[Development and validation of a nomogram combining clinical and ¹⁸F-FDG PET/CT parameters for prediction of...

Objective: To establish and validate a nomogram based on clinical characteristics and metabolic parameters derived from ¹⁸F-fluorodeoxyglucose positron emission tomography and computed tomography (¹⁸F-FDG PET/CT) for prediction of high-grade patterns (HGP) in invasive lung adenocarcinoma. Methods: The clinical and PET/CT image data of 311 patients who were confirmed invasive lung adenocarcinoma and underwent pre-treatment ¹⁸F-FDG PET/CT scan in Beijing Hospital between October 2017 and March 2022 were retrospectively collected. The enrolled patients were divided into HGP group (196 patients) and non-HGP group (115 patients) according to the presence and absence of HGP. The data were divided into training set and validation set at 7∶3 ratio using R statistical software and simple random allocation. A nomogram prediction model was constructed in training set. The area under the curve (AUC) of receiver operating characteristic (ROC) was depicted in the training and validation set respectively for assessing the prediction efficacy. The goodness of fit, consistency between predicted and observed probability and clinical usefulness of the model were evaluated by Hosmer-Lemeshow test, calibration curve and decision curve analysis (DCA). Results: The age of 311 patients were (65.6±10.9) years and included 148 males (47.6%). In training set of 217 patients, 141 (65.0%) contained HGP while in validation set of 94 patients, 55 (58.5%) contained HGP. Gender in training set, serum carcino-embryonic antigen (CEA) in validation set, smoking history, clinical stage, cytokeratin fragments (CYFRA21-1), maximum standardized uptake value (SUV_max), mean standardized uptake value (SUV_mean), metabolic tumor volume (MTV), total lesion glycolysis (TLG) and maximum diameter (D_max) in both sets showed significant differences between HGP and non-HGP groups (all P<0.05). The variables integrated in the model were gender, clinical stage, CYFRA21-1, SUV_mean and TLG. The AUC (95%CI) of the ROC curve in training and validation set were 0.888 (0.844-0.932) and 0.925 (0.872-0.977), the sensitivity and specificity were 85.1%, 79.0% and 83.6%, 89.7%, respectively. The model showed good goodness of fit (training set: χ²=8.247, P=0.410, validation set: χ²=1.636, P=0.990). Calibration curve and DCA also indicated good consistency and clinical net benefit of the nomogram model. Conclusion: The nomogram model based on clinical features and metabolic parameters derived from ¹⁸F-FDG PET/CT could effectively predict the presence of HGP in invasive lung adenocarcinoma and be beneficial to treatment planning.