The study retrospectively enrolled the first 36 patients who underwent FE-PLIF surgery consecutively in our hospital. The diagnosis was confirmed by two clinical professors based on a combination of clinical symptoms and imaging evidence including X-ray, computed tomography (CT), and magnetic resonance imaging (MRI). The patients were divided into three groups in the order of operation date, implying that Group A consisted of the initial 12 cases, Group B the subsequent 12 cases, and Group C the last 12 cases. The operations were performed by two fellowship-trained spine surgeons. The clinical demographic features and radiographic features of the patients are shown in Tables 1, 2.

The inclusion criteria were (1) discogenic low back pain, (2) single or double segmental degenerative lumbar stenosis mainly caused by lumbar disc herniation, facet joint hyperplasia, or hypertrophy of the ligamentum flavum, (3) degenerative lumbar spondylolisthesis up to Grade I according to the Meyerding standard, and (4) other single or double segmental degenerative lumbar diseases requiring stability and fusion. The exclusion criteria of this study were (1) the presence of a significant spinal deformity, (2) developmental or multi-segmental (over two) lumbar spinal stenosis, (3) severe spondylolisthesis hard to restore without an open procedure, (4) unclear location of the responsible segment or imaging data inconsistent with the patient’s symptoms, and (5) intolerance to FE-PLIF for other reasons such as severe cardiopulmonary disease and previous lumbar surgery.

All patients enrolled in our study were operated on using the iLESSYS Delta Endoscopic System (Joimax GmbH, Karlsruhe, Germany) for visualization. Under general anesthesia and neuromonitoring, the patient was placed in the prone position on a radiolucent table with the abdominal suspension to reduce abdominal pressure and therefore decrease bleeding. Routine disinfection, sterile towel sheet spread, and incision protective film affixation were performed. After confirming the surgical position and angle by “C-arm” (Figure 1A), a longitudinal incision of about 1.2 cm in length was made above the responsible segment with a No. 11 blade. Serial dilators were advanced step-by-step until palpating the lamina. A working cannula with an outer diameter of 13.7 mm and an inner diameter of 10.2 mm was inserted through the dilator. Next, the dilaters were removed, and the “C-arm” fluoroscopy was performed to confirm the location of the cannula.

Afterward, an endoscopic system with an irrigating channel, which had an outer diameter of 10 mm, an inner diameter of 6 mm and a view angle of 15-degree, was connected and placed into the working cannula. The ipsilateral laminectomy, removal of the ligamentum flavum and medial facetectomy were performed by endoscopic burrs, Kerrison punches and osteotomes (Figure 1B), and the nerve root inside the spinal canal was exposed layer by layer during this process (Figure 1C). With the nerve root carefully protected, a standard full-endoscopic discectomy was performed using endoscopic forceps.

The working channel was then replaced by a dedicated fork-shaped cannula with an inner diameter of 11.5 mm (Figure 1D) for cage implantation. The intervertebral space was further treated with serial reamers, curettes, and rasps until the cartilage endplate was completely peeled off and the bony endplate was entirely exposed for better fusion. Harvested local bones from laminectomy and arthrotomy performed earlier were inserted into the anterior disc space, and an expandable titanium cage (Shanghai Reach Medical Instrument Co., Ltd., Shanghai, China) was placed into the intervertebral space along the working channel. Its location was confirmed by “C-arm” fluoroscopy. The cage location, nerve root relaxation, residual nucleus tissue, and active bleeding inside and outside the spinal canal were rechecked under the endoscopic view. Next, the endoscope and working channel were withdrawn. Bilateral percutaneous pedicle screws and connecting rods with appropriate length were placed on the upper and lower vertebral bodies and fixed (Zina; Sanyou, Shanghai, China). If the patient was diagnosed with lumbar spondylolisthesis, the reduction was required simultaneously. The “C-arm” fluoroscopy was performed again to confirm the position of the internal fixation device. Next, the incision was sutured layer by layer. No drainage was required.

We used VAS which rages from 0 to 10 to evaluate the pre- and post-operative clinical results. Operative time, blood loss, times of intraoperative X-ray fluoroscopy, length of hospital stay, complications, and rate of conversion to an open procedure were recorded. The satisfaction of patients was scored using the Macnab criteria. The fusion rates were evaluated by using Bridwell’s fusion grading system on computer tomography scans or radiographs at the last follow-up. All radiographs in this study were assessed by two independent researchers. Through the discussion with another independent expert, different opinions on fusion healing were reconciled and reached a consensus.

SPSS version 22.0 (SPSS 109 Inc., Chicago, IL) was used to conduct statistical analysis. The data were analyzed using independent sample t-test, paired t-test, chi-square test, one-way analysis of variance, and regression analysis. Statistical significance was set at a p < 0.05 (two-sided).

All of the 36 patients enrolled in this study had undergone FE-PLIF between 2019 and 2021. The radiological measurement data revealed 27 cases of lumbar disc herniation, 5 cases of lumbar spinal canal stenosis, and 4 cases of lumbar spondylolisthesis. All patients underwent FE-PLIF successfully without conversion to open surgery. The blood loss was less than 70 mL in all patients; one patient required postoperative drainage in Group A due to intraoperative dural tear. The average number of intraoperative fluoroscopy performed was 6.97 ± 0.74. The patients enrolled in this study were followed up at an average of 1.41 ± 0.24 years (range: 1–2 years). The majority of patients had immediate relief in pain and dysesthesia. The incidence of complications was 8.3%. One case of dural tear and one case of incomplete reduction requiring open-access revision after 3-month follow-up in group A, one case of postoperative nerve root symptom in group B, and no complications in group C were reported. Patient satisfaction measured using the Macnab criteria showed the surgical outcomes were excellent in 27 (75%) patients, good in 8 (22.2%) patients, and fair in 1 (2.7%) patient with no poor assessment. There was no recurrence of clinical symptoms until the final follow-up (Figures 2 and 3).

The quantified clinical outcomes are shown in Table 3, Figure 4A,B. In Group A, the average postoperative hospitalization stay was 5.62 ± 1.69 days (range: 3–9 days). Case 1 stayed in the hospital after surgery for an especially longer duration than other patients because the complete observation was needed for the first case to ensure safety. The hospitalization duration of Case 5 was extended for the removal of the drainage tube and the recovery of dural tear. A significant improvement in the VAS of lumbar pain at day 1 after the surgery was recorded compared with preoperative VAS (p < 0.001). However, no statistical difference was found between day 1 after the surgery and the final follow-up (p = 0.137). As for the VAS of leg pain, the outcome at the final follow-up was significantly improved compared with that at day 1 after the surgery (p-value between preoperative and postoperative VAS was less than 0.001, and was 0.009 between postoperative and final follow-up VAS). In Group B, the average hospitalization stay was 5.50 ± 2.07 days (range: 3–8 days). The VAS of lumbar pain and leg pain were both significantly improved at day 1 after the surgery compared with the preoperative VAS (p < 0.001, both) and improved tremendously at the final follow-up compared with day 1 after the surgery (p = 0.008 and 0.026, respectively). Case 20 stayed in the hospital for a long time because she felt pain in the right leg that increased while walking. The pain did not subside immediately after the surgery and turned worse, which was believed to be related to the intraoperative traction for the nerve root. After hormone and dehydration therapy to relieve neuro edema, the pain reduced slightly 7 days after the surgery. In Group C, the average hospitalization stay was 4.38 ± 2.07 days (range: 2–9 days). There was a significant improvement on day 1 after the surgery in both VAS of lumbar and leg pain (p < 0.001, both). Statistical difference was only found in VAS of leg pain between day 1 after the surgery and the final follow-up (p = 0.003), but there was no significant difference in those of lumbar pain (p = 0.615). As shown in Figure 4, no statistical difference was found in the VAS of leg pain among all the three groups in each period. Furthermore, no statistical difference was found among them in the VAS of preoperative and final follow-up lumbar pain. The VAS of postoperative lumbar pain showed a different status; the VAS of postoperative lumbar pain in Group B was slightly higher than that of Groups A (p = 0.017) and C (p = 0.004).

According to Bridwell’s fusion grading system, there were 9 cases of definite fusion and 2 case of probable fusion in group A, 11 cases of definite fusion and 1 case of probable fusion in group B, and 8 cases of definite fusion in group C. At 1-year follow-up, the overall fusion rate with definite grade reached 77.7%. The fusion rate with definite grade reached 75% in group A, 91.6% in group B and 66.6% in group C (Table 3). No significant difference was observed in fusion rate among three groups at 1-year follow-up (p = 0.345).

The operative time was recorded to evaluate the learning curve of FE-PLIF. The average operative time was 410.00 ± 58.13 min in Group A (range: 305–535 min), 364.42 ± 37.42 min in Group B (range: 300–420 min), and 319.17 ± 42.90 min in Group C (range: 270–420 min). A statistical difference was found among the three groups (p < 0.001). Further analysis found that statistically significant differences existed not only between Group A and Group B (p = 0.032), but also between Group B and Group C (p = 0.012). The median of operative time appeared in Case 11. Case 1 in Group A experienced the longest operative time among all patients in that group. In Group B, Case 20 experienced a longer operative time than other patients because her herniated disc was extremely large, making it difficult to perform complete decompression. In Group C, Case 29 experienced the longest operative time in this group because the spondylolisthesis had lasted a few years and thus was hard to reduce. No statistical difference was present among the operative time of different diseases (p = 0.337). The learning curve shown in Figure 5 was characterized using a cubic regression analysis (y = –27.07x + 1.42x²–0.24x³ + 521.84, R² = 0.617, p = 0.000).