Nomograms to Predict Lung Metastasis in Malignant Primary Osseous Spinal Neoplasms and Cancer-specific Survival in Lung Metastasis Subgroup
Overview
Authors
Affiliations
Purpose: To construct and validate nomograms for predicting lung metastasis probability in patients with malignant primary osseous spinal neoplasms (MPOSN) at initial diagnosis and predicting cancer-specific survival (CSS) in the lung metastasis subgroup.
Methods: A total of 1,298 patients with spinal primary osteosarcoma, chondrosarcoma, Ewing sarcoma, and chordoma were retrospectively collected. Least absolute shrinkage and selection operator (LASSO) and multivariate logistic analysis were used to identify the predictors for lung metastasis. LASSO and multivariate Cox analysis were used to identify the prognostic factors for 3- and 5-year CSS in the lung metastasis subgroup. Receiver operating characteristic (ROC) curves, calibration curves, and decision curve analyses (DCA) were used to estimate the accuracy and net benefits of nomograms.
Results: Histologic type, grade, lymph node involvement, tumor size, tumor extension, and other site metastasis were identified as predictors for lung metastasis. The area under the curve (AUC) for the training and validating cohorts were 0.825 and 0.827, respectively. Age, histologic type, surgery at primary site, and grade were identified as the prognostic factors for the CSS. The AUC for the 3- and 5-year CSS were 0.790 and 0.740, respectively. Calibration curves revealed good agreements, and the Hosmer and Lemeshow test identified the models to be well fitted. DCA curves demonstrated that nomograms were clinically useful.
Conclusion: The nomograms constructed and validated by us could provide clinicians with a rapid and user-friendly tool to predict lung metastasis probability in patients with MPOSN at initial diagnosis and make a personalized CSS evaluation for the lung metastasis subgroup.