Development and Validation of a Pathomics-Based Prognostic Model for Patients with Lung Adenocarcinoma Undergoing First-Line EGFR-TKI Therapy.

BACKGROUND

Accurate prognostic prediction is crucial for personalized treatment of patients with lung adenocarcinoma (LUAD) receiving epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs). This study aims to develop and validate a pathomics-based prognostic model for EGFR-TKI-treated patients with LUAD.

PATIENTS AND METHODS

Data from 122 patients with LUAD who underwent first-line EGFR-TKI therapy were retrospectively analyzed. Pretreatment whole-slide images of hematoxylin and eosin (H&E)-stained biopsy specimens were collected for annotation and feature extraction. Maximum relevance minimum redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) Cox regression were applied to select features associated with disease progression. The selected features were used to construct the pathomicsScore, and its clinical relevance was assessed via Kaplan-Meier analysis. A predictive model incorporating both pathomicsScore and clinical risk factors was developed.

RESULTS

Five pathomics features associated with disease progression were identified, and a pathomicsScore was developed to stratify patients into low- and high-risk groups. PFS analysis revealed longer survival in the low-risk group. Both pathomicsScore and pathological stage were independent predictors of disease progression and were integrated into a predictive model. The model achieved area under the curve (AUCs) of 0.789 and 0.728, sensitivity of 0.909 and 1, and specificity of 0.677 and 0.714 in the training and validation cohorts. Time-dependent receiver operating characteristic (ROC) curves at 6, 12, and 18 months validated the model's predictive performance. Calibration curves showed excellent agreement between predicted and observed progression probabilities. Decision curve analysis confirmed the clinical utility of the model.

CONCLUSIONS

The pathomics-based model effectively predicts disease progression in patients with LUAD receiving EGFR-TKI therapy, enabling personalized treatment strategies.