Prediction Of Survival For Osteosarcoma Patients And Development Of Pulmonary Metastasis

Objective To reveal the pulmonary metastasis and overall survival in patients with osteosarcoma;to investigate the associated risk factors for development of pulmonary metastasis and prognostic factors for overall survival;to establish predictive models.Methods Patients’ records from the Surveillance,Epidemiology,and End Results(SEER),Therapeutically Applicable Research to Generate Effective Treatments(TARGET)and Tianjin Medical University Cancer Institute and Hospital were retrospectively analyzed.The Kaplan–Meier method was used to calculate the overall survival.The significant prognostic factors were identified by the univariable and multivariable Cox proportional hazard regression analysis.Seventy percentage of patients selected randomly from SEER database were used to investigate the prognostic factors and to construct predictive models.The receiver operating characteristic curve(ROC)and area under curve(AUC)were used to evaluate the sensitivity and specificity.The other 30% patients were used to internally valid the pre-constructed model.The discrimination and calibration were tested.Records from the native database were used to conduct external validation.Extra significant factors were integrated into the previous model to develop a new model.Risk factors for the development of pulmonary metastasis were investigated by logistic regression method.Similar methods were used in construction of the predictive model,internal and external validation,and domestic model construction.Results A total of 1,475 patients with primary osteosarcoma from the SEER database(2010.1-2016.12),184 patients from the TARGET program(2001.1-2017.12),and 203 patients from the native database(2009.1-2020.12)were collected in this study.The 5-year overall survival rate was 59.6%,56.1% and 70.0%,respectively.The significant prognostic factors for patients in the SEER database included the age,tumor site,historical types,grade,T stage,surgery,bone metastasis and pulmonary metastasis.The tumor site,necrosis,surgery,and pulmonary metastasis were associated with the survival for patients in the TARGET program.The significant prognostic factors in the native database included surgery,bone metastasis and pulmonary metastasis.The nomogram model based on SEER database showed AUC for 5-year survival prediction was 0.791.The C-index was 0.787,indicating a moderate predictive ability.The internal validation showed similar AUC performance.Better predictive ability was achieved in the external validation by native database.After including extra necrosis and ALP,the new nomogram subtly improved the AUC.The predictive model constructed by only significant factors(surgery,bone metastasis and pulmonary metastasis)in native database also showed good predictive performance.The prevalence of pulmonary metastasis in the SEER database and TARGET was 16.8%(248 out of 1475)and 21.7%(40 out of 184).We found 21(10.3%)patients with pulmonary metastasis at the diagnosis of osteosarcoma and 67(33%)metastases during the treatment.Training cohort revealed that the Grade III-IV,T2-3 stage,N1 stage and bone metastasis were risk factors for the pulmonary metastasis.The predictive nomogram showed AUC of 0.614 and C-index of 0.712.The internal validation showed similar performance with AUC of 0.628 and C-index of 0.65.N1 stage,tumor necrosis and Ki-67 were risk factors for metachronous pulmonary metastasis in the native cohort.The predictive model for metachronous pulmonary metastasis showed AUC was 0.669 and the C-index was 0.743.Conclusions The 5-year overall survival was about 70% for patients with osteosarcoma,while it was less than 30% for patients with pulmonary metastasis.The prevalence of the synchronous pulmonary metastasis was from 10% to 21.7%.Approximate 33% patients developed pulmonary metastasis during the treatment of osteosarcoma.Based on risk and prognostic factors,predictive models showed moderate performance.