Risk Factors for Persistent Neurological Deterioration After Biportal Endoscopic Posterior Decompression for Multilevel Cervical Myelopathy: The “Structural Mismatch” Concept
Article information
Abstract
Objective
This study aimed to evaluate clinical outcomes and identify risk factors for persistent neurological deterioration (PND) following unilateral biportal endoscopic (UBE) posterior decompression for multilevel degenerative cervical myelopathy (DCM).
Methods
We retrospectively analyzed 39 patients with multilevel DCM, including ossification of the posterior longitudinal ligament, who underwent selective UBE decompression involving 3 or fewer segments. All patients were followed for at least 1 year. Delayed neurological deterioration (DND) was defined as new or worsening myelopathic symptoms after initial postoperative improvement and was classified as transient neurological deterioration or PND. Clinical outcomes, including the Japanese Orthopaedic Association (JOA) score and visual analogue scale score; radiological outcomes, including cervical alignment, range of motion, and adjacent border stenosis; and surgical technique, categorized as lamina-preserving or laminectomy-based, were evaluated. Multivariable Firth penalized logistic regression was used to identify independent contributors to PND.
Results
At 1 year, JOA scores improved significantly, from 12.8 to 15.4 (p<0.001), and cervical alignment was maintained. DND occurred in 43.6% of patients, with 30.8% classified as PND. Multivariable analysis identified 3 factors independently associated with PND: residual stenosis at both the cranial and caudal borders (adjusted odds ratio [aOR], 21.4), use of lamina-preserving techniques (aOR, 7.44), and lower preoperative JOA score (aOR, 0.34).
Conclusion
Although selective UBE decompression provides effective, motion-preserving cervical decompression, PND was strongly associated with insufficient decompression at border levels and the use of lamina-preserving techniques in patients with severe myelopathy. These findings support the “structural mismatch” concept as a potential mechanism underlying clinical failure. To reduce the risk of PND, surgeons should ensure extended decompression at border levels and consider laminectomy-based techniques rather than lamina-preserving approaches in high-risk patients.
INTRODUCTION
Degenerative cervical myelopathy (DCM), including cases related to ossification of the posterior longitudinal ligament (OPLL), is a major cause of progressive neurological dysfunction in older adults [1]. For multilevel disease, posterior decompression (laminoplasty or laminectomy with fusion) is widely used, particularly when anterior compression is extensive. However, conventional posterior approaches can be associated with approach-related morbidity such as axial neck pain, muscle injury, reduced range of motion (ROM), C5 palsy, and alignment-related concerns [2-4].
Unilateral biportal endoscopic (UBE) posterior decompression has emerged as a minimally invasive option for cervical myelopathy [5-7]. Beyond reduced posterior soft-tissue disruption and magnified visualization, a key practical advantage is flexibility in level selection [5,8,9]. In multilevel stenosis, this technique enables the surgeon to select the most severely responsible segments among multiple diseased levels and perform precise endoscopic decompression through portal switching and tailored working trajectories [8]. However, this strategy introduces a critical trade-off—balancing minimal invasiveness against adequate decompression margin.
This trade-off is particularly pertinent to the phenomenon of delayed neurological deterioration (DND). While previous studies have attributed postoperative deterioration to mechanisms such as reperfusion injury [10], spinal cord edema [11], tethering effects [12], or residual stenosis [13], these risk factors are rarely distinguished by their reversibility. We hypothesize that DND is heterogeneous rather than a single entity: whereas transient neurological deterioration (TND) may stem from reversible physiological responses (e.g., edema or minor fluid collection [14]), persistent neurological deterioration (PND) likely represents a "structural failure"—such as residual border stenosis or inadequate expansion—resulting in persistent mechanical conflict.
Accordingly, this study investigated the 1-year longitudinal outcomes and complication profiles following UBE posterior cervical decompression for multilevel DCM, with a focus on safety and clinically relevant risk factors. Furthermore, we applied an operational classification for neurological worsening—distinguishing TND from PND—to describe postoperative trajectories and identify risk factors associated with PND following selective UBE decompression.
MATERIALS AND METHODS
1. Study Design and Patient Population
This study was approved by the Public Institutional Bioethics Committee (P01-202509-01-014). We conducted a retrospective analysis of a consecutive series of patients diagnosed with DCM who underwent multilevel UBE posterior cervical decompression at a single center between January 2023 and May 2024. All surgical procedures were performed by a single senior surgeon with 10 years of experience in endoscopic spine surgery, including 3 years of dedicated practice in endoscopic cervical laminectomy.
2. Inclusion and Exclusion Criteria
The inclusion criteria were as follows:
(1) DCM with myelopathic symptoms/signs refractory to conservative treatment. Symptoms included limb weakness, loss of fine motor skills, gait instability, sensory disturbance, urinary dysfunction, or muscle atrophy.
(2) Magnetic resonance imaging (MRI)-confirmed moderate to severe stenosis (Kang grade [15] 2 or 3), regardless of intramedullary signal change, with surgical levels determined by clinicoradiological correlation (Figure 1).
Surgical strategy and illustrative cases of biportal endoscopic posterior cervical decompression for degenerative cervical myelopathy. (A) Schematic illustration of surgical planning. Decompression was directed to all segments with Kang grade 3 stenosis (arrow) and intramedullary T2-weighted signal change (red area). Adjacent Kang grade 2 stenosis was not routinely included when the planned decompression would span more than three segments. (B and C) Single-segment decompression using unilateral laminotomy for bilateral decompression (ULBD). Preoperative (B) and postoperative (C) T2-weighted magnetic resonance imaging (MRI) demonstrate adequate decompression at C5–6, which had Kang grade 2 stenosis. (D and E) Two-segment decompression using endoscopic laminectomy. Preoperative (D) and postoperative (E) T2-weighted MRI show wide decompression of the C5–6 and C6–7 levels achieved by C6 laminectomy. (F–H) Postoperative 3-dimensional computed tomography reconstructions. (F) Two-segment decompression achieved by laminectomy. (G) Three-segment decompression achieved by laminectomy. (H) Three-segment decompression using a hybrid technique. Note the combined laminectomy and ULBD (red asterisk), with preservation of the intermediate lamina to reduce structural instability.
(3) Selected OPLL cases only when occupancy ratio <50% and K-line positive [16].
(4) Minimum follow-up ≥12 months.
Exclusion criteria included a history of prior cervical surgery, kyphosis beyond a predefined threshold (C2–7 Cobb angle ≥14.5° in the kyphotic direction on neutral lateral radiographs), segmental instability (anterior translation >3.5 mm on dynamic radiographs), OPLL with an occupancy ratio >50%, or long-segment OPLL extending beyond three levels. Prophylactic decompression of segments without clear myelopathic symptoms or significant stenosis was also excluded from the analysis.
3. Radiological Assessment and Grading
All patients underwent comprehensive preoperative imaging, including MRI, computed tomography (CT), and plain radiography. Cervical canal stenosis was classified according to the Kang grading system [15] using sagittal T2-weighted MRI sequences (Figure 1A):
Grade 0: Normal canal.
Grade 1: Obliteration of >50% of the subarachnoid space without cord deformity.
Grade 2: Spinal cord deformity without intramedullary signal change.
Grade 3: Increased intramedullary signal intensity at the compression level.
Preoperative cervical MRI was independently reviewed by 2 board-certified spine surgeons who were blinded to clinical outcomes. Any discrepancies were resolved by consensus between the 2 reviewers.
All patients underwent postoperative day 1 MRI to confirm the extent of decompression and rule out complications such as epidural hematoma. A 1-year follow-up MRI was recommended for all patients; however, not all patients underwent the 1-year MRI due to financial constraints. Cervical alignment was assessed using the C2–7 Cobb angle [17] on neutral lateral radiographs, and ROM [18] was evaluated using dynamic flexion-extension views preoperatively and 1 year postoperatively.
4. Clinical Outcomes and Complications Assessment
Clinical outcomes were evaluated using the visual analogue scale (VAS) for arm and neck pain, and the Japanese Orthopaedic Association (JOA) score for neurological function. The JOA recovery rate [19] was calculated as: recovery rate (%)=(postoperative JOA − preoperative JOA)/(17 − preoperative JOA) × 100. Any perioperative complication or postoperative DND was analyzed.
5. Postoperative DND
DND was defined as new or aggravated myelopathic symptoms (e.g., hand clumsiness, forearm cramping/atrophy, gait instability) that occurred after documented initial postoperative improvement, which was defined as patient-reported improvement in myelopathic symptoms recorded at the first postoperative outpatient visit (typically 2–3 weeks after discharge). All DND patients underwent conservative management initially, and patients who recovered with this were classified as TND. Patients who did not recover with conservative management by the end of follow-up, or for whom surgery was performed or recommended, were classified as having PND. Patients with perioperative complications were included in all statistical analyses of DND only if they achieved baseline neurological recovery at discharge.
6. Statistical Analysis
Clinical and radiological variables are presented as mean±standard deviation or number (%). Preoperative and 1-year outcomes were compared using the Wilcoxon signed-rank test (VAS, JOA) or paired t-test (Cobb angle, ROM). PND vs. Non-PND comparisons used Welch t-test for continuous variables and Fisher exact test for categorical variables; Pearson’s chi-square test was applied to 3×2 tables.
Factors associated with PND were evaluated using multivariable Firth penalized logistic regression, which was employed to address potential small-sample bias and separation issues given the limited number of PND events (n=12) relative to the number of candidate predictors. Adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were reported. All tests were 2-sided with p<0.05. Analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and R ver. 4.5.2 (R Foundation for Statistical Computing, Austria), and Python ver. 3.12.3 (Python Software Foundation, USA) with Matplotlib ver. 3.10.7. A swimmer plot was generated to descriptively illustrate the longitudinal clinical course of patients with DND.
7. Surgical Procedures
1) Surgical strategy and decision-making algorithm
UBE posterior cervical decompression was performed using one of 3 predefined approaches according to the target levels and preoperative planning. In this study, a “treated segment” was defined as an interlaminar level (e.g., C3–4).
(1) Unilateral laminotomy for bilateral decompression (ULBD; lamina-preserving) (Figure 1B and C)
• Indication: 1–2 contiguous stenotic levels with predominantly dorsal/dorsolateral compression.
• Technique: Resection of hypertrophied ligamentum flavum and undercutting decompression via unilateral laminotomy without complete laminectomy [5].
(2) Laminectomy (Figure 1D–G)
• Indication: When more extensive decompression was required.
• Technique: Complete removal of laminae across 2–3 segments [8].
(3) Hybrid strategy (classified as lamina-preserving) (Figure 1H)
• Indication: planned 3-segment decompression to minimize invasiveness.
• Technique: If MRI showed preserved CSF reserve at a given level despite stenosis, ULBD was applied at that level, while laminectomy was performed at the most severely compressed levels.
For analysis, surgical techniques were categorized as lamina-preserving (ULBD and hybrid) versus laminectomy, and the decompression extent was limited to a maximum of three segments (interlaminar spaces) targeting the primary pathological levels responsible for myelopathy.
2) Technique 1: biportal endoscopic ULBD
Under general endotracheal anesthesia, the patient was positioned prone with neck flexion secured using skin tape. The standard biportal endoscopic systems, a toolkit set for biportal endoscopic spine surgery, radiofrequency systems, working cannula, and a scope retractor (MD company, Korea) were utilized for this technique. Continuous saline irrigation at a controlled pressure (30 mmHg) was maintained throughout all endoscopic procedures using an automated fluid pumping system. High-speed endoscopic diamond-tip burrs (Primado2; NSK-Nakanishi Inc., Kanuma, Japan; 3.0 mm and 4.0 mm) were used for all bony work. A closed-suction drain (10-Fr Jackson-Pratt drain) was routinely placed in the epidural space. Two portals were established along the medial pedicle line at the target level. Upon accessing the interlaminar space, a unilateral laminotomy was performed. The bony resection extended cranially and caudally until the attachment sites of the ligamentum flavum were fully exposed. Following the bony work, the ligamentum flavum was detached from the laminar margins and resected en bloc to expose the dural sac. To ensure adequate lateral decompression and prevent nerve root tethering, an ipsilateral foraminotomy was typically performed prior to the central flavectomy when indicated. This sequence ensured safe manipulation of the dural sac during the removal of the contralateral flavum.
3) Technique 2: endoscopic laminectomy for multisegment decompression
For multilevel laminectomy (e.g., C4 and C5), a three-portal technique was utilized to facilitate broad triangulation. Three paramedian portals were created to expose the interlaminar spaces from C3–4 to C5–6. Using a high-speed diamond drill, the dorsal cortex and cancellous bone of the target laminae (C4, C5) were thinned. The ventral cortical bone was initially preserved to protect the underlying spinal cord ("egg-shell" drilling). To release the laminar complex, the inferior margin of the cranial lamina (C3) and the superior margin of the caudal lamina (C6) were drilled until the ligamentum flavum attachments were disconnected. Consequently, the bony-ligamentous complex—comprising the C4 and C5 laminae and the intervening ligamentum flavum—was removed en bloc as a single unit (endoscopic en bloc laminectomy using “spinous process floating technique”). This maneuver achieved broad decompression across three consecutive segments (C3–4, C4–5, and C5–6) while minimizing the risk of cord injury associated with piecemeal removal.
4) Technique 3: hybrid lamina-preserving technique
The hybrid approach combined ULBD and laminectomy as described above, with technique selection determined by the severity of stenosis at each level.
5) Postoperative management
Postoperative care focused on monitoring neurological status and managing potential complications (Figure 2). Surgical drains were typically removed on postoperative day 1 or 2. However, in cases where postoperative MRI suggested a significant fluid collection or hematoma, drainage was maintained for a minimum of 3 days. Patients were mobilized early with a rigid cervical orthosis, which was maintained for two weeks to facilitate soft tissue healing before gradual return to daily activities.
Representative postoperative complications. (A and B) Intraoperative spinal cord injury. Postoperative T2-weighted magnetic resonance imaging (MRI) (B) shows a new focal intramedullary high-signal lesion (yellow arrowhead) compared with preoperative MRI (A), consistent with cord contusion. (C) Asymptomatic epidural fluid collection. Postoperative MRI demonstrates a mixed epidural hematoma and irrigation saline collection without significant cord compression; the collection resolved with conservative management. (D) Symptomatic epidural hematoma originating from the subcutaneous layer. A subcutaneous hematoma (red asterisk) tracked into the epidural space, causing acute cord compression (white arrows) and requiring emergent evacuation. (E) Recurrent epidural hematoma from a muscular bleeder. A hematoma between the subcutaneous tissue and extensor musculature (yellow asterisk) extended into the epidural space, resulting in severe cord compression (white arrows) and necessitating 2 consecutive emergency evacuations.
RESULTS
1. Patient Characteristics and Surgical Profile
A total of 39 patients (male:female, 25:14) were included. Mean age was 66.1±12.3 years and mean follow-up was 14.4±2.5 months. Mean hospital stay was 6.4±3.3 days, and mean operative time per treated level was 41.8±13.9 minutes. Primary pathology was degenerative cervical spondylotic myelopathy in 23 patients and OPLL-induced myelopathy in 16 patients.
Regarding surgical technique, laminectomy-based decompression was performed in 28 patients, whereas lamina-preserving decompression was performed in 11 (ULBD, n=8; hybrid, n=3). In total, 82 segments were treated; most procedures addressed 2 segments (n=29), followed by 3 segments (n=7) and 1 segment (n=3) (Table 1).
Patient characteristics and operation details (n=39)
2. Clinical and Radiological Outcomes at 1 Year
Clinical outcomes improved significantly at 1 year. Neck pain VAS decreased from 6.3±1.1 to 2.3±1.0 (p<0.001), and arm pain VAS decreased from 7.2±0.8 to 2.6±1.3 (p<0.001). The JOA score improved from 12.8±0.8 to 15.4±1.7 (p<0.001), with a mean JOA recovery rate of 63.3%±38.3%.
Radiologically, cervical alignment and motion were maintained without significant change. Cobb angle was 15.0°±10.3° preoperatively and 15.4°±10.4° at 1 year (p=0.38), and ROM was 36.1°±12.4° and 34.1°±11.1°, respectively (p=0.17) (Table 2).
Clinical and radiological outcomes
3. Perioperative Complications and Postoperative Neurological Events
Overall, 14 patients (35.9%) experienced at least one perioperative complication. Neural complications occurred in 4 patients (10.3%), including transient C5 palsy (n=2), transient nerve root symptoms (n=1), and spinal cord contusion due to drill slippage (n=1). On postoperative day 1 MRI, postoperative surgical-site hematoma/fluid collection was identified in 8 patients (20.5%): 3 asymptomatic radiographic cases (7.7%) and 5 symptomatic cases (12.8%). Of the symptomatic patients, 3 (7.7% of the total cohort) required urgent evacuation; notably, all 3 recovered without significant residual neurological sequelae after timely intervention. Likewise, patients with transient C5 palsy, transient nerve root symptoms, or focal cord contusion had minimal additional neurological deficit during hospitalization and were discharged with improvement in their baseline myelopathic symptoms. Therefore, patients with perioperative events were retained in the analytic cohort for analysis (Table 3; Figure 2).
Details on postoperative complications (n=39)
4. Delayed Neurological Deterioration
After documented initial improvement, 17 patients (43.6%) developed DND. TND occurred in 5 patients (12.8%). PND occurred in 12 patients (30.8%); of these, additional surgery was performed in 3 (7.7%), additional surgery was recommended but not performed in 3 (7.7%), and 6 (15.4%) had persistent symptoms despite nonoperative management.
Swimmer plot (Figure 3) illustrates patient-level longitudinal trajectories of DND, including timing of onset, duration of conservative management, and escalation to surgery when applicable. The onset of DND occurred at a median of 12 weeks after surgery (range, 4–52 weeks). TND tended to occur earlier (median, 6 weeks) compared to PND (median, 13 weeks). All TND patients recovered with conservative management, whereas only 1 of 12 PND patients improved following revision surgery (Figures 4 and 5).
Swimmer plot illustrating the timeline and clinical courses of patients with delayed neurological deterioration (DND). Each horizontal bar represents the clinical course of an individual patient beginning on the day of the initial biportal endoscopic surgery. The timeline marks key milestones, including the onset of DND and initiation of conservative management, the timing of recommendation for revision surgery, and the performance of revision surgery. The endpoint of each bar reflects the final clinical status: for transient neurological deterioration, it marks the time of symptom resolution; for persistent neurological deterioration, it marks the final follow-up. Terminal markers indicate whether the patient eventually improved or had persistent symptoms.
Variable clinical outcomes after lamina-preserving decompression using unilateral laminotomy for bilateral decompression (ULBD) or a hybrid technique. (A) Successful decompression. Preoperative magnetic resonance imaging (MRI) shows central stenosis at C5–6 (red asterisk) without intramedullary T2-weighted signal change; postoperative MRI demonstrates adequate canal expansion (red arrow) with preservation of the posterior tension band. (B) Recurrence with kyphotic progression. Severe compression at C3–4 (red asterisk) was treated with ULBD (red arrow), but myelopathy persisted; at 1 year, imaging shows restenosis at the index level with progression of focal cervical kyphosis (yellow arrow). (C) Symptomatic impingement at a preserved level. After 4-level decompression with preservation of the C4 lamina, recurrent upper-extremity myelopathy developed; 1-year MRI and sagittal computed tomography show cord impingement at C3–4 (red arrows) between a ventral endplate spur and the retained C4 lamina (red asterisk). (D) Restenosis due to insufficient canal expansion. Long-segment decompression with preservation of the C5 lamina was followed by recurrent or persistent myelopathy; 1-year imaging shows the retained C5 lamina (red arrowheads) limiting posterior cord drift, leaving residual stenosis (yellow arrowheads), and resulting in clinical failure.
Variable clinical outcomes after biportal endoscopic multisegment laminectomy. (A and B) Successful comprehensive selective laminectomy including Kang grade 2–3 stenotic segments: 2 levels (C5–7) in panel A and 3 levels (C4–7) in panel B. One-year imaging confirms durable decompression without significant adjacent-level stenosis. (C) Transient neurological deterioration related to posterior cord shift after 3-level decompression (C4–7) with wide dome laminotomy. Symptoms resolved with neck immobilization. Early postoperative magnetic resonance imaging (MRI) shows posterior cord shift into the expanded canal (red asterisk), whereas follow-up MRI demonstrates stabilization within the remodeled canal (yellow asterisk). (D) Persistent recurrence and clinical failure after selective laminectomy at C5–7. One-year MRI shows progression of stenosis at both the cranial and caudal border levels (red arrows), with suspected intramedullary T2-weighted signal change (yellow asterisk). (E) Persistent recurrence or progression after limited selective laminectomy targeting only levels with preoperative signal change while sparing adjacent moderate stenosis. Despite early improvement, gait rapidly worsened, and 3-month MRI demonstrates aggravated myelomalacia with progression at both cranial and caudal adjacent segments; reoperation was recommended but not performed. (F) Progression of untreated distal stenosis requiring reoperation. Three-level laminectomy at C2–5 left Kang grade 2 stenosis at C5–6 untreated; at 1 year, MRI shows progression to severe stenosis with intramedullary T2-weighted signal change at C5–6, requiring salvage endoscopy-assisted laminoplasty of C5 and C6.
5. Factors Associated with PND (Univariable Analysis)
In univariable comparisons between the PND (n=12) and non-PND (n=27) groups, residual adjacent border stenosis (Kang grade 2) was significantly associated with PND (p=0.02). Relative to clear borders, stenosis at both cranial and caudal borders showed a markedly increased odds of PND (crude OR, 24.00; 95% CI, 1.69–340.69). Preoperative JOA score was lower in the PND group (12.4±0.9 vs. 13.0±0.7, p=0.04). Technique distribution also differed significantly (p=0.04), with lamina-preserving procedures being more frequent in the PND group. Other variables (OPLL diagnosis, preoperative cord T2 high signal, decompressed levels, operative time, postoperative hematoma/fluid collection) were not significantly different (Table 4).
Baseline, imaging, and surgical factors associated with PND
6. Multivariable Analysis (Firth Penalized Logistic Regression)
In the multivariable Firth penalized logistic regression model (PND events, n=12), 3 factors remained independently associated with PND. Lamina-preserving technique (vs. laminectomy-based) was associated with higher odds of PND (aOR, 7.44; 95% CI, 1.32–56.7; p=0.022). Residual stenosis at both adjacent borders was strongly associated with PND (aOR, 21.4; 95% CI, 2.20–398; p=0.008). Higher preoperative JOA score was protective (per 1-point increase: aOR, 0.34; 95% CI, 0.09–0.98; p=0.044) (Table 5).
Multivariable Firth penalized logistic regression model for PND during follow-up
DISCUSSION
1. Evolving Indications and the Challenge of Multilevel Decompression
Minimally invasive posterior cervical decompression has expanded from radiculopathy-focused procedures to a practical option for cervical myelopathy [20-22]. The biportal endoscopic approach offers magnified visualization with reduced posterior muscle injury and relative preservation of posterior stabilizers, which may facilitate earlier recovery. In our cohort, outcomes were consistent with these advantages: VAS and JOA improved significantly, while cervical alignment (C2–7 Cobb) and ROM were maintained at 1 year.
However, in multilevel stenosis and selected OPLL, minimally invasive decompression introduces a core trade-off. Limiting exposure and bone removal may reduce approach-related morbidity but can also narrow the effective decompression margin needed for reliable cord drift and durable symptom control. By limiting decompression to ≤3 segments and incorporating selective strategies, our study illustrates the practical difficulty of balancing preservation with adequacy. This trade-off is reflected in the observed PND rate, highlighting the need for refined selection criteria in multilevel UBE decompression.
2. DND: Mechanisms and Structural Drivers
Because this study focuses on DND after initial postoperative improvement, it is essential to frame these events within mechanisms established in the literature. Prior studies suggest several drivers: (1) reperfusion injury and cord edema [10,11]; (2) tethering effect from posterior cord shift [12]; (3) inadequate decompression preventing cord expansion [13]; and (4) postoperative epidural hematoma [14].
While these mechanisms explain general deterioration, our study sought to distinguish the drivers of PND—representing clinical failure—from transient setbacks (TND). Notably, acute perioperative events such as hematoma or fluid collection did not differ significantly between PND and Non-PND groups. Instead, our multivariable Firth penalized logistic regression model identified three factors independently associated with PND: residual adjacent stenosis (both sides), lamina-preserving technique, and lower preoperative JOA score (Table 5).
These findings support a "Structural Mismatch" theory as the primary mechanism for PND:
• The "Border-Level" Bottleneck (aOR, 21.4): The most potent predictor was residual stenosis (Kang grade 2) at both cranial and caudal borders. This suggests that even if the index level is decompressed, leaving "tight ends" creates a bottleneck [13,23]. This residual tethering likely prevents the cord from drifting into a neutral, tension-free position, sustaining mechanical irritation during neck motion [24,25] (Figure 5).
• The Trade-off of Lamina-Preserving Techniques (aOR, 7.44): Lamina-preserving strategies (e.g., ULBD), while protective of muscle attachments, inherently rely on undercutting. Our data suggests a critical trade-off: in complex cases, the limited bony decompression of these approaches may not provide a sufficient "safety margin" for cord expansion compared to the complete unroofing of a laminectomy. This limitation becomes particularly dangerous when combined with border stenosis (Figure 4).
• Neurological Reserve (preoperative JOA): The protective effect of a higher preoperative JOA score indicates that patients with severe myelopathy have cords intrinsically more vulnerable to minor residual compression or instability.
Synthesizing these factors, PND appears to occur when a vulnerable cord (low JOA) is subjected to insufficient spatial clearance caused by the combination of a limited decompression technique and residual border stenosis. Unlike TND, which may reflect transient physiological responses, PND reflects a durable failure to resolve this mechanical conflict.
3. Clinical Implications and Surgical Strategy
Based on our findings regarding both persistent failure (PND) and acute perioperative complications, we propose a comprehensive management strategy focusing on preoperative planning, meticulous intraoperative execution, and vigilant postoperative surveillance.
1) Preoperative planning: the "border-level" and technique selection
To minimize the risk of PND caused by structural mismatch, surgical planning may go beyond the index level:
• Border assessment: Surgeons must explicitly assess cranial and caudal borders. Our data indicates that residual stenosis (Kang grade 2) at both adjacent borders is a potent predictor of failure. In such cases, the decompression plan must be extended to clear these bottlenecks [20].
• Tailored technique: The choice between lamina-preserving and laminectomy-based techniques should be tailored to the patient's "structural risk." In patients with low neurological reserve (low JOA score) or severe multilevel stenosis involving borders, a laminectomy-based approach may be safer [8]. It guarantees a wider expansion margin to accommodate the vulnerable cord, whereas lamina-preserving techniques (like ULBD, hybrid) may carry a trade-off in decompression width.
2) Perioperative management: hematoma and early surveillance
Although 1-year clinical and radiographic outcomes were favorable, perioperative complications were common (35.9%), and postoperative day 1 MRI frequently demonstrated hematoma or fluid collection. Symptomatic collections occurred in 12.8% of patients, and 7.7% required urgent evacuation, notably recovering without major neurological sequelae after timely treatment. These findings emphasize 2 practical points: (1) close early neurological monitoring is essential, and (2) when symptoms evolve, clinicians should maintain a low threshold for repeat imaging and prompt intervention. It is important to acknowledge that multilevel laminectomy—even when performed endoscopically—inevitably creates a substantial posterior dead space. This space, combined with the venous oozing that may resume after discontinuation of irrigation pressure, likely contributes to the high rate of postoperative collection in our series. Future innovations in endoscopic hemostatic sealing of the posterior dead space may help reduce this complication.
Regarding hematomas, 3 patients in our cohort required urgent evacuation for acute epidural hematoma. In these cases, bleeding was presumed to originate from the muscular layer along the surgical tract [19]. Intraoperatively, this oozing can be masked by the hydrostatic pressure of continuous irrigation, only to manifest as postoperative rebound bleeding into the epidural space (Figure 2D and E). Therefore, meticulous hemostasis of the entire surgical tract is mandatory before closure [26], ensuring no active bleeding persists once water pressure is removed.
3) Rescue strategy for hematoma
If re-exploration is required for a symptomatic hematoma, we strongly recommend prompt evacuation using a microscopic open approach rather than an endoscopic approach. The water pressure and limited field of view in endoscopy may hinder the identification of the true bleeding focus and delay definitive hemostasis in an emergency setting. Importantly, minimally invasive access does not eliminate the risk of clinically meaningful postoperative collections in the cervical canal, especially where neurological reserve is limited.
4) Management protocol for patients who developed DND
(1) Initial conservative management: All DND patients were initially managed with a combination of pharmacological treatment (nonsteroidal anti-inflammatory drugs, neuropathic pain medications, and short-course corticosteroids where appropriate) and supervised physical therapy.
(2) Decision for surgery: Operative intervention was recommended when neurological deterioration showed clear progressive worsening despite adequate conservative management, combined with radiological evidence of stenosis progression or the appearance of new compressive lesions on follow-up MRI. Patients who showed objective neurological decline on serial examinations were prioritized for early surgical review.
(3) For patients in whom symptoms were manageable with conservative treatment and did not show objective progression: surgery was deferred. These patients were enrolled in a structured rehabilitation program consisting of regular cervical muscle-strengthening exercises, postural education, and physical therapy aimed at stabilizing the cervical spine and preventing further neurological compromise. Regular outpatient follow-up with repeat neurological examination and imaging (MRI where clinically indicated) was maintained.
5) How we address the paradox: structure preserving versus extended decompression
While extended decompression at border levels reduces the structural mismatch and thereby the risk of PND, broader laminectomy introduces the risk of epidural scar formation and postlaminectomy kyphosis, which can themselves lead to restenosis.
This paradox is precisely the reason why we defined narrow indications and maintained a strict limit of ≤3 treated segments in this study. Our rationale for employing biportal endoscopic decompression, even for multilevel disease, was that the muscle-preserving and structure-preserving nature of this approach—including preservation of the interspinous ligaments, facet joint capsules, and posterior musculotendinous complex—may reduce the risk of kyphotic deformity compared to conventional open laminectomy, even at equivalent or slightly extended decompression lengths.
However, we are careful not to overclaim this structural benefit in the absence of direct long-term comparative data. Our findings suggest that:
(1) For short-segment stenosis (1–2 levels), endoscopic decompression with adequate border coverage is likely achievable and structurally safe.
(2) For long-segment stenosis (≥4 levels), extension of endoscopic decompression carries increasing risk of the same structural complications—epidural scar, kyphosis—seen with conventional open approaches, without sufficient evidence that the endoscopic technique uniquely mitigates these risks at greater lengths.
Therefore, the principal conclusion of our study is one of caution: the indication for selective endoscopic decompression in long-segment cervical stenosis should not be extended beyond what our current data support. Prospective studies with long-term follow-up are needed to determine whether the structure-preserving characteristics of endoscopic laminectomy confer a genuine advantage over open procedures in longer-segment stenosis. Looking ahead, biportal endoscopic laminoplasty may represent a promising alternative that addresses this paradox directly: by maintaining the posterior tension band through a hinge mechanism while achieving reliable bilateral neural decompression, it may provide structurally more durable outcomes in patients at high risk of postlaminectomy kyphosis. In this sense, the present study may be viewed not only as a description of UBE multilevel laminectomy outcomes, but also as a build-up toward defining the boundaries of the technique and articulating the rationale for laminoplasty-based alternatives in appropriately selected patients. Prospective comparative studies between endoscopic laminectomy and laminoplasty approaches are warranted to directly test this hypothesis.
4. Limitations
This study has several limitations inherent to its design. First, the retrospective, single-center nature and small sample size (n=39) limit generalizability. While we employed Firth penalized logistic regression to mitigate small-sample bias in identifying predictors of PND, the resulting CIs—particularly for residual adjacent stenosis—remain wide. Thus, these estimates should be interpreted as strong clinical signals rather than precise risk quantifications.
Second, we utilized a Swimmer plot instead of the traditional Kaplan-Meier (KM) method. Given the small cohort size and the complex clinical trajectory of DND—where patients often experience initial improvement followed by either transient setbacks (TND) or persistent failure (PND)—standard KM analysis was insufficient to clearly visualize these fluctuating subgroups and their specific time courses.
Third, although postoperative day 1 MRI was routinely obtained to detect acute complications, 1-year MRI was not universal, limiting our ability to structurally correlate late persistent symptoms with radiographic findings in all cases. Additionally, the routine use of postoperative day 1 MRI may have led to a higher detection rate of asymptomatic collections compared to symptom-driven imaging protocols.
Finally, without a comparative cohort of open posterior procedures, technique-level conclusions should be interpreted cautiously. Furthermore, preoperative electrophysiological evaluation and intraoperative neuromonitoring (IONM) were not routinely employed in this cohort. Preoperative evoked potentials (motor evoked potentials/somatosensory evoked potentials) may have provided baseline cord function assessment, although their predictive value for DND remains to be established. IONM may serve as an adjunctive tool for assessing intraoperative decompression adequacy, and in patients with confirmed electrophysiological improvement following decompression, may have provided supportive evidence for distinguishing delayed structural deterioration from immediate surgical failure [27].
CONCLUSION
Biportal endoscopic posterior decompression is an effective, motion-preserving strategy for multilevel DCM. However, the occurrence of PND highlights a critical trade-off between minimal invasiveness and adequate decompression. Our findings suggest that PND is not merely a random event but the result of a "Structural Mismatch," driven by the combination of residual stenosis at border levels, lamina-preserving techniques that limit expansion, and low neurological reserve. To mitigate the risk of PND, surgeons should carefully tailor their surgical strategy, by considering extended decompression at border levels and favoring laminectomy-based strategy in high-risk patients.
Notes
Conflicts of interest
JY Kim, a member of the Editorial Board of Journal of Minimally Invasive Spine Surgery & Technique, is the author of this article. However, he played no role whatsoever in the editorial evaluation of this article or the decision to publish it. The other authors have nothing to disclose.
Funding/Support
This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.