This retrospective study was approved by the Institutional Committee for Clinical Studies. We retrospectively reviewed CSM patients with cervical kyphosis at our institution from May 2019 to April 2021. A total of 130 patients (63 men, 67 women) underwent two-level ACDF with either stand-alone cages (SA group, n = 64) or plate-cage systems (PC group, n = 66). Group allocation was based on surgeon preference and implant availability after ≥3 months of failed conservative treatment. To better appreciate the effects, we selected only patients who had undergone fusion surgery of two segments which included the apex of kyphosis between C3-6. Inclusion and exclusion criteria are listed in Table 1. Sixty-four patients underwent ACDF with stand-alone spacer (SA group) and 66 with plate cage system (PC group). Patient demographics at baseline are presented in Table 2. We quantified bone mineral density via dual-energy x-ray absorptiometry (DXA), with osteoporosis diagnosed when the lowest T-score among lumbar spine (L1-4), femoral neck, and proximal femur measurements reached ≤ −2.5 standard deviations (12).

All procedures were performed by a senior spinal surgeon with over 15 years of experience in spine surgery. Under general anesthesia, patients underwent a standardized anterior cervical discectomy and fusion (ACDF) via the Smith-Robinson approach. A right-sided anterolateral incision was utilized to expose the pathological disc levels. Following meticulous localization using intraoperative fluoroscopy, complete discectomy and neurological decompression were achieved. Cartilaginous endplates were meticulously decorticated using a high-speed pneumatic drill, while subchondral bone integrity was preserved to prevent graft subsidence. Osteophyte resection and posterior longitudinal ligament release were selectively performed when required for adequate cord decompression.

In the SA group, stand-alone cages with integrated screw fixation (Zero-P, DePuy Synthes) were implanted, whereas the PC group received polyetheretherketone (PEEK) interbody cages (Skyline, DePuy Synthes) combined with anterior cervical plating systems (Bengal, DePuy Synthes). All cages were pre-packed with allograft bone grafting material prior to insertion. Full ambulation was initiated on postoperative day 2, with external immobilization maintained using a rigid cervical collar for 4–6 weeks.

Operation time, blood loss, complications and length of hospital stay were collected perioperatively. The degree of dysphagia was evaluated by Bazaz et al. grading system (13) at 48 h postoperatively and at the 3-, 6-month follow-up. The Neck Disability Index (NDI) score (ranging from 0 to 50) was used as a primary outcome variable (14) and health-related outcomes were evaluated using the Visual Analogue Scale (VAS) for arm pain (VAS-arm) and for neck pain (VAS-neck), the Japanese Orthopedic Association (JOA) score (ranging from 1 to 17), and the EuroQol Five-Dimension Scale (EQ-5D). Assessments were conducted preoperatively, as well as 3 month, 6 months, 1 year, and 2 years postoperatively. The recovery rate (RR) of the JOA score was calculated by the following formula: [RR = (postoperative−preoperative JOA score)/(17−preoperative JOA score) × 100%] (15).

Cervical kyphosis was defined as cervical lordosis (C2-7 Cobb) <0° in C-spine lateral neutral x-ray and was not the compensation of global spine imbalance confirmed by whole spine x-ray (8, 11, 16). In this study, radiographs were taken preoperatively and during each follow-up session. Specifically, plain lateral radiographs in neutral, flexion, and extension positions were obtained for comprehensive assessment. Fusion rate, cage subsidence, and ASD were systematically assessed at the 2-year postoperative follow-up. On dynamic radiographs, the radiographic fusion criteria consisted of interspinous motion (ISM) < 1 mm and superjacent ISM ≥4 mm (17). Cage Subsidence was generally defined as ≥ 3 mm loss of height comparing the direct postoperative intervertebral height with the intervertebral height at the last follow-up moment (18, 19). Adjacent segment disease (ASD) was characterized by degenerative changes observable on x-rays, encompassing: (1) formation or increased anterior osteophytes; (2) new or increased disc space narrowing (>30%); (3) new or increased anterior longitudinal ligament calcification; (4) growth of radial osteophytes (20).

Lateral x-ray films were used to measure the sagittal alignment of the cervical spine. Cervical lordosis (C2-C7, CL) and fusion segmental lordosis (FSL) were measured in the lateral radiograph of neutral position (Cobb method). The fused segment disc wedge was calculated by adding up the Cobb angles of the three operated discs. C2 sagittal vertical axis (C2 SVA), which can be defined regionally using the distance between a plumb line dropped from the centroid of C2 and the posterosuperior aspect of C7, was obtained at the same time (Figure 1). Those radiographic metrics were repeated 3 times, and the results were averaged. To achieve blinding, identifying information on radiographs was removed, and they were randomly assigned to examiners, each with at least 3 years of clinical experience in spinal surgery.

The sample size was calculated with the Neck Disability Index (NDI) as the primary outcome variable in a superiority design. The standard deviation (SD) of NDI was set at 18, determined from our institutional data and relevant literature (21, 22). The minimal detectable difference (effect size) was set at 10. With a significance level of 0.05 and a power of 80%, the initial calculation indicated that 51 patients were needed per group. However, considering potential loss of follow - up, possible noncompliance of patients, and the complexity of surgical procedures that might introduce additional variability, we adjusted the sample size. In this study, a total of 130 patients were included, which is expected to provide sufficient statistical power to accurately assess the differences in NDI scores between the plate cage and stand—alone cage groups during anterior cervical discectomy and fusion (ACDF) for the treatment of cervical spondylosis.

To ensure the reliability and validity of the study results, a comprehensive statistical analysis was conducted on the data obtained from both groups. Descriptive statistics were used to analyze data in both groups. For group - to - group comparisons, an independent t test was applied to continuous variables meeting normality and variance—homogeneity assumptions, the chi—square test was used for categorical variables, and the Mann - Whitney U test or Fisher exact test was chosen when parametric test assumptions were violated.

To analyze ordinal data within groups, the Wilcoxon test and Friedman analysis of variance were employed. For between—group repeated—measurement data, the Kruskal - Wallis analysis of variance was used. A p-value < 0.05 was regarded as statistically significant. In subgroup analyses, the Bonferroni correction was used to adjust the significance level. SPSS 22.0 software (IBM, Armonk, New York, USA) was used for statistical analysis.

Operative data are summarized in Table 3. While no significant difference was observed in the length of hospital stay or surgical level between the SA and PC groups, the SA group demonstrated significantly shorter operative time (78.0 ± 11.5 vs. 86.5 ± 12.5 min, p < 0.001) and reduced intraoperative blood loss (193.0 ± 85.6 vs. 256.7 ± 110.3 mL, p < 0.001) compared to the PC group.

The complications of these patients are summarized in Table 4. In the SA group, postoperative complications included dysphagia (8 cases), subcutaneous hematoma (1 case), wound infection (1 cases) and hoarseness (3 case). In the PC group, postoperative complications included dysphagia (20 cases), cerebral fluid leakage (1 case), epidural hematoma (1 cases), subcutaneous hematoma (2 case), wound infection (1 cases), Horner syndrome (1 case) and hoarseness (3 cases). There was no implant failure or new neurological deficit in both groups.

The incidence of most postoperative complications did not differ significantly between the two groups. However, the PC group demonstrated a statistically higher rate of dysphagia compared to the SA group (p = 0.014). There were 8 patients developed mild dysphagia in the SA group and 20 patients developed dysphagia (12 mild, 8 moderate) in the PC group 48 h after surgery. At 3-month follow-up, 6 patients in the SA group showed mild dysphagia and only 1 at 6-month follow-up, however 15 patients showed dysphagia (9 mild, 6 moderate) in the PC group at 3-month follow-up, and the mild symptom still present in 8 patients at 6-month follow-up. Overall, the PC group exhibited a significantly higher incidence of postoperative dysphagia than the SA group throughout the follow-up period.

There were no significant differences between 2 groups preoperatively in NDI, VAS-arm, JOA, or EQ-5D score (p > 0.05). All outcome measures improved after surgery, regardless of the treatment strategy (Table 5). The NDI decreased significantly from 18.2 ± 1.8 points at baseline to 9.4 ± 0.8 and 9.3 ± 1.1 points 3-month and 2-year postoperative, respectively, in the SA group, from 17.8 ± 1.4 points to 9.7 ± 0.9 and 9.6 ± 1.2 points in the PC group, and there was no significant difference in the absolute value of NDI, nor in the decline of NDI among theme (Figure 2A). Likewise, secondary parameters, including JOA, VAS-arm, EQ-5D, also significantly improved after surgery and showed a comparable result in both groups (Figures 2B–D). At the 2-year follow-up, the SA group showed a more pronounced VAS-neck compared with the PC group (3.9 ± 1.0 and 2.1 ± 0.8, respectively), which, however, failed reach the minimal clinically important difference (MCID) of 2.6 points on the 10-point VAS scale (23) (Figure 2E).

Radiographic outcomes are detailed in Table 6. At the 2-year follow-up, fusion rates were comparable between the SA and PC groups (93.8% [60/64] vs. 95.4% [63/66], p = 0.667). Adjacent segment disease (ASD) occurred in 8 patients (12.3%) in the SA group and 7 patients (10.6%) in the PC group (p = 0.735), though none required revision surgery. Furthermore, cage subsidence was more frequently observed in the SA group than in the PC group (14.1% [9/64] vs. 9.1% [6/66]); however, no statistically significant intergroup difference was observed (p = 0.289).

Preoperative cervical sagittal alignment parameters—including CL, FSL, fused segment disc wedge, and C2 SVA—showed no significant differences between the SA and PC groups (p = 0.364, 0.528, 0.130, and 0.593, respectively; Table 6). Postoperatively, both groups achieved significant correction of sagittal alignment, with comparable improvements between cohorts (p > 0.05). However, at final follow-up, the PC group maintained superior cervical alignment: CL (13.6° ± 0.9° vs. 12.3° ± 2.4°; p < 0.001), FSL (13.2° ± 1.2° vs. 11.9° ± 2.9°; p = 0.001), and fused segment disc wedge (10.2 ± 1.3° vs. 8.9° ± 2.6°; p < 0.001). These findings demonstrate significantly greater preservation of sagittal correction in the PC group. No intergroup difference in postoperative C2 SVA was observed (p = 0.233).

A comparative analysis of radiographic outcomes between patients with osteoporosis (OP) and non-osteoporotic (non-OP) individuals revealed notable disparities. Non-OP patients exhibited a marked improvement in FSL, progressing from a baseline mean of −8.0° ± 0.9° to 12.8° ± 2.0° at the 2-year follow-up. In contrast, OP patients showed less pronounced gains, with FSL increasing from −8.3° ± 0.9° to 12.0° −8.3° ± 2.7°. Statistically significant difference in FSL (p = 0.044) was observed between OP and non-OP cohorts at final follow-up, underscoring the adverse impact of osteoporosis on postoperative alignment restoration.

Furthermore, preoperative stratification based on reducibility of fused segment lordosis yielded distinct postoperative outcomes (Table 7). In patients with non-reducible kyphosis (defined by persistent segmental malalignment on extension radiographs), the SA group demonstrated significantly lower cervical lordosis (CL), fusion segmental lordosis (FSL), and fused segment disc wedge compared to the PC group (p < 0.05 for all parameters). Conversely, in reducible cases (segmental lordosis restored on preoperative extension imaging), no significant differences in cervical sagittal alignment were observed between SA and PC groups (p > 0.05). These results highlight the critical influence of preoperative reducibility on the efficacy of SA in maintaining corrective alignment, particularly in biomechanically challenging non-reducible deformities.

This study has several limitations. First, the single-center retrospective design and relatively small sample size (n = 130) may restrict the statistical power and generalizability of the findings. Second, the mid-term follow-up period (median 28.5 months) may be insufficient to fully capture long-term complications, such as ASD or progressive alignment loss. Third, the absence of whole-spine radiographs limited analysis of global sagittal alignment parameters (e.g., thoracic kyphosis, lumbar lordosis), potentially confounding cervical alignment interpretation. Fourth, fusion rate was assessed via dynamic radiographs rather than CT scans—the diagnostic gold standard—which may overestimate fusion success. Fifth, bone mineral density (BMD) was evaluated using DXA based on the lowest T-score of the lumbar spine, femoral neck, and proximal femur, whereas cervical vertebral Hounsfield units (HU) measured by CT scans—a more reliable predictor of cage subsidence—were not analyzed (40). Sixth, although subgroup analyses were performed, unmeasured confounding factors may still influence outcomes, and additional stratification based on the magnitude of kyphosis may provide further insights and should be considered in future studies. Finally, the retrospective design inherently introduces potential selection bias, as construct selection was influenced by surgeon preference and implant availability. This may limit causal inference and affect the generalizability of the findings. A prospective multicenter trial with extended follow-up and advanced imaging protocols (e.g., CT-based BMD, whole-spine radiographs) is warranted to validate these findings.