428 – Ethical and Policy Issues in the Statistical Use of Big Data
Ethical Issues in Getting It Right: Using Big Data to Determine Medical Treatment
John Crowley
Cancer Research and Biostatistics
Pingping Qu
Cancer Research and Biostatistics
Bart Barlogie
University of Arkansas for Medical Sciences
Medical tests have long been used to determine treatment, either through eligibility criteria for a clinical trial, or to decide on treatment once a patient is on a trial. What is new with Big Data is that there are many more opportunities for getting it wrong. I will use examples to demonstrate approaches to minimize such risks. Important steps in getting it right include careful data management, adherence to principles of reproducible research, and validation, validation, validation. Data management methods and reproducible research will be discussed in the context of SWOG, one of the national cancer cooperative groups. Our work with gene expression profiles (gep) of the tumors of patients with multiple myeloma will be used to illustrate the principle of validation. Myeloma patients treated at the Myeloma Institute for Research and Therapy of the University of Arkansas for Medical Sciences are now treated with different protocols depending on a gep risk score developed on one set of patients at Arkansas, validated in another, and further validated using data from several institutions around the world.