TY - JOUR T1 - The value and efficacy of an <sup>18</sup>F-FDG metabolic nomogram for predicting EGFR gene mutations in non-small cell lung cancer JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 296 LP - 296 VL - 61 IS - supplement 1 AU - Yang Jiang AU - Chuning Dong AU - Xian Li AU - Xiaowei Ma AU - Yunhua Wang Y1 - 2020/05/01 UR - http://jnm.snmjournals.org/content/61/supplement_1/296.abstract N2 - 296Objectives: Epidermal growth factor receptor (EGFR) targeting therapy has been a very effective approach to treat non-small cell lung cancers (NSCLC). However, patient screening and treatment efficacy basically depend on the EGFR gene mutation status of the tumor cells. Thus, it is critical to determine EGFR mutation status in clinical practice. In this study, we demonstrated an 18F-FDG metabolic model for EGFR mutation prediction based on retrospective PET imaging data and evaluated its efficacy with prospective data. Methods: A predicting model was developed based on retrospectively reviewing 105 NSCLC patients (Training cohort) who were subjected for both EGFR gene expression and 18F-FDG PET/CT imaging prior to treatment from Jan 2017 to June 2018. The differences between subgroups were analyzed in eight clinical features and three 18F-FDG metabolic parameters, including the maximal standard uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) of the primary tumor. Multivariate logistic regression analysis was performed to identify predictors of EGFR mutations, and the prediction model and nomogram were constructed. The efficacy and accuracy of the prediction model were assessed using the Homsmer-Lemeshow test (HL test) and the receiver operating characteristics (ROC) analysis. The diagnostic performance of the model was further validated by a prospective review of 91 NSCLC patients (Validation cohort) from July 2018 to May 2019. Results: In the training cohort, the length of the tumor, MTV, and TLG (P&lt;0.05) were significantly negatively associated with EGFR mutations. Moreover, women, nonsmokers, and adenocarcinoma patients (P&lt;0.05) are more likely to have EGFR mutations. Multivariate analysis demonstrated that only gender, length of the tumor, SUVmax, and MTV of the eleven features were independent predictors of EGFR mutations. ROC analysis showed that the area under the curve (AUC) of SUVmax, MTV, and TLG for EGFR mutations prediction were 0.597, 0.643, and 0.639, respectively. The prediction model and nomogram were Po = 1/(1 + e^(-x) ), x = 1.338×gender + 0.518×Lt - 0.075×SUVmax - 0.079×MTV - 0.718, whereas Lt is length of tumor, male = 0, female = 1. The sensitivity, specificity, and AUC of the predictive model was 80.0%, 66.2%, and 0.775 (95%CI: 0.687-0.864), respectively. In the validation cohort, the overall sensitivity, specificity, and AUC were 78.4%, 68.5%, and 0.792 (95%CI: 0.699-0.884), respectively. The HL test showed the nomogram has excellent accuracy (X2= 3.872, P=0.869). Conclusions: Consequently, the quantitative model incorporating FDG signatures and clinical features have great potential to be applied for individual accurate detecting EFGR mutations. The model and nomogram may benefit decision-making much when genetic tests are not available. Keywords: Epidermal growth factor receptor; Non-small cell lung cancer; Positron emission tomography; ER -