Introduction

Degenerative cervical myelopathy (CM) is a leading cause of acquired spinal cord (SC) compression [21]. Based on epidemiological data, 90% of population in sixth decade of life or older have evidence of degenerative changes in the cervical spine [30]. Although a mechanical compression seems to play an important role in the pathogenesis of CM, other causes such as dynamic factors have been suggested. The pathophysiology of CM is based on the repeated traumatic injuries to SC being caused by both static and dynamic mechanical factors [16]. A dynamic compression of SC can occur in flexion (F) or extension (E) and may be missed on standard imaging and radiography [33]. Therefore, an application of dynamic magnetic resonance imaging (dnMRI) in F and E positions provides additional information on dynamic cord compression that may not be visible on neutral (N) MRI [31]. Differences in the SC diameter and cervical cord lengths in different positions have been shown previously [15, 19]. Recent studies report that the cervical canal is widened by 10–15% on flexion but narrowed by 10–25% on extension during dnMRI [4, 6].

A dynamic model of cervical myelopathy with stretch-associated injury is becoming widely accepted as the principal etiological factor of CM [8]. Reid et al. found that at the cervicothoracic junction, SC elongates up to 24% of its length by nearly 18 mm during neck flexion. Additionally, when the ventral-dorsal diameter of SC is reduced by 20 to 30%, the axial tension forces exceed the compliance of the spinal cord fibres [26]. This data shows that SC is physiologically subjected to significant biomechanical forces in the extreme positions and becomes particularly susceptible to injury in the presence of coexisting cervical stenosis.

This study aims to compare advanced morphological measurements of the cervical canal and spinal cord in patients with cervical stenosis on dynamic and static MRI. Additionally, a separate analysis was performed for predictors of the severity of myelopathy using morphometric parameters.

Methods

Patients with cervical canal stenosis who had MRI (Philips 1.5 T scanner) in neutral, flexion and extension positions in 2016/2017 at Royal Victoria Infirmary, Newcastle upon Tyne, UK, were retrospectively reviewed. The measurement technique is presented in Fig. 1. A dynamic imaging was performed in F and E positions depending on the patient’s comfort with stabilizing sponges. Standard T1- and T2-weighted sequences in the sagittal plane and T2-weighted sequences alone in the axial plane in all three positions were obtained. All enrolled patients were scanned with dnMRI as a part of standard protocol due to symptomatic cervical canal stenosis and presented with radiculopathy, myelopathy or myeloradiculopathy. A sagittal spinal canal diameter of less than 12 mm on T2-weighted MRI was defined as cervical stenosis. The diagnosis of cervical stenosis was made by a multidisciplinary team consisted of a spinal surgeon and neuroradiologist. Patients with a congenital deformity in the cervical spine or cranio-cervical junction, any form of spondyloarthritis including ankylosing spondylitis, diffuse idiopathic skeletal hyperostosis (DISH), ossification of the posterior longitudinal ligament (OPLL), Chiari malformation and history of cancer were excluded from the analysis.

Fig. 1
figure 1

Measurements technique of the morphometric parameters on neutral, sagittal T2-weighted magnetic resonance image. aLCC anterior length of the cervical cord, pLCC posterior length of the spinal cord. Cervical lordosis Cobb angle (a) and cervical cord angle (b)

Full size image

A four-stage classification of Muhle et al. [18] was used to score the degree of cervical stenosis on axial and sagittal 2 T-weighted MRI (Table 1). The severity of myelopathy was evaluated using Nurick classification system for myelopathy on the basis of gait abnormalities [22], grade 0 was used for patients with radiculopathy alone and grades 1–5 for those with signs of myelopathy.

Table 1 Muhle classification
Full size table

A morphometric analysis was performed in three positions: neutral, flexion and extension. The tightest area of the spinal canal was defined as compression level on the neutral MRI, normal canal appearance above at C3 and below at C7 was used as reference level. The mid-cord distance (L value) was measured by the method previously described by Miyata et al. [17]. The length of the cervical cord (LCC) was defined as the distance between a line crossing the cord at the upper edge of the anterior (aLCC) and posterior (pLCC) arches of C1 and the line along the lower endplate of C7 [15]. The cervical cord angle (CCA) was measured as a longitudinal cervical spinal cord angle on sagittal MRI, and a negative value was assigned to kyphotic and positive to lordotic SC positions. The difference between F and E CCA was calculated with the following formula:

$$ {dif}_{F-E} CCA=\pm \left({180}^{{}^{\circ}}-\left|{CCA}_F\right|\right)\hbox{--} \left[\pm \left({180}^{{}^{\circ}}-\left|{CCA}_E\right|\right)\right]. $$

Absolute values of CCA in F//E were used to calculate the difference from straight angle with ‘minus’ for kyphosis and ‘plus’ for lordosis.

Cervical lordosis was measured as a sagittal Cobb angle of the lower endplate of C2 and C7 (C lordosis) with a negative value for kyphotic and positive for lordotic spine curvature. A correlation between the difference in the angulation of spine and cervical cord (S/C angle ratio) in F and E was calculated using the following formula:

$$ S/C\ angle\ ratio={dif}_{F-E}C\ lordosis/{dif}_{F-E} CCA. $$

Other morphometric parameters were measured at compression level on axial T2-weighted MRI in all three positions: spinal cord (SC) area, cerebrospinal fluid (CSF) area and CSF reserve ratio (CSF/CSF + SC).

All the measurements were performed using INFINITT (INFINITT Healthcare Co. Ltd., Seoul, South Korea).

Statistical analysis

All the results were statistically analysed using the computer software Statistica 9 (TIBCO Software Inc., Paulo Alto, California). Statistical significance was assumed for p < 0.05. Normal distribution of variables was checked with the Shapiro-Wilk normality test. Repeated measures analysis of variances (ANOVA) with post hoc Scheffé test was used to compare the mean results on neutral, flexion and extension MRI at compression level. Univariate multiple regression analysis for Nurick grade as dependent factor was used. All morphometric parameters as well as the differences between them in F and E positions were analysed as independent factors. A comparison of Muhle grade between F, E and N positions was performed with a chi-square test. All results were presented as mean ± standard deviation (SD).

Results

Sixty-three patients and 34 men, with the mean age of 58.2 ± 11 years, were analysed. The distribution of cases in the Nurick classification was as follows: grade 0–8 (13%), grade 1–14 (22%), grade 2–18 (29%), grade 3–8 (13%), grade 4–9 (14%) and grade 5–6 (10%). The compression level was found at C5/6 in 28 (44%) patients, C6–7 in 19 (36%) and in 16 (25.4%) cases at C3/4 and C4/5. The summary of the mean values of morphometric parameters in three positions is presented in Table 2.

Table 2 Comparison of means for morphometric parameters in neutral, flexion and extension positions
Full size table

Significant differences were found for pLCC, SC area, C lordosis angle and CCA. pLCC was 112.7 ± 10.3 mm in F, 110.3 ± 10.6 mm in N and 102.5 ± 10.7 in E. The differences in all positions were significant (F vs. N, p = 0.031; F vs. E, p < 0.001; N vs. E, p = 0.028). A prominent difference was noted between SC area in E and N (59.9 ± 14.3 vs.71.2 ± 20.1 mm2, p < 0.001, respectively) and F and E with the difference of 8.2 ± 4.8 mm2, p = 0.001. There was no difference for SC area between N and E positions. The mean C lordosis angle was significantly different between three positions and was − 17.1° ± 11.8 (kyphosis) in F, 6.1° ± 9.3 in N and 25.7° ± 11.7 in E. Similarly, the results of CCA showed significant difference in static and dynamic positions with − 169.91° ± 8.54 (kyphosis), 163.54° ± 9.55 and 159.67° ± 12.73 in F, N and E, respectively. The difference between F and E for C lordosis angle was 42.80° ± 14.4 and for CCA 30.42° ± 9.6. The mean S/C angle ratio was calculated for 1.4 ± 1.3.

There was no significant difference between L value at compression level between F (6.90 ± 1.02 mm) and E (6.28 ± 1.19 mm). Similar comparison of L value above the compression at C3 reference level showed a significant difference between F and E (8.5 ± 3.34 vs.7.10 ± 1.2 mm; p = 0.036). At the reference level below compression at C7, the mean L value was similar in F and E (7.2 ± 1.19 vs. 7.24 ± 1.36 mm, respectively; p = 0.96) (Fig. 2).

Fig. 2
figure 2

A chart bar presenting measurements of L value in flexion and extension at compression, C3 and C7 levels. L value indicates mid-cord distance and asterisk indicates difference statistically significant

Full size image

The distribution of Muhle grades in different scanning position revealed discrepancy between N and E (chi-square test; 15.2; p = 0.016) (Table 3). The difference was especially marked at grade 0 where only 4 (6%) patients were assigned in E, but 11 (17%) in N. On the contrary, only 8 (13%) cases were classified to grade 3 in N, but 13 (21%) in E. There was no difference in Muhle grades between F and N positions.

Table 3 Distribution of cases in Muhle grade depending on the scanning position
Full size table

A univariate regression analysis was used to explain the relationship between the severity of myelopathy in Nurick grade and cervical morphometric parameters. Thirty-eight independent factors were tested in the model including measurements in three positions as well as the differences between F and E (Table 4, Fig. 3). Age and S/C angle ratio positively correlated with Nurick classification. An increase by one grade in Nurick classification correlated with an increment in age by 2.4 years above 56.9 (p = 0.041) and in S/C angle ratio by 0.06 above 1.27 (p = 0.011). A significant negative correlation between Nurick grade and SC area (p = 0.006) in E and F-E difference for L value (0.004) was also found. A decrease of both SC area in E by 2.2 mm2 below 60.6 and dif L value by 0.04 mm below 0.7 correlates with an increase of Nurick grade by one.

Table 4 Univariate regression analysis predicting Nurick grade
Full size table
Fig. 3
figure 3

Scatter diagram and regression line for the correlation between Nurick grades and difference of L value in flexion-extension (a), spine/cord angle ratio (b), age (c) and spinal cord area in extension (d). Regression line (solid line) and 95% confidence interval (dashed line). SC spinal cord, E extension, F flexion, N neutral

Full size image

Discussion

The biomechanical theory with a dynamic model of cervical myelopathy appears to more fully address the clinical course and pathophysiology leading to the spinal cord injury in patients with or without static cord compression [8]. Stretch-associated injury is considered as the main etiological factor of cervical myelopathy, and significant tissue strain in the presence of cervical stenosis can exceed the compliance of SC inducing axonal and myelin injury [9, 13, 19, 25]. Physiologically, in the absence of cervical stenosis, the natural elongation of the cord in flexion/extension is distributed over the entire length of SC. However, in segmental tethering of SC with reduced cord mobility due to compression, the axial strain cannot be distributed throughout the cord leading to the accumulation of distracting forces and axonal injury.

Biomechanical studies clearly show that any motion at the cervical spine causes subsequent contribution of coordinated motion of the cord [27]. The spinal cord follows the changes in length and angulation by the physiological motion of the spinal canal. The mechanism was reported that SC folds during extension and unfolds during flexion [5, 15].

This study provides objective analysis of radiological changes in the stenotic cervical spine in static and dynamic positions. At first stage, all the measurements were compared in three positions, and then a model with 38 independent factors including also differences in extreme positions (flexion-extension) was built to look for a possible correlation with the severity of myelopathy in the Nurick classification. This model incorporated also parameters describing cervical spine and cord angulations in different positions. The comparison of means for all parameters showed significant differences for pLCC, SC area, C lordosis angle and CC angle. The posterior length of the cervical cord in F was the longest and the shortest in E. There was no prominent difference in aLCC between all positions. The elongation of the pLCC (flexion-extension) was 10.2 mm (9.9%) and 6.3 mm (5.9%) for aLCC. Kuwazawa et al. [15] also found smaller elongation of aLCC (9.3 mm; 8.4%) than pLCC (14.4 mm; 13.9%). The same authors found posture-dependent differences of the length of the cervical cord in the recumbent and erect series. In the study of Breig on human cadavers, the mean elongation of the cervical cord during F and E was 8–10 mm [3]. In the study of Koschorek et al. [14] performed on flexion/extension MRI, the average lengthening of the cervical cord was 12 mm. One explanation for significant elongation of the pLCC is greater change of the CC angle in F than in E from N position, which primarily affects pLCC. In our series, the flexion-neutral CC angle difference was 26.5° and extension-neutral only 3.8°. This data correlates with the meaningful lengthening of the pLCC occurring typically in F. Elongation and narrowing of SC in F result in the reduction of SC area in this position, whereas in E cord folds and thickens. The reflection of this can be found in the measurement of SC area, which was significantly smaller in F than in N and E by 23 and 18% retrospectively. Zhang et al. [33] also found that cervical cord sagittal diameter was smaller in F at C3-T1 levels than in N and E. This data shows that SC is physiologically subjected to significant biomechanical forces in the extreme positions and becomes particularly susceptible to injury in the presence of coexisting cervical stenosis. This is especially pronounced when the ventral-dorsal diameter of SC is reduced by 20 to 30% and axial tension forces exceeded the compliance of the spinal fibres [2, 26].

The present study showed that the distance between the vertebral body and the spinal cord (L value) was similar in all positions indicating no substantial cord movement at the tightest level. Interestingly, a comparison of L value in extreme positions (flexion-extension) at compression level, above at C3 and below at C7 showed a significant difference only at C3 of 20% (8.5 vs. 7.10 mm; p < 0.05), but neither at compression level nor at C7. Miura et al. reported that L value in myelopathic patients was significantly smaller than in control subjects in both N and maximum F/E positions. They also found that an average cord distances at C4/5 and C5/6 were significantly smaller than those at C2/3 and C7/Th1 in all positions. This data confirms the concept of segmental cervical cord tethering in myelopathy with the limitation of the normal cord displacement during spine movements. Pathological forces within the cervical cord at stenotic levels might result in the transmission of the longitudinal stretching force and shear injury to myelin and neural elements even remotely to the compression segments [13] with symptoms arising distally from the point of the cord tethering [24, 28].

Dynamic imaging of the cervical spine seems to be more sensitive in detecting spinal cord compression. Data form this study showed that 50% of patients were classified to high grades (grade 2 and 3) in Muhle scale in E, but only 35% in neutral. Grade 0 was three times more often in N than in E indicating that neutral MRI may underestimate the degree of cervical cord compression. Zeitoun et al. [32] reported that 22.5% of cases in grade 3 on extension MRI were classified to grade 1 in neutral position. According to Zhang et al., the ratio of grade 3 to grade 1 at compression levels doubled with dnMRI comparing to neutral MRI [33]. The difference between E and static MRI was most significant between C3 and C6 levels, especially among those older than 65 years of age [7]. Higher rate of cervical stenosis revealed on extension MRI may be associated with shortening of the subarachnoid space, shortening and thickening of the spinal cord, and folding of the ligamentum flavum increasing of number of spinal cord impingements that is invisible on neutral MRI.

A univariate regression showed that higher Nurick grade positively correlates with age and S/C angle ratio. Age was postulated as predictor of cervical stenosis in previous studies [12, 23]. Interestingly, a contrary finding was presented by Wang et al. [30], who reported an increasing incidence of cervical spondylosis with aging before age 50 years, and decreasing with aging after age 50 years, especially in the elderly after 60 years. They explain that the volume and inflammation of the nucleus get lesser since chronic degeneration contributes to atrophy of the nucleus with the aging process [1, 29]. During aging, the inflammatory effect of the nucleus is weaker than the degeneration of the nucleus after 60 years of age, and the incidence of cervical spondylosis will decrease with aging.

An interesting finding from this study is the relationship between Nurick grade and S/C angle ratio. As the cervical cord follows the motion of cervical spine, a proportional change of the cervical cord angle is expected in F and E. However, there was a mismatch between the cord and spine angulation change in the extreme positions (flexion-extension). The S/C angle ratio was 1.27 for grade 1 and proportionally increased up to 1.52 in grade 5. A lesser increment of the cord angle in F and E positions may cause a ‘hammer effect’ by the cervical spine resulting in repetitive microtrauma of the cord. Mullin et al. [20] presented a concept of the sagittal bowstring effect when an ‘effective kyphosis’ results in spinal cord tethering over the kyphosis in the sagittal plane during flexion. This is also consistent with another finding in this analysis showing a negative correlation between Nurick grade and difference in L value between F and E positions. The mean difference in L value was only 0.4 mm for grade 4 and 0.44 mm for grade 5. In other words, reduced mobility of the cervical cord at compression level affects the severity of myelopathy.

The predictive value of SC area in the cervical myelopathy has been reported in the literature [10, 11]. However, previous studies focused mainly on the static measurements and did not analyse dynamic values of SC area in the predictive models. In our analysis, a correlation between SC area and myelopathy was found, but only is extension. Patients with myelopathy had SC area of 60 mm2 or smaller with 58.1 mm2 in grade 4 and 49.6 mm2 in grade 5. The natural tendency of the cord to fold and thicken in E is distorted in the presence of stenosis, which results in the reduction of SC area and potential SC injury [4, 6, 15].

The main limitation of the present study is its retrospective nature and the limited sample size. The criteria for dynamic scanning with MRI were not unified, and not all the patients with myelopathy went through dnMRI. Because our sample size was limited, thereby preventing us from arriving at more conclusive results, further research could investigate a larger number of patients. The distribution of patients in Nurick grades was not balanced and could potentially affect the power of statistical results. Nurick grade was devised primarily to assess ambulatory status and does not focus on the upper limbs function, which may be the first manifestation of the myelopathy.

Patients did not have a routine brain MRI to rule out other potential causes (white matter lesions, hydrocephalus or brain atrophy) contributing to gait impairment regardless of cord compression.

Conclusions

Cervical myelopathy is a complex pathology, and recent data indicates that dynamic compression with increased strain and shear stress in the cervical cord is responsible for clinical symptoms. This study, focusing on anatomical parameters, elucidates a relationship between dynamic morphometric parameters and severity of myelopathy in Nurick classification. Reduced mobility of the cervical cord in flexion-extension positions was found to be a predictor of severe myelopathy. This is consistent with a tethered cord effect in the cervical stenosis. Another interesting finding was spine/cervical cord angle ratio mismatch correlating with higher Nurick grade. A biomechanical explanation of this phenomenon is that different ranges of angulation between the spine and cervical cord in the dynamic positions may cause repetitive microtrauma of the spinal cord.

Especially extension MRI is useful to determine more accurately the number of levels and severity where the spinal cord is compromised. Further clinical studies that are more systematic in design, combining functional imaging, electrophysiology and ds MRI, are warranted.