Development of an innovative nomogram of risk factors to predict postoperative recurrence of gastrointestinal stromal tumors

BACKGROUND: There are many staging systems for gastrointestinal stromal tumors (GISTs), and the risk indicators selected are also different; thus, it is not possible to quantify the risk of recurrence among individual patients. AIM: To develop and internally validate a model to identify the risk factors for GIST recurrence after surgery. METHODS: The least absolute shrinkage and selection operator (LASSO) regression model was performed to identify the optimum clinical features for the GIST recurrence risk model. Multivariable logistic regression analysis was used to develop a prediction model that incorporated the possible factors selected by the LASSO regression model. The index of concordance (C-index), calibration curve, receiver operating characteristic curve (ROC), and decision curve analysis were used to assess the discrimination, calibration, and clinical usefulness of the predictive model. Internal validation of the clinical predictive capability was also evaluated by bootstrapping validation. RESULTS: The nomogram included tumor site, lesion size, mitotic rate/50 high power fields, Ki-67 index, intracranial necrosis, and age as predictors. The model presented perfect discrimination with a reliable C-index of 0.836 (95%CI: 0.712-0.960), and a high C-index value of 0.714 was also confirmed by interval validation. The area under the curve value of this prediction nomogram was 0.704, and the ROC result indicated good predictive value. Decision curve analysis showed that the predicting recurrence nomogram was clinically feasible when the recurrence rate exceeded 5% after surgery. CONCLUSION: This recurrence nomogram combines tumor site, lesion size, mitotic rate, Ki-67 index, intracranial necrosis, and age and can easily predict patient prognosis.