The Development and Performance of Alternative Criteria for Lung Cancer Screening

Abstract

Background: The recommendation for lung cancer screening (LCS) developed by the U.S. Preventive Services Task Force (USPSTF) may exclude some high-benefit people.

Objective: To determine whether alternative criteria can identify these high-benefit people.

Design: Model-based projections.

Setting: United States.

Participants: People from the 1997-2014 National Health Interview Survey (NHIS) to develop alternative criteria using fast-and-frugal tree algorithms and from the 2014-2018 NHIS and the 2022 Behavioral Risk Factor Surveillance System for comparisons of USPSTF criteria versus alternative criteria.

Measurements: Life-years gained from LCS were estimated using the life-years gained from screening computed tomography (LYFS-CT) model. "High-benefit" was defined as gaining an average of at least 16.2 days of life from 3 annual screenings, which reflects high lung cancer risk and substantial life gains if lung cancer is detected by screening.

Results: The final alternative criteria were 1) people who smoked any amount each year for at least 40 years, or 2) people aged 60 to 80 years with at least 40 pack-years of smoking. The USPSTF and alternative criteria selected similar numbers of people for LCS. Compared with the USPSTF criteria, the alternative criteria had higher sensitivity (91% vs. 78%; P < 0.001) and specificity (86% vs. 84%; P < 0.001) for identifying high-benefit people. For racial and ethnic minorities, the alternative criteria provided greater gains in sensitivity than the USPSTF criteria (Black: 83% vs. 56% [P < 0.001]; Hispanic: 95% vs. 73% [P = 0.086]; Asian: 94% vs. 68% [P = 0.171]) at similar specificity. The alternative criteria identify high-risk, high-benefit groups excluded by the USPSTF criteria (those with a smoking duration of ≥40 years but <20 pack-years and a quit history of >15 years), many of whom are members of racial and ethnic minorities.

Limitation: The results were based on model projections.

Conclusion: These results suggest that simple alternative LCS criteria can identify substantially more high-benefit people, especially in some racial and ethnic groups.

Primary funding source: U.S. Department of Veterans Affairs Lung Precision Oncology Program.

Figure 1.

Simple criteria for identifying people for LCS. FFT ¼ fast-and-frugal tree; LCS ¼ lung cancer screening; USPSTF ¼ U.S. Preventive Services Task Force. Top. 2021 USPSTF criteria (single path). Bottom. Alternative simple criteria (2-path) trained on the outcome of ≥16.2 days of life gained with LCS (FFT16.2): a long smoking duration path (≥40 smoke-years), and a higher smoking intensity and older age path (≥40 pack-years and age 60 to 80 years).

Figure 2.

Comparisons of sensitivity, specificity, and population size of FFT16.2 and USPSTF2021. FFT_16.2 = fast-and-frugal trees trained on the outcome of ≥16.2 days of life gained with lung cancer screening; LCS = lung cancer screening; LYFS-CT ¼ life-years gained from screening computed tomography; NHIS ¼ National Health Interview Survey; USPSTF₂₀₂₁ = U.S. Preventive Services Task Force 2021 criteria. Top. Comparison of the sensitivity, specificity, and population size of FFT_16.2 and USPSTF₂₀₂₁ for the entire population and based on smoking status. The analysis was done in the FFT evaluation test set and was based on sensitivity and specificity to identify high-benefit people for LCS, defined on the basis of the LYFS-CT threshold of ≥16.2 days of life gained. Numerical sensitivity and specificity values, CIs, and P values are provided in Supplement Table 4a (available at Annals.org). Bottom. Comparison of the sensitivity, specificity, and population size of FFT_16.2 and USPSTF₂₀₂₁ based on race and ethnicity. The analysis was done in the NHIS evaluation test set and was based on sensitivity and specificity to identify high-benefit people for LCS, defined on the basis of the LYFS-CT threshold of ≥16.2 days of life gained. Numerical sensitivity and specificity values, CIs, and P values are provided in Supplement Table 4b (available at Annals.org).