Abstract
11020
Background: The multidisciplinary approach in treatment of osteosarcoma has been well established and well adopted nationwide. This study combined the clinical prognostic factors at initial presentation into a nomogram to predict local control (LC), metastasis free survival (MFS), and overall survival (OS) for patients with non-metastatic bone osteosarcoma. Methods: We reviewed 397 osteosarcoma patients treated from 1995 to 2014. Patients with metastatic disease at diagnosis or limited follow up were excluded, resulting in 283 cases for analysis. Clinical and pathologic variables were recorded. Predictive variables included age at diagnosis, gender, previous radiation history, site, tumor size, histologic subtype, histologic grade, extra-osseous extension (EOE), lymphovascular invasion (LVI), necrosis rate and margin. The multivariate Cox proportional hazards regression was used to analyze survival outcomes and risk variables. Results: At 10 years, LC was 70.4% (95% confidence interval, CI: 64.0%-76.7%), MFS was 64.7% (95% CI: 58.1%-71.3%), and OS was 61.5% (95% CI: 54.9%-68.1%). Multivariate Cox model identified age (p = 0.033), site (p = 0.020), EOE (p = 0.017), LVI (p = 0.011), and margin (p = 0.039) were correlated with LC; age (p = 0.028), tumor size (p < 0.001), histologic grade (p = 0.039), and LVI (p = 0.014) were correlated with MFS; whereas age (p < 0.001), prior radiation history (p = 0.010), tumor size (p = 0.002), histologic subtype (p = 0.012), EOE (p = 0.002), and LVI (p = 0.001) were associated with OS. The nomograms were drawn on the basis of the Cox regression model and were internally validated using bootstrapping, with predictive accuracy of ±7.2% for 10-year LC, ±7.4% for 10-year MFS, and ±7.5% for 10-year OS. Conclusions: Nomograms have been developed to predict the 10-year local-control failure, recurrence and death. We suggest that this tool at presentation may be useful for individualized risk evaluation, patient counseling, and making clinical decisions.
