Abstract PR001: A comparison of modeling approaches to assess the interplay of social, behavioral, and environmental factors on lung cancer racial disparities

Abstract Purpose: While risk factors for lung cancer incidence are well established, our understanding of factors driving an increased risk of lung cancer among Black/African American populations remains incomplete. Methods: We assembled data from five U.S. lung cancer case-control studies matched on race, gender and age: the Southern Community Cohort Study, the Inflammation, Health, Ancestry and Lung Epidemiology study, the Northern California Lung Cancer Study, three Detroit area lung studies, and the NCI-Maryland Lung Cancer Study. Incident lung cancers were ascertained from state and regional cancer registries or National Death Index mortality records. Social, behavioral, and environmental factors were ascertained by in-person interviews or mailed-in questionnaires. Participant's residential addresses were geocoded and area-level area deprivation and environmental justice variables were linked at the census block. We evaluated interactions between known lung cancer risk factors and race using multilevel logistic regression analyses and novel tree-based approaches to identify potential factors contributing to racial disparities. Tree-based approaches included a conditional matched classification tree analysis designed to detect complex interactions, as well as a specialized partially recursively induced structured moderation (PRISM) analysis to study variation in disparity in lung cancer risk across levels of educational status. Results: Among 5,829 lung cancer cases and 10,671 controls (64% self-identified Black/African American and 36% White), the average age was 60 years, 50% were female, and 24% had less than a high school education. Conditional logistic regression models fit after imputation of missing data identified increased risk of lung cancer associated with personal history of cancer, family history of lung cancer, current smoking status, smoking duration, smoking pack-years, and higher neighborhood deprivation; whereas, reduced lung cancer risk was associated with higher education and increased BMI. Similar patterns were observed among Black/African Americans only, with the addition of secondhand smoke exposure. Assessment for interactions with race identified significant interactions between race and body mass index (BMI) (p-value 0.0001), PM2.5 (p= 0.010) and gender (p=0.001) and lung cancer risk. A conditional matched classification tree analysis identified both increased and decreased risk subgroups defined by multi-way interactions of similar variables that were identified as significant main effects in the conditional logistic regression analysis. Smoking pack-years and BMI were identified as the most important determinants of heterogeneity of disparity in lung cancer risk across educational levels in our study population. Conclusions: The use of both traditional and novel tree-based based approaches provide approaches for identifying factors contributing to racial disparities. Future studies should assess the interplay of lung cancer risk factors to inform development of targeted interventions. Citation Format: Melinda C. Aldrich, Chen Zhao, Christine M Lusk, Lucie McCoy, James Jaworski, Michael Mumma, Catherine Pichardo, Michael Betti, John K Wiencke, Brid Ryan, Ann G Schwartz, Sunil Rao. A comparison of modeling approaches to assess the interplay of social, behavioral, and environmental factors on lung cancer racial disparities abstract. In: Proceedings of the 17th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2024 Sep 21-24; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2024;33(9 Suppl):Abstract nr PR001.