• OBJECTIVE
    • Currently, there is a lack of good clinical tools for evaluating the effect of chemotherapy preoperatively on primary high-grade bone sarcomas. Our goal was to investigate the predictive value of the clinical findings and establish a scoring system to predict chemotherapy response.
  • METHODS
    • We conducted a retrospective multicenter cohort study and reviewed 322 patients with primary high-grade bone sarcomas. Patients who routinely received neoadjuvant chemotherapy and underwent primary tumor resection with an assessment of tumor necrosis rate (TNR) were enrolled in this study. The medical records of patients were collected from November 1, 2011, to March 1, 2018, at Peking University People's Hospital (PKUPH) and Peking University Shougang Hospital (PKUSH). The mean age of the patients was 16.2 years (range 3-52 years), of whom 65.5% were male. The clinical data collected before and after neoadjuvant chemotherapy included the degree of pain, laboratory inspection, X-ray, CT, contrast-enhanced magnetic resonance (MR), and positron emission tomography-computed tomography (PET-CT). Several machine learning models, including logistic regression, decision trees, support vector machines, and neural networks, were used to classify the chemotherapy responses. Area under the curve (AUC) of the scoring system to predict chemotherapy response is the primary outcome measure.
  • RESULTS
    • For patients without events, a minimum follow-up of 24 months was achieved. The median follow-up time was 43.3 months, and it ranged from 24 to 84 months. The 5 years progression-free survival (PFS) of the included patients was 54.1%. The 5 years PFS rate was 39.7% for poor responders and 74.9% for good responders. Features such as longest diameter reduction ratio (up to three points), clear bone boundary formation (up to two points), tumor necrosis measured by magnetic resonance (up to two points), maximum standard uptake value (SUVmax ) decrease (up to three points), and significant alkaline phosphatase decrease (up to 1 point) were identified as significant predictors of good histological response and constituted the scoring system. A score ≥4 predicts a good response to chemotherapy. The scoring system based on the above factors performed well, achieving an AUC of 0.893. For nonmeasurable lesions (classified by the revised Response Evaluation Criteria in Solid Tumors [RECIST 1.1]), the AUC was 0.901.
  • CONCLUSION
    • We first devised a well-performing comprehensive scoring system to predict the response to neoadjuvant chemotherapy in primary high-grade bone sarcomas.