(SLB-019) Predicting the Risk of Non-Small Cell Lung Cancer through Polygenic Approach: A US Based Study

Background: Non-small cell lung cancer (NSCLC) remains a major health challenge due to its high mortality rate and late-stage diagnosis. ​
Traditional risk assessment methods often fail to capture the complex genetic factors involved. Polygenic risk scores (PRS) offer a promising solution by integrating multiple genetic variants to predict individual risk more accurately. This approach can enhance early detection, personalize treatment plans, and ultimately improve patient outcomes by addressing the limitations of current diagnostic tools.​

Objectives: The objective of this study was to develop and validate a predictive model for NSCLC risk by integrating PRS from Optum’s Clinicogenomics data (OCD) i.e. Broad-panel next generation sequencing (NGS) results and demographic variables with scope to identify key features that enhance early detection.​

Methods: The Optum® de-identified Market Clarity Dataset was utilized to assess NSCLC patients in the US between January 1, 2020, and October 30, 2024, with 1) ICD9 and ICD10 codes for Lung cancer i.e. 2312, 162 % and C34%, D022% respectively 2) NSCLC cohort in OCD and 3) Continuous eligibility.​ Similar inclusion and exclusion criteria were followed to create control cohort to statistically validate PRS scoring. Allele frequency used as weights to calculate relative risk contribution and sum weighted scores were used to evaluate PRS. Logistic regression (LR) was performed to validate the association between the PRS and the NSCLC status. Data was randomly split into 80:20 ratios for model training and testing purpose. ​
Statistical validations were performed using ROC/AUC curve and f1-score. Scope is to further improve the methodology with inclusion of demographic variables, smoker, non-smoker status and co-existing comorbidity. ​

Results: The study cohort consisted of 20,339 patients. NSCLC cohort showed EGFR, MET and ALK showed highest PRS (270), and L858R in EGFR showed maximum association affecting 32% of NSCLC patients. The PRS of 270 in NSCLC suggests that patients have higher cumulative genetic risk compared to control cohort, indicating greater genetic predisposition to NSCLC based on their allele frequency. Whereas control group showed BRCA gene with highest PRS (150). Association of PRS with NSCLC was established using LR (AUC: 0.65). ​

​

Conclusions: Allele frequency percentages for evaluating polygenic risk score and combined with logistic regression for validation, provides a robust framework for NSCLC risk prediction. This approach can be valuable for genetic risk stratification in clinical and research settings.​