Abstract A106: Are we there yet? Performance of risk-model based eligibility for lung cancer screening

Abstract

Abstract Introduction Lung cancer screening rates in the US remain low. Evidence suggests that risk-model based approaches may outperform current risk-factor based criteria in determining screening eligibility. We compared the USPSTF 2021 lung cancer screening eligibility criteria to three existing risk prediction models in a large and racially diverse population of smoking individuals. Methods We identified current and former smokers from a predominantly Black American study population recruited from community health centers across 12 southeastern US states from 2002 to 2009 and followed until 2019. Smoking behaviors were collected by in-person interviews or mailed-in questionnaires using three follow-up surveys. Incident lung cancers were ascertained from state cancer registries and National Death Index mortality records. Screening eligibility was based on USPSTF 2021 criteria or on pre-specified risk thresholds for each risk prediction model. Performance was assessed by sensitivity (the percentage of individuals who developed lung cancer that were deemed eligible for screening) and specificity (the percentage of those without lung cancer excluded from screening). Disparities by race and sex were evaluated in each subgroup. Results Of 52,911 ever smokers (64% Black, 31% White, 4% Other), a total of 1,705 developed lung cancer. More than a third of all persons with a history of smoking were screening eligible across all models: USPSTF 41%; Prostate, Lung, Colorectal and Ovarian Cancer (PLCO) m2012 39%; Lung Cancer Risk Assessment Tool (LCRAT) 43%; Lung Cancer Death Risk Assessment Tool (LCDRAT) 43%. All models exhibited nearly identical sensitivity (correctly identifying persons with lung cancer as eligible for screening) and specificity (excluding participants without lung cancer from screening): USPSTF Sensitivity: 0.61, Specificity: 0.59; PLCOm2012 Sensitivity: 0.62, Specificity: 0.60; LCDRAT Sensitivity: 0.62, Specificity: 0.58; LCDRAT Sensitivity: 0.60, Specificity: 0.58. Compared to Black persons with lung cancer, White persons with lung cancer were more likely to be eligible for screening across all models. Only 54-57% of Black persons with lung cancer were eligible for screening compared to 70-73% of White persons with lung cancer. The smallest racial disparities were observed using the USPSTF criteria (USPSTF sensitivity ratio for Whites to Blacks (SR [95% CI]): 1.27 [1.17-1.36]; PLCOm2012: 1.28 [1.17-1.39]; LCRAT: 1.29 [1.18-1.40]; LCDRAT: 1.33 [1.22-1.45]). There was no evidence of sex disparities greater than 1.1 RR in any of the models, with most estimated effects being less than 1+/- 0.05 RR. Conclusion In a racially diverse population of smoking individuals, we found similar performance of risk models vs USPFTF 2021 criteria for lung cancer screening. Racial disparities in screening eligibility persisted across all models, with the USPSTF 2021 criteria resulting in the smallest disparity. Given the poor performance of these models, there remains significant room for improvement. Citation Format: Adoma A. Manful, Megan H. Murray, Sarah F. Mercaldo, Jeffrey D. Blume, Melinda C. Aldrich. Are we there yet? Performance of risk-model based eligibility for lung cancer screening [abstract]. In: Proceedings of the 15th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2022 Sep 16-19; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2022;31(1 Suppl):Abstract nr A106.