the 180-minute group did (
Table IV ).
This study demonstrated that the duration of sustained compression of the spinal cord is a crucially important factor in the secondary injury process. Sustained compression for 180 minutes resulted in long-term decreases in the amplitudes of somatosensory evoked potentials and greater lesional volumes as determined with both magnetic resonance imaging and histological evaluation (
Table IV ). The larger size of the lesions was associated with a greater functional deficit. As a group, the dogs who had had decompression after 180 minutes of sustained compression had greater cord damage and greater long-term functional deficits. Those dogs never regained normal use of the hindlimbs. In contrast, decompression after thirty minutes of sustained compression resulted in early recovery of somatosensory evoked potentials after decompression that coincided with histological findings of minimal long-term spinal cord damage and early functional improvements. Those dogs had rapid recovery of hindlimb motor function. Taken together, these data provide strong evidence of a sequence of secondary injury processes that are causally related to time-dependent sustained cord compression.
Our findings are similar to those of Delamarter et al.
13 , who observed a close association between the duration of spinal cord compression and the extent of functional recovery, amplitude of somatosensory evoked potentials, and neurohistological characteristics. They used rapid compression of 50% of the cord followed by static ligature and examined the results following decompression performed immediately, after one hour, after six hours, after twenty-four hours, or after six weeks. Together, the results of the two studies clearly demonstrate a temporal window wherein early decompression may result in neurological preservation.
A number of experimental studies and techniques have been used in an attempt to correlate sustained cord compression with progressive injury
2,20-23 . Studies of sustained cord compression have demonstrated a wide window of opportunity wherein decompression may alter the secondary injury pathophysiology
2,15,20,24,25 . We found that models that apply spinal cord compression with an unremitting interface force, such as those that rely on weights, balloons, or dynamic clips, create high spinal cord interface pressures that do not rapidly decline. Clinical causes of sustained cord compression result in spinal cord displacement, with the greatest interface pressures sustained at the time of initial trauma and interface relaxation occurring thereafter.
Previous studies have demonstrated a rapid increase in interface pressure during the short period before the decline in the amplitude of the somatosensory evoked potentials, and this has been attributed to the viscoelastic tissue response to dynamic loading
17 . This relatively narrow range of maximum pressures, which coincides with loss of somatosensory evoked potentials, may be related to a vascular threshold. Pressures above this limit may diminish regional spinal cord blood flow. This hypothesis is supported by our previous findings of almost complete loss of regional spinal cord blood flow at the end of dynamic compression
14,16 . While pressure phenomena may contribute to early neurological loss and diminished blood flow, the relatively rapid viscoelastic relaxation of the spinal cord within the early phases of sustained cord compression suggests that there are other mechanisms of secondary injury that are linked to tissue displacement, such as ischemia
16 . Although, in our previous studies, tissue relaxation decreased pressures to 12% to 25% of the maximum within one hour during sustained cord compression, regional spinal cord blood flow remained significantly below baseline levels (p < 0.05)
14,16,26 . Upon decompression there was a rapid rebound and hyperemic spinal cord blood-flow response; however, regional spinal cord blood flow diminished throughout the three-hour postdecompression monitoring period
14,16 . Hence, regional hypoperfusion and ischemia probably contribute to the progressive secondary injury. This does not rule out the possibility of other secondary injury pathways.
The histopathological changes found in the present study represent the classic morphology of spinal cord injury. The greatest injury was at the epicenter of compression, with relative centripetal damage cephalad and caudad. The histochemical results corresponded with the findings on magnetic resonance imaging, which showed a region of high signal intensity within the central region of the cord. The lesional volume increased with a longer duration of sustained cord compression. The magnetic resonance imaging and histological findings showed that our model produced consistent pathological changes and that those changes were associated with the duration of compression and with long-term functional impairment. This suggests that, with similar patterns of spinal cord injury, magnetic resonance imaging may be a correlative and reliable predictor of neurological damage and recovery.
In summary, the spinal cord undergoes viscoelastic relaxation during sustained compression; nonetheless, a longer duration of compression injury is associated with reduced electrophysiological recovery, increased pathological changes, and significantly greater functional impairment. Hence, sustained displacement is an important factor in the secondary injury process. This finding indicates that damage to the spinal cord depends strongly on the duration of displacement and the timing of treatment. The results underscore the importance of timely decompression to improve long-term functional recovery.
Note: The authors thank Aileen J. Anderson, PhD, for her assistance with the analysis of the histological specimens and the lesional volume. They also express gratitude to Chris Biro for the technical assistance provided in the set-up and performance of the somatosensory evoked potentials.