Improving implementation of lung cancer screening with risk prediction models
Menée aux Etats-Unis à partir de données portant sur 409 726 fumeurs, cette étude compare la sensibilité et la spécificité de neufs modèles pour prédire le risque de cancer du poumon et identifier ainsi les personnes nécessitant un examen de dépistage par tomographie numérique
In this issue, Katki and colleagues compare the performance of 9 lung cancer risk models in the National Institutes of Health–AARP Diet and Health Study and the CPS-II (Cancer Prevention Study II) Nutrition Survey cohort (1). Each model used National Health Interview Survey data from 2010 to 2012 to estimate the number of U.S. persons eligible for screening. Four models (the Bach model; the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial Model 2012 [PLCOM2012]; the Lung Cancer Risk Assessment Tool [LCRAT]; and the Lung Cancer Death Risk Assessment Tool [LCDRAT]) demonstrated superior performance, selected similarly sized screening populations (7.6 million to 10.9 million ever-smokers), and generally agreed on which ever-smokers to screen. This study confirms findings from previous studies that also found superior performance of the PLCOM2012 (2) and the Bach model (3, 4).
Annals of Internal Medicine , éditorial, 2017