Studies to use extended NLST follow-up data
1. Validate all the lung cancer prediction models in the longer NLST follow-up, especially by race/ethnicity. This is part of our current work on seeing if use of prediction models might reduce health disparities in lung screening eligibility. We will use the NLST results as part of a meta-analysis of 4 cohorts as the best current answer to this question. We will examine the AUC and the Expected-to-Observed (E/O) ratio, to examine the discrimination and calibration, respectively, of the prediction models.
2. Examine the statistical properties of NLST CT screening for reducing mortality, and for examining natural history parameters. We plan to examine several issues:
1. We plan to calculate the per-screen time-dependent hazard-ratio for the mortality reduction using the methods of: Hanley JA, Njor SH. Disaggregating the mortality reductions due to cancer screening: model-based estimates from population-based data. Eur J Epidemiol. 2018 May;33(5):465-472.
2. We plan to calculate average lead time for lung cancer by different factors (histology, gender, etc)
3. We plan to calculate transition rates from early-stage to late-stage cancer, hopefully by different factors (histology, gender, etc)
4. We are developing our own simulation model of the natural history of lung cancer and want to examine its fit to NLST data
Anil K. Chaturvedi
Li C. Cheung