Impact of cage type on subsidence following anterior cervical discectomy and fusion: a retrospective study

Article information

Asian Spine J. 2025;.asj.2025.0197

Publication date (electronic) : 2025 September 22

doi : https://doi.org/10.31616/asj.2025.0197

Pierce J. Ferriter, Suhas K. Etigunta, Akiro H. Duey, Christopher Gonzalez, Katrina Nietsch, Ashley M. Rosenberg, Bashar Zaidat, Avanish Yendluri, Daniel Berman, Junho Song, Jun S. Kim, Samuel K. Cho

Icahn School of Medicine at Mount Sinai, New York, NY, USA

Corresponding author: Samuel K. Cho, Department of Orthopaedic Surgery, Icahn School of Medicine at Mount Sinai, Mount Sinai West, 425 West 59th St, New York, NY 10019, UEA, Tel: +1-212-636-8250, Fax: +1-212-636-3102, E-mail: samuel.cho@mountsinai.org

Received 2025 April 19; Revised 2025 June 6; Accepted 2025 June 16.

Abstract

Introduction

Anterior cervical discectomy and fusion (ACDF) is a commonly performed procedure for treating cervical myelopathy and radiculopathy. Historically, bone grafts have been used to fill the void created after disk removal, promoting fusion between adjacent vertebral bodies [1]. Autografts, typically harvested from the iliac crest, have demonstrated high fusion rates with relatively few associated complications [1]. While autograft was once the gold standard for ACDF, complications include increased surgical time and blood loss, donor site pain, and infection [1–3]. To mitigate these risks, structural allografts and synthetic cages have gained popularity as alternatives [4,5]. However, the optimal cage material for ACDF remains a topic of ongoing debate and research.

Subsidence is commonly defined as a postoperative loss of disk height by 2–3 mm, resulting from the interbody cage settling into the adjacent vertebral body, primarily due to a mismatch in the elastic modulus between the cage material and surrounding bone. Other risk factors for cage subsidence include advanced patient age, low bone mineral density, and specific cage geometries [6,7]. When the interbody implant settles into adjacent vertebral bodies, it can lead to reduced disk height, disrupted sagittal alignment, pseudarthrosis, and foraminal stenosis, potentially causing radiculopathy and pain [8]. Although cage subsidence is common, its impact on clinical outcomes is not well characterized. Severe subsidence often results in loss of segmental lordosis, but it typically does not significantly impact global cervical alignment parameters [8]. Various cages are currently used, including titanium, carbon fiber, ceramic, and polyetheretherketone (PEEK) [4,9,10]. Given that cage subsidence is a common hardware complication in patients undergoing ACDF, identifying the cage type with the lowest subsidence risk is crucial. This retrospective study aimed to identify the relationships between allograft or interbody cage types and the rate of subsidence in patients following ACDF surgery.

Materials and Methods

Statistical analysis

ADH, PDH, anterior subsidence, and posterior subsidence measurements were compared across spacer types using analysis of variance (ANOVA) tests. Multivariable logistic regression models evaluated the relationship between subsidence and spacer type (structural allograft vs synthetic cages) (PEEK, titanium, and ceramic), adjusting for potential confounders, including the level within the fusion (categorized as top, middle, or bottom for multi-level fusions), the involved cervical level (C3–4, C4–5, C5–6, C6–7), age, sex, smoking history (current or former smoker), and osteopenia diagnosis. The regression analyses provided beta coefficients and 95% confidence intervals (CIs) for each spacer type, quantifying the extent of subsidence associated with each type (Fig. 1).

Fig. 1

Multivariable regression of subsidence and lordosis by cage type. forest plot showing β coefficients and 95% confidence intervals from multivariable linear regression models evaluating the relationship between cage type and anterior subsidence, posterior subsidence, and segmental lordosis. PEEK, polyetheretherketone.

Results

After exclusion, the cohort consisted of 120 patients (50 males [41.7%] and 70 females [58.3%]; mean age, 53.6±10.9 years) with a total of 223 levels fused (mean levels per patient: 1.86±0.69). Fifty-seven patients (47.5%) had a history of smoking, and 25 patients (20.8%) were diagnosed with osteopenia or osteoporosis. Surgical procedures included 38 one-level fusions (31.7%), 61 two-level fusions (50.8%), and 21 three-level fusions (17.5%). Spacer distribution included 62 structural allografts (51.7%), 27 PEEK cages (22.5%), 20 titanium cages (16.7%), and 11 ceramic cages (9.2%) (Table 1). Radiographic assessments were conducted at two time points: immediate postoperative radiographs with a median follow-up of 11 days (range, 0–61 days), and long-term postoperative radiographs with a median follow-up of 12.0 months (range, 6.0–41.3 months) (Table 1). Intraclass correlation coefficients (ICCs) showed excellent agreement for immediate postoperative disk height (ICC, 0.973; 95% CI, 0.94–0.99), and good agreement for final disk height (ICC, 0.870; 95% CI, 0.75–0.94) and subsidence (ICC, 0.856; 95% CI, 0.72–0.93).

Table 1

Demographics and implant characteristics by spacer type in ACDF

Subsidence and disk height

ANOVA tests revealed significant overall differences in both anterior (p<0.001) and posterior (p=0.004) subsidence across spacer types (Table 2). The differences in the distribution of anterior subsidence measurements across cage types are visually illustrated in Fig. 2. Structural allografts exhibited the highest mean anterior subsidence (1.9±1.3 mm) and posterior subsidence (1.5±1.1 mm). Anterior subsidence values were lowest for PEEK cages (0.9±1.4 mm), followed by ceramic (1.3±1.5 mm) and titanium cages (1.7±1.9 mm). Similarly, posterior subsidence values were lowest for PEEK cages (0.9±1.2 mm), followed by titanium (1.0±1.1 mm) and ceramic cages (1.3±1.4 mm). The trends in posterior subsidence are depicted in Fig. 3. The largest observed mean differences occurred between structural allograft and PEEK cages.

Table 2

Mean disc heights, subsidence, and segmental lordosis by spacer type (analysis of variance results)

Fig. 2

Anterior subsidence by cage type. Distribution of anterior subsidence (in mm) stratified by cage type. Polyetheretherketone (PEEK) cages demonstrate the lowest mean anterior subsidence, followed by ceramic and titanium cages, with structural allograft showing the highest subsidence. Each dot represents a single fused level.

Fig. 3

Posterior subsidence by cage type. distribution of posterior subsidence (in mm) stratified by cage type. Polyetheretherketone (PEEK) and titanium cages show lower posterior subsidence anterior subsidence compared to structural allograft and ceramic cages. Each dot represents a single fused level.

The incidence of clinically significant subsidence (>2 mm and >3 mm) was analyzed using chi-square testing, which showed significant differences in subsidence rates across cage types for both anterior (p=0.014) and posterior (p=0.008) subsidence >2 mm. Post hoc pairwise comparisons revealed PEEK cages had significantly lower rates of anterior and posterior subsidence >2 mm compared to structural allografts (corrected p=0.047 and 0.017, respectively), with no significant differences between PEEK and titanium or ceramic cages (Table 2).

Immediate ADH was similar across spacer types (p=0.504), but final ADH differed significantly (p=0.018). PEEK cages had the highest mean final ADH (7.3±1.4 mm), followed by titanium (6.8±1.4 mm), ceramic (6.7±1.3 mm), and structural allograft (6.5±1.5 mm). PDH also varied significantly over time (p<0.001). Structural allografts showed the largest decrease (from 6.9±1.4 to 5.4±1.3 mm), with ceramic cages showing a comparable reduction (from 5.7±1.0 to 4.5±1.0 mm). Titanium cages had a moderate observed decrease in PDH (from 6.5±0.8 to 5.4±0.9 mm), while PEEK cages showed the least change (from 6.6±1.3 to 5.8±1.1 mm). These observed trends were not tested for pairwise statistical significance.

Discussion

In ACDF, selecting an appropriate interbody device material is crucial for achieving successful fusion and restoring normal cervical alignment and lordosis. Subsidence, a known complication of ACDF, is characterized by a reduction in disk height due to settling of the interbody cage into the adjacent vertebral bodies [11]. This can lead to neck pain, higher nonunion rates, and worse clinical outcomes [6]. Our retrospective study of 120 patients and 223 fusion levels compared the impact of various spacer types (structural allografts, PEEK, titanium, and ceramic) on subsidence and segmental lordosis in ACDF. The results showed that structural allografts were associated with a significantly greater subsidence than PEEK cages. PEEK cages showed the lowest subsidence rates and best preserved segmental lordosis (Tables 2, 3).

Table 3

Multivariable regression results comparing cage types to structural allograft for subsidence and segmental lordosis

The reported incidence of cage subsidence after ACDF varies widely, ranging from <10% to >50%, depending on how subsidence is defined and measured [6,12–14]. This variability reflects differences in study design, imaging protocols, and follow-up intervals. Factors associated with increased subsidence risk include low bone mineral density, smoking, age, and cage characteristics such as size, geometry, and material composition [7]. A key biomechanical explanation for subsidence is the mismatch in modulus of elasticity between the implant and surrounding vertebral bone. This mismatch can create stress concentrations, microfractures, and subsequent settling of the implant into the vertebra [15].

Differentiating between radiographically and clinically significant subsidence is important. Radiographic subsidence is defined as a measurable reduction in disk height (typically >2–3 mm), but not all cases translate to clinically significant outcomes. Obermueller et al. reported subsidence rates ranging from 27% to 62% depending on the assessment method, but found no significant association between subsidence and worse clinical outcomes, such as Visual Analog Scale (VAS) scores for neck pain [14]. Clinically significant subsidence may manifest as increased neck pain, radiculopathy due to foraminal stenosis, loss of segmental lordosis, and potentially higher rates of pseudarthrosis. In a study by Lee et al. [12], 46.8% of patients had subsidence at the 12-month follow-up, but clinical outcomes (VAS scores for neck and arm pain) were not significantly different from those without subsidence.

Subsidence can adversely impact surgical outcomes by compromising the structural integrity of the fusion construct, potentially leading to nonunion [11]. Loss of disk height and segmental lordosis can alter cervical spine biomechanics, causing abnormal load distribution and further degeneration of adjacent segments [16]. Severe subsidence can require revision surgery, increasing risks and costs. However, Ryu et al. [17] found no significant difference in radiological and clinical outcomes (minimum follow-up: 5 years) between subsidence and non-subsidence groups if foraminal decompression was also performed.

Cage type likely contributes to subsidence incidence and requires further investigation [6]. Studies evaluating the impact of cage type on subsidence have yielded inconsistent results. Seaman et al. [18], Igarashi et al. [19], and Cabraja et al. [20] found that titanium cages are associated with greater subsidence than PEEK cages, while Godlewski et al. [21] found no significant difference in the subsidence rate between different implant types. Similar controversy exists when comparing structural allograft to PEEK cage types, with conflicting findings regarding the optimal cage type and design [22–26]. While our study did not include zero-profile cages, prior studies have shown higher subsidence rates with these implants compared to traditional constructs, without consistent clinical correlation [22].

Structural allografts showed the highest mean anterior and posterior subsidence, while PEEK cages had the lowest. In addition to continuous measurements, analysis of clinically significant subsidence (>2 mm and >3 mm) revealed significantly lower rates of both anterior and posterior subsidence with PEEK cages compared to structural allografts. These differences were confirmed by post hoc pairwise testing. However, no significant differences were observed between PEEK and titanium or ceramic cages. These findings align with the biomechanical principle that the lower modulus of elasticity of structural allografts compared to synthetic materials results in increased risk of implant settling into adjacent vertebral bodies [15].

Titanium cages showed intermediate results in terms of subsidence, with lower anterior subsidence than allografts (though not statistically significant on multivariate regression) and significantly reduced posterior subsidence compared to structural allografts (Table 3). Titanium cages are known for their high strength and biocompatibility, but have yielded mixed results regarding subsidence in the literature [27]. A meta-analysis by Seaman et al. [18] found similar fusion rates between PEEK and titanium cages but higher subsidence rates with titanium cages. However, Milczynska et al. [28] reported lower subsidence rates with titanium cages. Given titanium’s significantly higher modulus of elasticity, the literature should consistently show lower subsidence rates, but discrepancies highlight the need for further research to elucidate the optimal conditions under which titanium cages can be used to minimize subsidence.

Ceramic cages, although less frequently used, were associated with an intermediate risk of subsidence. Notably, anterior subsidence was relatively low, and posterior subsidence was comparable to that of titanium spacers, although multivariable analysis did not reveal significant differences compared to structural allografts (Table 3). The inherent brittleness and high modulus of elasticity of ceramic materials pose a risk for subsidence. Nonetheless, their biocompatibility and osteoconductive properties render them a viable option in certain clinical scenarios. Further research is warranted to elucidate the long-term outcomes associated with ceramic cages and their potential role in minimizing subsidence.

Among the synthetic options evaluated, PEEK cages exhibited the lowest anterior and posterior subsidence, with significantly less subsidence than structural allografts in multivariable analysis (Table 3). After controlling for potential confounders, PEEK was associated with reduced anterior and posterior subsidence. These findings are consistent with previous studies suggesting that PEEK cages provide a more favorable biomechanical environment for fusion, reducing the risk of subsidence. Balakumar et al. [29] reported PEEK cage subsidence rates ranging from 15% to 33.8%, noting that this did not consistently affect fusion rates or clinical outcomes. Our results support the notion that PEEK cages may offer a balance between mechanical stability and compatibility with the vertebral body, leading to satisfactory long-term outcomes despite some degree of subsidence.

Since the primary goal of ACDF is spinal fusion, implant materials must also support robust bony ingrowth. While PEEK’s modulus of elasticity more closely approximates that of bone, providing a mechanical advantage, its hydrophobic nature and lack of surface bioactivity may limit osseointegration, potentially contributing to higher rates of nonunion or pseudoarthrosis. Despite this, PEEK remains a widely used interbody material in ACDF due to its favorable properties, including radiolucency, lower modulus, and a favorable complication profile. Recent large-scale meta-analyses have reported lower overall complication rates with PEEK cages compared to titanium and structural allografts, as well as significantly lower subsidence rates compared to structural allografts, supporting their continued popularity [30].

Some limitations of this study should be acknowledged. As a retrospective analysis from a single high-volume academic institution, our findings may not be generalizable to other populations due to potential differences in patient demographics, comorbidities, surgical techniques, and instrumentation across institutions that could impact subsidence rates and outcomes. Additionally, bone mineral density data, obtained qualitatively from medical records, lacked quantitative measurements, such as Hounsfield units, which may have masked differences in bone mineral density between cohorts. Individual surgeon technique, particularly aggressive endplate preparation and potential subchondral bone violation, can also significantly impact subsidence; however, this factor cannot be accounted for in a retrospective review. Moreover, fusion status was not assessed, limiting our ability to evaluate pseudarthrosis as a contributor to subsidence. Determining solid fusion would have required standardized postoperative flexion-extension radiographs or thin-slice computed tomography imaging, which were not consistently obtained across this retrospective cohort. Pseudarthrosis can contribute to implant settling, often asymptomatically, and may be underdiagnosed without uniform imaging.

Despite efforts to minimize radiographic measurement error, consistency in measurement remains a challenge, particularly due to differences in spacer opacity. To address this, we employed standardized measurement protocols, single-observer data collection with review by a senior orthopaedic surgeon, and an interobserver reliability analysis. A second researcher remeasured a subset of cases, showing good-to-excellent agreement for ADH, PDH, and subsidence, supporting the reproducibility of our measurements.

Another major limitation was the limited sample size, as only a small subset of patients met the imaging criteria for inclusion, resulting in relatively small cohorts. Although the structural allograft and synthetic cage cohorts were almost evenly split, the synthetic cage subtypes were unevenly distributed, reflecting the real-world usage at our institution and strict radiographic follow-up criteria. Despite this limitation, several key comparisons reached statistical significance, underscoring the robustness of the observed effects. The median follow-up period of 12 months may not fully capture long-term subsidence and its clinical implications. Additionally, the lack of consistently available patient-reported outcome measures (PROMs) limits the ability to correlate radiographic findings with clinical outcomes. Future prospective studies should incorporate PROMs, extend follow-up durations, and utilize multicenter data to improve generalizability and assess the functional impact of cage selection.

Finally, while PEEK cages showed the lowest subsidence rates, the biomechanical explanation likely involves factors beyond material composition, such as cage geometry and the surface area in contact with the vertebral endplates that may influence the distribution of axial loads across the fused segment. Recent studies suggest that cage surface area and cage-to-body contact ratio may be equally important determinants in reducing stress concentrations, and, in turn, subsidence risk [21]. Future research should assess cage dimensions alongside material composition and fusion success to better understand the clinical implications of implant selection.

Conclusions

Our study underscores the importance of interbody spacer selection in anterior cervical fusion procedures. PEEK cages were found to be associated with reduced subsidence and better preservation of disk height and lordosis, while structural allografts showed the highest risk of subsidence. Further research with longer follow-up and larger patient populations is required to understand the long-term clinical implications of these findings, including the potential for adjacent segment disease and need for revision surgery.

Article information Continued