Clinical model of pulmonary metastasis in patients with osteosarcoma: A new multiple machine learning-based risk prediction

Abstract

Background Metastasis is one of the most significant prognostic factors in osteosarcoma (OS). The goal of this study was to construct a clinical prediction model for OS patients in a population cohort and to evaluate the factors influencing the occurrence of pulmonary metastasis. Methods We collected data from 612 patients with osteosarcoma (OS), and 103 clinical indicators were collected. After the data were filtered, the patients were randomly divided into training and validation cohorts by using random sampling. The training cohort included 191 patients with pulmonary metastasis in OS and 126 patients with non-pulmonary metastasis, and the validation cohort included 50 patients with pulmonary metastasis in OS and 57 patients with non-pulmonary metastasis. Univariate logistics regression analysis, LASSO regression analysis and multivariate logistic regression analysis were performed to identify potential risk factors for pulmonary metastasis in patients with osteosarcoma. A nomogram was developed that included risk influencing variables selected by multivariable analysis, and used the concordance index (C-index) and calibration curve to validate the model. Receiver operating characteristic curve (ROC), decision analysis curve (DCA) and clinical impact curve (CIC) were employed to assess the model. In addition, we used a predictive model on the validation cohort. Results Logistic regression analysis was used to identify independent predictors [N Stage + Alkaline phosphatase (ALP)+Thyroid stimulating hormone (TSH)+Free triiodothyronine (FT3)]. A nomogram was constructed to predict the risk of pulmonary metastasis in patients with osteosarcoma. The performance was evaluated by the concordance index (C-index) and calibration curve. The ROC curve provides the predictive power of the nomogram (AUC = 0.701 in the training cohort, AUC = 0.786 in the training cohort). Decision curve analysis (DCA) and clinical impact curve (CIC) demonstrated the clinical value of the nomogram and higher overall net benefits. Conclusions Our study can help clinicians effectively predict the risk of lung metastases in osteosarcoma with more readily available clinical indicators, provide more personalized diagnosis and treatment guidance, and improve the prognosis of patients. Mini Abstract A new risk model was constructed to predict the pulmonary metastasis in patients with osteosarcoma based on multiple machine learning.