Importance of Quantitative Imaging Biomarkers’ Technical Performance Characteristics in Designing and Analyzing Clinical Trials
Quantitative imaging biomarkers (QIBs) are being used increasingly to diagnose disease, predict patients’ outcomes, and monitor and adapt treatment. In clinical trials, QIBs are used to identify patients eligible for trials and often serve as non-invasive endpoints, providing results earlier than traditional patient outcomes. Yet biomarker measurements are inherently biased and imprecise. When their bias and imprecision are overlooked, measurements can be mis-interpreted, studies can be underpowered, and treatment effects can be over- or under-estimated.
In this presentation I first review the key technical performance characteristics of a QIB, namely precision, bias, linearity, and the regression slope against true values. I then examine four scenarios using QIBs in clinical practice and in trials. I illustrate how the technical performance characteristics of a QIB affect the interpretation of pulmonary nodule volume doubling time in ruling out malignancy, the sample size of a trial investigating ultrasound elastography to diagnose stage F4 liver cirrhosis, the eligibility of Parkinson’s disease subjects for a trial of a new intervention, and the treatment effect of the rate of brain atrophy in a randomized clinical trial of progressive multiple sclerosis. Methods for correcting for the QIB measurement error are presented for each scenario.
Funding Support, Disclosures, and Conflict of Interest: Dr. Obuchowski is the statistician for the Quantitative Imaging Biomarker Alliance, through a contract between RSNA and the Cleveland Clinic
Quantitative Imaging, Statistical Analysis, Clinical Trials
Not Applicable / None Entered.