Prediction of EGFR mutation status in NSCLC using nomogram model based on CT radiological and clinical features

ObjectiveTo investigate the value of constructing a combined CT imaging and clinical features prediction model for evaluating epidermal growth factor receptor (EGFR) mutations in patients with non-small cell lung cancer (NSCLC).

MethodsA total of 258 patients with pathologically confirmed NSCLC (182 in the training group and 76 in the validation group) were enrolled. Single factor and multiple factor logistic regression analysis were used to analyze and screen out independent factors affecting EGFR mutation to build CT imaging-clinical prediction model. A nomogram was drawn to visualize the model. Finally, receiver operating characteristic curve, calibration curve and decision curve were used to evaluate the practicability of the model.

ResultsFinally, gender, smoking status, burr sign, and pleural indentation sign were screened as independent predictive factors for NSCLC EGFR mutations, and a CT imaging-clinical prediction model was established. In the training and validation groups, the predicted area under the curve of the model was 0.799 and 0.797, respectively (P < 0.01), indicating good predictive performance. The calibration curve showed that the prediction models of the training and validation groups had good consistency with the observed results. Decision curve showed that the model had achieved good clinical benefits in predicting EGFR mutations.

ConclusionsThe CT imaging-clinical feature prediction model has well predictive value for EGFR mutations in NSCLC patients and can be served as noninvasive tool for preoperative clinical prediction.