Post date: Feb 4, 2019 3:58:01 PM
Title: False Positives in Cardiotoxicity Reporting in Cancer Clinical Trials
Speaker: Jay Herson, PhD., Department of Biostatistics, Johns Hopkins University
Abstract
In clinical trials for new cancer therapies cardiotoxicity is defined by change in the left ventricle ejection fraction (LVEF) difference from baseline. However two types of random deviations can be observed from the true value of LVEF. LFEV is measured by ultrasound and there is a reading error. Also there is physiologic deviation caused by stress, over the counter medications or some characteristic of the tumor. A clinical trial of perfectly normal patients is assumed who have no disease. They are assumed to have a normal LVEF of 55 at baseline. A model is assumed for the two types of variation and computer simulation is used to simulate 1000 clinical trials of 1500 patients each. The patients have a baseline visit and four visits while under treatment. The two types of random variation are applied by sampling from the normal distribution. The median and interquartile range of false positives is presented. Sample size requirements are presented to detect cardiotoxicity incidence of greater than the false positive level. A test-retest strategy is evaluated for patients who enter Visit 2 with a compromised LVEF. The literature for breast cancer clinical trials is reviewed to see how many trials have had cardiotoxicity reported within the false positive range. Finally, the BCIRG 006 clinical trial is revisited to explain the role false positives had in avoiding premature termination of the trial and in the discovery of two types of cardiotoxicity—each applying to different types of drugs.