Fusion Assessment, Reoperation Rates, and Imaging Modality Agreement after Four-Level Anterior Cervical Discectomy and Fusion: A Retrospective Single-Center Study Authors: Bruno Verna, Pedro Rocha Torres, Tom Folkerts, Julia Wimmer, Anna-Maria Mielke, Steven M Maurer, Jiaqi Zhu, Jennifer Shue, Marco D Burkhard, Andrew A Sama, Frank P Cammisa, Alexander P Hughes, Federico P Girardi Published: 2026 ABSTRACT AbstractStudy DesignRetrospective single-center cohort study.ObjectiveTo evaluate fusion and reoperation rates after four-level anterior cervical discectomy and fusion (ACDF) using a hierarchical imaging strategy, and to assess agreement between imaging modalities.MethodsPatients undergoing primary four-level ACDF with postoperative imaging for fusion assessment and ≥2-year follow-up were included. Fusion status by ≥12 months was determined using a hierarchical imaging strategy, with computed tomography (CT) prioritized, followed by dynamic flexion-extension (F-E) radiographs and plain radiographs. Early fusion (6-12 months) was analyzed separately. Reoperation within 24 months was recorded. Agreement between imaging modalities was assessed using Fleiss’ kappa for tri-modality comparison and Cohen’s kappa for pairwise analyses.ResultA total of 152 patients had imaging available for fusion assessment; 140 had ≥12-month follow-up. Fusion by 12 months was observed in 81.4% (114/140). Early fusion occurred in 70.1% (101/144). Nonunion most commonly involved a single level, predominantly at C6–7. Within 24 months, reoperation occurred in 7.7% of patients with nonunion and 2.6% of fused patients. In a tri-modality subset, overall agreement was high (κ=0.823), with near-perfect agreement between CT and F-E radiographs (κ=0.938).ConclusionFour-level ACDF demonstrated high fusion rates and low reoperation rates in one of the largest single-center cohorts reported to date. CT and F-E radiographs showed strong agreement, while plain radiographs tended to overestimate fusion. These findings suggest that acceptable outcomes may be achieved in appropriately selected patients and that construct length alone should not preclude consideration of an anterior-only approach.Keywords: anterior cervical discectomy and fusion, multilevel cervical spine surgery, fusion rate, imaging modality agreement, radiographic assessment Introduction Anterior cervical discectomy and fusion (ACDF) is a well-established surgical procedure for the treatment of degenerative cervical spine disease, including cervical radiculopathy and myelopathy. ACDF not only provides decompression of neural structures but also stabilizes affected motion segments.1,2 Outcomes following single- and two-level ACDF have been extensively studied and consistently demonstrate high fusion rates, reliable symptom improvement, and favorable safety profiles.3-5 As a result, ACDF remains one of the most widely performed surgical procedures for the treatment of degenerative cervical spine pathology.5-7 However, the role of multilevel ACDF, particularly constructs involving three or more levels, remains more controversial. As the number of fused segments increases, concerns regarding pseudarthrosis, construct failure, and the potential need for revision surgery become more prominent, with symptomatic nonunion representing a major cause of reoperation in these constructs.8-13 Several clinical series have reported variable fusion and nonunion rates with longer constructs, leading some surgeons to favor posterior approaches in patients requiring extensive multilevel decompression.14-16 Nevertheless, when performed in appropriately selected patients and with meticulous proper surgical technique, multilevel ACDF may still provide effective decompression, restoration of cervical alignment, and favorable clinical outcomes.8,17,18 Interpretation of the existing literature on fusion following four-level ACDF is further complicated by substantial variability in radiographic fusion assessment. Studies have used a range of imaging modalities, including plain radiographs, dynamic flexion-extension (F-E) radiographs, and computed tomography (CT), often applying different fusion assessment criteria across modalities.14,19,20 Because these approaches differ in their evaluation of radiographic fusion, reported fusion rates vary across studies. Furthermore, many published series include relatively small patient cohorts, limiting the ability to draw consistent conclusions regarding true fusion rates and revision risk.10,19,21 A recent systematic review of four-level ACDF studies highlighted the heterogeneity in reported outcomes and imaging assessment methods across the literature. 20 In addition, variability in surgical technique and surgeon experience across studies further contribute to differences in outcomes, particularly in long-segment constructs, where technical factors such as endplate preparation, graft placement, and fixation can have a cumulative effect. Fusion rates are also influenced by differences in the timing of postoperative imaging assessments. Although radiographic fusion is commonly evaluated at one year following ACDF, follow-up intervals vary across studies, which also contribute to differences in reported fusion outcomes8,21,22 Given these limitations, further investigation is needed to better characterize radiographic fusion following four-level ACDF using clearly defined imaging strategies. The purpose of the present study was therefore to evaluate fusion outcomes following four-level ACDF in a large single-center cohort using a hierarchical imaging assessment strategy incorporating CT, dynamic F-E radiographs, and plain radiographs. Agreement between imaging modalities was evaluated, and reoperation outcomes were analyzed separately according to fusion status. Methods Study Design and Setting This retrospective cohort study evaluated patients who underwent four-level ACDF at a single tertiary spine center. All procedures were performed by four fellowship-trained spine surgeons. Institutional review board approval (#2020-1877) was obtained prior to study initiation. Patient Selection All cervical spine operations performed by the four participating surgeons between 2016 and 2023 were reviewed to identify patients who underwent anterior cervical surgery. Cases involving posterior cervical procedures, circumferential cervical fusion, corpectomy, cervical disc replacement, or hybrid cervical procedures were excluded. Patients undergoing four-level ACDF for degenerative cervical spine disease were then identified. Additional exclusion criteria included age younger than 18 years, nondegenerative surgical indications (including infection, tumor, or trauma), prior cervical spine surgery (fusion or decompression), standalone interbody constructs without anterior plate fixation, less than two years of clinical follow-up, and absence of postoperative imaging sufficient for fusion assessment. Since reoperation within 24 months was included as a secondary outcome, the minimum follow-up requirement was applied to ensure uniform ascertainment of revision events. The detailed patient identification and screening process is illustrated in Figure 1. Flow diagram illustrating patient identification, exclusion criteria, and cohort selection for analysis of fusion outcomes and reoperation following four-level anterior cervical discectomy and fusion (ACDF) Surgical Technique All procedures were performed using a standard anterior cervical approach. A transverse skin incision was made, followed by vertical division of the platysma. The interval between the sternocleidomastoid and strap muscles was developed, with division of the omohyoid as needed, and blunt dissection carried down to the anterior cervical spine. The longus colli muscles were elevated to allow placement of self-retaining retractors. At each operative level, an annulotomy was performed and discectomy carried out using pituitary rongeurs. A high-speed burr was then used to perform a thorough decompression extending to the posterior longitudinal ligament (PLL), which was completely resected to ensure adequate central and foraminal decompression and to allow for uniform disc space distraction. Bilateral foraminotomies and uncinate resections were performed as needed to confirm adequate neural decompression. Endplate preparation was performed using curettes and a high-speed burr, including partial removal of the cartilaginous endplate and limited resection of the adjacent vertebral body to expose bleeding trabecular bone while avoiding excessive violation of subchondral bone. This technique facilitated placement of a large interbody graft and enhanced the biological environment for fusion. Interbody reconstruction included titanium cages, PEEK cages, or structural allografts. Graft materials included local autograft, iliac crest autograft, and adjunctive biologic materials, with combinations used when appropriate. Following sequential trialing, appropriately sized interbody grafts with lordotic profile were placed at each level. A rigid anterior cervical plate was then applied using the shortest construct feasible to maintain distance from adjacent levels. Screw fixation was performed using the largest appropriate screws based on preoperative imaging to optimize construct stability. Intraoperative fluoroscopy was used to confirm implant positioning and alignment. Fusion Assessment Fusion status by 12 months was the primary endpoint, evaluated using a hierarchical, best-available imaging strategy. Imaging modalities were assessed in this sequence, with CT used when available, followed by dynamic F-E radiographs and plain radiographs. Fusion was defined according to strict modality-specific radiographic criteria consistent with prior literature.11,19,22 On CT, fusion was defined as the presence of continuous bridging trabecular bone across the interbody space without evidence of hardware loosening. On dynamic F-E radiographs, fusion was defined as less than 1 mm of interspinous motion at the operative level on magnified images. On plain radiographs, fusion was defined by the presence of continuous bridging bone across the fused segment without evidence of a persistent radiolucent gap. Fusion status was determined at the patient level. Patients were classified as having nonunion if lack of fusion was present at any operative level. Fusion was additionally evaluated at the individual operative level to permit calculation of a per-level fusion rate. Fusion assessments were performed by two trained spine research fellows using the predefined modality-specific criteria. The primary cohort was divided between the two readers, with each study independently assessed by one reader. Prior to formal assessment, the readers reviewed representative cases together to standardize application of the fusion criteria. Early Fusion Cohort Patients with postoperative imaging obtained between 6 and 12 months were included in an exploratory early-fusion cohort. Early fusion was defined as radiographic evidence of fusion within this interval according to the same modality-specific criteria. Because imaging within this timeframe may not reflect definitive arthrodesis, absence of fusion was not classified to define nonunion, and early fusion rates were reported descriptively. Interobserver Reliability To evaluate interobserver reliability, a post hoc blinded reassessment was performed in a subset of 30 studies from each imaging modality (CT, dynamic F-E radiographs, and plain radiographs). Selected studies were independently reassessed by the reader who had not performed the original assessment and was blinded to the original fusion classification, clinical outcomes, and radiology reports. Reoperation Assessment Reoperation was defined as any subsequent cervical surgical procedure performed after the index ACDF. Procedures occurring within 24 months of the index operation were recorded. Reoperations were categorized by procedure type and analyzed descriptively, with rates reported according to fusion status. Imaging Modality Agreement Agreement between imaging modalities was assessed in patients with multiple postoperative imaging studies available. A subset of patients with CT, F-E radiographs, and plain radiographs was used to evaluate tri-modality agreement. Overall agreement was assessed using Fleiss’ kappa statistic. Pairwise agreement between modalities (CT vs F-E radiographs, CT vs plain radiographs, and F-E radiographs vs plain radiographs) was evaluated using Cohen’s kappa. Statistical Analysis Descriptive statistics were used to summarize patient characteristics and surgical variables. Fusion and early fusion rates were reported with 95% confidence intervals. Imaging modality distribution was summarized descriptively. Interobserver reliability for binary patient-level fusion status was assessed using Cohen’s κ with 95% confidence intervals and overall percentage agreement. Agreement between imaging modalities was assessed using kappa statistics, including Fleiss’ kappa for overall tri-modality agreement and Cohen’s kappa for pairwise comparisons. Analyses were performed using R Statistical Software (version 4.2.2; The R Foundation for Statistical Computing). Results Patient Selection and Cohort Characteristics A total of 237 patients underwent primary four-level ACDF during the study period and met surgical eligibility criteria. Of these, 152 patients had postoperative imaging available for fusion assessment and comprised the analytic cohort. All 152 patients were evaluated for reoperation within 24 months, while definitive fusion status by ≥12 months was available in 140 patients. The mean age at surgery was 64.4 ± 8.5 years, 45.4% of patients were female, and the mean body mass index was 28.9 ± 5.2 kg/m2. Baseline demographic and surgical characteristics are summarized in Table 1. Table 1.Baseline Characteristics and Surgical InformationVariableN (152)Age (years)64.382 (8.540)Female sex, n (%)69 (45.4%)BMI (kg/m2)28.936 (5.194)Smoking status, n (%) Current (daily)3 (2.0%) Current (some days)2 (1.4%) Former61 (40.1%) Never85 (55.9%) Unknown1 (0.7%) Surgical Characteristics Charlson Comorbidity Index0.789 (1.238)ASA class, n (%)​I2 (1.3%)II117 (77.0%)III33 (21.7%)PDM (minutes)156.033 (51.033)EBL (mL)70.757 (35.956)LOS (days)3.8 (2.4) Implant type, n (%) Titanium cage60 (39.5%) PEEK cage43 (28.3%) Structural allograft49 (32.2%)Graft material, n (%)* Local autograft109 (71.7%) Iliac crest autograft68 (44.7%) Adjunctive biologic35 (23.0%)Continuous variables are presented as mean ± standard deviation; categorical variables as n (%). BMI: body mass index; ASA: American Society of Anesthesiologists; PDM: procedure duration; EBL: estimated blood loss; LOS: length of stay; PEEK: polyetheretherketone.*Graft material categories are not mutually exclusive, as multiple graft materials could be used in the same patient. Baseline Characteristics and Surgical Information Continuous variables are presented as mean ± standard deviation; categorical variables as n (%). BMI: body mass index; ASA: American Society of Anesthesiologists; PDM: procedure duration; EBL: estimated blood loss; LOS: length of stay; PEEK: polyetheretherketone. *Graft material categories are not mutually exclusive, as multiple graft materials could be used in the same patient. Fusion Outcomes by 12 Months Fusion status by 12 months was determined for 140 patients using the predefined hierarchical imaging strategy. Overall, 114 patients (81.4%: 95% CI, 73.9%–87.5%) demonstrated radiographic fusion, while 26 patients (18.6%: 95% CI, 12.5%–26.1%) were classified as having nonunion. Based on the hierarchical imaging strategy, final fusion status was determined by CT in 57 patients (40.7%), F-E radiographs in 50 (35.7%), and plain radiographs in 33 (23.6%). The corresponding modality-specific fusion rates were 78.9% (45/57; 95% CI, 66.7%–87.5%) for CT, 82.0% (41/50; 95% CI, 69.2%–90.2%) for dynamic F-E radiographs, and 84.8% (28/33; 95% CI, 69.1%–93.3%) for plain radiographs. Among patients with nonunion, the majority involved a single operative level, with C6–7 representing the most frequently affected segment. At the individual operative level, 528 of 560 levels (94.3%) demonstrated fusion by 12 months. Representative examples of radiographic fusion and nonunion on CT are shown in Figure 2, while corresponding dynamic F-E radiographs illustrating fused and nonunion cases are shown in Figure 3. Detailed fusion outcomes are summarized in Table 2. Representative computed tomography (CT) images demonstrating radiographic fusion and nonunion following four-level anterior cervical discectomy and fusion (ACDF). (A-B) Sagittal and coronal CT images showing solid fusion with continuous bridging trabecular bone across the interbody spaces (yellow arrows). (C-D) Sagittal and coronal CT images demonstrating nonunion, with absence of continuous bridging bone across the interbody space at the C5-6 and C6-7 levels (red ovals highlight regions of discontinuity) Representative dynamic flexion–extension radiographs demonstrating fusion and nonunion following four-level anterior cervical discectomy and fusion (ACDF). (A-B) Flexion and extension views of a fused patient demonstrating no measurable interspinous motion (≤1 mm) across operative levels. (C-D) Flexion and extension views of a patient with nonunion, with interspinous motion exceeding 1 mm at the C3-4 and C6-7 levels (red arrows), consistent with radiographic nonunion Table 2.Fusion OutcomesCategoryVariablen (%) [95% CI]Patient-level outcomesFusion by 12 months114 (81.4) [73.9–87.5]Nonunion26 (18.6) [12.5–26.1]Nonunion characteristicsSingle-level nonunion21 (80.8)Two-level nonunion4 (15.4)≥3-level nonunion1 (3.8)Level distribution of nonunionC3–43 (11.5%)C4–51 (3.8%)C5–62 (7.7%)C6–726 (100%)Imaging modalityCT57 (40.7)F-E radiographs50 (35.7)Plain radiographs33 (23.6)Early fusion (exploratory)Fusion at 6–12 months101 (70.1) [62.2–77.0]Values are presented as n (%), with 95% confidence intervals reported for fusion outcomes. Percentages are calculated based on the relevant subgroup denominator. Patient-level outcomes and imaging modality are based on the fusion-assessed cohort (n=140). Nonunion characteristics and level distribution are reported relative to the nonunion cohort (n=26). Nonunion characteristics describe the number of nonunion levels per patient, whereas level distribution reflects the specific operative levels involved. Patients with multilevel nonunion therefore contribute to more than one segmental row; all patients with nonunion had C6–7 involvement, including 21 with isolated C6–7 nonunion and 5 with additional nonunion levels. Early fusion is reported for the exploratory cohort (n=144). CT: computed tomography; F-E: flexion-extension. CI: confidence interval. Fusion Outcomes Values are presented as n (%), with 95% confidence intervals reported for fusion outcomes. Percentages are calculated based on the relevant subgroup denominator. Patient-level outcomes and imaging modality are based on the fusion-assessed cohort (n=140). Nonunion characteristics and level distribution are reported relative to the nonunion cohort (n=26). Nonunion characteristics describe the number of nonunion levels per patient, whereas level distribution reflects the specific operative levels involved. Patients with multilevel nonunion therefore contribute to more than one segmental row; all patients with nonunion had C6–7 involvement, including 21 with isolated C6–7 nonunion and 5 with additional nonunion levels. Early fusion is reported for the exploratory cohort (n=144). CT: computed tomography; F-E: flexion-extension. CI: confidence interval. Early Fusion Outcomes Postoperative imaging between 6 and 12 months was available for 144 patients. Early fusion was observed in 101 patients (70.1%; 95% CI, 62.2%–77.0%). As this represents an exploratory cohort, absence of fusion at this time point was not used to define nonunion. Interobserver Reliability In the interobserver reliability subset, agreement was high across all imaging modalities. For CT, overall agreement was 93.3%, with κ = 0.814 (95% CI, 0.429–1.000). Agreement for dynamic F-E radiographs was 93.3%, with κ = 0.830 (95% CI, 0.348–1.000), while agreement for plain radiographs was 90.0%, with κ = 0.706 (95% CI, 0.268–1.000) (Table 3). Table 3.Interobserver Reliability of Fusion AssessmentModalityAgreement (%)Cohen’s κ95% CINCT93.30.8140.429–1.00030F-E radiographs93.30.8300.348–1.00030Plain radiographs90.00.7060.268–1.00030Agreement represents the percentage of concordant binary patient-level fusion classifications between readers. Cohen’s κ and corresponding 95% confidence intervals are reported for each imaging modality. CT: computed tomography; F-E: flexion-extension; CI: confidence interval. Interobserver Reliability of Fusion Assessment Agreement represents the percentage of concordant binary patient-level fusion classifications between readers. Cohen’s κ and corresponding 95% confidence intervals are reported for each imaging modality. CT: computed tomography; F-E: flexion-extension; CI: confidence interval. Imaging Modality Agreement Agreement between imaging modalities was evaluated in patients with multiple postoperative imaging studies available. In a tri-modality subset of 38 patients, overall agreement among CT, F-E radiographs, and plain radiographs was high (Fleiss’ κ = 0.823; 95% CI, 0.606–0.954), indicating strong agreement across modalities. Pairwise agreement analysis demonstrated the highest concordance between CT and F-E radiographs (κ = 0.938; 95% CI, 0.656–1.000), followed by CT and plain radiographs (κ = 0.791; 95% CI, 0.437–0.938) and F-E radiographs and plain radiographs (κ = 0.732; 95% CI, 0.406–0.932). Plain radiographs tended to classify a greater proportion of cases as fused compared with CT and F-E imaging. An example of discrepant fusion assessment across imaging modalities is shown in Figure 4. Agreement statistics are summarized in Table 4. Representative example of discrepant fusion assessment across imaging modalities following four-level anterior cervical discectomy and fusion (ACDF). (A) Plain radiograph suggesting apparent fusion at the operative levels. (B) Coronal computed tomography (CT) image demonstrating nonunion at the C6-7 level (red oval). (C-D) Flexion and extension radiographs of the same patient showing interspinous motion exceeding 1 mm at C6-7 (red arrows), consistent with radiographic nonunion Table 4.Imaging Modality AgreementCategoryComparisonκ value95% CINOverall agreement (tri-modality)CT, F-E, and plain radiographs (Fleiss’ κ)0.8230.606–0.95438Pairwise agreementCT vs F-E radiographs0.9380.656–1.00038CT vs plain radiographs0.7910.437–0.93838F-E radiographs vs plain radiographs0.7320.406–0.93238Agreement between imaging modalities was assessed using Fleiss’ kappa (κ) for overall tri-modality agreement and Cohen’s kappa for pairwise comparisons. κ values are reported for each comparison. CT: computed tomography; F-E: flexion-extension. Imaging Modality Agreement Agreement between imaging modalities was assessed using Fleiss’ kappa (κ) for overall tri-modality agreement and Cohen’s kappa for pairwise comparisons. κ values are reported for each comparison. CT: computed tomography; F-E: flexion-extension. Reoperation Outcomes Within 24 months of the index ACDF procedure, 5 of 152 patients (3.3%) underwent cervical revision surgery. Among the 140 patients with definitive fusion assessment, reoperation occurred in 2 of 26 patients (7.7%) with nonunion and in 3 of 114 patients (2.6%) with radiographic fusion. Posterior cervical decompression and fusion was the most common revision procedure, while the remaining cases involved surgical wound irrigation and debridement. The most common indication for reoperation was adjacent segment disease with central stenosis, followed by symptomatic pseudarthrosis and infection. Detailed revision indications and procedures are summarized in Table 5. Table 5.Reoperation OutcomesCategoryVariablen (%)Overall reoperationReoperation within 24 months5 (3.3)Reoperation by fusion statusFused3 (2.6)Nonunion2 (7.7)Revision procedure typePosterior cervical decompression and fusion4 (80.0)Surgical wound irrigation and debridement1 (20.0)Indication for revisionAdjacent segment disease with central stenosis3 (60.0)Symptomatic pseudarthrosis1 (20.0)Infection1 (20.0)Levels of revision surgeryC4-T11C6-T11C6-C71C7-T11Other1Time to revisionMean ± SD (days)395.2 ± 269.3Median (IQR), days336 (322-573)Values are presented as n (%) unless otherwise specified. Overall reoperation rate was calculated for the full analytic cohort (n=152), whereas reoperation rates stratified by fusion status were calculated among patients with definitive ≥12-month fusion assessment (n=140). Percentages for revision procedure type and indication for revision are calculated based on the total number of revision surgeries (n=5). Time to revision is reported as mean ± standard deviation and as median (interquartile range). Reoperation Outcomes Values are presented as n (%) unless otherwise specified. Overall reoperation rate was calculated for the full analytic cohort (n=152), whereas reoperation rates stratified by fusion status were calculated among patients with definitive ≥12-month fusion assessment (n=140). Percentages for revision procedure type and indication for revision are calculated based on the total number of revision surgeries (n=5). Time to revision is reported as mean ± standard deviation and as median (interquartile range). Discussion Multilevel ACDF, particularly more than three-level constructs, remains an area of ongoing debate, largely driven by concerns regarding pseudarthrosis, construct durability, and the potential need for revision surgery.10,13,18,20 As a result, some surgeons favor posterior approaches for extensive cervical pathology, while others continue to advocate for an anterior approach in appropriately selected patients.8,13,16 In the present study, which represents one of the largest single-center cohorts of primary four-level ACDF for degenerative pathology 20 , we evaluated fusion outcomes using a structured hierarchical imaging strategy and found that radiographic fusion was achieved in 81.4% of patients by 12 months, with a low revision rate of 3.3% within two years. Importantly, agreement between CT and dynamic F-E radiographs was near perfect, while plain radiographs demonstrated a modest tendency to overestimate fusion. These findings support four-level ACDF as a reliable surgical strategy for multilevel cervical degenerative disease when performed in experienced hands and evaluated using clearly defined radiographic criteria. Reported fusion rates following four-level ACDF vary widely across the literature, ranging from approximately 40% to over 90%, reflecting differences in study design, patient selection, surgical technique, fusion criteria, imaging modality, and duration of follow-up.21,22 In a radiographic series of 63 patients, White et al reported complete four-level fusion in only 41.3% of patients, with pseudarthrosis present in 58.7%, although most cases involved a single nonfused level and the per-level fusion rate exceeded 80%. 9 Similarly, Wewel et al observed pseudarthrosis in 45.8% of patients, with higher rates in four-level constructs and approximately 11% requiring revision surgery. 11 In contrast, other series have demonstrated more favorable outcomes, with fusion rates exceeding 80% and low rates of reoperation for symptomatic pseudarthrosis. 8 In the present study, radiographic fusion by 12 months was observed in 81.4% of patients, which aligns with these more favorable reports, particularly given the use of a strict patient-level definition of nonunion and a structured, modality-specific imaging assessment strategy. When evaluated at the individual operative level, however, the fusion rate was 94.3%, allowing more direct comparison with prior studies reporting per-level outcomes, including White et al 9 . Importantly, fusion following multilevel ACDF is known to evolve over time. Chang et al demonstrated that fusion rates increased from 26.4% at 4 months to 64.3% at 1 year and 81.6% at final follow-up, highlighting the time-dependent nature of arthrodesis in long-segment constructs. 22 A similar pattern was observed in the present study, where early fusion was identified in 70.1% of patients between 6 and 12 months, increasing to 81.4% at one year. In this context, the fusion rate observed at 12 months likely represents a conservative estimate of definite fusion. Consistent with prior reports, nonunion in our cohort most commonly involved a single level and was predominantly observed at the caudal segment, particularly C6-7, which has been identified as the most frequent site of pseudarthrosis in multilevel ACDF constructs.9,11,12 The reliability of fusion assessment also warrants consideration, particularly when dynamic F-E radiographs are interpreted using a strict <1-mm interspinous process motion (ISPM) threshold. Martin et al recently demonstrated variability in ISPM measurements among practicing spine surgeons, highlighting the potential influence of reader methodology on precise motion measurements. 23 In the present study, however, interobserver agreement for binary fusion classification was high across all imaging modalities, including dynamic F-E radiographs, which demonstrated 93.3% agreement (κ = 0.830; 95% CI, 0.348–1.000). CT similarly demonstrated 93.3% agreement (κ = 0.814; 95% CI, 0.429–1.000), while agreement for plain radiographs was 90.0% (κ = 0.706; 95% CI, 0.268–1.000). Although the wide confidence intervals reflect the limited size of the reliability subset and small number of discordant classifications, the high observed agreement suggests that standardized application of predefined criteria can provide reproducible fusion classification. These findings complement those of Martin et al by suggesting that variability in precise ISPM measurement does not necessarily translate into poor reliability of binary fusion assessment. A key strength of the present study is the evaluation of agreement between imaging modalities used for fusion assessment. While prior studies have reported fusion rates using a single imaging modality or a best-available approach, few have directly compared CT, dynamic F-E radiographs, and plain radiographs within the same patient cohort. In our analysis, agreement between CT and F-E radiographs was near-perfect, suggesting that dynamic radiographs represent a modality comparable to CT for assessing fusion in appropriately selected patients. In contrast, agreement between plain radiographs and the other modalities was lower, with plain radiographs demonstrating a tendency to overestimate fusion. These findings have important implications for both clinical practice and research methodology. The high concordance between CT and F-E radiographs supports the use of a hierarchical imaging strategy in which CT is reserved for equivocal cases, while dynamic radiographs may serve as an effective primary modality for routine follow-up. At the same time, the observed discrepancy with plain radiographs highlights the limitations of relying solely on static imaging for fusion assessment, particularly in multilevel constructs where subtle motion may not be detected. Although differences in imaging modality may contribute to variability in reported fusion rates across the literature, they likely represent only one of several factors influencing these outcomes. Despite the presence of radiographic nonunion in a subset of patients, the rate of revision surgery in the present study remained low. Within 24 months, reoperation occurred in 3.3% of patients overall, including 7.7% of patients with nonunion and 2.6% of those with radiographic fusion. These findings demonstrate that despite the presence of radiographic nonunion, symptomatic pseudarthrosis requiring revision remained relatively uncommon. This observation is consistent with prior studies demonstrating a discordance between radiographic nonunion and clinical outcomes. White et al reported high rates of radiographic pseudarthrosis, yet only a small proportion of patients required revision for symptomatic nonunion. 9 Similarly, Wewel et al observed that although pseudarthrosis was common, only 11.1% of patients underwent revision surgery. 11 At the same time, reported reoperation rates following multilevel ACDF vary widely across the literature. Laratta et al reported a reoperation rate of 35%, 21 while De la Garza-Ramos et al observed reoperation rates of approximately 19–21% in three- and four-level constructs. 8 In this context, the reoperation rates observed in the present study support the notion that four-level ACDF can achieve low rates of revision, even in the presence of radiographic nonunion. Together with the predominance of single-level nonunion in our cohort, these findings suggest that radiographic pseudarthrosis alone may not be sufficient to warrant revision and should be interpreted in conjunction with clinical presentation. The findings of the present study have important implications for surgical decision-making in patients requiring multilevel cervical decompression. Although increasing construct length has traditionally raised concerns regarding pseudarthrosis, complications, and the need for revision surgery, the results of this study demonstrate that four-level ACDF can achieve acceptable fusion rates and low reoperation rates. We believe that one of the fundamental factors contributing to these outcomes is the use of a consistent and meticulous surgical technique. In this context, a standardized approach emphasizing thorough decompression with complete resection of the posterior longitudinal ligament, uniform disc space distraction, preparation of a bleeding trabecular bone surface for fusion, and restoration of appropriate cervical lordosis may play a critical role in promoting successful arthrodesis in multilevel constructs. In our series, this was achieved through a systematic technique including wide discectomy, partial endplate and vertebral body preparation to facilitate placement of large interbody grafts, use of local autograft combined with allograft material, and rigid anterior plate fixation with appropriately sized screws to maximize construct stability and maintain segmental alignment. In recent years, there has been a trend toward favoring posterior approaches for extensive multilevel cervical pathology, in part due to concerns regarding outcomes with long anterior constructs. However, posterior surgery is also associated with its own set of limitations, including greater muscle dissection, higher blood loss, increased postoperative pain, and higher rates of certain perioperative complications, with some studies demonstrating higher odds of adverse events in multilevel posterior constructs.16,24-27 In contrast, the anterior approach allows for direct ventral decompression, restoration of cervical alignment, and preservation of the posterior paraspinal musculature through an intermuscular dissection plane.3-5,7 Preservation of posterior cervical musculature may be particularly relevant, as increasing evidence suggests that muscle quality, especially fatty infiltration of the posterior paraspinal muscles, is associated with worse clinical and radiographic outcomes following cervical spine surgery.28-36 In this context, avoiding unnecessary posterior disruption may represent an additional advantage of an anterior-only approach in appropriately selected patients. Rather than being defined by construct length alone, surgical strategy should be guided by patient-specific pathology, surgeon experience, and a balanced consideration of the advantages and limitations of each approach. This study has several limitations. First, its retrospective design and single-center setting may limit generalizability, although the inclusion of a relatively large cohort of four-level ACDF procedures performed by multiple experienced surgeons represents a strength compared with prior series. Second, radiographic follow-up was not uniform across all patients, and fusion assessment was based on the best available imaging within predefined time intervals, which may introduce variability in the timing of fusion assessment. It is also possible that patients with early radiographic and clinical improvement were less likely to undergo continued imaging follow-up, which may lead to underrepresentation of patients with successful early fusion. Third, although fusion was assessed using a hierarchical imaging strategy, differences in imaging modality may still influence classification in some cases. Finally, this study was designed as a radiographic evaluation of fusion outcomes and did not include detailed assessment of patient-reported outcomes or symptom-specific measures; therefore, the clinical significance of radiographic nonunion among patients who did not undergo revision cannot be determined from the present analysis. Future studies incorporating clinical outcome measures and comparative analyses between surgical strategies, including anterior and posterior approaches, would provide a more comprehensive understanding of the relative benefits and limitations of these treatment options that could help optimize decision-making in multilevel cervical spine surgery. Conclusion In one of the largest single-center cohorts of patients undergoing four-level ACDF, fusion rates at one year exceeded 80% and reoperation rates within two years were 3.3%. Fusion assessment using a hierarchical imaging strategy demonstrated high agreement between CT and dynamic flexion-extension radiographs, while plain radiographs showed a tendency to overestimate fusion. Although radiographic nonunion was observed in a subset of patients, revision surgery for symptomatic pseudarthrosis was infrequent. These findings suggest that four-level ACDF can provide reliable radiographic outcomes with low rates of reoperation when performed in an experienced setting with a consistent and meticulous surgical technique. Construct length alone should not preclude consideration of an anterior-only approach in appropriately selected patients. Footnotes Funding: The authors received no financial support for the research, authorship, and/or publication of this article. The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. ORCID iDs Bruno Verna https://orcid.org/0009-0008-2823-6724 Tom Folkerts https://orcid.org/0009-0008-6597-8348 Julia Wimmer https://orcid.org/0009-0008-7294-4445 Steven M. Maurer https://orcid.org/0000-0003-4346-0528 Marco D. Burkhard https://orcid.org/0000-0003-1501-1952 Andrew A. Sama https://orcid.org/0000-0002-3014-7287 Alexander P. Hughes https://orcid.org/0000-0001-7293-9672 Federico P. Girardi https://orcid.org/0000-0002-3269-7602 Ethical Considerations This study was approved by our hospital’s institutional review board (IRB# 2020-1877). Informed consent was waived by the IRB because of the retrospective nature of the study. This study follows the ethical principles of the Helsinki Declaration. Data Availability Statement The datasets generated and analyzed during the current study are not publicly available due to confidentiality restrictions but are available from the corresponding author on reasonable request.* References Associated Data Data Availability Statement The datasets generated and analyzed during the current study are not publicly available due to confidentiality restrictions but are available from the corresponding author on reasonable request.*