This study was conducted in compliance with the principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Institutional Review Board (IRB) of the Ethics Committee of the Second Xiangya Hospital of Central South University (IRB no., 2023-043-XYEY). Written informed consent was obtained.

This study was approved by the local ethics committee, and informed consent was obtained at the final follow-up. The study was a retrospective analysis of 62 patients who underwent three-level ACDF at a single institution between January 2014 and December 2019. The patients were divided into two groups based on the surgical technique used: the PC group, which used a cage-with-plate system, and the SC group, which used self-locking cages. The inclusion criteria for the enrolled patients were as follows: (1) diagnosis of cervical radiculopathy or myelopathy, (2) spinal cord ventral compression primarily due to three-level cervical disc herniation on magnetic resonance imaging (MRI), and (3) unsatisfactory response to conservative treatment for at least 8 weeks. The exclusion criteria were as follows: (1) use of hybrid surgical techniques, (2) history of spine surgery, (3) severe osteoporosis, and (4) presence of specific conditions such as fracture, tumor, infection, ankylosing spondylitis, developmental stenosis, ossification of the posterior longitudinal ligament, diffuse idiopathic skeletal hyperostosis, or significant segmental instability.

During each surgery, neuroelectrophysiological monitoring, including evoked potential and electromyographic monitoring, was performed. Under general anesthesia, horizontal incisions were made, and all surgical procedures were performed by the same senior surgeon (L.K.) using a standard Smith-Robinson approach [6]. In the first 10 cases, double transverse incisions (each 4–6 cm) were adopted, while all subsequent patients were treated with a single transverse incision (approximately 8–10 cm). The disc and cartilaginous endplates were carefully removed to avoid excessive damage to the bony endplate. Residual posterior osteophytes and hypertrophic uncovertebral joints were meticulously removed using curettes and Kerrison rongeurs for neural foramen decompression. The foraminal space was visually inspected and palpated using a blunt nerve hook to confirm adequate decompression. In cases with severe foraminal stenosis or lateral disc herniation compressing the nerve root (based on preoperative MRI and computed tomography [CT] assessment), targeted uncoforaminotomy was performed to achieve direct nerve root decompression. Complete decompression of the spinal cord and nerve roots was achieved, and the ideal cage size (ROI-C or ROI-MC+; LDR, SainteSavine, France) was selected by radiographic-assisted trials. The cage width was determined by the distance between the two Luschka joints, and the cage height was determined by radiographic trials to ensure a tight fit without over-distracting the disc space or facet joints. All patients received allogeneic bone (Osteo-link; Hubei Osteolink Biomaterial Co. Ltd., Wuhan, China) to fill the cage. In the SC group, self-retaining anchoring clips were inserted along the axis of the disc using the implant holder following cage insertion. These clips were then passed through the cage and impacted into the vertebrae. In the PC group, an anterior plate (ATLANTIS; Medtronic Sofamor Danek, Memphis, TN, USA) was fixed in front of the vertebral body of the fusion segment after cage insertion. The ACDFs were performed level by level, starting with the most prominent segment of compression, followed by the lower and upper segments. The wound was closed with absorbable undyed sutures (SXMD2B406; Johnsons & Johnsons, Shanghai, China), and a 15-mm-diameter silicone drain (Ande; Shandong Ande Healthcare Apparatus Co. Ltd, Zibo, China) was placed postoperatively, typically removed when drainage was less than 30 mL per day. The patients were advised to wear a neck brace for 4–6 weeks.

The clinical outcomes, including the Visual Analog Scale (VAS), modified Japanese Orthopedic Association (mJOA) scores, and Neck Disability Index (NDI), were assessed independently by two evaluators. As the study aimed to assess mid-to-long-term results, statistical analysis was only conducted on preoperative, 2-day postoperative, and 2- and 5-year follow-up data. Any disagreements between the residents were resolved through discussion with another independent expert.

Complications such as dysphagia, hematoma, hoarseness, infection, and implant-related complications were also documented. Dysphagia was specifically assessed at 2 days, 3 months, 6 months, and 12 months after surgery using the Bazaz et al. [4] grading system, which categorizes dysphagia as none (no swallowing difficulties), mild (infrequent episodes), moderate (occasional difficulty with specific foods), or severe (frequent difficulty with most foods).

Radiological evaluation was conducted using standard anteroposterior, lateral, and flexion-extension radiographs at various time points: preoperatively; 2 days postoperatively; 1-, 2-, and 5-year postoperatively; and during the final follow-up.

To determine CL, Cobb’s angle was measured between the lower endplate of C2 and the lower endplate of C7 on a lateral radiograph taken in a neutral position. The fusion segmental angle (FSA) was defined as the Cobb’s angle between the upper endplate of the cranial fusion segment and the lower endplate of the caudal fusion segment. DH was calculated as the average of the anterior and posterior DHs [7].

Fusion was confirmed on 150% magnified flexion-extension radiographs if the interspinous motion (ISM) was <1 mm and superjacent ISM was ≥4 mm [8]. Cage subsidence was diagnosed when a reduction of >3 mm in surgical DH was observed on follow-up radiographs compared with the immediate postoperative radiograph [9]. ASD was diagnosed radiographically based on the presence of anterior osteophytes, increased narrowing of the intervertebral space, and calcification of the anterior longitudinal ligament on cervical radiographs [10].

To minimize bias, both clinical and radiological evaluators were blinded to the follow-up time points and patient identity. Prior to assessment, clinical data and radiographs were randomly assigned unique identifiers to ensure that the evaluators were unaware of the time periods or patients associated with the data or images. This blinding process aimed to maintain objectivity in the evaluation of clinical outcomes and radiographic measurements.

Statistical analysis was conducted using IBM SPSS ver. 27.0 statistical software (IBM Corp., Armonk, NY, USA). Measurement data are expressed as mean and standard deviation, while count data are expressed as number of cases and percentages. The Shapiro-Wilk test was used to assess data normality. Independent t-tests were used to compare normally distributed continuous variables between the groups, while the Mann-Whitney U test was used for non-normally distributed data. Paired t-tests were used for within-group comparisons. Categorical variables were analyzed using chi-square or Fisher’s exact tests. Statistical significance was set at p<0.05.

Independent t-tests were used to compare symptom improvements (VAS, NDI, and mJOA) between the subsidence and non-subsidence groups from postoperative day 2 to the 5-year follow-up. Linear regression analyses were conducted for outcomes that exhibited significant differences between these groups to quantify the strength and direction of the association.

To evaluate the intraobserver and interobserver variability in the measurement of CL, FSA, and DH, the intraclass correlation coefficient (ICC) was calculated separately for the SC and PC groups. Reliability was assessed using the criteria by Shrout and Fleiss [11] for reliability testing (poor, ICC<0.40; fair to good, ICC 0.40 to 0.75; excellent, ICC>0.75). Higher ICC values indicated better reliability.

Thirty-two patients in the SC group and 30 patients in the PC group were followed up for 64.8±2.9 months and 65.6±3.0 months, respectively (total 62 patients: 35 males and 27 females). In the SC group, 19 patients underwent C3–6 fusion and 13 patients underwent C4–7 fusion. Similarly, in the PC group, 17 patients underwent C3–6 fusion and 13 patients underwent C4–7 fusion. No statistically significant differences were observed between the two groups regarding sex, age, body mass index (BMI), surgical level, or follow-up duration. Baseline patient characteristics are summarized in Table 1.

The operative time was significantly shorter in the SC group than in the PC group. Blood loss was significantly lower in the SC group than in the PC group. Regarding fusion rates, the SC group showed increasing fusion rates over time, reaching 100% at 5 years. The PC group also displayed a similar trend, with fusion rates reaching 100% at 5 years. There were no significant differences between the two groups at any follow-up point (Table 2).

The SC and PC groups exhibited significant improvements in neck pain VAS, arm pain VAS, mJOA scores, and NDI at all postoperative time points compared with the preoperative values. No significant differences were observed between the two groups at any time point. The specific data for each clinical outcome are presented in Table 2.

Regarding arm pain VAS, no statistically significant differences were observed between the 2-year postoperative values and the 2-day postoperative values in both groups, as well as between the 5-year and 2-year postoperative values. Regarding neck pain VAS scores, mJOA scores, and NDI, significant differences were observed between the 2-year and 2-day postoperative values. However, no significant differences were noted between the 5-year and 2-year postoperative values. The trends of these changes are shown in Fig. 1.

At the 5-year follow-up, compared to postoperative day 2, independent t-tests revealed no significant differences between the subsidence and non-subsidence groups regarding improvements in VAS scores for neck pain (p=0.202), VAS scores for arm pain (p=0.765), or mJOA (p=0.712). However, a significant difference was observed in NDI improvement between the two groups (p=0.045). Further linear regression analysis showed that subsidence had a significantly negative impact on NDI improvement, as indicated by both the unstandardized coefficient (B=−0.827; 95% confidence interval [CI], −1.62 to −0.03) and standardized coefficient (β=−0.256; 95% CI, −0.49 to −0.02). Nevertheless, with an adjusted R² value of 0.050, although there was a statistically significant relationship between subsidence and NDI improvement, subsidence alone accounted for only 5% of the variance in NDI improvement, indicating limited explanatory power.

CL, FSA, and DH were evaluated at multiple time points, including preoperatively, immediately postoperatively, and at 1-year, 2-year, and 5-year follow-ups.

Preoperatively, no significant differences in CL, FSA, or DH were observed between the SC and PC groups. Postoperatively, both groups showed significant improvements in CL, FSA, and DH. However, the PC group had better outcomes over time. At the 1-year follow-up, the PC group had superior DH preservation than the SC group. At the 2-year follow-up, the PC group showed significantly higher CL, better FSA maintenance, and superior DH preservation, with these trends continuing at the 5-year follow-up. The SC group experienced a more rapid decline in these outcomes over time. The specific values are listed in Table 1, and the trend of the changes is presented in Fig. 2.

In the SC group, the intraobserver ICC for CL, FSA, and DH were 0.822, 0.785, and 0.793, respectively. The interobserver ICCs for these measurements were similarly high, with values of 0.810, 0.772, and 0.781, respectively. In the PC group, the intraobserver ICCs for CL, FSA, and DH were 0.831, 0.804, and 0.814, respectively, with interobserver ICCs of 0.818, 0.792, and 0.803, respectively. These values indicate excellent intraobserver and interobserver reliability in both groups for all measurements (all ICC>0.75).

Spearman correlation analyses were performed to assess the relationship between the 5-year postoperative loss of CL (relative to postoperative day 2) and improvements in clinical outcomes, including VAS neck pain, VAS arm pain, mJOA scores, and NDI. The results showed no statistically significant correlations. To explore whether other postoperative alignment changes might be associated with clinical improvements, we analyzed the relationship between the loss of FSA from postoperative day 2 to the 5-year follow-up and symptom improvement measures. Similar to CL loss, none of these correlations were statistically significant, suggesting that the loss of CL and FSA from postoperative day 2 to 5 years was not significantly associated with any clinical outcome at the 5-year follow-up. The results of the Spearman correlation analyses are presented in Table 3.

The 95% CIs for our correlation and regression analyses provided a critical context for interpreting the radiographic-clinical relationships. The CIs for all Spearman correlations, including CL and FSA loss versus symptom improvement (e.g., CL loss vs. NDI: ρ=−0.025; 95% CI, −0.27 to 0.22; FSA loss vs. NDI: ρ=−0.154; 95% CI, −0.39 to 0.10), crossed zero, reinforcing the absence of statistically or clinically meaningful associations between alignment loss and long-term outcomes. Similarly, the regression model for subsidence versus NDI improvement, although nominally significant (B=−0.827; 95% CI, −1.62 to −0.03; standardized β=−0.256; 95% CI, −0.49 to −0.02), explained minimal variance (R²=0.05) and highlighted marginal clinical relevance. The proximity of CIs to zero across all analyses suggests that postoperative subsidence and alignment changes possibly represent radiographic phenomena with limited impact on patient-reported recovery. Nevertheless, the directional trend in subsidence-related disability warrants surveillance in high-risk cohorts, pending validation in larger studies with the power to detect subtle effects.

Reconstructed multiaxial CT is highly reliable for predicting pseudoarthrosis and is preferred for evaluating extragraft bone bridging owing to its superior diagnostic qualities [12]. However, assessing fusion status can be subjective, lacks specific parameters, and is limited to static moments, potentially missing dynamic cases of pseudoarthrosis. Song et al. [13] reported that using a cutoff value of ISM ≥1 mm on radiographs magnified by 150% and taken with ≥4 mm superjacent ISM had comparable accuracy to CT scans for detecting anterior cervical pseudarthrosis. In our study, some patients who appeared to have fusion on CT did not meet ISM fusion standards (Fig. 5). Furthermore, some patients declined to undergo CT because of its high cost and time requirements. Therefore, to ensure consistent evaluation standards, we adopted the ISM criterion for assessing pseudarthrosis instead of CT.

Our study found that both self-locking cages and the cage-with-plate system are highly effective in achieving fusion in multilevel ACDF, with all patients achieving 100% fusion by the final follow-up. This aligns with previous research, including a 5-year follow-up study by Sun et al. [14], which compared zero-profile spacers with the traditional cage-with-plate system in three-level ACDF and found a high fusion rate in both groups. Similarly, Zhu et al. [15] conducted a 3-year follow-up study on 62 patients with multilevel cervical spondylotic myelopathy undergoing three-level ACDF and found high fusion rates in both self-locking cage and cage-with-plate groups, with no significant difference between them. A retrospective study by Chen et al. [16] comparing zero-profile devices and the traditional cage-with-plate system in three-level ACDF also found high fusion rates in both groups, with no statistically significant difference. Overall, these studies indicate that both self-locking cages and the cage-with-plate system are effective promoters of fusion in multilevel ACDF.

The zero-profile anchored cage utilized in our study featured two anchoring clips that provided a fixation mechanism akin to that of a plate and screws. A systematic review of ACDF outcomes using various implants reported bone fusion rates of 88.6% in 5,738 patients treated with a stand-alone cage, 91.4% in 3,971 patients treated with a screw-plate system, and 96.6% in 499 patients treated with a zero-profile anchored cage [17]. Our 5-year follow-up study demonstrated a fusion rate of 100% with the zero-profile cage, indicating that its fusion outcomes were non-inferior to those achieved with the cage-with-plate system. This further supports the effectiveness of the zero-profile anchored cage in promoting fusion in ACDF procedures.

Subsidence is a crucial clinical concern in ACDF because it can lead to reduced intervertebral height and sagittal malalignment and result in unfavorable clinical outcomes and adjacent segmental disease [18]. A systematic review of 4,784 patients using single- and multilevel cages reported a mean subsidence rate of 21.1% (range, 0 to 83%) [18]. A meta-analysis of nine observational studies found that in two-level ACDF, zero-profile spacers significantly increased the risk of subsidence compared to plate and cage constructs, with follow-up durations of 6–36 months [19]. Consistent with these findings, our study reported a significantly higher subsidence rate in the SC group (16.7%) than in the PC group (5.6%) at the 5-year follow-up. Our previous study and a prospective study by Igarashi et al. [20] reported that subsidence primarily occurs in the early postoperative period, possibly due to substantial pressure exerted on the interior endplate during patient mobilization [7]. Finite element modeling and simulation by Galbusera et al. [21] highlighted that anterior plate constructs provide substantial segmental stability during the early postoperative period, facilitating effective load redistribution and reducing endplate contact pressures, thereby mitigating the risk of cage subsidence. Our finding of no new occurrences of subsidence during the 2–5 years postoperative period further supports this perspective.

The endplate may contribute to subsidence. In the SC group, the insertion of anchored clips through the cage may have caused partial endplate damage, which could have increased the incidence of subsidence. Lowe et al. [22] found that vertebral bodies with intact endplates had a significantly higher ultimate compressive strength than those with resected endplates. Moreover, endplate injury significantly increased the risk of subsidence and displacement of interbody fusion devices. Therefore, it is important to carefully handle the endplate during surgical procedures to minimize injury and reduce the incidence of subsidence.

The dimensions of the cage, including its height and size, are also associated with subsidence. Larger grafts have been reported to experience higher distractive and compressive forces, which may increase the risk of subsidence. Specifically, cages with a height of 6.5 mm or 7.5 mm have a higher risk of subsidence compared to those with a height of 4.5 mm or 5.5 mm, due to the increased stress on the vertebral endplates [23]. Similarly, cages with a smaller anteroposterior diameter, such as 12 mm, have been associated with a higher risk of subsidence than wider cages, such as 14 mm, owing to the smaller surface area distributing pressure on the endplate [24]. Based on these findings, it is recommended to minimize over-distraction during surgery and use the widest possible cage to reduce the risk of subsidence.

Zhu et al. [15] attributed lower VAS scores to the occurrence of subsidence, whereas Lee et al. [25] reported poor clinical outcomes in groups with high rates of subsidence. However, our study did not find any statistically significant differences in clinical outcomes (VAS, mJOA scores, and NDI) between the SC group with a high subsidence rate and the PC group with a low subsidence rate. This finding is consistent with those of numerous previous studies [14,16,19]. To further investigate the relationship between subsidence and clinical outcomes, we categorized the patients into two groups: those with subsidence and those without. Although a significant difference in NDI was detected between the two groups, linear regression analysis showed that subsidence accounted for only 5% of the variation in NDI. Therefore, we propose that subsidence is a multifactorial radiographic phenomenon that may not necessarily have a significant effect on the clinical outcomes.

The lack of clinical deterioration despite cage subsidence can be explained by two interrelated mechanisms. First, the direct surgical management of foraminal stenosis through osteophyte removal and selective uncoforaminotomy achieved sufficient decompression, creating a protective buffer that prevented recurrent neural compression, even when subsidence occurred. Second, neurological recovery (evidenced by early postoperative improvements in the mJOA and NDI scores) was largely complete before significant subsidence developed, creating a temporal dissociation where radiographic changes were no longer associated with symptom progression. This sequence of events, where maximal neurological recovery precedes structural compromise, explains why subsidence did not translate to worsened clinical outcomes.

A crucial aspect of anterior cervical spinal surgery is the restoration and maintenance of CL. In our study, both CL and FSA improved in both groups in each follow-up period compared to pre-surgery measurements. However, the SC group exhibited a more pronounced decline in CL and FSA during the last two follow-ups. We concur that the presence of an anterior cervical plate facilitates load redistribution through the plate, alleviating contact stress at the graft-bone interface and aiding in preserving CL and FSA loss while preventing cage subsidence during the fusion process [15]. This observation is consistent with numerous previous studies, including a systematic review and meta-analysis by Cheung et al. [26], which found that cage-plate constructs outperform stand-alone cages in restoring and maintaining CL owing to the additional support provided by the anterior plate, thereby enhancing segmental stability.

Although the SC group showed inferior cervical curvature maintenance compared to the PC group in our study, no significant differences in clinical outcomes were observed between the two groups at any follow-up period. Both groups demonstrated significant improvements compared to preoperative levels, suggesting that cervical curvature loss is not correlated with clinical symptoms. This finding aligns with previous studies by Spanos et al. [27], Godlewski et al. [28], and Meng et al. [29], who reported that changes in cervical sagittal alignment and lordosis following ACDF had no significant impact on symptom recovery or clinical outcomes. To further investigate this correlation, Spearman correlation analyses showed no significant correlation between the loss of CL and FSA and the improvement in mJOA, VAS, or NDI in all patients at the last follow-up compared with immediately after surgery. These findings suggest that the manifestation of clinical consequences resulting from CL loss may require several years to develop and may not be evident within relatively short-term follow-up periods.

Postoperative dysphagia is a common complication of ACDF. In our study, we observed a significantly lower incidence of dysphagia immediately after surgery in the SC group (6.25%) than in the PC group (50%). This finding aligns with a systematic review by Cho et al. [30], which suggested that the use of anterior locking plates is associated with higher rates of postoperative dysphagia. The potential mechanisms include mechanical impingement on the esophagus, inflammation, and prevertebral swelling caused by the plates. Additionally, Gowd et al. [31] reported that postoperative dysphagia after ACDF is predominantly caused by mechanical compression of laryngeal structures rather than by direct injury to the recurrent laryngeal nerve. Increased endotracheal cuff pressure, retractor placement during surgical procedures, advanced age, previous history of similar dysfunction, prolonged surgical procedures, and multilevel surgery were identified as risk factors for dysphagia. In our study, prolonged surgery in the PC group (129.1±9.58 minutes vs. 100.28±11.99 minutes, p<0.001) and the resulting intraoperative compression could have contributed to the higher incidence of dysphagia observed in this group compared to the SC group.

The surgical principles of anterior cervical plating, including minimizing plate length and optimizing screw trajectory, are established strategies for reducing adjacent segment stress and subsequent degeneration. As highlighted in a previous systematic review and meta-analysis [26], the use of excessively long plates or suboptimal screw placement (e.g., excessively medial or parallel angulation) may concentrate biomechanical stress at transitional segments, accelerating disc degeneration and osteophyte formation. In our cohort, the PC group demonstrated a significantly higher ASD incidence (13.3% vs. 1.56% in the SC group), aligning with these mechanistic observations. Although our surgical protocol adhered to standard techniques (e.g., selecting the shortest plate spanning only the fused levels and positioning screws obliquely near the vertebral margins), the inherent rigidity of plate fixation itself, even when technically optimized, appears insufficient to fully mitigate the risk of ASD in multilevel constructs. This contrasts sharply with the lower ASD rates in the SC group, likely attributable to the absence of anterior hardware-induced stress concentration. Thus, our findings reinforce the importance of implant selection in balancing segmental stability against adjacent segment preservation, particularly in multilevel procedures in which cumulative biomechanical impacts are magnified. This study does not directly evaluate novel plate configurations, but it suggests critical long-term evidence that zero-profile systems have inherent advantages regarding ASD prevention without compromising fusion efficacy, a consideration increasingly relevant in contemporary practice.

This study has several limitations. First, its retrospective design introduces inherent risks of selection and confounding biases, which may influence the validity of our conclusions. Second, the single-center nature of the data limits the generalizability of our findings to broader populations with diverse demographics or surgical practices. Most critically, the modest sample size (N=62) significantly restricted the statistical power. Post-hoc analyses using G*Power ver. 3.1 (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany; http://www.gpower.hhu.de/) revealed that for Spearman correlations, our cohort achieved only 58% power to detect a medium effect (ρ=0.30) and <35% power for small effects (ρ=0.15–0.20) at α=0.05. Similarly, the observed regression model linking subsidence to NDI improvement (R²=0.05) would require N>200 to achieve 80% power, leaving our analysis susceptible to type II errors (i.e., failing to detect true associations). Finally, while the 5-year follow-up duration is a strength, even longer-term data (>10 years) may be necessary to fully evaluate the durability of cervical alignment maintenance and ASD. Subgroup analyses (e.g., age, sex, BMI) were infeasible due to the limited sample size, potentially obscuring nuanced differences in outcomes. These limitations underscore the need for cautious interpretation of non-significant correlations and small effect sizes. External validation in a larger cohort or a meta-analysis combining similar single- and multilevel ACDF studies seems necessary to more definitively characterize the clinical relevance of subsidence and alignment changes.