Multidimensional tumor anatomy, intentional incomplete resection, and long-term outcomes in cervical dumbbell tumors: an exploratory cohort study

Objective To evaluate the reproducibility of multidimensional imaging features in cervical dumbbell tumors and examine their associations with intentional incomplete resection, observed surgical management, and long-term radiographic and functional outcomes. Methods We retrospectively analyzed 69 consecutive patients treated surgically from 2014 through 2023. Tumor anatomy was characterized using the Toyama-Asazuma classification and reproducible quantitative measurements of tumor dimensions, foraminal anatomy, and vertebral artery (VA) relationships. Exploratory preoperative associations were controlled using the Benjamini-Hochberg false-discovery rate. Because only 14 incomplete resections occurred, adjusted analysis was limited to a prespecified two-variable Firth penalized model for intentional incomplete resection. Longitudinal tumor control, function, alignment, instability, and fusion were also assessed. Results Gross-total resection was achieved in 55 patients (79.7%); all 14 incomplete resections were intentional. The principal quantitative measurements showed good-to-excellent reproducibility, including the extraspinal-to-intervertebral foramen ratio (interobserver ICC, 0.915). Intentional incomplete resection remained associated after false-discovery-rate correction with previous surgery, an intradural component, VA encasement, and a greater extraspinal-to-intervertebral foramen ratio. In the Firth model, the ratio showed an adjusted association with intentional incomplete resection (OR, 2.19 per 1-unit increase; bootstrap 95% CI, 1.31-7.87; approximate p=0.048), as did VA encasement (OR, 21.94; bootstrap 95% CI, 4.08-177.47; approximate p=0.004). Combined approach, facetectomy, and fusion were analyzed as observed surgeon-selected management choices rather than objective procedural requirements. Radiographic progression occurred in 10 of 14 patients (71.4%) after intentional incomplete resection and in none of 55 after gross-total resection (log-rank p<0.001). Functional scores were similar between resection groups at 2 years and final follow-up, although these observational comparisons do not establish equivalence. No delayed radiographic instability was identified, and all 28 instrumented patients met 1-year fusion criteria. C1–2 tumors demonstrated a distinct management and alignment profile. Conclusions Reproducible multidimensional anatomy provides information beyond topographic classification in cervical dumbbell tumors. Greater extraspinal burden relative to the foraminal corridor and VA encasement were associated with intentional incomplete resection, which carried substantially poorer long-term radiographic control. These findings support individualized anatomical assessment and structured surveillance rather than fixed surgical thresholds.