Statistical models to predict future subject’s lung cancer risk: application to NLST and PLCO data: Extended Incidence and Mortality
We fitted the NLST data relating to the lung cancer outcome with its baseline covariates. For each fitted model, we create a scoring system for predicting potential lung cancer risk and obtain a corresponding optimal stratification rule. The subpopulation of participants satisfying any given level of risk score can be identified accordingly. Then, all the resulting stratification strategies are evaluated via a conventional cross-validation process.
We illustrate the proposed methods using NLST chest X-ray group as the training and test set, and PLCO lung component as the independent validation set.
The aim of this study is to develop a quantitative stratification procedure for predicting potential cancer risk to identify individuals at higher risk of specific cancers.
SuChun Cheng, ScD, Dana-Farber Cancer Institute
Lu Tian, PhD, Standford University
L.J. Wei, PhD, Harvard University