Analytic diagnosis methods for disease ruling

PROJECT SUMMARY/ABSTRACT Disease ruling, that is, ruling-out or ruling-in, is a main objective of cancer diagnosis to aid in medical decision making. Taking prostate cancer as our primary example, treatment of intermediate and high risk, early stage prostate cancer has shown survival benefits, but not so for low risk cancer. Meanwhile, prostate cancer screen- ing via prostate-specific antigen (PSA) testing has led to excess unnecessary prostate biopsies and over-detection of low risk cancer. New strategies are urgently needed for diagnosing aggressive cancer, while avoiding un- necessary prostate biopsies and over-treatment of indolent cancer. Ruling out or ruling in aggressive prostate cancer at various clinical decision points in patient care are thus essential. However, the lack of a robust ana- lytic framework and methodology for disease ruling is one major roadblock. To address this pressing analytic need, this project will develop a novel performance metric directly tied to clinical utility so as to establish an analytic paradigm for disease ruling. Furthermore, robust and efficient statistical and computational methods will be developed for the formulation and simultaneous performance estimation of multiplex diagnostic rules. Five specific aims will be pursued. Aim 1 will focus on the conceptual development of the proposed analytic paradigm as well as disease ruling methods with single predictors. Aim 2 considers disease ruling with multiple predictors where these predictors are combined via a standard procedure, e.g., logistic regression. Aims 3 and 4 investigate multiplex disease ruling via empirical optimization in linear and Boolean classes, respectively. As Aims 1–4 target either ruling-out or ruling-in, Aim 5 investigates simultaneous ruling-out and ruling-in. These proposed analytic methods will be thoroughly investigated through rigorous asymptotic studies and extensive simulations. They will be applied to a number of our prostate cancer biomarker studies, which motivated this project, from the National Cancer Institute's Early Disease Research Network (EDRN). User-friendly computer software will be made available to the research community. These proposed methods will facilitate more effective and clinically relevant diagnostics research for cancers as well as other diseases.