Prediction of recurrence in stage I EGFR mutation-positive NSCLC: Combination of CT appearance and selected co-occurring gene alterations by machine learning.

Le résumé fourni par la source

8053 Background: Predicting the risk of postoperative recurrence is becoming increasingly important in patients (pts) with resectable EGFR mutation positive ( EGFRm) non-small cell lung cancer (NSCLC) after the ADAURA trial. The presence of non-solid ground-grass opacity (GGO) component on computed tomography (CT) images is known to affect the prognosis of stage I lung cancer. Only a few reports investigating the relationship between gene alterations and CT images in stage I EGFRm NSCLC. Methods: We have developed a machine learning-based model to predict recurrence within five years in TNM stage I (UICC 8th) EGFRm NSCLC pts who underwent surgery between 1985 and 2019 using whole exome sequencing in the PRISM project. We evaluated the pts’ characteristics, recurrence-free survival (RFS), the CT appearance (pure GGO and part solid [GGO]), without GGO [pure solid]), and consolidation tumor ratio (CTraio). Then, we analyzed the correlation between the image findings and gene alterations predicting a high risk of recurrence. Results: Among 1351 pts, stage I, EGFRm were 308 (22.8%). The median RFS for stage I EGFR-m pts was 123.2 months (m). In the prediction model, TP53 and RBM10 genes were among the gene alterations that had a high impact on the high-risk recurrence within five years. Among the 302 stage I, EGFRm pts for whom CT images were available, 166 (55.6%) pts had GGO, and 137 (45.4%) pts had pure solid appearance. The median CTratio was 0.87. The median RFS was not reached (NR) for GGO, 86.5m for pure solid (hazard ratio [HR] 2.42 [1.58-3.73], p<0.0001). There was a negative correlation indicating that the larger the CTratio, the shorter the RFS (p<0.0047). The proportion of TP53 mutation (TP53m) was lower in GGO and higher in pure solid (p<0.0001, Table). The median RFS was NR for GGO plus TP53m-negative pts, was 119.7m for GGO plus TP53m-positive pts, 84.8m for pure solid plus TP53m-negative pts, and 98.8m for pure solid plus TP53m-positive pts, respectively. Among pts with 0≦CTratio<0.5 46 (15.2%), with 0.5≦CTratio<1 119 (39.4%), and with CTratio=1 137 (45.4%), the proportion of TP53m increased as the CTratio increased. Conclusions: We found a linear relationship between CTratio and proportion of co-occurring TP53m with EGFRm. Combination of CT appearance (GGO or pure solid) and co-occurring TP53m can predict recurrence in stage I, EGFRm NSCLC. [Table: see text]

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.