Development And Validation Of A Prognostic Model For Gastrointestinal Stromal Tumors Patients And Analysis Of Its Clinical Utilities

ObjectiveGastrointestinal stromal tumor(GIST)are rare tumors with a malignant tendency.Although the emerge of tyrosine kinase inhibitors largely improved the prognosis of GISTs,tumor recurrence is still a common clinical phenomenon largely due to the unreliability of individual risk assessment.This study aimed to develop and validate an accurate prognostic model for GISTs.MethodSEER*stat software was used to collect data from Surveillance,Epidemiology,and End Results database(SEER database)extracted the relevant data of 1788 patients with gastrointestinal stromal tumor diagnosed by histopathology from 2010 to 2016,and randomly divided them into training set(1168 cases)and verification set(501 cases)according to the ratio of 7:3.Through univariate analysis and multivariate analysis of the training set,the prognostic factors were determined and included the prognosis model.The prognosis related factors were visualized by using forest map.The discrimination and predictability of the model were evaluated by the AUC,C-index,and calibrate curve and comparison to that of other factors on GISTs.The decision curve analysis was applied to predict the clinical net benefit(NB)of the nomogram under different risk threshold probabilities.The verification set is utilized to verify the model.ResultBased on the multivariate analysis of the training cohort,seven independent survival-related factors were age,sex,marital status,site,grade,stage and surgery,which were all integrated into the nomogram.The AUC and C-index of the model were 0.770(95%CI,0.734-0.806)and 0.752(95%CI,0.734-0.770),respectively,which was statistically higher than that of the following factor:age(0.601,0.614),site(0.567,0.553),grade(0.649,0.624),and stage(0.662,0.651).In the validation set,the nomogram probability was superior to the four other factors(AUC:0.770v0.533 to 0.656;C-index:0.777v0.539 to 0.658).All calibrate curves demonstrated an excellent consistency in the probability at 3 and 5year OS between the nomogram prediction and actual observation.The DCA indicated the most beneficial clinical effect with the nomogram for OS under threshold probabilities of 0-0.45.ConclusionThe prognostic model was developed and verified to assess the prognosis of GISTs patients,demonstrating better prediction performance and clinical net benefits.