2024 Asia Conference on Lung Cancer (ACLC) - Poster & ePosters-PP02.25

PP02.25 - Yan Li

Pdf Summary

The study describes the development and validation of a predictive model for small-cell lung cancer (SCLC) transformation in patients with EGFR-mutant lung adenocarcinoma (LUAD) who have relapsed after treatment with EGFR-TKI. This transformation accounts for 3-14% of resistance cases and results in a more aggressive cancer. Transcriptomic analysis was conducted on pre-treatment samples from LUAD cases that experienced transformation (LUAD-BT) and those that did not (LUAD-NT). These analyses utilized the nCounter Pan-Cancer Pathways gene expression panel.<br /><br />In the study, immunohistochemistry (IHC) was applied to a broader cohort to validate candidate mRNAs. Model construction involved a LASSO-penalized logistic regression analysis, identifying six significant mRNAs: COL6A6, CASP12, HHIP, ZBTB16, BIRC3, and GATA2. These mRNAs exhibited distinct expression patterns between transformed and non-transformed LUADs.<br /><br />IHC analyses were performed to assess the H-score of the tumor region and all regions for each marker using QuPath-0.4.3. After further LASSO analysis, four H-scores (tumor H-score of BIRC3, COL6A6, CASP12, and total H-score of GATA2) were selected for the final model. Three machine learning algorithms were used for model training, with a Randomforest Model showing the best performance, achieving a sensitivity of 87.5% and a specificity of 100% with an AUC of 0.988 in an independent test cohort.<br /><br />The study concludes that predicting SCLC transformation remains a critical unmet need. The constructed model, employing four IHC markers, promises a reliable, cost-effective technique suitable for clinical practice, enhancing early detection and management of transformation in LUAD patients.

Keywords

small-cell lung cancer

EGFR-mutant lung adenocarcinoma

predictive model

transcriptomic analysis

immunohistochemistry

LASSO regression

machine learning

Randomforest Model

biomarkers

clinical practice