Higher Risk Threshold Better for Selecting Ever-Smokers for Lung Cancer Screening
, by DCEG Staff
A new study by investigators from the National Cancer Institute used population data to model three different scenarios for selecting ever-smokers for lung cancer screening with low-dose CT. They determined that a new risk threshold could lower the ratio of individuals needed to screen in order to prevent one lung cancer death, improving on the performance of current United States Preventive Services Task Force (USPSTF) guidelines. The USPSTF is considering recommending externally validated risk assessment models for screening. The findings were published June 3, 2019, in Annals of Internal Medicine.
The researchers compared guidelines from the USPSTF, which uses smoking history and age, the National Comprehensive Cancer Network (NCCN), which uses a threshold of 1.3 percent 6-year risk of incident lung cancer, as well as a third approach with a higher risk threshold of 2.19 percent. The 2.19 percent risk threshold would select the same total number of ever-smokers in 2015 as the USPSTF guidelines, but with a higher risk profile. As a result, the 2.19 percent threshold was projected to prevent 14 percent more lung cancer deaths than the USPSTF guidelines, while screening the same number of people. In contrast, the NCCN threshold was projected to screen 57 percent more people while preventing 37 percent more deaths than USPSTF. The authors project that 194 would need to be screened by the USPSTF guidelines to prevent one death. This improves to 169 by the 2.19 percent threshold but worsens to 222 by the 1.3 percent NCCN threshold.
The authors conclude that their findings indicate the importance of regularly evaluating risk thresholds/models for screening guidelines to ensure they perform as expected.
Reference:
Landy R, Cheung LC, Berg CD, Chaturvedi AK, Robbins HA, Katki HA. Contemporary implications of USPSTF and risk-based guidelines for lung cancer screening eligibility in the US. Annals of Internal Medicine. June 4, 2019. DOI: 10.7326/M18-3617. [Epub ahead of print].