Thoracic Myelopathy in Ossified Ligamentum Flavum: Surgical Management and Long-Term Outcome Following 2 Different Techniques of Surgical Decompression

Abstract

Study Design:

Retrospective cohort study.

Objective:

Thoracic ossified ligamentum flavum (TOLF) has been reported to present with varying degrees of neuro-deficit and multiple factors have been purported to affect its outcome. Purpose of study was to analyze factors affecting outcome and impact of ultrasonic osteotome (UO).

Methods:

We retrospectively reviewed patients treated for thoracic myelopathy secondary to OLF between 2010 and 2017. 77 patients with complete clinico-radiological records and 2 years follow-up were included. Initial 45 patients, conventional high-speed burr (HSB-group A) was used for decompression. In others, UO was used in combination with HSB (group B). Myelopathy was graded using modified Japanese orthopaedic association grading pre-operatively and each postoperative visit. At final follow-up, recovery rate was calculated. Radiological details including location, morphology, dural ossification, signal change and spinal ossifications were recorded.

Results:

Mean mJOA at presentation and final follow-up were 4.3±1.8 and 7.6±1.9 respectively (p = 0.001). HRR was 49.9±23 at final follow-up. A significant reduction in dural tear (12.5%; 29%) and surgical time (125.8±49.5; 189.4±52.5) were observed in group B (p = 0.00). However, there was no statistically significant difference (p = 0.18) in recovery rates between groups A (44.8±26.1) and B (52.8±24.3). Symptom duration (p = 0.00), severity of myelopathy (p = 0.04) and cord signal changes on MRI (p = 0.02) were important predictors of outcome.

Conclusion:

Use of UO significantly reduced operative time and dural tears, although resulted in similar recovery rate as compared with HSB. Pre-operative severity of myelopathy, symptom duration and presence of cord signal change were the most significant predictors of outcome.

Introduction

The ligament flava are paired spinal ligaments which connect adjacent laminae extending across the spinal column from C2 to S1.^1,2 Pathologically, these ligaments may undergo hypertrophy, calcification or ossification. ³ Ossification of the ligamentum flavum (OLF) is rare pathology, which was initially described by Polgar in 1926 on lateral radiographs. ⁴ It is characterized by endochondral ossification of hypertrophied fibrous tissues within the ligament. ⁵ A majority of our understanding on this topic has been based on the literature from the Far-East, especially Japan. ⁶ It is well-recognized as a rare cause for myelopathy, and most lesions involve lower thoracic or thoraco-lumbar spine. ⁷ The exact etiology is still unclear, although metabolic, genetic and biomechanical theories have been reported. ⁸

Surgical decompression is the ideal management for OLF patients presenting with myelopathy.^1-8 Early decompression not only helps in stabilizing the neurological status, but also offers the best chance of recovery.^2,6 Although options for surgical decompression include laminectomy, laminoplasty and laminectomy and instrumented stabilization; the last option has generally been considered to offer the best opportunity to enable complete excision of lesion and cord decompression.^2,6 Surgical decompression is technically challenging due to dural adhesions; and has traditionally been performed using a combination of high-speed burrs (HSB) and rongeurs. ⁶ Recently, tissue-selective, ultrasonic osteotomes (UO) have enabled safer removal of these ossifications. ⁹ In view of its rarity and wide variations in available treatment options, ideal line of management and long-term outcome are largely ambiguous. ¹⁰

The current study was planned to retrospectively analyze overall outcome in patients with symptomatic TOLF and identify prognostic factors which determine neurological recovery. An additional subgroup analysis of patients was performed to compare outcomes in patients who underwent decompression with UO and those where only HSB were employed.

Material and Methods

After obtaining Institutional Review Board clearance (IRB) records of patients diagnosed with myelopathy secondary to thoracic OLF (T2 and below) and treated surgically at our Institution between 2010 and 2017 were retrospectively reviewed. Only patients with isolated thoracic myelopathy and minimum follow up of 2 years were included. Patients with incomplete clinico-radiological records were excluded. The details of 117 patients were initially retrieved. Thirty-two patients were excluded due to compression secondary to other pathologies (thoracic disc herniation, ossification of the posterior longitudinal ligament, syringomyelia, tumors etc.), tandem stenosis in cervical/ lumbar spines, inadequate preoperative imaging and revision surgery. Five patients died due to other medical reasons and 3 were lost to follow up. Overall, 77 patients were finally included for this study (Figure 1). Being a retrospective study waiver of consent was approved by IRB (IRB-No.2018/10/03).

Figure 1.

Flow chart depicting the strategy to include patients for our study.

These patients were further divided into 2 groups, groups A and B. In the initial 45 patients (group A), only conventional HSB and rongeurs were used for spinal decompression. UO was installed in our hospital in 2014, which was used either alone or in combination with HSB and rongeurs in latter 32 patients (group B) to complete the decompression. In all patients, decompression was performed under microscopic guidance.

Demographic data including age, sex, body mass index and medical co-morbidities [graded using American Society of Anesthesiologists (ASA) grade and Charlson Comorbidity Score] of all patients was recorded. Detailed clinical history and examination findings were also obtained. The myelopathy was objectively graded for using Nurick and modified Japanese orthopaedic association (mJOA) grading systems pre-operatively and postoperatively at 6th month, 1 year and final follow-up time points. At final follow-up, Hirabayashi recovery rate (HRR%) was calculated using the formula:

$$\frac{\text{Postoperative JOA score}-\text{Preoperative JOA score}}{11-\text{Preoperative JOA score}}\times 100$$

The recovery outcome was ranked based on HRR as excellent (100–75%), good (74–50%), fair (49–25%), unchanged (24–0%), or deteriorated (<0%). ⁸ Additional details including estimated blood loss, duration of procedures and any peri-/ post-operative adverse events or complications were recorded.

Only patients with complete pre-operative radiological workup including plain radiographs, MRI and CT were included. Details regarding relevant radiological findings including location, morphology [axial (Sato classification) and sagittal] of TOLF, extent of involvement, dural ossification, presence and extent of cord signal change, canal compromise and associated spinal ossifications at the other levels were recorded (Figure 2). All radiological measurements were performed on PACS software by a spine fellow and further ratified by all senior spine surgeons independently.

Figure 2.

A, Mid-sagittal CT cuts showing associated spinal ligament ossification (Diffuse Idiopathic Spinal Hyperostosis – DISH) (arrow). B, Mid-sagittal T2WI MRI showing Beak type OLF, C, Round/mountain type OLF, D, Measurement of canal compromise on sagittal MRI, E, Length of cord signal change on T2 W MRI. F, Axial CT cuts showing Tram track type dural ossification (arrow).

Surgical Technique

Posterior decompressive laminectomy and resection of TOLF at the involved levels with instrumented stabilization were performed in all patients by 1 of the senior spine surgeons. In both the groups, laminectomy was started by uniform thinning of lamina using HSB up to lamina-facet junction. The lateral gutter was made using the burr and thinned out to leave a thin wafer of bone. In group A, ossified flavum was released from lateral margin using dural hook and 1 mm Kerrison’s rongeur. In group B, final cutting was done using ultrasonic osteotome.

Lamina over the dura was sharply excised and in cases where dura was adherent, it was left as thin wafer or carefully excised along with a portion of duramater making an attempt to keep the underlying arachnoid intact (Figure 3). In case of an accidental durotomy, it was repaired with an appropriate-sized dural patch positioned over the defect. The wound was closed in layers over a single sub-fascial drain. All procedures were performed under intra-operative (IONM) guidance. Throughout the procedure and during early post-operative period, adequate care was taken to maintain hemodynamic stability (mean arterial pressure goals ≥75-80 mmHg).

Figure 3.

Intra-operative images showing A. Thinning of the lamina with the help of a high-speed burr and preparing a lateral gutter, B. Final completion of the bone cut at the lateral gutter using ultrasonic bone scalpel, C. Final island of lamina before removal, D. Removal of the island of bone over the cord.

Post-Operative Protocol

In patients without dural tear, surgical drains were removed on second or third post-operative days (POD2/3); while in patients with incidental dural tear, the drains were retained until POD4/5 (so as to allow the fascial layer to heal). Post-operatively, patients were usually encouraged to lie on lateral decubitus position in order to prevent wound complications. Patients without dural tear were mobilized on POD1/2; while those with tear were mobilized on POD 4/5 after removal of drain. The patients were discharged in accordance with their pain status and general recovery.

Statistical Analysis

The statistical analysis was done using SPSS software for windows (Ver. 24.0, SPSS Inc, Chicago, IL, USA). The chi-square and ANOVA test were used for categorical values between 2 or 3 groups. Independent t-test was performed to compare the statistical significance of the association for continuous data. Pearson’s correlation coefficient (univariate analysis) was used to test the correlations between various clinico-radiological factors and recovery rate. Multiple linear regression analysis was further performed to determine the association between the independent or regressor variables [clinico-radiological factors including age, sex, body mass index (BMI), diabetes mellitus, duration of symptoms, type of OLF, presence of dural ossification, spinal cord signal change on T2WI-MRI, spinal canal compromise on MRI and pre-operative mJOA score] and the recovery rate (RR-dependent or criterion variable). Results are expressed as the mean±standard deviation, with a p value of <0.05 considered statistically significant.

Results

The mean age at surgery was 57.9±9.5 years (range:38-93 years). Mean Charlson comorbidity index was 2.7±1.2 (range:0-6). Twenty seven (35.1%) and 50 (64.9%) patients belonged to ASA grades 1 and 2, respectively. Mean follow-up duration was 60.8±26.0 months (range:12-114 months). A majority of patients presented with gait instability (85.7%). The other presenting symptoms (in decreasing order of frequency) were numbness over lower extremities (57.1%), back pain (35%), bladder/bowel disturbances (11.6%) and lower limb weakness (7.8%). Twenty-six percent of patients presented within 1 month after the onset of symptoms, while 33.8% and 40.3% presented between 1 to 6 months and ≥6 months, respectively. A total of 17 (22.1%) patients were Nurick 3, 41 (53.2%) were grade 4 and 19 (24.7%) grade 5 on presentation. The mean surgical time was 152.6±51.4 minutes (range:60 to 390 minutes) and mean intra-operative blood loss was 500.6±259.8 ml (range:150ml-1400 ml).

Most common site of OLF was T10-11 (62.3%), followed by T9-10 (14.0%) and T11-12 (9.1%). On the basis of Sato nomenclature, 46.7%, 29.9%, 16.9% and 6.5% of OLFs were morphologically classified as enlarged, tuberous, extended and fused types, respectively. On sagittal CT, round (51.9%) and beak (48.1%) configurations showed fairly similar distribution. Dural ossification was found in 34 (44.1%) patients, of which “tram-track” variety was the most common (70.5%). This was followed by “comma” (20.5%) and “bridge”–type (9.0%) ossifications. Ossified posterior longitudinal ligament (OPLL-20.8%) and diffuse idiopathic skeletal hyperostosis (DISH-9.1%) were the other associated spinal ossifications (at other spinal levels).

The mean canal compromise on MRI was 66.9±12.7% (range:35.5-99.3%). Hyper-intense cord signal changes on T2WI-MRI were seen in 24(31.2%) patients, with mean length of signal being 12.5± 8.8 mm (range:3.5-59.9 mm).

There was a statistically significant improvement in mJOA at the final follow-up [7.6±1.9 (range:3-11)], as compared with pre-operative status [4.3±1.8 (range: 1-9)] (p = 0.00) in our cohort. The mean recovery rate (Hirabayashi score) was 49.9±23.9% (range:-20-100) at the final follow-up (Table 1). Although there was a trend toward positive correlation between pre-operative severity of Nurick grade and number of patients having signal intensity change on T2 W MRI, it was not statistically significant (p = 0.45) (Figure 4).

Table 1.

Distribution of Patients According to Recovery Rate.

Recovery rate (RR)	Total patients	Percent (%)
Excellent (75-100%)	12	15.5
Good (50-75%)	33	42.8
Fair (25-50%)	20	25.9
Unchanged (0-25%)	12	15.5
Deteriorated (<0%)	2	2.5

Figure 4.

Correlation between the number of patients with signal changes on T2WI-MRI and Nurick grades.

Dural tear was the most common complication observed in 17 (22%) patients, among whom only 34% had dural ossification (p = 0.78). Among the patients with dural tear, tuberous lesion (Sato classification) was most commonly observed (11 patients–64.7%), while enlarged, extended and fused types were observed in 3 (17.6%), 1 (5.9%) and 2 (11.8%) patients, respectively. Ten (58.9%) patients had round-type lesion. In 11 patients, dural tear was managed with dural patch and water tight closure. All dural tears healed with no major sequalae. There was no statistically significant association between the occurrence of dural tear and final recovery (p = 0.48). Four patients developed neurological worsening following surgery. While 1 of these patients (T8-T10 OLF, fused, mountain-type, comma-shaped ossification) deteriorated to paraplegia (ASIA-A), 1 (T10-T11, fused, mountain-type, comma-shaped ossification) had Brown-Sequard-type deficit and 2 (T2-T3, enlarged, mountain-type without dural ossification and T10-T11, extended-, beak-type without dural ossification) had MRC grade 2 deterioration of motor power post-operatively. The deterioration in both patients with partial deficit (1 of whom had incidental dural tear) was transient and they completely recovered at the 6th week follow-up. The other 2 patients underwent immediate re-exploration and wide laminectomy. The patient with complete deficit showed partial recovery and her neurological status remained stable at Nurick 3, mJOA 9 at final follow-up. In the patient with Brown-Sequard deficit (who also sustained incidental dural tear), there was no post-operative recovery. While 1 of these patients (patient with deterioration to ASIA-A) belonged to group B, others belonged to group A. There were 3 early re-admissions [2 patients with urinary tract infection (UTI) and 1 with acute renal failure (ARF)]. All 3 patients were managed conservatively with medical therapy and recovered well.

Based on univariate analysis, patients aged ≤60 years (p = 0.04), duration of symptoms ≤ 1 month (p = 0.00), milder grades of myelopathy at presentation(p = 0.000) and absent cord signal intensity on T2-weighted MR images(p = 0.05) showed statistically significant correlation with better post-operative recovery (Table 2). On multiple linear regression analysis, pre-operative duration of symptoms (≥1 month - p = 0.0001), positive cord signal changes on T2WI-MRI (p = 0.0167) and pre-operative severity of myelopathy (moderate or severe grades-p = 0.0433) were observed to be the most important predictors of outcome (p<0.05).

Table 2.

Univariate Analysis of Correlation between Recovery Rate (RR) and Various Factors.

Factors	Classification	No.	Mean RR	Standard deviation	t	p value
Sex	Male	54	46.3	25.2	-0.858	0.39
	Female	23	51.7	25.9
Age	<60 years	48	52.5	23.9	2.055	0.04
	>60 years	29	40.5	26.3
BMI	<25	24	40.9	22.9	-1.165	0.10
	>25	53	51.2	26.1
Diabetes mellitus	Yes	30	45.6	28.6	0.641	0.52
	No	47	49.5	23.3
Duration of Symptom	<1 month	19	66.5	24.3	12.557	0.00
	1-6 months	27	50.5	21.1
	>6 months	31	34.4	21.9
Type of OLF(Axial)	Enlarged	35	52.5	22.9	1.782	0.16
	Tuberous	23	50.5	21.4
	Extended	14	36.3	33.3
	Fused	5	37.2	28.7
Type of OLF (sagittal)	Beak	37	51.1	25.3	1.034	0.31
	Round	40	45.1	25.4
Dural Ossification	Yes	34	52.5	23.9	-1.410	0.16
	No	43	44.4	26.2
Signal change on T2 W MRI	Yes	32	58.7	18.8	0.65	0.05
	No	45	42.1	29.4
Canal compromise on MRI	<50%	10	50.2	24.7	1.076	0.91
	50-75%	44	50.9	25.3
	>75%	23	41.5	25.8
Preoperative mJOA score	>7	2	67.5	21.3	13.632	0.000
	4-7	50	51.5	20.4
	<4	25	33.9	23.7

Bold denotes significant p-value (<0.05)

A subgroup analysis was performed between groups A and B (Table 3). Both groups were comparable in terms of age distribution (p = 0.63), gender (p = 0.45), Charlson index (p = 0.87), co-morbidities (p = 0.92), duration of symptoms (p = 0.78), preoperative mJOA (p = 0.23), levels (p = 0.27), types of OLF (p = 0.49) and intra-medullary signal change on MRI (p = 0.55). A significant reduction in the incidence of dural tear and surgical time were observed following the use of UCD for decompression, as compared with HSB and rongeurs (p = 0.000). However, there was no statistically significant difference (p = 0.18) in the recovery rates between the groups A(44.8±26.1) and B(52.8±24.3).

Table 3.

Analysis Between Group A and Group B Patients.

Parameter	High-speed burr (group A)	Ultrasonic device (group B)	p-value
Age(years)	57.6 ± 7.2	58.6 ± 12.6	0.65
Male Sex	31 (55.6%)	23 (44.4%)	0.76
Diabetes Mellitus	17 (56.6%)	13 (43.3%)	0.08
Preoperative mJOA score	4.27 ± 1.7	4.25 ± 1.9	0.96
Recovery rate (%)	44.8 ± 26.1	52.8 ± 24.3	0.18
Surgical time(min.)	189.4 ± 52.5	125.8 ± 49.5	0.029
Blood loss(ml.)	521.7 ± 308.3	493.1 ± 217.7	0.638
Dural tear	13 (29%)	4 (12.5%)	0.000

Bold denotes significant p-value (<0.05).

Discussion

Eighteen years after the initial description, Anzai ¹¹ described the first case of OLF presenting with neurodeficit. In 1960, Yamaguchi ¹² reported a case of severe myelopathy, which was managed by surgical decompression. Since then, various reports of thoracic myelopathy secondary to TOLF have been published. In view of its relative rarity, various aspects regarding its global prevalence, ideal line of management, and long-term outcome are still largely unclear. The current study was thus planned to retrospectively evaluate the long-term outcome in OLF patients, who underwent instrumented decompression. To our knowledge, ours is the largest series published till date.

Thoracic myelopathy due to OLF usually demonstrates an insidious onset and slow progression. The prevalence of symptomatic OLF is commonly reported to peak at around 50-60 years of age. ⁵ In our study, 35.4% of patients were older than 60 years. While the literature has reported mixed results with regard to sexual predilection;^5,13 males (70.1%) were significantly more affected as compared to their female counterparts in our study. Gait instability and sensory disturbances were the most common presenting symptoms in our cohort.

The pathogenesis of OLF is multifactorial;^14-16 and broadly includes intrinsic and extrinsic factors. While intrinsic causes include genetic and dietary factors; biomechanical alterations are the major extrinsic factors leading to its development.^3,15,17 It has been demonstrated that when tensile forces act across ligamentum flavum, there are local elevations in BMP-2, TGF-beta, and SOX. ¹⁸ These factors lead to enhanced differentiation of fibroblasts into chondroblasts and osteoblasts; thereby leading to OLF. ¹⁸

We could observe a significantly higher prevalence of OLF in the lower thoracic spine (85%); with T10-T11 being the most commonly involved level (62.3%), followed by T9-T10. This predilection could be secondary to greater tensile stresses acting across the posterior columns at the lower thoracic segments.^7,19

The most common morphological configuration in our study was the enlarged–type. Other studies had reported a higher prevalence of round-type morphology.^20-23 It has also been reported that other spinal ossification pathologies can be observed in association with OLF. ²⁴ In our study, OPLL was the most commonly associated pathology, followed by DISH. The ossification begins in the capsular portion and progresses toward interlaminar part of LF. On axial plane, it progresses from posterior to anterior direction. Based on this progression, Sato^20-22 classified the lesions into 5 types: lateral, extended, enlarged, fused and tuberous. Enlarged and tuberous morphologies were most commonly observed in our cohort.

Chronic compression of the cord leads to infarction, avascular necrosis, gray matter chromatolysis and white matter demyelination. A high-intensity cord signal on T2WI-MRI indicates edema, demyelination, myelomalacia, cavitation, necrosis and related to the severity of myelopathy. ²⁵ Cord signal change was seen in 41.5% of our patients, which is relatively higher than other studies.^13,26 This could be due to greater number of patients (40%) presenting with symptoms ≥6 months. A slight trend of positive correlation between signal intensity change on T2WI-MRI and high pre-operative Nurick grade could be observed, although it was not statistically significant.

Surgical treatment is the norm rather than an exception in symptomatic OLF. ²⁷ The mean improvement in mJOA score was 3.3, with a mean mJOA at final follow-up of 7.6. The overall RR was 49.9%, which is consistent with previous literature (mean improvement rates of 40-50%).^28,29

Multiple prognostic factors determining outcome in patients with symptomatic OLF have been studied. Younger age and shorter duration of symptoms have been purported to be major positive predictive variables.^25,26,30 In our study, 75% of patients with excellent recovery presented with symptoms of less than 1 month duration (which was significant on both uni- and multi-variate analyses). In our study, there was statistically significant correlation between intra-medullary T2-weighted signal change and surgical outcome on both uni- and multi-variate analyses. Similar observation has been reported in previous studies too.^19,31 A low pre-operative mJOA at presentation also significantly correlated with poor outcome in our study. This observation was also in accordance with previous studies by Sanghvi., Yu and Li.^7,32,33

Durotomy (22%) was the most common complication (similar to reports in the existing literature on prevalence between 10 and 32% ³³ ). Certain morphological types, namely tuberous (64.7%) and round-type (58.9%) demonstrated relatively greater association with intra-operative durotomies. Sixty six percent of patients with durotomy had no dural ossification. In the series by Li, the incidence of dural ossification, durotomies and CSF leakage were 24.7%, 20% and 10.6%, respectively. ¹³

Another catastrophic complication is an acute neurological deterioration during early postoperative period.^31,34 The significance of strict intra-operative neuromonitoring (IONM) and maintenance of hemodynamic stability (strict adherence to mean arterial pressure goals) in these challenging surgeries can therefore never be understated. Four of our patients developed post-operative neuro-deterioration; of whom 2 sustained only transient deficits and 1 recovered partially after re-exploration. The introduction of UO has provided major benefits including decreased incidence of mechanical injury, limited heat generation, reduced bleeding which may allow enhanced visibility, and shorter time for laminectomy over HSB.^35,36 Studies have shown significant decrease in surgical time and blood loss with use of UO in cervical and thoracic laminectomies.^35,37 To our knowledge, this is first study which looks into outcomes in TOLF where UO has used been employed for decompression.

There was no significant difference in RR at final follow up between the groups A and B. A significant decrease in durotomies was noted in group B. This difference in durotomies could have also been due to the experience gained by the surgical team over years. Li compared the efficacy of UO and HSB in cervical laminoplasty and reported a low incidence of durotomies in the former group. ³⁵ Chen reported no durotomies during spinal decompression with UO in thoracic OLF. ³⁸ One of the major limitations of our study was the retrospective nature of data collection. In a study by Kawaguchi et al, ³⁹ impaired joint position sense was reported as an important prognostic factor in patients with OLF. However, since the current study was of retrospective nature (based on the available hospital records), we could not include specific clinical examination details like intactness of position sense in our outcome analysis. Nevertheless, this is one of the largest studies on patients with OLF and myelopathy discussing long-term outcome with 2 different surgical modalities.

Conclusion

Symptomatic thoracic OLF treated with posterior decompression and stabilization showed satisfactory outcomes. Though there was no significant difference in neurological recovery following use of UO, its use significantly reduced operative time and dural tears. Pre-operative neurological status, duration of symptoms and presence of cord signal changes on T2WI-MRI were significant predictors of outcome in patients with thoracic OLF.