Jae Lee

PhD

Moffitt Cancer Center

Senior Member and Interim Chair of Biostatistics and Bioinformatics

PLCO (Learn more about this study)

PLCO-248

Jan 12, 2017

Absolute Risk Model Evaluation and Integration of Imaging Biomarkers for Early Detection and Risk Assessment of Cancer

Early detection of cancer is well recognized as one of the most important ways to reduce mortality from cancer. As such, absolute risk models have been developed and widely used for more effective screening and detection of cancer. Some of the most widely used risk models include Gail model (Gail et al., JNCI, 1989), BRCAPro model (Parmigiani et al., AJHG, 1998), and Tyrer-Cuzick model (Tyrer et al., Stat. in Med., 2004) for breast cancer and the ones reported in Tammemagi et al. (NEJM, 2013) and Katki et al. (JAMA, 2016) for lung cancer.
The National Cancer Institute has established the Early Detection Research Network (EDRN) program to promote and expedite the development and clinical translation of techniques and diagnostics assays for early risk assessment and detection of cancer. We are one of the recently-funded Clinical Validation Centers, Moffitt Imaging Biomarker Validation Center (MIBVAC), under the EDRN program for the coming five years (U01-1018693, 2016-2021; Jae Lee, PhD, Overall PI). For our MIBVAC study, we plan to extensively evaluate and utilize risk models for an integrated modeling with lung and breast imaging biomarkers in order to obtain enhanced biomarker models for early risk assessment and detection of cancer. We have two aims for our investigation on the PLCO data.
Aim 1. To evaluate current absolute risk models of lung, breast, and other cancer for their accuracy such as sensitivity and specificity and clinical utility such as positive predictive value (PPV) and negative predictive value (NPV), and cost-effectiveness, considering various cost factors from screening and reduced clinical care expenditure from early detection on target screening population.
Aim 2. To integrate lung LDCT and breast DBT imaging biomarkers with validated risk models for improved early detection of lung, breast, and other cancer.
In order to perform our investigation on these aims, we first plan to examine and evaluate current risk models for their important performance indices such as sensitivity, specificity, PPV, and NPV as well as for different degrees of impact among different risk factors at varying risk levels. Also, variability of these performance criteria will need to be understood among different sub-populations of cancer screening. We plan to extensively use the large PLCO cohort data to evaluate current risk models for the above criteria. We will then expand our study to integrate our imaging biomarkers with some of validated risk models for enhanced early detection and risk assessment of lung, breast, and other cancer.