INTRODUCTION
Lumbar degenerative disease represents a significant burden on healthcare systems worldwide, with spinal fusion procedures increasingly utilized to address symptomatic disc degeneration, spondylolisthesis, and spinal stenosis. Among the various fusion techniques, transforaminal lumbar interbody fusion (TLIF) has emerged as a widely accepted surgical treatment that provides effective decompression while maintaining spinal stability through interbody support and circumferential fusion. The evolution of surgical approaches has led to the development of minimally invasive and endoscopic TLIF techniques, which offer substantial advantages over traditional open procedures by reducing approach-related morbidity, preserving paraspinal muscle integrity, and facilitating faster patient recovery [1-6]. Recent meta-analyses have demonstrated that endoscopic TLIF achieves comparable fusion rates and clinical outcomes to conventional approaches while minimizing soft tissue trauma and postoperative pain [2,4,6]. Consequently, endoscopic TLIF has become an increasingly attractive alternative to conventional open procedures for treating lumbar degenerative conditions.
The selection of an appropriate interbody cage remains central to successful TLIF outcomes, as the cage serves multiple critical functions including restoration of disc height, expansion of foraminal space, and facilitation of solid bony fusion. Polyetheretherketone (PEEK) cages have traditionally been favored due to their radiolucent properties that enable clear radiographic assessment of fusion progression, along with an elastic modulus that closely approximates that of human bone [7-9]. However, PEEK cages are associated with limitations including restricted footprint due to the narrow transforaminal corridor and increased risk of subsidence, particularly in osteoporotic bone [8,10]. In contrast, expandable cages have been developed to address these constraints by allowing insertion through small portals followed by in situ expansion, potentially optimizing segmental lordosis restoration and maximizing endplate coverage [11-17]. Despite these theoretical advantages, expandable cages introduce mechanical complexity and may generate higher expansion forces that could paradoxically increase subsidence risk [11-13]. The ongoing debate regarding the ideal cage design reflects the complex interplay between biomechanical requirements and technical considerations in achieving optimal fusion outcomes.
Although both expandable and PEEK cages are widely utilized in endoscopic TLIF procedures, comparative evidence remains limited, particularly regarding long-term radiographic parameters and patient-reported clinical outcomes. Existing studies have primarily focused on minimally invasive TLIF techniques, with sparse data specifically addressing endoscopic approaches [7,15,16,18]. Key outcomes of interest include fusion rates, cage subsidence, disc height restoration, sagittal alignment correction, and clinical results as measured by Numerical Rating Scales (NRS) for pain and functional outcome scales such as the MacNab criteria. Understanding the comparative performance of these cage types in the endoscopic setting is crucial for optimizing surgical decision-making and patient outcomes. Therefore, the aim of this study was to compare 1-year radiographic and clinical outcomes between expandable and PEEK cages in patients undergoing endoscopic TLIF.
MATERIALS AND METHODS
1. Study Design and Data Collection
This retrospective cohort study included patients who underwent endoscopic TLIF at Daejeon Woori Hospital from January 2015 to December 2024. All procedures were performed by a single surgeon to eliminate intersurgeon variability and ensure consistency in surgical technique. Patients were included if they had radiographic (computed tomography [CT]) and clinical evidence of degenerative disease of the lumbar spine, complete preoperative imaging, and follow-up imaging (CT) at a minimum 1 year postoperatively.
This study was approved by the Institutional Review Board (IRB No: P01-202511-01-022). The requirement for informed consent was waived due to the retrospective nature of the study.
2. Inclusion and Exclusion Criteria
Inclusion criteria comprised patients ≥18 years of age with degenerative lumbar spine disease requiring surgical intervention, endoscopic TLIF procedure performed by the study surgeon, complete preoperative and minimum 1-year postoperative radiographic imaging, minimum 1-year clinical follow-up data available, and single-level surgery. Exclusion criteria included patients <18 years of age, absence of appropriate preoperative or follow-up imaging, malignant indication for surgery, infectious indication for surgery, revision surgery cases, incomplete clinical follow-up CT data at 1 year and multi-level surgery.
3. Cage Classification
Two groups were defined based on the type of interbody cage used during the endoscopic TLIF procedure: the PEEK cage group and the expandable-cage group. These cage types represent the primary interbody spacer options compared in this study. Cage selection was based on surgeon preference, patient anatomy, and specific pathology characteristics. The expandable cages used in this study measured 28 mm in length, 12 mm in width, and 8 mm in height. The PEEK cages were substantially larger, with dimensions of 45 mm in length, 15 mm in width, and 9–11 mm in height. The PEEK cages incorporated a built-in anterior lordotic angle of 6°. The expandable TLIF cages used included inherent pre-lordotic angulation; contemporary designs typically offer 6°–12° (device-dependent ranges of 0°–18°), and biplanar systems may allow intraoperative lordotic expansion up to approximately 22°. The specific PEEK and expandable-cage designs used in this study are illustrated in Figure 1.
4. Surgical Technique Considerations
Patients who required direct decompression, such as those with central or lateral recess stenosis necessitating unilateral laminotomy for bilateral decompression (ULBD), underwent endoscopic TLIF combined with ULBD. In contrast, patients for whom indirect decompression alone was sufficient underwent Kambin triangle lumbar interbody fusion (KLIF) through Kambin triangle. TLIF and KLIF cases were not separately analyzed in the statistical comparisons, which represents a methodological limitation of the present study.
In all cases, the interbody space was packed with autobone mixed with demineralized bone matrix (DBM). The same fusion materials and graft preparation technique were used in both the PEEK and expandable-cage groups.
Insertion of large PEEK cages required a working corridor of approximately 20 mm. Given the cage width of 12–15 mm, insertion was performed under direct endoscopic visualization with continuous C-arm guidance. The endoscopic working sleeve was used to protect the exiting and traversing nerve roots during cage passage. Even with large PEEK cages, no insufficiency of the insertion corridor was encountered, as careful root retraction and controlled advancement allowed safe cage placement. In cases treated with expandable cages, distraction was titrated under fluoroscopy to restore disc height and achieve a modest increase in the segmental disc angle. Depending on device mechanics (height-only vs. height-plus-lordosis expansion), the intended goal was a small early increase in segmental lordosis rather than aggressive angular correction.
Due to its larger footprint, the PEEK cage was introduced in a slightly transverse orientation and rotated into position. In contrast, the expandable cage, with its slim preexpanded profile, was inserted straight along the disc axis.
5. Data Collection Variables
Patient demographic and clinical variables collected included age at surgery, sex, surgical level, and degree of spondylolisthesis (when present). Radiographic parameters assessed at preoperative and 1-year postoperative time points included anterior and posterior disc height, foraminal height, segmental lordosis, presence and degree of cage subsidence, cage length, and cage-to-spine area ratio. Clinical outcome measures evaluated at 1-year follow-up included fusion assessment using Bridwell grading criteria and functional status according to the MacNab criteria.
6. Radiographic Assessment
All radiographic measurements were performed on thin-slice CT images obtained preoperatively and at 1-year follow-up. Disc height measurements included anterior and posterior disc heights measured at the midline on sagittal reconstructions. Foraminal height was measured as the maximum vertical dimension of the neural foramen on sagittal images. Segmental lordosis was defined as the angle between the superior endplate of the upper vertebra and the inferior endplate of the lower vertebra. Cage subsidence was measured as the total vertical displacement of the cage into the adjacent vertebral bodies. Cage length and cage-to-spine area ratio were calculated to assess cage sizing relative to patient anatomy. Fusion status was evaluated using the Bridwell grading system, with grades 1–2 considered successful fusion. Figure 2 details the radiographic measurement protocol and definitions, including anterior/posterior disc height, foraminal height, segmental lordosis, total subsidence, cage distance, and the cage-to-vertebral body area ratio.
7. Statistical Analysis
Descriptive statistics were calculated to summarize patient demographics, radiologic parameters, and clinical outcomes. Continuous variables were expressed as means with standard deviations, and categorical variables as counts with percentages. Comparisons between the 2 cage types (PEEK vs. expandable) were assessed using Welch t-tests for continuous variables with unequal variances and chi-square tests for categorical variables.
Multivariate analyses were performed to identify factors associated with fusion outcomes and subsidence. For fusion assessment, ordinal logistic regression was used to model Bridwell grade as an ordinal outcome. For subsidence analysis, ridge-penalized logistic regression with bootstrap standard errors (B=400) was employed to model clinically significant subsidence (≥5 mm). Variables with univariate p-values <0.10 were included in multivariate models. Statistical significance was defined as p<0.05. All statistical analyses were performed using R ver. 4.3.1 (R Foundation for Statistical Computing, Austria). For the ≥5 mm subsidence endpoint, we used ridge-penalized logistic regression (α=0) with predictors standardized prior to model fitting. The penalty parameter (λ) was selected via 10-fold cross-validation to balance bias and variance. Bootstrap resampling (B=400) was used to derive standard errors and percentile 95% confidence intervals (CIs). All analyses were conducted in R software using the glmnet package.
RESULTS
1. Patient Demographics and Radiologic Outcomes
A total of 84 patients were included, with 32 in the expandable-cage group and 52 in the PEEK cage group. Baseline demographics, including age (65.8±7.2 years, p=0.834) and sex distribution (21.9% vs. 32.7% male, p=0.414), were comparable between groups (Table 1).
Most procedures were performed at the L4–5 level (72.6%), followed by L5–S1 (13.1%), L3–4 (11.9%), and L2–3 (2.4%). The expandable-cage group had a higher proportion of L5–S1 cases (25.0% vs. 5.8%). Degenerative spondylolisthesis was the most common indication (89.3%), with no significant difference in pathology distribution.
Radiologic measurements showed that disc and foraminal height restoration were comparable between groups (all p>0.4). At the 1-year follow-up CT, however, segmental lordosis significantly increased in the PEEK cage group compared with the expandable-cage group (2.6°±3.8° vs. -0.3°±4.6°, p=0.003). Total subsidence was greater in the expandable-cage group (4.6±2.4 mm vs. 2.2±1.7 mm, p=0.001). Expandable cages were shorter in length (29.5±2.0 mm vs. 34.9±3.1 mm, p=0.001) and had a smaller cage-to-vertebral body area ratio (0.2 vs. 0.4, p=0.001). Representative postoperative radiographs are shown in Figure 3, and subsidence measurement examples in Figure 4.
2. Fusion Outcomes (Bridwell Grade)
Ordinal logistic regression was used to identify predictors of fusion quality. Age, sex, subsidence, and cage-to-vertebral body area ratio were not independently associated with Bridwell fusion grade. Cage length demonstrated a marginal association (odds ratio [OR], 1.90; 95% CI, 0.99–3.64; p=0.053). Cage type (expandable vs. PEEK) did not significantly predict inferior fusion outcomes (OR, 25.49; 95% CI, 0.06–10,351.33; p=0.291). The extremely wide confidence interval for cage type suggests unstable estimates due to sparse-data effects, warranting cautious interpretation (Table 2).
3. Risk Factors for Severe Subsidence (≥5 mm)
In the ridge-penalized logistic regression model, a higher cage-to-vertebral body area ratio was significantly protective against severe subsidence (OR, 0.34 per 0.1 increase; 95% CI, 0.12–0.92; p=0.033). Age, sex, fusion grade, and cage type were not associated with subsidence ≥5 mm. Adjusted regression estimates are summarized in Table 3.
4. Clinical Outcomes (NRS Improvement)
Linear regression analysis demonstrated that neither fusion grade nor total subsidence correlated significantly with leg or back NRS improvement. For leg pain, cage type had no effect (β=0.32, p=0.568). For back pain, however, expandable cages were independently associated with greater NRS improvement (β=1.71; 95% CI, 0.35–3.07; p=0.014). Between-group and within-group NRS changes are shown in Figure 5, and corresponding regression estimates in Table 4.
According to the MacNab criteria, 96.4% of patients achieved excellent or good outcomes, with no significant difference between cage types (Table 5). Representative preoperative and postoperative magnetic resonance imaging findings demonstrating disc height restoration and neural decompression are shown in Figure 6.
5. Complications
No cage insertion-related complications such as root impingement or neural injury occurred. Two patients undergoing TLIF with concomitant ULBD experienced incidental dural tears, both repaired intraoperatively with TachoSil without postoperative cerebrospinal fluid leakage. Two cases of postoperative osteomyelitis occurred and resolved with intravenous antibiotics. Five patients developed transient radicular pain suspected to be related to DBM-induced chemical neuritis; symptoms improved with 1 week of intravenous dexamethasone. No screw failure or other hardware-related complications were identified.
DISCUSSION
This retrospective cohort study comparing expandable and PEEK cages in endoscopic TLIF revealed significant differences in both radiographic and clinical outcomes at 1-year follow-up. PEEK cages demonstrated superior segmental lordosis correction (2.6° vs. -0.3°, p=0.003) and significantly lower subsidence (2.2 mm vs. 4.6 mm, p=0.001) compared with expandable cages. However, despite experiencing greater subsidence, patients receiving expandable cages showed significantly greater improvement in back pain scores (1.71 NRS points more improvement, p=0.014). Multivariate analysis identified the cage-to-vertebral body area ratio as a protective factor against clinically significant subsidence (OR, 0.34; p=0.033), while cage length demonstrated a borderline association with fusion outcomes.
The magnitude of lordosis gain with expandable cages is primarily determined by cage mechanics and endplate preparation. Height-only designs typically yield a modest early increase in lordosis that may diminish over time, whereas biplanar (height-plus-lordosis) devices can provide greater initial correction. However, the clinically realized gain is often well below the device’s nominal maximum and should be interpreted in the context of the final 1-year angle. In our series, the PEEK cages incorporated a 6° built-in anterior lordotic angle, which likely contributed to the greater 1-year segmental lordosis observed in the PEEK group [14,17,19].
In addition, variation in insertion trajectory between the 2 cage types may also have influenced the difference in segmental lordosis. The PEEK cage, which enters slightly transversely and is subsequently rotated, tends to seat more anteriorly on the sagittal plane, whereas the expandable cage generally occupies a more central anterior-posterior position. This difference in final cage positioning could have contributed to the greater lordosis gain observed in the PEEK group. Prior work has also demonstrated that anterior cage placement facilitates greater segmental lordosis restoration in posterior lumbar interbody fusion procedures [20].
The observed differences between cage types can be attributed to fundamental design and biomechanical characteristics. Expandable cages, while offering the advantage of insertion through smaller portals, have inherently smaller footprints that reduce endplate contact area and generate higher contact pressures during expansion. This expansion mechanism may produce microtrauma and predispose to subsidence, as demonstrated in our cohort. A recent meta-analysis by Lin et al. similarly reported no clear radiologic or clinical superiority of expandable cages over static designs [14]. These findings are consistent with the significantly smaller cage-to-vertebral body area ratio in the expandable group (0.2 vs. 0.4, p<0.001).
A greater proportion of cases in the expandable-cage cohort were located at L5–S1, and the unique biomechanics of this level may have contributed to the radiographic differences observed between groups. The lumbosacral junction typically demonstrates larger native disc height, greater shear forces, and a steeper sacral slope than upper lumbar levels, which can limit the degree of additional angular correction and create a ceiling effect on postoperative lordosis gain. These anatomic characteristics also increase endplate loading and have been associated with a higher risk of subsidence in prior TLIF analyses [8]. Furthermore, achieving solid fusion at L5–S1 is biomechanically more demanding for posterior-only constructs, supporting the possibility that level-specific factors influenced both the lordosis and subsidence patterns observed in our expandable cohort [21].
In contrast, PEEK cages provide a larger surface area for load distribution and possess an elastic modulus closer to that of human bone, facilitating more physiologic stress transfer. The superior lordosis correction observed with PEEK cages likely reflects their ability to maintain structural integrity during insertion and seating, thereby minimizing endplate violation and maintaining dorsal height restoration. These biomechanical advantages contribute to lower subsidence rates and more durable segmental alignment.
A notable and paradoxical finding of this study was that, despite significantly greater radiographic subsidence in the expandable-cage group, patients treated with expandable cages experienced greater improvement in back pain at 1 year. This observation contrasts with the traditional assumption that increased subsidence correlates with inferior clinical outcomes and suggests a more nuanced radiographic-clinical relationship in the endoscopic TLIF setting.
Several mechanisms may explain this discrepancy. First, expandable cages may provide greater initial distraction of the disc space, which reduces mechanical strain on the posterior annulus and facet joint structures more closely associated with axial back pain than radicular symptoms. Although disc height restoration primarily produces indirect neural decompression, the unloading of pain-sensitive posterior spinal elements during cage expansion may contribute to axial back pain improvement. This effect may persist even after subsequent subsidence, as early tensioning of the posterior longitudinal ligament and annular fibers can diminish discogenic pain generators. Second, despite their smaller initial footprints, expandable cages may achieve stable endplate engagement after settling, promoting fusion in a position that reduces micromotion and associated back pain. Third, subsidence itself does not invariably worsen symptoms; in some cases, controlled settling of the cage into stronger cancellous bone may enhance construct stability and result in additional pain relief.
Our findings align partly with previous comparative studies evaluating expandable versus static cages in TLIF. Stickley et al. [11] reported significantly higher intraoperative subsidence rates with expandable cages, while Canseco et al. [10] found comparable clinical outcomes between expandable and static PEEK cages at 1 year. Chang et al. [13] demonstrated similar long-term radiographic outcomes between cage types despite differences in subsidence patterns. Collectively, these results suggest that increased subsidence with expandable cages does not necessarily result in inferior short-term clinical outcomes.
Other studies have reported contradictory findings. Jha et al. [12] observed improved clinical and radiographic outcomes with expandable cages in a propensity-matched analysis, while Hawasli et al. [15] noted superior sagittal parameter restoration with expandable devices in minimally invasive TLIF. These discrepancies likely reflect differences in surgical technique, cage design, patient selection, and follow-up duration. Wu et al. [22] found comparable subsidence rates between expandable and static cages in unilateral biportal endoscopic fusion, contrasting with our findings, possibly due to technical differences between biportal and uniportal endoscopic approaches.
Most existing comparative studies have focused on minimally invasive or open TLIF techniques, with limited data specifically addressing endoscopic approaches. Our study contributes to this gap by providing dedicated endoscopic TLIF comparative data, which is particularly relevant given the rapid expansion of full-endoscopic lumbar fusion. The narrow working corridor, restricted visualization, and unique implant trajectories in endoscopic TLIF may influence cage biomechanics differently than in conventional approaches.
Recent systematic reviews have demonstrated the efficacy of endoscopic TLIF compared with minimally invasive techniques, but comparative data on cage selection within the endoscopic subset remain limited [2,4,6]. Our findings expand this limited evidence base and highlight the importance of considering both radiographic stability and clinical improvement when selecting cages for endoscopic fusion.
Although several perioperative complications were observed, none were directly related to insertion of the large PEEK cages. Incidental dural tears occurred only in cases requiring additional ULBD and were successfully repaired without postoperative sequelae, suggesting that the risk was associated with decompression rather than cage placement. Osteomyelitis and transient radicular pain were effectively treated with medical therapy, and no hardware-related failures occurred.
Importantly, the absence of insertion-related neural injury underscores the technical feasibility of placing large PEEK cages through an endoscopic working corridor. The combination of direct endoscopic visualization, C-arm guidance, and protective working sleeves allowed safe implant advancement despite the larger footprint. These findings support the safe use of large PEEK cages in endoscopic TLIF and demonstrate their radiographic advantages without added procedural morbidity.
This study benefits from several methodologic strengths, including the single-surgeon design that eliminates inter-operator variability and ensures consistent surgical technique throughout the cohort. The complete radiographic follow-up at 1 year with standardized CT assessment provides robust objective outcome measures. The inclusion of both radiographic and clinical parameters allows for comprehensive evaluation of cage performance across multiple domains.
However, several limitations must be acknowledged. The retrospective design introduces potential selection bias in cage choice, and the relatively small sample size, particularly in the expandable group (n=32), may limit statistical power for detecting smaller effect sizes. The 1-year follow-up, while adequate for assessing early outcomes, may not capture long-term complications or fusion progression. The absence of comprehensive functional outcome measures beyond NRS pain scores limits the clinical assessment scope. Additionally, the lack of randomization in cage selection may introduce unmeasured confounding factors that could influence the observed outcomes. Moreover, because TLIF and KLIF cases were not analyzed separately, differences in cage trajectory and working corridor may have confounded the magnitude of lordosis correction between groups. In addition, the ordinal fusion model yielded unstable estimates with extremely wide CIs for cage type, suggesting sparse-data/separation and limiting the strength of inference for this endpoint.
CONCLUSION
This study demonstrates important trade-offs between expandable and PEEK cages in endoscopic TLIF. PEEK cages achieved superior segmental lordosis correction and significantly lower subsidence, likely due to larger endplate contact areas and more favorable load distribution. Conversely, expandable cages provided greater improvement in back pain despite higher subsidence, possibly through enhanced early disc height restoration and indirect neural decompression. The protective effect of larger cage-to-vertebral body area ratios against subsidence further supports the biomechanical advantages of PEEK cage design. These radiographic and clinical trade-offs should guide individualized cage selection in endoscopic TLIF, considering factors such as bone quality, deformity correction goals, and symptom predominance. Further prospective studies with longer follow-up are needed to refine optimal cage selection criteria in endoscopic lumbar fusion.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
