Immunohistochemical markers predicting long-term recurrence following clival and spinal chordoma resection: a multicenter study

OBJECTIVE Chordomas are rare tumors from notochordal remnants and account for 1%–4% of all primary bone malignancies, often arising the clivus sacrum. Despite margin-negative resection postoperative radiotherapy, chordomas recur. Further, immunohistochemical (IHC) markers have not been assessed as predictive chordoma recurrence. The authors aimed to identify IHC that long-term (≥ 1 year) recurrence by using trained multiple tree-based machine learning (ML) algorithms. METHODS reviewed records patients who had undergone treatment clival spinal between January 2017 June 2021 across Mayo Clinic enterprise (Minnesota, Florida, Arizona). Demographics, type treatment, histopathology, other relevant clinical factors were abstracted each patient record. Decision tree random forest classifiers tested predict based on unseen data an 80/20 split. RESULTS One hundred fifty-one diagnosed treated identified: 58 clivus, 48 mobile spine, 45 sacrococcygeal in origin. Patients with cervical oldest among groups (58 ± 14 years, p = 0.009). Most male (n 91, 60.3%) White 139, 92.1%). underwent or without radiation therapy 129, 85.4%). Subtotal followed 51, 33.8%) was most common modality, gross-total then 43, 28.5%). Multivariate analysis showed S100 pan-cytokeratin more likely increase risk (OR 3.67, 95% CI 1.09–12.42, p= 0.03; OR 3.74, 0.05–2.21, 0.02, respectively). In decision analysis, a follow-up > 1897 days found 37% encounters 90% chance being classified (accuracy 77%). Random 500 trees) age, surgical location tumor, S100, pan-cytokeratin, EMA predicting CONCLUSIONS clinicopathological variables combined ML tools successfully demonstrated high capacity patterns accuracy 77%. drivers This shows power algorithms analyzing outcomes conditions small sample size.