Predicting the prognosis of patients with lung adenocarcinoma treated with third-generation EGFR-TKI alone using nomograms based on CT radiomic and clinicopathological factors.

Background: Considerable differences exist in the prognosis of patients suffering from advanced non-small cell lung cancer. In fact, they undergo third-generation epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKI) single-agent targeted treatment. This study aimed to recognize the predictive factors influencing the prognosis of such patients, and identify patients at high risk of relapse to instruct clinical therapy and ameliorate prognosis. Methods: Our research included 255 patients suffering from advanced lung adenocarcinoma (LUAD) treated with third-generation EGFR-TKI monotherapy alone, categorized into the training and validation cohorts. In univariate and multivariate analyses, clinicopathologic features and radiomic features were included. Moreover, statistically significant predictive factors were applied to formulate a nomogram. Area under the curve (AUC) of receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) were utilized to verify the model. Results: Neutrophil-lymphocyte ratio (NLR) ≥4.6, EGFRex21 L858R mutation, brain metastasis, and radiomic characteristics worked as independent risk factors for progression-free survival (PFS). Age ≥60 years, EGFRex21 L858R mutation, brain metastasis, monocyte-lymphocyte ratio (MLR) ≥0.3, and radiomic features acted as independent risk factors for overall survival (OS). Incorporating this into the nomogram, the AUC of the nomogram for forecasting 6-, 12-, and 24-month PFS and 1-, 2-, and 3-year OS were 0.810, 0.862, 0.873, and 0.886, 0.881, and 0.839, respectively, in the training cohort, and 0.885, 0.858, 0.847 in the validation cohort, and 0.804, 0.824, 0.806, all showing excellent discrimination. The calibration curves showed great consistency between predicted and actual observations. DCA exhibited a gratifying positive net benefit in most threshold probabilities, indicative of a beneficial clinical result. As claimed by PFS and OS survival analysis, the nomogram could distinguish low-risk group from high-risk group. Conclusions: Our research formulated and corroborated a nomogram, according to clinicopathologic factors and radiomic features. Such a prediction model could be utilized as a strong tool for forecasting the prognosis of patients.