Dorsal and ventral horn atrophy is associated with clinical outcome after spinal cord injury

Our study protocol was designed in accordance with the Declaration of Helsinki and was approved by the local ethics committee of Zurich (KEK-ZH-Nr. 2012-0343, PB_2016-00623). All participants gave their written informed consent before participation.

We recruited 17 patients with SCI (mean age 48.7 ± 14.1 years, 3 female patients) between November 2014 and May 2016 who were preciously admitted to the University Hospital Balgrist (Zurich, Switzerland). Twenty-one healthy controls (mean age 41.7 ± 11.3 years, 7 female controls) from the local neighborhood served as a control dataset that was acquired and used in a previous study.⁸

Inclusion criteria for patients with SCI were traumatic cervical SCI, no other neurologic or mental disorders affecting clinical outcome, age between 18 and 70 years, MRI compatible, and no pregnancy.

All patients were examined with comprehensive clinical protocols to assess neurologic and functional impairment. These included the International Standards for Neurological Classification of Spinal Cord Injury protocol for motor, light-touch, and pinprick score and completeness of injury⁹; the Spinal Cord Independence Measure (SCIM) to measure daily life independence¹⁰; the Graded Redefined Assessment of Strength, Sensibility and Prehension (GRASSP) for assessing upper limb function¹¹; and the Walking Index for Spinal Cord Injury (WISCI).¹² All patients completed the full protocol, except GRASSP score was not available for 1 patient.

Contact heat evoked potentials (CHEPs) and somatosensory evoked potentials (SSEPs) were acquired bilaterally in patients at the dermatomes C4, C6, and C8 to measure the integrity of the spinothalamic tract (i.e., CHEPs) and the dorsal column (i.e., SSEPs). For the acquisition of CHEPs¹³ and SSEPs,¹⁴ the same protocols were applied as previously described.

A contact heat stimulator (PATHWAY Pain & Sensory Evaluation System, Medoc, Ramat Yishay, Israel) was used to deliver contact heat stimuli from a baseline temperature of 35°C to a peak temperature of 52°C with a heating rate of 70°C/s and a cooling rate of 40°C/s. For each dermatome, we first assessed heat perception and pain thresholds within 2 consecutive trials. For the CHEPs recording, scalp recording sites were prepared with Nuprep (D.O. Weaver & Co, Aurora, CO) and alcohol. Three 9-mm Ag/AgCl surface disk electrodes were positioned according to the international 10-20 system with the active electrode at the Cz position and referenced to linked earlobes (A1–A2); impedances were kept <5 kΩ. Ten to 15 contact heat stimuli were applied (interstimulus interval 8–12 seconds). Two seconds after each stimulus, an audio cue appeared, and patients rated their perceived intensity according to a numeric rating scale. All signals were sampled from 100 milliseconds before the trigger to 1,500 milliseconds after the trigger at a sampling rate of 2,000 Hz with a preamplifier (20,000× bandpass filter = 0.25–300 Hz; ALEA Solutions, Switzerland). Data were recorded in a LabView-based program (V1.43 CHEP; ALEA Solutions, Zurich, Switzerland) with a 100-millisecond period before the trigger and 1-second posttrigger period. Raw data were bandpass filtered from 0.5 to 30 Hz.

For dermatomal SSEPs, Key Point (Medtronic, Mississauga, ON, Canada) was used to record and deliver electric stimulation of 3 Hz. Stimuli were elicited by single 0.2-millisecond, repetitive, square-wave electric stimulation. We first assessed electric perception and pain thresholds for each dermatome (not exceeding 40 mA) within 2 consecutive trials. For the recording of SSEPs, surface gel electrodes (10 mm) were used on each dermatome after the skin was prepared with Nuprep (D.O. Weaver & Co) and alcohol. Disposable needle electrodes (Spes Medica, Srl, Genova, Italy) were positioned according to the international 10-20 system with the active electrode positioned at the contralateral side for the stimulated dermatome (C3-4) referenced to Fz; impedances were kept <5 kΩ. The stimulation intensity was individually set as 3-fold electric perception threshold. Averages of 2 traces of 300 cortical responses were obtained for each dermatome. Raw data were bandpass filtered from 2 to 2,000 Hz.

We determined amplitudes and latencies of each dermatome for each patient after averaging all single-trial waveforms for CHEPs (i.e., N2P2, N2, P2) and SSEPs (i.e., N1P1, N1, P1).

Furthermore, CHEPs and SSEPs were classified as normal (onset latency ≤2 SDs from control dermatome recording), pathologic (onset latency >2 SDs from control dermatome recording), or absent (not recordable).¹⁴ The CHEPs protocol was acquired fully in 14 patients and partially in 1 patient. For SSEPs, 12 patients received the full protocol and 2 patients participated in part of the protocol.

All imaging was performed on a clinical 3T Skyra^Fit scanner (Siemens Healthcare, Erlangen, Germany) equipped with a 16-channel radiofrequency receive-only head and neck coil and a radiofrequency body transmit coil. A stiff neck (Laerdal Medicals, Stavanger, Norway) was used in all participants to minimize motion artifacts. As a result of motion artifacts, 1 patient was excluded from macrostructural analysis, and 3 patients had to be excluded from microstructural analysis.

At the lesion epicenter, a sagittal T1-weighted (repetition time 600 milliseconds, echo time 9.9 milliseconds, flip angle 150°, in-plane resolution 0.57 × 0.57 mm, slice thickness 3.3 mm), a sagittal T2-weighted (repetition time 3,500 milliseconds, echo time 84 milliseconds, flip angle 160°, in-plane resolution 0.34 × 0.34 mm, slice thickness 2.75 mm), and an axial T2-weighted image (repetition time 5,510 milliseconds, echo time 93 milliseconds, flip angle 150°, in-plane resolution 0.5 × 0.5 mm, slice thickness 3.6 mm) were acquired to assess the lesion size.

At the cervical cord above the level of injury (centered at C2-3), 5 volumes were acquired with a T2*-weighted 3-dimensional multiecho gradient recall echo sequence (multiple echo data image combination¹⁵) in the oblique axial plane (i.e., perpendicular to the cord) to assess gray and white matter atrophy. Each of the 5 volumes acquired consisted of 20 partitions with a resolution of 0.5 × 0.5 mm² (field of view 192 × 162 mm², slice thickness 2.50 mm [10% gap], repetition time 44 milliseconds, echo time 19 milliseconds, flip angle 11°, readout bandwidth 260 Hz/pixel). Each volume took 2.13 minutes to acquire. Application of zero-filling interpolation doubled the nominal in-plane resolution (0.25 × 0.25 mm²).

At the identical level, a high-resolution diffusion tensor imaging (DTI) dataset was acquired with a cardiac-gated reduced-FOV single-shot spin-echo echo planar imaging sequence with outer volume suppression¹⁶ to assess microstructural changes of the whole spinal cord. Four measurements of 6 b = 0 (T2-weighted) and 30 b = 500 s/mm² volumes were acquired, resulting in 144 images per participant and a nominal acquisition time of 6.17 minutes. The following parameters were applied: repetition time of 350 milliseconds; echo time of 71 milliseconds; slice thickness of 5 mm (10% interslice gap); resolution of 0.76 × 0.76 mm²; FOV of 133 × 30 mm²; phase oversampling of 50%; 5/8 partial-Fourier imaging in the phase-encoding direction; cardiac trigger delay of 200 milliseconds; and minimal time between triggers of 1,800 milliseconds. After acquisition, zero-filling interpolation was used to double the in-plane resolution (0.38 × 0.38 mm²).

Statistical analysis of all macrostructural MRI, neurophysiologic, and clinical data was performed with Stata13 (StataCorp LP, College Station, TX). The mean age was not statistically different between healthy controls and patients (Mann-Whitney U test: z = −1.61, p = 0.10). All images were visually inspected for artifacts, and the analysis was conducted on 3 slices from each modality at the same level.

First, we assessed the morphometric differences in SCA, GMA, WMA, DCA, VHA, and DHA between patients and healthy controls by means of analysis of covariance, adjusted for age. For microstructural differences between patients and healthy controls, we used SPM12 for voxel-based analysis of the different DTI indexes (FA, MD, AD, RD) by means of analysis of covariance, adjusted for age. All statistical parametric maps were initially thresholded with a cluster-defining threshold of p < 0.01 (uncorrected) and clusters surpassing a cluster threshold of p < 0.05 (family-wise error corrected) are reported. Next, we used linear regression analysis to investigate the relationship between changes at the lesion site (midsagittal lesion area, length and width, and size of midsagittal tissue bridges) and remote cord macrostructural and microstructural changes. We then determined associations between macrostructural (SCA, GMA, WMA, DCA, VHA, and DHA) and microstructural (DTI indexes within lateral corticospinal tract, dorsal column, and spinal lemniscus) parameters and tract-specific clinical measures (motor, pinprick, and light-touch score, GRASSP, SCIM) using linear regression models, adjusted for age and lesion area. Finally, we investigated associations between macrostructural and microstructural MRI indexes and neurophysiologic outcome measures using linear regression models, adjusted for age and lesion area. Note that only patients with both recordable electrophysiologic potentials and available MRI data entered this regression analysis, resulting in a total number of 8 patients. For all microstructural associations, we extracted mean values of DTI indexes within anatomic regions of interest (lateral corticospinal tract, dorsal column, and spinal lemniscus [containing spinothalamic and spinoreticular tracts]) as embedded in the Spinal Cord Toolbox.²² The level of significance was set to p < 0.05.

Patients were scanned 6.7 ± 7.8 years after injury. An area of hyperintense signal was visible on the T2-weighted sagittal images in 16 patients (figure 1, A and B); 13 patients showed hyperintensities in their dorsal column, covering on average 41.4 ± 21.0% of the whole dorsal column, and 2 patients showed hyperintensities in the dorsolateral funiculus (e.g., corticospinal tract). The radiologic level of injury (hyperintense T2-weighted signal) covered the vertebral level C3-5 in 1 patient, C3 in 1 patient, C4 in 2 patients, C4-5 in 1 patient, C5 in 2 patients, C6 in 2 patients, C6-7 in 4 patients, and C7 in 2 patients. Two patients showed no signal alteration within the cord. The average lesion area was 45.4 ± 66.6 mm² with a lesion length of 11.3 ± 9.4 mm and a lesion width of 4.3 ± 3.5 mm. In 2 patients, the lesion occupied the full cord area, and no midsagittal tissue bridges could be identified. In the remaining 15 patients, the midsagittal tissue bridges had an average width of 2.9 ± 1.9 mm. No magnetic resonance abnormalities were identified in the control group.

Two patients were motor and sensory complete; 2 patients were motor complete and sensory incomplete; and the remaining 13 patients were motor and sensory incomplete. The motor score (maximum 100) was 68.1 ± 30.4; the light-touch score was (mean ± SD) 66.3 ± 32.7 (maximum 112); and the pinprick score (maximum 112) was 52.7 ± 35.0. Manual dexterity was impaired as assessed by the GRASSP score (149.8 ± 66.3 [maximum 232]), and functional independence was impaired as assessed by the SCIM score ([63.1 ± 31.3 [maximum. 100]). Eight patients were able to walk independently (20 of 20 points in the WISCI score); 2 patients were dependent on walking aids (5 of 20 and 9 of 20 points in the WISCI score); and 7 patients were not able to walk (0 of 20 points). All data are summarized in table 1.

All patients had neurophysiologic impairment of the spinothalamic tract, and a majority had impaired function of the dorsal column below the level of lesion as assessed by CHEPs and SSEPs, respectively. The mean ± SD perception/pain thresholds and the amplitudes and latencies of the recorded signals are shown in table 2.

Compared to healthy controls, patients showed a decreased SCA of 20.2% (p < 0.001, healthy controls 92.30 ± 8.49 mm², patients 73.71 ± 20.04 mm²). In patients, WMA was decreased by 16.9% (p = 0.001, healthy controls 75.34 ± 8.06 mm², patients 62.64 ± 18.22 mm²), and GMA was decreased by 30.0% (p < 0.001, healthy controls 16.96 ± 1.25 mm², patients 11.93 ± 2.73 mm²). In the white matter, DCA was decreased by 21.4% (p < 0.001, healthy controls 23.73 ± 2.99 mm², patients 18.65 ± 4.76 mm²). Within the gray matter, the bilateral VHA showed a 34.4% decrease in patients compared to healthy controls (left: p < 0.001, healthy controls 3.84 ± 0.29 mm², patients 2.56 ± 0.62 mm²; right: p < 0.001, healthy controls 3.95 ± 0.40 mm², patients 2.61 ± 0.58 mm²). In patients, the DHA was decreased bilaterally by 33.4% (left: p < 0.001, healthy controls 3.63 ± 0.41 mm², patients 2.42 ± 0.66 mm²; right: p < 0.001, healthy controls 3.60 ± 0.51 mm², patients 2.37 ± 0.66 mm²) (figure 1C), and smaller DHA was associated with smaller DCA (p < 0.001, R² = 0.74, 95% confidence interval [CI] 2.21–4.28).

Voxel-based analysis of the cervical cord revealed a 16.6% decrease in FA in the left dorsolateral funiculus (e.g., containing spinothalamic and lateral corticospinal tracts, p = 0.003; localization [x, y, z] −4.2, −18.5, 26; z score 4.42; cluster extent 154), 14.9% decrease in the right dorsolateral funiculus (p = 0.025; localization [x, y, z] 6, −18.5, 37; z score 4.34; cluster extent 85), and 17.0% decrease in the posterior funiculus (i.e., containing dorsal columns; p = 0.004; localization [x, y, z] 0.7, −22.3, 37; z score 3.80; cluster extent 145) in patients compared to healthy controls. AD was also decreased in patients compared to healthy controls in the same regions, namely by 12.8% in the left dorsolateral funiculus (p = 0.014; localization [x, y, z] −3.1, −19.2, 26; z score 3.72; cluster extent 58), 12.8% the right dorsolateral funiculus (p = 0.002; localization [x, y, z] 4.1, −18.8, 26; z score 4.70; cluster extent 94), and 9.9% in the posterior funiculus (p = 0.020; localization [x, y, z] 0.7, −19.2, 32; z score 3.69; cluster extent 52). RD increased by 31.8% in the dorsal column (p = 0.022; localization [x, y, z]: 0.3, −20.7, 37; z score 3.47; cluster extent 70) and by 34.0% in the left dorsolateral funiculus (p = 0.023; localization [x, y, z] −5, −19.2, 32; z score 3.22; cluster extent 69) in patients compared to healthy controls. MD was not significantly different between patients and healthy controls (figure 2).

Greater lesion area and length were associated with greater SCA decrease above the level of injury (lesion area: p = 0.048, R² = 0.25, 95% CI −3.64 to −0.23 1/mm³; lesion length: p = 0.006, R² = 0.42, 95% CI −0.55 to −0.11 1/mm²) independently of age. The width of total midsagittal tissue bridges was associated with less SCA decrease (p = 0.007, R² = 0.39, 95% CI 0.02–0.11 1/mm²). Greater lesion length was associated with smaller GMA (p = 0.012, R² = 0.40, 95% CI −3.74 to −0.57 1/mm²), VHA (p = 0.039, R² = 0.29, 95% CI −8.33 to −0.25 1/mm²), and DHA (p = 0.004, R² = 0.49, 95% CI −8.36 to −2.03 1/mm²), while midsagittal tissue bridges were positively associated with GMA (p = 0.035, R² = 0.28, 95% CI 0.33–0.781/mm²) and DHA (p = 0.011, R² = 0.38, 95% CI 0.26–1.74 1/mm²) (figure 3A). Greater lesion length and preserved midsagittal tissue bridges were associated with WMA (lesion length: p = 0.014, R² = 0.38, 95% CI −0.61 to −0.08 1/mm²; tissue bridges: p = 0.011, R² = 0.38, 95% CI 0.02–0.12 1/mm²) above the level of injury (figure 3B).

The width of total midsagittal tissue bridges was associated with DCA above the level of lesion (p = 0.019, R² = 0.29, 95% CI 0.25–2.39 1/mm²). Neither lesion size nor midsagittal tissue bridges were associated with microstructural changes above the level of lesion.

The size of the cross-sectional area of the dorsal columns identified those patients with bilateral recordable SSEPs of the dermatomes C6 and C8 (figure 4). This relationship was not evident for the WMA and CHEPs. Higher AD values within the dorsal column were associated with shorter SSEP N1P1 latency at the C4 dermatome (p = 0.0024, R² = 0.83, 95% CI: −0.00007 to −0.00001 10⁻³ × s²/mm²), corrected for age and lesion area. DTI metrics within the spinothalamic tracts were not associated with CHEPs recordings.

GMA was associated with motor score (p = 0.007, R² = 0.72, 95% CI 1.77–9.26) and pinprick score (p = 0.003, R² = 0.58, 95% CI 3.48–13.90); VHA area was associated with motor score (p = 0.001, R² = 0.78, 95% CI 6.74–21.93); and DHA was associated with pinprick score (p = 0.004, R² = 0.57, 95% CI 7.43–31.52) when corrected for lesion area and age (figure 5A).

To quantify tract-specific associations with appropriate clinical outcome, we used the extracted mean values of DTI indexes within the regions of interest (i.e., corticospinal tract, dorsal column, and spinothalamic tract). FA and RD within corticospinal tract and the dorsal columns were associated with SCIM score (corticospinal tract: FA: p = 0.002, R² = 0.80, 95% CI 105.82–361.55; RD: p = 0.001, R² = 0.83, 95% CI −169829.70 to −60547.48; dorsal columns: FA: p = 0.002, R² = 0.80, 95% CI 106.07–341.42; RD: p = 0.003, R² = 0.79, 95% CI −216944.1 to −60854.36) independently of lesion extent and age (figure 5B). AD within the dorsal columns was associated with GRASSP score independently of lesion extent and age (p = 0.30, R² = 0.60, 95% CI 40298.4–646999.7).