Development and Validation of Novel Nomograms for Predicting Specific Distant Metastatic Sites and Overall Survival of Patients With Soft Tissue Sarcoma

Abstract

Purpose: The goal of this study is to construct nomograms to effectively predict the distant metastatic sites and overall survival (OS) of soft tissue sarcoma (STS) patients. Methods: STS case data between 2010 and 2015 for retrospective study were gathered from public databases. According to the chi-square and multivariate logistic regression analysis determined independent predictive factors of specific metastatic sites, the nomograms based on these factors were consturced. Subsequently, combined metastatic information a nomogram to predict 1-, 2-, and 3-year OS of STS patients was developed. The performance of models was validated by the area under the curve (AUC), calibration plots, and decision curve analyses (DCA). Results: A total of 7001 STS patients were included in this retrospective study, including 4901 cases in the training group and the remaining 2,100 patients in the validation group. Three nomograms were established to predict lung, liver and bone metastasis, and satisfactory results have been obtained by internal and external validation. The AUCs for predicting lung, liver, and bone metastases in the training cohort were 0.796, 0.799, and 0.766, respectively, and in the validation cohort were 0.807, 0.787, and 0.775, respectively, which means that the nomograms have good discrimination. The calibration curves showed that the models have high precision, and the DCA manifested that the nomograms have great clinical application prospects. Through univariate and multivariate COX regression analyses, 8 independent prognosis factors of age, grade, histological type, tumor size, surgery, chemotherapy, radiatiotherapy and lung metastasis were determined. A nomogram was then constructed to predict the 1-, 2-, and 3-years OS, which has a good performance in both internal and external validations. Conclusion: The nomograms for predicting specific metastatic sites and OS have good discrimination, accuracy and clinical applicability. The models could accurately predict the metastatic risk and survival information, and help clinical decision-making.