(Learn more about this study)
Initial CDAS Request Approval
Sep 28, 2021
Evaluation of Bach lung cancer risk prediction model in a lung screening cohort
Unlike several other lung cancer risk prediction models that were derived and validated in large measure on data from trials of lung cancer screening, the Bach model was derived on data from a randomized intervention trial (CARET) and validated in population based cohorts. There are important potential differences between these cohorts that could affect the performance of the Bach model. This project will assess the performance of the Bach model in the NLST CT and CXR arms and attempt to measure to what extent it provides reliable prediction of lung cancer incidence in a screening context.
-Derive Bach risk prediction estimates for participants in the NLST study
-Measure performance of the model separately in the CT and CXR arms of the study, applying conventional measures of model performance including the AUC and measures of calibration.
-Examine whether risk prediction shortcomings, if identified, are explained by stage distribution
-Test whether model variables can be made more parsimonious without loss of predictive accuracy, specifically testing whether the asbestos variable, sex variable, and three separate measures of smoking history are required for prediction.