Surgical results and prognostic factors following percutaneous full endoscopic posterior decompression for thoracic myelopathy caused by ossification of the ligamentum flavum

Minimally invasive surgery (MIS) has shown satisfactory surgical results for the treatment of thoracic myelopathy (TM) caused by ossification of the ligamentum flavum (OLF). This study investigated the prognostic factors following MIS and was based on the retrospective analysis of OLF patients who underwent percutaneous full endoscopic posterior decompression (PEPD). Thirty single-segment OLF patients with an average age of 60.4 years were treated with PEPD under local anaesthesia. Clinical data were collected from the medical and operative records. The surgical results were assessed by the recovery rate (RR) calculated from the modified Japanese Orthopaedic Association (mJOA) score. Correlations between the RR and various factors were analysed. Patients’ neurological status improved from a preoperative mJOA score of 6.0 ± 1.3 to a postoperative mJOA score of 8.5 ± 2.0 (P < 0.001) at an average follow-up of 21.3 months. The average RR was 53.8%. Dural tears in two patients (6.7%, 2/30) were the only observed complications. Multiple linear regression analysis showed that a longer duration of preoperative symptoms and the presence of a high intramedullary signal on T2-weighted MRI (T2HIS) were significantly associated with poor surgical results. PEPD is feasible for the treatment of TM patients with a particular type of OLF. Patients without T2HIS could achieve a good recovery if they received PEPD early.

To the best of our knowledge, all the published studies on prognostic factors in OLF patients have been based on traditional open surgery. However, MIS has been advantageous in treating OLF with satisfactory surgical results 12,[15][16][17] . This study investigated the prognostic factors following MIS and was based on the retrospective analysis of OLF patients who underwent percutaneous full endoscopic posterior decompression (PEPD).

Materials and Methods
The Ethics Committee of Tangdu Hospital approved the study, and all patients signed informed consent. The methods described were performed in accordance with relevant guidelines and regulations. This was not a study commissioned or funded by any manufacturer. Patient population. Between April 2016 and January 2018, thirty consecutive patients with TM caused by single-segment OLF underwent PEPD under local anaesthesia using the endoscopic system (iLESSYS ® , Joimax ® GmbH, Karlsruhe, Germany). The indications for surgery were progressive muscle weakness in the lower extremities, gait disturbance, dorsal pain, and sphincter dysfunction. Neurological examinations and preoperative MRI/ CT were performed to determine the location of OLF and the target area for decompression. OLF was classified into lateral, extended, enlarged, fusion, and tuberous types based on axial CT as well as round and beak types based on sagittal MRI 19,20 . Fusion and tuberous types of OLF were excluded based on the surgical difficulties in endoscopic decompression 15 . In addition, patients with ventral compression (thoracic disc herniation, ossification of the posterior longitudinal ligament), concomitant cervical or lumbar lesions and severe cardiopulmonary disease were also excluded. The clinical characteristics were recorded, including sex, age, the duration of preoperative symptoms, a history of smoking and comorbidities (diabetes mellitus or hypertension). The presence of a high intramedullary signal on T2-weighted MRI (T2HIS) was recorded. The location of the lesions was classified as upper (T1-T4), middle (T5-T8), and lower (T9-T12) on the basis of the thoracic spine level. The presence of intraoperative dural adhesion/dural ossification (DA/DO), operation time and estimated blood loss (EBL) were also recorded.
Clinical and radiologic assessments. All patients were followed for at least one year. Pre-and postoperative neurological statuses were evaluated using the modified Japanese Orthopaedic Association (mJOA) score ( Table 1). The recovery rate (RR) = (postoperative mJOA-preoperative mJOA)/(11-preoperative mJOA) × 100% 21 . According to the RR, the surgical results were divided into good (50-100%), fair (25-49%), unchanged (0-24%), or deteriorated (<0%) 22 . The cross-section area (CSA) and the anteroposterior diameter (APD) of the spinal canal (Fig. 1) were measured on axial CT and sagittal T2-weighted MRI using the Picture Archiving and Communication Systems (PACS). The CSA was measured as follows 23 : The widest distance between two pedicles as viewed on a CT scan was measured as the transverse spinal canal diameter (Fig. 1a), equal to the transverse spinal canal diameter at the maximally compressed CT scan. A vertical line extending through the endpoints of the transverse diameter determined the boundary of the spinal canal and was used to measure the compressed CSA (Fig. 1b). The normal CSA was measured on the pedicle section of the same vertebrae (Fig. 1c). The ratio of CSA = (the compressed CSA)/(the normal CSA) × 100%. The APD was measured at the compressed level and at two normal levels above and below the compressed level (Fig. 1d) 6 . The average value of the APD just above and below the affected segment was the normal APD. The ratio of APD = (the compressed APD)/(the normal APD) × 100%. www.nature.com/scientificreports www.nature.com/scientificreports/ Surgical techniques. The patient was carefully arranged in the prone position. According to anaesthesiologists, dexmedetomidine (0.2-0.7 μg/kg/min) and sufentanil (0.1 μg/kg) were used to alleviate pain and maintain the sober situation. Patient feedback could be received promptly during the operation. The entry point in the skin was 5-6 cm from the midline. After the infiltration of local anaesthetics (0.5% lidocaine), an 18-gauge spinal needle was introduced under fluoroscopic guidance to the lamina on the root of the spinous process. Then, an approximately 10-mm skin incision was made, and the dilation catheter was inserted in sequence. A specially designed 8.5-mm diameter bevelled working cannula was placed, and a specially designed 7.5-mm diameter circular saw was placed through the cannula (Fig. 2a-d). Finally, the endoscope was placed through a circular saw, and laminotomy was achieved under the view of the endoscope (Fig. 2e). The unossified ligamentum flavum (LF) and soft tissue were removed by forceps and radiofrequency. Diamond abrasors were used to grind the contralateral ossified LF into a thin and translucent shape (Fig. 2f), and then Endo-Kerrison punches were used to remove the remnant ossified LF (Fig. 2g). The ipsilateral lesion was treated in the same manner. The whole procedure utilized the technique of "over-the-top" decompression 15 . Finally, the dural sac was exposed, and pulsation of the dural sac improved (Fig. 2h). Notably, tight DA/DO should be maintained but completely isolated from the surrounding LF to avoid a dural tear 15 . After complete decompression, the cannula was removed, and the incision was closed without the use of a suction drain.

Neurological status Score
Statistical analysis. The data were analysed using Statistical Package for the Social Sciences (SPSS) software (version 18; IBM Corp., Armonk, NY). Student's t-tests and one-way analysis of variance were used to compare the statistical significance of the association for continuous data. Pearson's rank correlation coefficients were used to test the correlations between various factors and the RR. Multiple linear regression analysis was conducted to determine the quantitative variables that best correlated with the surgical results. A P value less than 0.05 was considered statistically significant. Qualitative data were converted to numbers for quantitation as follows 24   Radiographical findings. There were 24 lesions (80%) located at the lower thoracic spine, 4 lesions (13.3%) at the middle, and 2 lesions (6.7%) at the upper. According to axial CT images, 3 cases (10.0%) were classified as lateral type, 3 (10.0%) cases were extended type, and 24 (80.0%) cases were enlarged type. According to the sagittal MRI images, 20 cases (66.7%) were classified as round type, and 10 (33.3%) were classified as beak type. T2HIS was observed in 16 cases (53.3%). The mean APD and CSA of the normal thoracic canal were 11.6 (7.7-17.4) mm and 167.0 (110.1-296.0) mm 2 , and the mean APD and CSA of the compressed level were 5.4 (3.3-8.9) mm and 109.1 (66. 3-195.9) mm 2 . The mean ratios of the APD and CSA were 47.3% and 66.2%, respectively. According to the pre-and postoperative CT and MR images, decompression was completed successfully by PEPD with a dome-shaped laminotomy through limited laminectomy and flavectomy (Fig. 3).
Surgical outcome. In general, patients' neurological status improved from a preoperative mJOA score of 6.0 ± 1.3 to a postoperative mJOA score of 8.5 ± 2.0 points (P < 0.001) at an average follow-up time of 21.3 months, yielding an average RR of 53.8%. According to the RR, 16 (53.3%) cases were classified as good, 7 (23.3%) cases were fair, 7 (23.3%) cases were unchanged, and 0 cases were deteriorated. The mean operation time was 167.0 (100-240) minutes. The mean EBL was 36.2 (25.0-50.0) ml. During PEPD, we found DA/DO in 11 (36.7%) patients, in whom 2 patients (cases 8 and 18) experienced dural tears, yielding an overall incidence of 6.7% (2/30). However, the two patients did not receive repair or indwelled drainage. They healed after staying in a prone position for one week with a pressure dressing. Neither cerebrospinal fluid cysts nor incision dehiscence occurred during their follow-up. No neurological deficits or other complications occurred in this study.

Relationship of the RR to various factors.
Univariate analysis revealed that a poor RR was significantly related to a longer duration of symptoms, a lower preoperative mJOA score, a lower ratio of APD, a lower ratio of CSA, the presence of DM, the presence of T2HIS, and the presence of intraoperative DA/DO (Table 2).  (Table 3).

Discussion
Percutaneous endoscopic surgery, which is the most minimally invasive spinal surgery, can achieve decompression of spinal stenosis based on improvements in equipment and optical technology 25 . To the best of our knowledge, this is the first study about the prognostic factors of OLF treated by MIS and the largest study about the surgical results of PEPD for the treatment of OLF. We found that PEPD was feasible for the treatment of thoracic OLF, and OLF patients without T2HIS could achieve a good recovery if they received PEPD early. OLF mostly occurs in the lower thoracic spine and mostly affects male adults aged 40 to 60 years 7,24,26-28 . We found similar results, and our data showed that 56.7% (17/30) of the patients were male, 53.3% (16/30) were less than 60 years of age, and 80% (24/30) had decompression at the lower thoracic spine. In this study, patients' neurological status improved significantly after PEPD, and the average RR was 53.8%, which is comparable to the 16-58.7% reported by several studies 6,24,27,29,30 .
Patients may suffer kyphosis after traditional open laminectomy because of the removal of bone tissue and destruction of the posterior tension band 18,31,32 . Posterior instrument fusion is recommended but still controversial. Yu et al. found that the overall mean increase in kyphosis was only 2.9° at a minimum of 1 year after surgery in 49 OLF patients, and no patient required additional surgery due to spinal deformity 24 . Aizawa et al. suggested that there was no relationship between surgical outcomes and an increased kyphotic angle after laminectomy 31 . In addition, thoracic kyphosis might be influenced by changes in back extensor strength in particular, and providing strong, natural extrinsic support for the spine seems to be important for decreasing the incidence of spinal deformities 33 . Percutaneous endoscopic surgeries minimize damage to the paraspinal muscles, facet joints, lamina and posterior ligamentous complexes 15 . We believe that OLF patients who undergo PEPD without fusion may have a much lower risk of kyphosis, and we will continue to conduct a control study to verify this point.
There is great interest in surgical techniques that can minimize complications. Recently, a meta-analysis showed a moderately high rate of perioperative complications after laminectomy for OLF, and the incidence of dural tears, cerebrospinal fluid (CSF) leaks, infections, and early neurological deficits were 18.4%, 12.1%, 5.8%, and 5.7%, respectively 34 . It is promising that the incidence of complications reported in MIS has been much lower 12,15 . The incidence of dural tears was only 6.7% (2/30) in this study. Since the diameter of punches and forceps used in PEPD was very small and the endoscopic visualization of anatomical structures in a liquid environment was very clear, endoscopic manipulation was so gentle that dural tears could be avoided as much as possible. However, tight DA/DO should remain in situ as a floating adherent fragment 12,15 .
Several studies have reported that some factors might affect the surgical results of OLF, including the duration of preoperative symptoms, preoperative neurological status, intramedullary signal changes in T2WI, age, sex, type of OLF, and dural tears 3,7,20,27,30,[35][36][37] . However, it is still uncertain whether these factors are predictive of the surgical results following MIS. www.nature.com/scientificreports www.nature.com/scientificreports/ The duration of preoperative symptoms was confirmed to be significantly correlated with the surgical results in our study, which is consistent with many studies 3,19,29,35,38,39 . We believe that the exacerbation of preoperative neurological status over a long time is mostly irreversible. Thus, earlier surgical intervention might result in better clinical outcomes, regardless of traditional laminectomy or MIS.
The presence of T2HIS was confirmed to be significantly correlated with surgical results in our study, which is similar to some studies 7,30 . Intramedullary signal changes in the spinal cord occur during pathologic changes, such as the loss of nerve cells, oedema, gliosis, demyelination, and Wallerian degeneration 40,41 . We believe that patients with intramedullary signal changes have irreversible spinal cord changes. As the symptoms are subjective, objective findings, such as radiologic examinations, are important for evaluating the factors that affect surgical results 30 .
The severity of myelopathy before surgery, which is mostly indicated by the preoperative mJOA score, was shown to be the most important predictor of surgical results in many studies 3,24,36,38,[41][42][43] . However, few studies found a significant correlation between the preoperative mJOA score and RR according to multiple regression analysis 3,36,41 , and one study even showed no significant correlation between them in the univariate analysis 1 . Of note, we did not find a significant correlation between the preoperative mJOA score and RR in multiple regression analysis, similar to some studies 24,29,38 . A lower preoperative mJOA score may result in worse recovery, but it is not a predictor of the surgical results following PEPD.
The morphology of OLF might affect the surgical results but is still controversial. Kuh et al. suggested that the beak type of OLF with high intramedullary signal changes might be a poor prognostic factor 20 . However, Kang et al. found that patients with beak type OLF could achieve satisfactory RRs and considered that localized compression of the spinal cord could be corrected better than diffuse compression of a round type OLF 42 . Previous studies showed that the RR was significantly better in non-fused compression patients than in fused patients according to axial CT images 6 . However, Li et al. found that the axial CT configurations (unilateral, bilateral or bridged) were not significantly related to the surgical results 41 . Ando et al. also found that there were no statistically significant differences in the RR among the five OLF types (lateral, extended, enlarged, fused, and tuberous) 27 , which was similar to our results.
As the degree of compression increases, the stress distribution on the spinal cord increases. Our data showed that the ratio of APD and CSA had a negative influence on recovery in the univariate analysis, which is consistent with some studies 24,29 . However, we did not find any statistical correlation between the RR and the ratio of APD/CSA in multiple regression analysis. According to the surgical criteria of PEPD 15 , we excluded the fusion and tuberous types of OLF, which both tend to have poor surgical results because of the severe compression to the spinal cord and the low APD/CSA ratio 6,19,24 . The insufficient types of OLF in this study might make it difficult for us to draw significant correlations between the RR and the APD/CSA ratio, as well as between the RR and the types of OLF.
OLF located in the middle thoracic spine may be a predictor of poor outcome 24 . The middle thoracic spine, as the watershed area of the spinal circulation, has the narrowest canal and the maximal kyphotic angle. Thus, the spinal cord in the middle thoracic spine had insufficient compensatory space during posterior decompression, and there is a much greater risk of iatrogenic injury and ischaemic damage 44 . However, we did not find a significant correlation between the location of the lesion and RR. In our study, all four patients (cases 8, 17, 23, 28) with   www.nature.com/scientificreports www.nature.com/scientificreports/ OLF located in the middle thoracic spine achieved satisfactory surgical results (good, fair, good, and good). PEPD showed features including small trauma, high safety (local anaesthesia), and fine and accurate manipulation 15 . It is advantageous to use PEPD to treat OLF, especially to treat OLF located in the middle thoracic spine.
Thoracic OLF is commonly associated with other spinal disorders, such as disc herniation and ossification of the posterior longitudinal ligament (OPLL). Vascular injury is more likely to occur in anterior lesions such as OPLL, which could result in more severe myelopathy and worse recovery regardless of the less severe stenosis 28,44 . Therefore, we excluded patients with coexisting spinal conditions because their therapeutic decisions were more complicated and the surgical results were more unpredictable 6,41 .
The main limitation of this study was the lack of a control group which considered as the gold standard, such as the laminectomy. If the control group was established, the significance of this research would be greatly strengthened. We had started to conduct a retrospective case-control study in which laminectomy is used as a control group, and relevant data were being collected and analyzed. In addition, it was a retrospective study with a relatively small number of patients, and only included single-segment and incomplete types of OLF patients, which could have an impact on the adequacy of our results and could have weakened the statistical significance. Nonetheless, we believe that our results may provide preliminary data in support of the guidelines for PEPD to treat OLF. We will continue to refine this minimally invasive approach.
conclusions PEPD, as the most minimally invasive spinal decompression surgery, is feasible for the treatment of TM patients with a particular type of OLF. A longer preoperative duration of symptoms and the presence of T2HIS were significantly associated with poor surgical results following PEPD. Patients without T2HIS could achieve a good recovery if they received PEPD early.