Molecular Markers Predict Long-term Recurrence in Patients Treated for Spinal and Skull Base Chordomas: A Mayo Clinic Multi-center Study

Introduction: Chordomas are rare tumors from notochordal remnants and account for 1% to 4% of all primary bone malignancies. Despite margin-negative resection, chordomas often recur. Further, molecular markers have not been assessed as predictive markers for chordoma recurrence. We aimed to identify molecular markers predictive of long-term chordoma recurrence (≥ 1 year) using machine learning(ML)algorithms.

Methods: We identified patients who had treatment for clival and spinal chordomas between January 2017 and June 2021 across Mayo Clinic enterprise. Demographics, type of treatment, follow-up duration and other relevant clinical factors were abstracted patients' records. Decision tree(DT) and Random Forest(RF) classifiers were trained and tested to predict the long-term recurrence based on unseen data using an 80/20 split.

Results: One hundred fifty-one patients diagnosed and treated for chordomas were identified:58 chordomas from the clivus,48 chordomas of the mobile spine and 45 sacrococcygeal of origin. Patients diagnosed with cervical chordomas were the oldest among all groups (58±14 years; p=0.009). Most patients were males(60.3%) and white(92.1%). Subtotal resection followed by radiation therapy(33.8%), was the most common treatment modality, followed by gross total resection, then radiation therapy only. Multivariate analysis shows that the presence of S100 and pan-cytokeratin is more likely to increase the risk of recurrence(OR=3.67;CI=[1.09,12.42],p=0.03). DT showed a clinical follow-up>1897 days was found in 37% of patients and 90% chance of being classified for recurrence(Accuracy=77%). RF (n =500 trees) predicted all factors with the highest variable importance based on mean minimal depth. Factors predicting long-term recurrence are age,type of treatment,location of the tumor,S100,pan-cytokeratin and EMA.

Conclusion : Our molecular variables combined with tree-based ML tools successfully demonstrate a high capacity to individually identify the patient's recurrence pattern with an accuracy of 77%. S100, pan-cytokeratin, and EMA were the molecular drivers of recurrence. This shows the power of ML algorithms in analyzing and predicting outcomes even with small sample sizes.