Development and Validation of Nomogram-Based Prognosis Tools for Patients with Extremity Osteosarcoma: A SEER Population Study

Abstract

Objective. Osteosarcoma, usually occurring in the extremities, is the most common malignant bone tumour. The purpose of this study is to develop and validate nomogram-based prognosis tools for survival (OS) and cancer special survival (CSS) of patients with osteosarcoma of the extremities via the application of survival analysis. Materials and Methods. A total of 1427 patients diagnosed with osteosarcoma of the extremities during 2004–2015 were selected from the National Cancer Institute’s (NCI) Surveillance, Epidemiology, and End Results- (SEER-) Medicare database. The samples were randomly assigned to either the training set ( $n=856$ ) or the validation cohort ( $n=571$ ). Kaplan–Meier (K–M) survival analysis was conducted to calculate patients’ 1-, 3-, and 5-year OS and CSS rates. Cox proportional hazard ratio (HR) regression models were employed to identify and examine the factors that have a significant impact on OS and CSS with data from the training cohort. Results. The results of univariate and multivariate analyses performed in the training cohort indicated that older age, increased tumour size, higher grade, distant tumour extension, amputation, or no surgery (all $\text{HR}>1$ , $P<0.05$ ) were risk predictors of poor OS and CSS. Subsequently, the independent prognosis signatures were utilised to construct nomograms. The concordance index (C-index), calibration plot, and decision curve analysis (DCA) were simultaneously used to validate the nomograms. The internally validated C-index values of the OS and CSS prediction models for the training set were 0.752 (95% confidence interval [CI]: 0.738–0.765) and 0.754 [95% CI: 0.740–0.768], respectively. Then, the models were validated in the validation cohort population, which also demonstrated the models had good reliability for prognostication. Conclusions. The SEER cohort of patients with osteosarcoma of the extremities can be employed to produce effective tools that can assist in prognosis modelling.