POLYGENIC RISK SCORES IMPROVE NON-SMALL CELL LUNG CANCER PREDICTION IN REAL-WORLD US DATA
Author(s)
Vikash Kumar Verma, MBA, PharmD1, Louis Brooks Jr, MS2, Marissa Seligman, PharmD3, Abhimanyu Roy, MBA4, Abhinav Nayyar, MBA, MBBS5, Ankitkumar Arora, MPharm6, Anuj Gupta, MSc7, Ram Kumar Mishra, PhD8, Pallavi Mohanty, PhD9, Vishan Khatavkar, MBA10.
1Optum Lifesciences, Boston, MA, USA, 2Optum, Bloomsbury, NJ, USA, 3Optum, Winchester, MA, USA, 4Optum, Gurgaon, India, 5Optum Life Sciences, Gurugram, India, 6Optum Global Solutions, Gurgaon, India, 7Optum Lifesciences, Noida, India, 8Optum Global Solution, Gurugram, India, 9Optum Global Solutions, Gurugram, India, 10Optum Lifesciences, Gurugram, India.
1Optum Lifesciences, Boston, MA, USA, 2Optum, Bloomsbury, NJ, USA, 3Optum, Winchester, MA, USA, 4Optum, Gurgaon, India, 5Optum Life Sciences, Gurugram, India, 6Optum Global Solutions, Gurgaon, India, 7Optum Lifesciences, Noida, India, 8Optum Global Solution, Gurugram, India, 9Optum Global Solutions, Gurugram, India, 10Optum Lifesciences, Gurugram, India.
OBJECTIVES: Non-small cell lung cancer (NSCLC) remains a leading cause of mortality; early risk stratification is essential for targeted screening. We accessed the clinical utility of a polygenic risk score (PRS)-based prediction model for NSCLC in a US real-world cohort and its incremental value over clinical risk factors.
METHODS: We conducted a retrospective cohort study using Optum® Market Clarity data between January-2013 to June-2024. Adult patients with confirmed NSCLC were identified, and the index date was defined as the diagnosis of NSCLC. A 12-month pre-index (baseline), and post-index (follow-up) were used. Genotype data were available for a subset of patients, and PRS was constructed using cumulative allele frequency from patient level variant information. Logistic regression models assessed associations between PRS and NSCLC risk, adjusting for demographics, smoking, comorbidities, and top genomic alterations. Model performance was evaluated using AUC, precision, recall, and odds ratios (OR).
RESULTS: From an initial NSCLC cohort of 43,104 patients, 6,227 met inclusion criteria for PRS analysis. The PRS-based logistic regression model achieved an AUC of 0.82, with precision 63.5% and recall of 65.3%. The confusion matrix exhibited strong predictive accuracy (true positives = 800, true negatives = 2,053). Higher PRS quintiles were associated with increased NSCLC risk. Odds ratio analysis indicated significant associations for clinical features such as lung nodules (OR = 7.0) and hemoptysis (OR = 2.0). Genomic alterations including KRAS G12C, EGFR I858R, and EGFR E746_A750deletion exhibited modest associations (OR ≈ 1.1 each). The most frequent alterations were TP53 splice variants (n = 4,045), KRAS G12C (n = 1,897), and PIK3CA E545K (n = 1,080). Median effect allele sizes were highest for PIK3CA H1047R (3.1) and EGFR E746_A750 deletion (2.9).
CONCLUSIONS: PRS-based prediction significantly improved NSCLC risk stratification beyond clinical and genomic factors. Combining PRS with mutation profiles may support personalized screening strategies and inform cost-effective prevention approaches.
METHODS: We conducted a retrospective cohort study using Optum® Market Clarity data between January-2013 to June-2024. Adult patients with confirmed NSCLC were identified, and the index date was defined as the diagnosis of NSCLC. A 12-month pre-index (baseline), and post-index (follow-up) were used. Genotype data were available for a subset of patients, and PRS was constructed using cumulative allele frequency from patient level variant information. Logistic regression models assessed associations between PRS and NSCLC risk, adjusting for demographics, smoking, comorbidities, and top genomic alterations. Model performance was evaluated using AUC, precision, recall, and odds ratios (OR).
RESULTS: From an initial NSCLC cohort of 43,104 patients, 6,227 met inclusion criteria for PRS analysis. The PRS-based logistic regression model achieved an AUC of 0.82, with precision 63.5% and recall of 65.3%. The confusion matrix exhibited strong predictive accuracy (true positives = 800, true negatives = 2,053). Higher PRS quintiles were associated with increased NSCLC risk. Odds ratio analysis indicated significant associations for clinical features such as lung nodules (OR = 7.0) and hemoptysis (OR = 2.0). Genomic alterations including KRAS G12C, EGFR I858R, and EGFR E746_A750deletion exhibited modest associations (OR ≈ 1.1 each). The most frequent alterations were TP53 splice variants (n = 4,045), KRAS G12C (n = 1,897), and PIK3CA E545K (n = 1,080). Median effect allele sizes were highest for PIK3CA H1047R (3.1) and EGFR E746_A750 deletion (2.9).
CONCLUSIONS: PRS-based prediction significantly improved NSCLC risk stratification beyond clinical and genomic factors. Combining PRS with mutation profiles may support personalized screening strategies and inform cost-effective prevention approaches.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR163
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
SDC: Oncology