Establishment of multi-factor integrated EGFR mutation prediction model in non-small cell lung cancer

Abstract

Objective To establish a predictive model that is more sensitive, specific, simple, and economical for epidermal growth factor receptor (EGFR) mutations. Methods A total of 172 patients with advanced non-small cell lung cancer who underwent EGFR mutation detection were enrolled in this experiment and the relationship between EGFR mutation status, imaging features and serum carcinoembryonic antigen (CEA) as well as CYFRA21-1 levels was analyzed retrospectively, besides, a predictive model of EGFR mutation was established. Results Located around the lung (χ2=4.592, P=0.032), pleural indentation (χ2=12.071, P=0.001) and the appearance of bilateral intrapulmonary metastasis (χ2=13.389, P 6.005, tumor maximum diameter <5.050 cm, CYFRA21-1<2.72 ug/L, CEA≥8.050 μg/L were more prone to EGFR mutations (P<0.05), and their AUC were 0.671, 0.609, 0.602, 0.665 respectively.When there was comprehensive seven factors above predict EGFR mutation, the AUC value is 0.845. Conclusions The prediction accuracy of EGFR by single factor is low, however the prediction model based on seven factors could improve the prediction ability of EGFR.As a supplement to EGFR detection, predictive models may be considered in areas where biopsy is under-sampling, EGFR detection is false negative or genetic testing is not available.Nevertheless, a larger and multi-agency prospective study is needed to further validate. Key words: Carcinoma, non-small-cell lung; Receptor, epidermal growth factor; Carcinoembryonic antigen; CYFRA21-1