Abstract C038: Addition of clinical demographic covariates to arsenic-associated lung cancer risk model improves performance

Abstract Background: Four robust genetic biomarkers were recently discovered from a pool of 88 arsenic-associated genes significant for lung cancer prediction and implemented in a risk model created using logistic regression which achieved high accuracy with an AUC of 0.886 (95% CI: 0.838-0.932) on a curated lung cancer gene expression dataset using receiver operating characteristic (ROC) curve analysis. These genes are significantly modulated by arsenic exposure, suggesting their relevance in toxicogenomic-based lung cancer detection. This research examines whether the incorporation of clinical covariates including age, gender, and smoking status affect the performance of the predictive model. Methods: Logistic regression modeling was performed on this dataset with the addition of the clinical covariate metadata associated the patient samples, which included smoking status, age, and gender, with predictive accuracy evaluated using receiver operating characteristic (ROC) curve analysis. Eight clinical covariate combinations were assessed with the model: 1. The 4 genes alone 2. The 4 genes + smoking status 3. The 4 genes + age 4. The 4 genes + gender 5. The 4 genes + smoking status + age 6. The 4 genes + smoking status + gender 7. The 4 genes + age + gender 8. The 4 genes + smoking status + age + gender Results: The model performance based on each case is as follows: The 4 genes alone result in an AUC of 0.894 (95% CI: 0.857 – 0.931). The 4 genes + smoking status yield an AUC of 0.926 (95% CI 0.898 – 0.953). The 4 genes + age give an AUC of 0.911 (95% CI 0.879 – 0.942). The 4 genes + gender express an AUC of 0.929 (95% CI 0.902 – 0.955). The 4 genes + smoking status + age have an AUC of 0.916 (95% CI 0.886 – 0.947). The 4 genes + smoking status + gender output an AUC of 0.926 (95% CI 0.898 – 0.953). The 4 genes + age + gender show an AUC of 0.929 (95% CI 0.903 – 0.955). The 4 genes + smoking status + age + gender produce an AUC of 0.929 (95% CI 0.903 – 0.955) Conclusion: This study evaluated the predictive performance of an arsenic-associated gene expression lung cancer risk model with the inclusion of the clinical covariates age, gender, and smoking status. the gene-only model yielded an AUC of 0.894 (95% CI: 0.857–0.931), indicating strong baseline discriminative power. Incorporation of clinical variables progressively improved model performance, with the combined model (genes + smoking + age + gender) achieving the highest AUC of 0.929 (95% CI: 0.903–0.955). All models were statistically significant (p < 0.001), supporting the relevance of clinical variables to lung cancer risk prediction. Citation Format: Joshua D. Kennedy, Sonalika Singhal, Jack Pendleton, Jace Howard, Dhanush Tummala, Sandeep Singhal. Addition of clinical demographic covariates to arsenic-associated lung cancer risk model improves performance [abstract]. In: Proceedings of the 18th AACR Conference on the Science of Cancer Health Disparities; 2025 Sep 18-21; Baltimore, MD. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2025;34(9 Suppl):Abstract nr C038.