Anil Chaturvedi

PhD

National Cancer institute

Investigator

PLCO (Learn more about this study)

PLCO-269

Mar 31, 2017

Efficient estimation of AUC in nested case-control studies

Our group at the NCI/DCEG recently completed two nested case-control studies (EEMS 2009-00516 and EEMS 2012-00356) for the discovery and validation of the association of circulating levels of inflammation markers with risk of lung cancer within the PLCO Trial screening arm. We found that levels of four markers—CRP, SAA, sTNFRII, and CXCL9/MIG were reproducibly associated with risk of lung cancer. Additionally, through an analysis re-weighted to the PLCO screening arm cohort to account for selection probability into our study, we showed that the incorporation of circulating inflammation marker levels into risk prediction models with standard demographic and behavioral risk factors did not significantly improve model performance (e.g. sensitivity, specificity, and AUC). These analyses utilized both inflammation marker data and demographic/behavioral factor data on only the selected cases and controls in our study.

In the current application, we would like to utilize the same data to evaluate novel methods for more efficient estimation of the AUC in nested case-control studies. Briefly, the novel methodology proposes to use inflammation marker data on the cases and controls selected into our study, but utilize demographic and behavioral actors on the entire PLCO screening arm cohort. This project will be done in collaboration with Dr. Nilanjan Chatterjee and Dr. Parichoy Pal Choudhry at the Johns Hopkins University. The current CDAS application is for approval to share the data from our studies (EEMS 2009-00516 and EEMS 2012-00356 + the PLCO Screening arm cohort data) with our extramural collaborators.

Specific aims:

1. To develop novel methodology for efficient estimation of AUC in nested case-control studies through the use of marker data on selected cases and controls and questionnaire-based confounder data on the entire cohort.