Risk-based models like LCRAT demonstrated significantly higher sensitivity (0.632 vs 0.451; p=0.01) and discrimination (AUC 0.720 vs 0.629; p=0.01) for lung cancer screening than USPSTF-2021 criteria.
Cross-Sectional (n=30,163)
Sí
Do risk-based and life-years-gained models improve lung cancer detection and life years gained compared to USPSTF 2021 criteria in adults aged 50-80 who ever smoked?
Risk-based and life-years-gained models for lung cancer screening eligibility offer improved cancer detection, better discrimination, and more life years gained compared to the USPSTF 2021 criteria.
Tasa de eventos absoluta: 0.72% vs 0.629%
valor p: p=0.01
Abstract Rationale Risk based and life-years gained models to determine lung cancer screening (LCS) eligibility have more optimal test characteristics than previous LCS guidelines but there is little direct comparison to United States Preventive Services Task Force 2021 (USPSTF-2021) criteria. This study compares the performance of risk-based and life-years-gained models with the USPSTF-2021 criteria using data from the National Institute of Health All of Us Research Program, a demographically diverse cohort that reflects the LCS eligible population of the United States. Methods We performed a cross-sectional study of adults aged 50-80 years who ever smoked in the All of Us (AoU) cohort. We compared the performance of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial Model 2012 (PLCOm2012), Lung Cancer Death Risk Assessment Tool (LCDRAT), Lung Cancer Risk Assessment Tool (LCRAT), and Life-years Gained From Screening (LYFS-CT) models to USPSTF 2021 eligibility criteria. We calculated risk thresholds for each model that resulted in the same number of LCS eligible patients as USPSTF-2021. We used lcrisks (R-software) to calculate life years gained from screening and compared models to USPSTF-2021 using descriptive statistics, lung cancer sensitivity using the Z score test, discrimination with area under the curve (AUC) analysis, and number needed to screen (NNS). Results Of the 30,163 cohort participants who met inclusion criteria from 5/6/2018 through 6/15/2025, 144 (0.5%) were diagnosed with lung cancer and 5,816 (19.2%) were eligible for screening by USPSTF 2021 criteria. The sensitivity estimates of LYFS-CT (0.569; p = 0.04), LCDRAT (0.625; p = 0.01) and LCRAT (0.632;p=0.01) models were significantly higher, while the sensitivity of PLCOm2012 (0.535;p0.15) was not different, in comparison to USPSTF 2021(0.451). The AUC of all four models PLCOm2012 (0.671;p=0.04), LCDRAT (0.717;p=0.01), LCRAT (0.720;p=0.01), LYFS-CT (0.689; p = 0.01) were significantly higher than the AUC of USPSTF-2021 (0.629). The total life years gained (LYG) by screening the eligible population were also higher for all models LYFS-CT 456.7 LYG; 95% CI 448.5, 464.9, LCDRAT 441.7 LYG; 95% CI 433.2, 450.2, LCRAT 441.1 LYG; 95% CI 435.6,452.5, and PLCOm2012 411.8 LYG; 95% CI 403.0,420.5 compared to USPSTF-2021 361.6 LYG, 95% CI 352.3, 370.8. The NNS to detect one lung cancer was 74 for LCRAT, 75 for LCDRAT, 87 for LYFS-CT, 95 for PLCOm2012, and 126 for USPSTF-2021. Conclusions In this diverse contemporary cohort of US patients, model-based eligibility for LCS resulted in improved cancer detection and discrimination, more life years gained, and a lower NNS compared to USPSTF-2021 eligibility. This abstract is funded by: NIH
Caruso et al. (Fri,) conducted a cross-sectional in Lung cancer screening (n=30,163). Risk-based and life-years-gained models (PLCOm2012, LCDRAT, LCRAT, LYFS-CT) vs. USPSTF 2021 eligibility criteria was evaluated on Discrimination with area under the curve (AUC) (p=0.01). Risk-based models like LCRAT demonstrated significantly higher sensitivity (0.632 vs 0.451; p=0.01) and discrimination (AUC 0.720 vs 0.629; p=0.01) for lung cancer screening than USPSTF-2021 criteria.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: