Learning Curve Analysis: Impact of Ligamentum Flavum Removal Methods on Unilateral Biportal Endoscopic Laminectomy for Lumbar Spinal Stenosis

1. Patient Selection and Surgeon Expertise

This research received ethical approval from the Institutional Review Board of Keimyung University Dongsan Hospital (approval number: 2023-01-009). We conducted a retrospective analysis of 54 patients who were diagnosed with LSS and treated with UBE laminectomy between 2021 and 2022 by a single surgeon. These patients had persistent symptoms despite undergoing conservative treatment for more than 3 months. The inclusion criteria focused on those who underwent UBE decompression at a single level. We excluded patients with more than 2 surgical levels, lumbar spondylolisthesis indicating instability, a history of previous lumbar surgery, or the presence of spinal infections or tumors. Before performing UBE decompressive laminectomy, the senior author (YSK) attained considerable experience with open microscopic unilateral laminectomy for bilateral decompression and discectomy for 5 years and acquired 2 years of experience with minimally invasive laminectomy for bilateral decompression and discectomy via a tubular retractor (METRx, Medtronics, Minneapolis, MN, USA). However, the senior author did not perform endoscopic spinal surgery prior to this study. This specific background allows for a detailed observation of the learning curve from the perspective of an experienced microscopic spine surgeon.

2. Data Collection

We collected clinical, surgical, and radiological data from electronic medical records and surgical video recordings. The parameters analyzed included patient demographics (age, sex), the presence of diabetes mellitus, body mass index (BMI), diagnosis, and preoperative and postoperative scores from the visual analogue scale (VAS) and Oswestry Disability Index (ODI). Additionally, the duration of hospitalization, surgical level, operative time, and preoperative and postoperative imaging data (computed tomography and magnetic resonance imaging [MRI]) were recorded. We reviewed the magnetic resonance images and assessed the degree of spinal stenosis according to the Schizas et al. [11] spinal stenosis grading system as measured by the morphology of the dural sac on MRI.

3. Surgical Technique

All procedures were performed under general anesthesia, with patients positioned prone on a radiolucent surgical table. The UBE technique necessitated continuous normal saline irrigation; thus, appropriate waterproof draping and a continuous drainage system were employed to maintain the sterility of the surgical field [9]. We utilized fluoroscopic guidance to identify and mark the midline intervertebral space and pedicles of the operative spinal level. UBE decompression was performed through 2 incisions called portals, through which an arthroscope and surgical instruments were inserted. In all the patients, UBE decompression was performed via a left-sided approach at our center. The scope portal was made with a 1.5-cm incision at the lower margin level of the upper pedicle, corresponding to the intended lumbar spine level for surgery. Additionally, the working portal was created with a 2-cm incision at the upper one-third level of the lower-level pedicle, 2.5–3 cm from the scope portal (Figure 1). The operative technique involved the use of a serial dilator for multifidus muscle dissection, soft tissue detachment, and trocar insertion to establish a clear operative field. Surgical instruments entering through the working portal and arthroscope, referred to as “triangulation,” are important. Afterward, normal saline was infused to create a surgical space. During UBE decompression, the flow and pressure of normal saline are crucial. Excessive pressure can lead to increased intracranial pressure-related complications and potential neural tissue compression. In contrast, insufficient pressure results in a less clear operative field. Therefore, it is advisable to position normal saline approximately 70–100 cm above the patient to facilitate its inflow [10]. Once the potential working space is created by water flow, the spinolaminar junction can be easily identified after dissecting the soft tissue, and the bone work for the laminectomy procedure begins. Soft tissue bleeding control during the procedure involved the use of a radiofrequency wand, and bone bleeding at the laminectomy site was managed using bone wax. Laminectomy was performed until the midline cleft of the ligamentum flavum (LF) was discernible [12]. The LF was removed on the ipsilateral side first and on the contralateral side sequentially. To remove the LF, detachment and dissection from the bone window are important preliminary steps. Additionally, partial removal of the facet joints was conducted, if needed, to decompress the traversing nerve root at the bilateral lateral recess [9,10,13,14]. In early cases, referred to as the cranial group, removal of the LF was performed from the cranial to caudal direction without any detachment of the LF from the superior portion of the caudal lamina; in later cases, referred to as the caudal group, removal was conducted from the caudal-to-cranial direction with earlier detachment of the LF from the superior portion of the caudal lamina.

Figure 1.

Surgical incision for the scope portal and working portals. For the scope portal, a 1.5-cm incision is made at the lower margin level of the pedicle of the upper lumbar spine. For the working portal, a 2-cm incision is performed approximately one-third above the pedicle of the lower level of the lumbar spine. The medial pedicular line should be located midline of the 2-portal incision. Additionally, the 2 portals should be 3 cm apart to prevent fighting between the scope and instruments.

4. Statistical Analysis

We conducted comparative analyses of clinical and surgical factors, including complications, between patients grouped by the direction of LF removal (cranial vs. caudal). The data are reported as the means±standard deviations. Categorical variables were compared via the chi-square test, whereas continuous variables were analyzed via Student t-test. For clinical outcomes, simple regression analysis was performed to exclude the effect of increasing proficiency over time (IBM SPSS Statistics ver. 26.0, IBM Co., Armonk, NY, USA).

In this study, we analyzed the learning curve via learning curve cumulative summation analysis (CUSUM), a modification of the traditional CUSUM designed to indicate when an adequate level of competence has been achieved from an initial state of incompetence. When a surgeon begins a new surgical procedure, the process is considered "out of control" until the surgeon achieves sufficient competency. The CUSUM monitors each subsequent surgery and identifies when the surgical process becomes "in control."

In our study, the learning curve was evaluated based on operative time and occurrence of dural tears, which are the most common complication for beginners. For the learning curve of the operative time, the target was set at 92.5 minutes, representing the mean operation time (92.3±31.7 minutes). Failure was defined as "the operation lasting longer than 92.5 minutes." An acceptable failure rate was defined as the probability of an operation exceeding 92.5 minutes being less than 0.2, whereas an unacceptable failure rate was defined as this probability being more than 0.4. For the learning curve regarding the occurrence of dural tears, failure was defined as "an intraoperative dural tear occurring." An acceptable failure rate was defined as the probability of a dural tear occurring being less than 0.2, and an unacceptable failure rate was defined as this probability being more than 0.4, on the basis of previous studies [15,16].

Before analysis, we determined the unacceptable failure rate (P0), acceptable failure rate (P1), type I error (α), and type II error (β). The type I error (α) and type II error (β) rates were set at 0.05 and 0.20, respectively. The equations presented in Table 1 were used to calculate the CUSUM score, resulting in a decrease of 0.29 (S) for each successful operation and an increase of 0.71 (1-S) for each failure. A decision limit of h=-2.83 was chosen on the basis of the equation. All analyses were performed via R ver. 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria).

Table 1.

Values used to plot the learning curve cumulative summation test

1. Patient Demographics and Baseline Characteristics

Table 2 presents the baseline demographics of the 54 enrolled patients, who were divided into a cranial group (n=22) and a caudal group (n=32) on the basis of the direction of LF removal. Both groups had comparable average ages (72.23±9.61 years for cranial, 70.19±10.19 years for caudal) and similar characteristics regarding BMI, diabetes incidence, and surgical level distribution, ensuring a consistent baseline for analysis. The majority of patients in both groups had grade C and D spinal stenosis according to the Schizas et al. [11] grading system. Grade C indicates obscured rootlets and cerebrospinal fluid (CSF) with visible epidural fat, whereas grade D signifies a more severe condition with nonvisible epidural fat. Statistical analysis revealed no significant differences in demographics or spinal stenosis distribution between the cranial and caudal groups (p=0.49 for surgical levels, p=0.39 for stenosis grade).

Table 2.

Demographic data for the cranial and caudal groups

2. Comparison of Clinical Outcomes Between the 2 Groups

The clinical outcomes and postoperative complications are summarized in Table 3. Operative duration showed a significant distinction, averaging 110.86±32.63 minutes for the cranial group and 79.56±24.21 minutes for the caudal group (p<0.01). This reduction highlights the efficiency gains with experience and adjustment in the direction of LF removal. Despite these operational differences, both groups demonstrated notable improvements in clinical metrics postoperatively, with VAS and ODI scores significantly decreasing in both groups, indicating effective pain relief and functional recovery. However, no statistically significant difference in these clinical outcomes was found between the groups, suggesting that both removal approaches are equally effective from a patient outcome perspective.

Table 3.

Clinical outcomes and postoperative complications

Among all 54 patients, approximately 25.93% had complications (14 patients). All 14 patients had intraoperative dura mater tears. The complication rates were significantly different, with the cranial group experiencing a total of 9 patients with dural tears (40.91%), 2 of whom required revision surgery, compared with the total of 5 patients with dural tears in the caudal group (15.63%, p=0.04). This difference underscores the technical challenges associated with the cranial group and highlights the potential for improved safety with the caudal group.

3. LC-CUSUM Analysis of the Operating Time and Complications

The cumulative number of failures for operative time and the occurrence of dural tears are shown (Figures 2A and 3A). learning curve cumulative summation (LC-CUSUM) analysis indicated that 26 patients were required to achieve surgical competency in terms of operative time (Figure 2B). With respect to the occurrence of dural tears, 29 patients were needed to reach competency (Figure 3B). This result suggests that the learning curve for operative time is not significantly impacted by the technical change implemented after the 23rd case. However, the learning curve for intraoperative dural tear occurrence appears to be influenced by technical changes, indicating that this modification may contribute to overcoming the learning curve.

Figure 2.

(A) Cumulative number of failures in terms of total operative time. (B) Cumulative summation test of the learning curve for the total operative time. Performance can be judged to be proficient from the 26th case. LC-CUSUM, learning curve cumulative summation.

Figure 3.

(A) Cumulative number of failures in terms of the incidence of dural tears. (B) Cumulative summation test of the learning curve for intraoperative dural tear occurrence. Dural tear occurrence decreased to a level indicative of proficiency after the 29th case. LC-CUSUM, learning curve cumulative summation.

1. Comparison With Existing Literature With Respect to the Learning Curves for UBE Decompressive Laminectomy

There are several reports related to the learning curve of UBE decompressive laminectomy. Choi et al. [5] reported the learning curve for UBE spine surgery, whereby the surgical duration remained consistent for 14 patients with lumbar disc herniation. However, they did not provide demographic information about the enrolled patients. Additionally, their study included various patient groups, including patients with lumbar disc herniation and LSS, which could have added heterogeneity to their results. After 58 patients underwent UBE decompressive laminectomy, Park et al. [16] reported that proficiency improved. However, in this study, surgical levels and approach directions were divided into left and right sides; thus, there might be less standardization in explaining the actual learning curve in a more controlled setting.

Our study demonstrated that, compared with other surgeons, novice surgeons could overcome the learning curve with fewer patients. This may be attributed to our controlled surgical environment, which included variable control, a left-sided approach, single-level surgeries, and procedures performed on the L4–5 segment. However, when assessing complications related to the learning curve, we observed a rate of approximately 25%, which is higher than the 10% previously reported in other studies [5,10,16]. This discrepancy likely stems from the unique context of our study, including the senior author's initial lack of experience with endoscopic methods and a meticulous review process that identified even asymptomatic dural tears.

2. Impact of Dural Tears on the Learning Curve and Dural Tear Management

While most other studies have analyzed the learning curve in terms of operative time [10,16,22], this study also analyzed the incidence of dural tears, the most common complication among novice surgeons. Our analysis revealed that overcoming the learning curve for reducing complications such as dural tears requires more cases than for overcoming the learning curve for operative time, which has not been reported in other studies. Achieving proficiency in minimizing dural tears requires more cases due to the technical skills involved. In our series, we could overcome the learning curve by modifying the surgical technique for the direction of LF removal.

The importance of the direction of LF removal is closely related to its anatomical structure. When the lateral portion of the LF is removed via a Kerrison Rongeur without first dissecting the superior portion from the caudal lamina, the base of the LF remains firmly attached to the caudal lamina. This necessitates forceful removal, which can lead to a dural tear, detaching the epidural ligament and dura mater. Even when the lateral portion of the LF is correctly detached, many dural tears still occur because of the procedure of dissecting the LF from the superior portion of the caudal lamina with a Kerrison Rongeur, especially when the Rongeur is placed in the interstitial space without direct visualization of the dura mater. For these reasons, we changed the direction of LF removal.

Our approach for managing dural tears is as follows [23,24]: for dural tears smaller than 1 cm, we use a sealing technique with TachoSil (Takeda Pharmaceutical Co., Tokyo, Japan) applied both inside and outside the dura. For dural tears larger than 1 cm, we opt for open revision surgery. In our cases, all the tears were smaller than 1 cm, making TachoSil repair sufficient. However, there were 2 instances of revision surgery. One patient required endoscopic revision following dural repair with TachoSil, and another patient necessitated microscopic dural repair due to CSF leakage. Postoperatively, patients are required to rest for 2 days to ensure sufficient time for epithelialization of the dura mater. After bed rest, once patients can maintain a sitting position and walk without any symptoms of CSF leakage, such as postural headache, nausea and vomiting [25], they can be discharged on postoperative day 3 or 4.

3. Strategies for Overcoming the Learning Curve of UBE Laminectomy

To overcome the learning curve, the most crucial aspect is to initiate bone work as quickly as possible without causing any soft tissue damage, which involves promptly docking the working portal to enable direct bony contact at the spinolamina junction. To accomplish this goal, it is crucial to gain an understanding of the fascicular anatomy of the multifidus muscle and utilize the interfascicular plane. This can be achieved through an oblique incision of the thoracodorsal fascia, which is distinct from the vertical skin incision. Using a dilator with pivoting motion to detach the muscle and create a working portal allows contact at the spinolaminar junction [26].

The removal direction of the LF is a crucial aspect of surgery. The inner layer of the LF is broadly attached to the inferolateral border of the lower-level pedicle [27], making resection difficult without first splitting this area. Early detachment of the LF from the caudal lamina facilitates easier entry into the epidural space, allowing for sequential detachment of the LF from the cranial direction (Figure 4). This approach can prevent dura mater tear and reduce the operative time required to manage dural tear. Additionally, it may facilitate LF removal, making it more accessible to novice surgeons and helping them overcome the learning curve. Our data revealed a significant reduction in the incidence of dura mater tear when the direction of LF removal was changed (Figure 5). Another study suggested that performing en bloc resection of the LF from the caudal end reduces the risk of CSF leakage [28].

Figure 4.

Caudal-to-cranial directional approach to remove the ligamentum flavum. (A) The lamina has been made transparent to highlight the direction of removal of the ligamentum flavum. As explained, starting the removal from the caudal aspect of the ligamentum flavum and proceeding cranially not only minimizes complications but also allows a shorter operative time. (B) Intraoperative photograph demonstrating detachment of the ligamentum flavum (star) from the caudal lamina (arrow). As the ligamentum flavum is dissected, the underlying epidural space is observed (square).

Figure 5.

Distribution of cases with complications between the cranial and caudal group. This figure shows a schematic diagram of operative time and case numbers divided into cranial and caudal groups. The red circles represent cases where complications occurred. There were significant differences in the complication rate and mean operative time between the 2 groups (complication rate p=0.04, mean operative time p<0.01).

4. Limitations

Our study has several limitations. First, this was a retrospectively designed study with a relatively small sample size, which may affect the generalizability of the results. Additionally, there was a change in the surgical technique after the 23rd case, specifically in the direction of LF removal. This change might have introduced bias into the individual results, as it may have influenced the learning curve and outcomes. To better distinguish the contributions of surgical technique versus experiential learning, future research should employ a controlled study design that includes multiple surgeons with varying levels of endoscopic experience.