Abstract
Purpose: Our study aimed to develop and validate a predictive model based on chest CT radiomics features combined with clinical variables to evaluate the efficacy of anlotinib and its prognostic value in patients with advanced lung cancer. Methods: This single-center retrospective study included 68 patients with advanced lung cancer who received anlotinib monotherapy at a tertiary grade-A hospital in China between January 2021 and July 2024. All patients had experienced disease progression after receiving standardized chemotherapy, targeted therapy, or immunotherapy prior to enrollment, had not undergone radiotherapy, and had completed all prior treatments at our hospital. Chest CT scans were performed before anlotinib treatment, and radiomics features were extracted. Based on treatment response, patients were grouped, and a radiomics model, a clinical model, and a combined model were constructed. Model performance was evaluated using receiver operating characteristic (ROC) curves, the Hosmer-Lemeshow test, calibration curves, and decision curve analysis (DCA), with internal validation performed via bootstrap resampling (500 iterations). Additionally, Kaplan-Meier survival curves were generated for the high- and low-risk groups predicted by the combined model, and survival differences were compared using the log-rank test. Results: The areas under the ROC curve (AUCs) for the radiomics, clinical, and combined models were 0.721, 0.812, and 0.866, respectively, with the combined model significantly outperforming the other two models (DeLong test, P < .05). Internal validation showed an AUC of 0.866 (95% CI: 0.751-0.967) for the combined model. The integrated model demonstrated good calibration via Hosmer-Lemeshow testing (χ² = 7.81, P = .553) and showed significant clinical net benefit within the threshold probability range of 0.15-0.85 on decision curve analysis. Survival analysis revealed statistically significant differences between the actual treatment-responsive and non-responsive groups (log-rank P < .05), as well as between the model-predicted high-risk and low-risk groups (log-rank P < .05). Multivariable Cox regression confirmed the nomogram score derived from the integrated model as an independent predictor of overall survival (HR = 1.263, 95% CI: 1.090-1.463, P = .002). Conclusion: The combined model incorporating chest CT radiomics features and clinical characteristics demonstrated high accuracy and clinical utility in predicting the efficacy and prognosis of anlotinib treatment in patients with advanced lung cancer.