Preoperative Decreased Hounsfield Unit Values of Cervical Vertebrae and the Relative Cross-Sectional Area of Flexion/Extension Paraspinal Muscles Are Novel Risk Factors for the Loss of Cervical Lordosis after Open-Door Laminoplasty

Open-door laminoplasty is widely used for patients with cervical spondylotic myelopathy (CSM). However, the loss of cervical lordosis (LCL) seems to be unavoidable in the long-term follow-up after surgery, which may affect the clinical outcomes. The risk factors for this complication are still unclear. In this study, patients who underwent open-door laminoplasty between April 2016 and June 2021 were enrolled. Cervical X-rays were obtained to measure the C2–7 Cobb angle, C2–7 sagittal vertical axis (SVA), T1 slope (T1S) and ranges of motion (ROM). Cervical computed tomography (CT) scans and magnetic resonance imaging (MRI) were collected to evaluate the cervical Hounsfield unit values (HU) and the relative cross-sectional area (RCSA) of paraspinal muscles, respectively. A total of 42 patients were included and the average follow-up period was 24.9 months. Among the patients, 24 cases (57.1%) had a LCL of more than 5° at a 1-year follow-up and were labeled as members of the LCL group. The follow-up JOA scores were significantly lower in the LCL group (13.9 ± 0.6 vs. 14.4 ± 0.8, p = 0.021) and the mean JOA recovery rate was negatively correlated with LCL (r = −0.409, p = 0.007). In addition, LCL was positively correlated to the preoperative T1S, flexion ROM, flexion/extension ROM and the RCSA of flexion/extension muscles, while it was negatively correlated to extension ROM and the HU value of cervical vertebrae. Furthermore, multiple linear regression showed that preoperative T1S, mean HU value of cervical vertebrae, flexion/extension ROM and the flexion/extension RCSA were independent risk factors for LCL. Spine surgeons should consider these parameters before performing open-door laminoplasty.


Introduction
Cervical open-door laminoplasty has been widely used for patients with cervical spondylotic myelopathy (CSM) since its invention by Kiyoshi Hirabayashi in the 1970s [1,2]. Compared with cervical laminectomy, open-door laminoplasty can preserve the posterior structure and maintain cervical motion to some extent. However, the loss of cervical lordosis (LCL) or kyphotic deformity can still be observed in the long-term follow-up after surgery [3][4][5]. Lee et al. showed that postoperative LCL occurred in 80% of patients within a 1-year follow-up [6]. The cervical sagittal imbalance might affect the clinical outcomes. Xus et al. reported that the postoperative Japanese Orthopaedic Association scores (JOA scores) and JOA recovery rates were related to LCL after surgery [7]. It was also reported that patients with cervical kyphosis after laminoplasty had a worse prognosis [8]. 2 of 13 Previous studies have investigated risk factors of LCL. Regarding parameters of preoperative cervical alignment, high T1S has been proven to be a risk factor for LCL [9][10][11]. Other studies showed that cervical ranges of motion (ROM) were also related to LCL [12,13]. Several studies found that a smaller relative cross-sectional area (RCSA) of the deep extensor muscles may lead to LCL [10,14]. However, whether the flexor muscles and bone quality of the cervical vertebrae affected the LCL has rarely been investigated.
Therefore, the risk factors of LCL are still controversial and require more comprehensive investigation. The purpose of the present study is to investigate preoperative parameters including cervical alignment, cervical bone quality and paraspinal muscles for predicting LCL after open-door laminoplasty.

Patients
This study is a retrospective study. A total of 56 patients who underwent open-door laminoplasty in our institution from April 2016 to June 2021 were enrolled. The exclusion criteria were as follows: (1) diagnoses of tumors, fractures, infections or spine deformity; (2) previous history of spine surgery; (3) follow-up of less than 1 year; and (4) lack of radiographic imaging or poor image quality. Two cases were excluded for diagnoses cervical tumors and four cases for a revision surgery. In total, 50 cases met the inclusion criteria. During the follow-up, 3 cases were excluded for lack of image data or poor quality of radiographs and 5 cases did not return a visit for personal reasons. Finally, 42 cases completed the follow-up. All cases were treated by the same medical group and received homogenous treatment, and they were all required to avoid strenuous exercise and trauma after surgery, while no other changes to the patients' lifestyle were required. All patients received short-term NSAID painkillers to relieve neck pain from the incision after surgery, but none of them received any specialized physiotherapy after surgery and during follow-up.

Radiographic Assessment
All patients underwent standard cervical spine X-ray (lateral, flexion and extension radiographs) preoperatively and lateral radiographs postoperatively at the last follow-up. CT scans and MRIs of the cervical spine were also obtained within 1 month before surgery. In the first year, postoperative patients were required to return to visit at 3, 6, 9 and 12 months after surgery, every 6 months in the second year and once every year after that.
Cervical alignment parameters were measured as follows ( Figure 1): cervical lordosis (CL), represented by C2-7 Cobb angle, was defined as the angle formed by the parallel lines of inferior endplates of C2 and C7. LCL was defined as a decrease in CL greater than 5 • at the last follow-up [7]. C2-7 SVA was defined as the horizontal distance between the vertical line from the center of the C2 body and the posterior superior aspect of C7. T1S was defined as the angle between parallel lines of superior endplates of T1 and a horizontal line. The fROM was defined as the difference in C2-7 angles in the neutral and flexion positions while the eROM was defined as the difference in C2-7 angles in extension positions and neutral positions. The ROM was defined as the difference in C2-7 angles in extension positions and flexion positions. The ratio of fROM/eROM was calculated to represent the asymmetry of cervical mobility.
The HU values of cervical vertebrae were measured through the hospital picture and archiving system (PACS) by a standard method published by Schreiber et al. [15]. We performed a 3D reconstruction of the CT film and chose three slices on the axis plane: in the middle of the vertebral body, 2 mm below the superior endplate and 2 mm above the inferior endplate. Then, we drew an oval region of interest (ROI) of the medullary space as large as possible (excluding the cortical margins to avoid volume averaging) on each slice to calculate the HU value ( Figure 2). The HU value of each vertebral body was defined as the average HU value of three represented slices and the HU value of cervical spine was defined as the mean value of C2-C7. The HU values of cervical vertebrae were measured through the hospital picture a archiving system (PACS) by a standard method published by Schreiber et al. [15]. W performed a 3D reconstruction of the CT film and chose three slices on the axis plane: the middle of the vertebral body, 2 mm below the superior endplate and 2 mm above t inferior endplate. Then, we drew an oval region of interest (ROI) of the medullary spa as large as possible (excluding the cortical margins to avoid volume averaging) on ea slice to calculate the HU value ( Figure 2). The HU value of each vertebral body was d fined as the average HU value of three represented slices and the HU value of cervi spine was defined as the mean value of C2-C7.   The HU values of cervical vertebrae were measured through the hospital picture a archiving system (PACS) by a standard method published by Schreiber et al. [15]. W performed a 3D reconstruction of the CT film and chose three slices on the axis plane: the middle of the vertebral body, 2 mm below the superior endplate and 2 mm above t inferior endplate. Then, we drew an oval region of interest (ROI) of the medullary spa as large as possible (excluding the cortical margins to avoid volume averaging) on ea slice to calculate the HU value ( Figure 2). The HU value of each vertebral body was d fined as the average HU value of three represented slices and the HU value of cervi spine was defined as the mean value of C2-C7. ImageJ Software (version 1.52α, National Institutes of Health, Bethesda, MD, US was used to measure the cross-sectional area (CSA) of the paraspinal muscles. All d ImageJ Software (version 1.52α, National Institutes of Health, Bethesda, MD, USA) was used to measure the cross-sectional area (CSA) of the paraspinal muscles. All data were collected on the level of C5/6, because previous research indicated that the muscles on these levels had the strongest correlation to the cervical paraspinal muscles and also because the surgical segment in all cases contains C4-C6. As shown in Figure 3, paraspinal muscles could be divided into flexion muscles and extension muscles. Flexion muscles included longus colli and longus capitis (LCo + LCa) and sternocleidomastoid (SCM). Extension muscles included multifidus (Mult), semispinalis cervicis (SeCe), semispinalis capitis (SeCa), splenius cervicis and splenius capitis (SpCe + SpCa) and levator scapulae (LSc). The CSAs of the C5 vertebral body were also measured as a reference point. The CSAs of muscles were manually outlined around the outer fascia and the software could calculate the area automatically [16,17]. The ratio of paraspinal muscle/vertebral body was calculated as the relative cross-sectional area and the flexion muscles/extension muscles ratio's CSA was calculated to represent the asymmetry of cervical paraspinal muscles.
were collected on the level of C5/6, because previous research indicated that the muscles on these levels had the strongest correlation to the cervical paraspinal muscles and also because the surgical segment in all cases contains C4-C6. As shown in Figure 3, paraspinal muscles could be divided into flexion muscles and extension muscles. Flexion muscles included longus colli and longus capitis (LCo + LCa) and sternocleidomastoid (SCM). Extension muscles included multifidus (Mult), semispinalis cervicis (SeCe), semispinalis capitis (SeCa), splenius cervicis and splenius capitis (SpCe + SpCa) and levator scapulae (LSc). The CSAs of the C5 vertebral body were also measured as a reference point. The CSAs of muscles were manually outlined around the outer fascia and the software could calculate the area automatically [16,17]. The ratio of paraspinal muscle/vertebral body was calculated as the relative cross-sectional area and the flexion muscles/extension muscles ratio's CSA was calculated to represent the asymmetry of cervical paraspinal muscles. All radiographic parameters were measured by two independent experienced observers and the average values were taken into consideration.

Clinical Function Assessment
Japanese Orthopaedic Association scores (JOA scores) were used to assess the cervical neurological function preoperatively and at the last follow-up. This self-rating scale includes 4 aspects: upper-limb function, lower-limb function, sensory function and bladder function. The score ranges from 0 (worst) to 17 (normal). The JOA recovery rate was calculated by the following formula: (postoperative JOA score-preoperative JOA score)/(17 − preoperative JOA score) × 100% [18].

Statistical Analysis
We used an independent sample t-test and a chi-square test to compare variables between groups and a paired sample t-test for the preoperative radiographic parameters and in the last follow-up. Pearson's correlation analysis and Spearman correlation analysis were used to assess correlations between parameters and LCL. Multiple linear regression and the receiver operating characteristic curve (ROC) were used to investigate the risk factors for predicting LCL. Intra-and inter-rater reliability were assessed using the intraclass correlation coefficient (ICC) and ICC values greater than 0.8 was considered good. All data were analyzed via SPSS and a significant difference was defined as p < 0.05. All radiographic parameters were measured by two independent experienced observers and the average values were taken into consideration.

Clinical Function Assessment
Japanese Orthopaedic Association scores (JOA scores) were used to assess the cervical neurological function preoperatively and at the last follow-up. This self-rating scale includes 4 aspects: upper-limb function, lower-limb function, sensory function and bladder function. The score ranges from 0 (worst) to 17 (normal). The JOA recovery rate was calculated by the following formula: (postoperative JOA score-preoperative JOA score)/(17 − preoperative JOA score) × 100% [18].

Statistical Analysis
We used an independent sample t-test and a chi-square test to compare variables between groups and a paired sample t-test for the preoperative radiographic parameters and in the last follow-up. Pearson's correlation analysis and Spearman correlation analysis were used to assess correlations between parameters and LCL. Multiple linear regression and the receiver operating characteristic curve (ROC) were used to investigate the risk factors for predicting LCL. Intra-and inter-rater reliability were assessed using the intraclass correlation coefficient (ICC) and ICC values greater than 0.8 was considered good. All data were analyzed via SPSS and a significant difference was defined as p < 0.05.

Results
A total of 42 (28 male and 14 female) patients were included in the present study. All the demographic data and clinical outcomes are shown in Table 1. The average age was 56.8 ± 8.4 years and the average BMI was 25.7 ± 2.3 kg/m 2 . The mean follow-up period was 24.9 ± 21.7 months. For the operation levels, 26 cases received laminoplasty at level C3-6 and the other 16 cases involved laminoplasty at C2 or C7. Improvements in JOA scores were observed in all patients at the last follow-up with an average recovery rate of 37.8 ± 14.7%. The mean C2-7 Cobb angle was 18.0 ± 9.7 • preoperatively, and was 12.9 ± 10.8 • at the last follow-up, with an average decrease of 5.1 • . An increase in C2-7 SVA and decrease in T1S was also observed after surgery, which indicated that not only had the cervical lordosis decreased, but also that the whole cervical sagittal imbalance would appear after open-door laminoplasty. Other parameters of cervical alignment are also shown in Table 1. As presented in Tables 2 and 3, all parameters were compared between the LCL and NLCL groups. Baseline data including age, gender, BMI and whether the operation level involved C2/C7 or not were not significantly different between the two groups. Regarding the cervical alignment, preoperative C2-7 lordosis did not differ between the two groups while the LCL group had significantly smaller C2-7 Cobb angles at the last follow-up with a lower JOA score. For the preoperative parameters, higher T1S, larger fROM, smaller eROM and greater fROM/eROM were observed in the LCL group. As for the comparison of paraspinal muscles, the RCSA of LCo + Lca, total flexion muscles and flexor/extensor muscles were significantly larger in the LCL group. Additionally, the HU values of the C2-7 vertebral body respectively and the average values were significantly lower in the LCL group.  The results of correlation analysis between preoperative parameters and LCL are shown in Table 4. The JOA mean recovery rate was negatively correlated with the LCL (r = −0.430, p = 0.005). The preoperative T1S (r = 0.437, p = 0.004), fROM (r = 0.485, p = 0.001) and fROM/eROM (r = 0.545, p < 0.001) values were positively correlated with LCL while the eROM (r = −0.386, p < 0.011) showed a significant negative correlation. The average HU values of the C2-7 vertebral body showed a significant negative correlation with LCL (r = −0.352, p = 0.022). Regarding the RCSA of paraspinal muscles, the RCSA of LCo + Lca (r = 0.317, p = 0.041) and the RCSA of flexion/extension muscles (r = 0.421, p = 0.006) were positively correlated with LCL. Above all parameters, the fROM/eROM (r = 0.545) showed the strongest correlation with LCL.
Multiple linear regression showed that preoperative T1S, mean cervical HU value, fROM/eROM and flexion/extension RCSA were independent risk factors for LCL, as shown in Table 5. The ROC curve showed that the corresponding AUC values were 0.719, 0.801, 0.794 and 0.787, respectively, as shown in Figure 4. Multiple linear regression showed that preoperative T1S, mean cervical HU valu fROM/eROM and flexion/extension RCSA were independent risk factors for LCL, shown in Table 5. The ROC curve showed that the corresponding AUC values were 0.71 0.801, 0.794 and 0.787, respectively, as shown in Figure 4.

Discussion
Open-door laminoplasty is a common procedure for multiple-level cervical spondylotic myelopathy. Indirect decompression is achieved with a posterior shift of the spinal cord at the expense of the spinal canal. However, the loss of cervical lordosis (LCL) seems to be inevitable after surgery, which may lead to decompression failure and poor clinical outcomes [5][6][7][8]. A decrease in C2-7 Cobb angle was observed in 32 (76.2%) patients in the 1-year follow-up and, among them, 24 cases (57.1%) had a loss of cervical lordosis of more than 5 • . The mean C2-7 Cobb angle was reduced from 18.0 • preoperatively to 12.9 • .
Several risk factors for LCL after laminoplasty have been investigated in previous studies. It was reported by Zhang [9] and Kim [10] that a higher preoperative T1S was an independent risk factor of LCL. Lee et al. found that patients with preoperative TIS > 29 • faced a higher risk for a decrease in lordosis of more than 5 • [19]. In the present study, patients in the LCL group had a higher preoperative T1S and this was moderately related to a decrease in C2-7 Cobb angle. Multiple linear regression showed that it was an independent risk factor for LCL. Patients with high T1S may also have large cervical lordosis as compensation to maintain a horizontal gaze and a balanced sagittal alignment, which requires extra work provided by the cervical paraspinal muscles. The primary degeneration or asymmetry of paraspinal muscles or damage from operation might lead to weakness of paraspinal muscles after surgery, resulting in LCL.
The roles of cervical range of motion and extension muscles in the occurrence of postoperative LCL were also recently noticed. Previous studies focused more on the role of extension range of motion (eROM). A study by Lee showed that a smaller extension range of motion (eROM) was an independent predictor of LCL and the cutoff value was 14 • [12] and another study demonstrated that eROM was related to LCL after laminoplasty [20]. However, recent studies have suggested that the flexion range of motion (fROM) should not be ignored. The fROM could be used to distinguish the LCL group from the NLCL group and a greater fROM was related to LCL [13]. A recent study by Masashi showed that a composite indicator, which was the gap between fROM and eROM, was a highly useful indicator for LCL [21]. Smaller fROM or ROM may be caused by degenerative structures including ossification of the posterior longitudinal ligament, osteophytes or degenerative intervertebral disc, which limits the cervical alignment change from lordosis to kyphosis. Conversely, patients with larger fROM or eROM may have more potential to change cervical lordosis. In the present study, the fROM, eROM and the ratio of fROM/eROM, which represents the asymmetry of cervical mobility, showed correction with LCL. Additionally, fROM/eROM was investigated to be an independent risk factor for LCL by multiple linear regression analysis.
The paraspinal muscles play an important role in maintaining cervical lordosis and balance [22,23]. The semispinalis cervicis was found to be the most important muscle in maintaining cervical lordosis [24], and preoperative CSA of the semispinalis cervicis was related to LCL [14]. In addition, the CSA of deep extensor muscles was smaller in the LCL group according to a study conducted by Kim [10]. A recent study found that the asymmetry of cervical extension muscles at the C6 level was associated with cervical sagittal imbalance after laminoplasty [25]. However, little attention has been paid to the anterior flexion muscles. We hypothesized that the cervical flexion muscles also played an important role in maintaining spinal alignment balance and that the asymmetry of paraspinal muscles may be related to LCL. In the present study, the RCSA of longus colli and longus capitis and the RCSA of total flexion muscles were significantly larger in the LCL group. Correlation analysis showed that the flexion/extension muscles ratio's RCSA was positively related to LCL. Additionally, multiple regression analysis confirmed that it was a risk factor for LCL. As mentioned above, extension muscles provide the main work in maintaining cervical lordosis. The damage from surgery caused degeneration and weakness of the flexion muscles, which may further aggravate the imbalance of the flexion and extension muscles, resulting in LCL. A previous retrospective study mentioned that for patients who underwent expansive laminoplasty, cases in a postoperative No Pain Group (VAS < 3) had stronger neck muscle strength and higher muscle strength recovery rate at follow-up compared to a postoperative Pain Group (VAS ≥ 3) [26]. Proper neck muscle exercises may help to maintain cervical alignment and relieve neck pain.
Whether the cervical bone quality was related to LCL after open-door laminoplasty has scarcely been investigated. Previous studies have shown that the HU values of the vertebral body in CT scans was reliable for assessing bone mineral density (BMD) [15,27]. In the present study, the HU values of each cervical of C2-7 and the average value were all significantly lower in the LCL group. In addition, the average cervical HU value showed negative correlation with the LCL. Multiple linear regression showed that the HU value of C2-7 was another independent risk factor for LCL. Several studies demonstrated the relationship between cervical sagittal alignment and cervical HU values. Lovecchio et al. showed that the cervical HU values were positively correlated with cervical kyphosis [28]. Another study investigated whether vertebral wedging was more likely to occur in patients with osteoporosis and aging people [29]. Loss of horizontal trabeculae in the vertebrae increases the risk of minor anterior vertebral body fracture and wedge deformity [30]. Additionally, a negative effect on BMD might occur in the early phase after cervical or lumbar surgery [31,32]. Thus, patients with low cervical HU values may have higher risk of vertebral wedge change, which results in unstable alignment and a decrease in cervical lordosis.
This study had some limitations. Firstly, it is a retrospective study from a single center. A prospective study with a larger sample size is needed to provide a higher level of evidence. Secondly, only a few cases obtained a whole-spine X-ray so we could not investigate the effect from thoracic and lumbar vertebrae on cervical alignments. Thirdly, because all of the radiographic parameters, the ROI of HU and RCSA of muscles were measured manually, it was difficult to recognize the boundary of vertebral bodies and muscles in some cases, which may cause errors.

Conclusions
The loss of cervical lordosis is a common complication after open-door laminoplasty and results in poor prognosis. Preoperative T1S, C2-7 HU values and the asymmetry of fROM/eROM and flexion/extension paraspinal muscles were risk factors for LCL. Spine physicians should evaluate these parameters before performing open-door laminoplasty and pay attention to pre-postoperative cervical muscle-strengthening exercises.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Registration ID: SYSKY-2022-184-01).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The datasets generated and analyzed during the current study are not publicly available due to the fact that they constitute an excerpt of research in progress but are available from the corresponding author on reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.