Keywords: outliers, efficacy, winsorization, inter-quartile range, multiple imputation
Evaluating efficacy in a clinical trial is often based on estimating the average treatment effect within the study population and assuming that this represents the true treatment effect of the investigational compound. Outlying data values in the study sample may arise for a number of reasons including measurement or input errors and may not represent the true variation of the efficacy. Thus, it is desirable to evaluate the impact of outliers on estimates of efficacy to confirm that conclusions are valid and not based on potentially erroneous values.
During this lunch roundtable we will discuss methods to identify outliers (eg, interquartile range, z-score [i.e., standard deviation]) as well as valid ways to evaluate the influence of outliers (eg, winsorization, multiple imputation).
Come prepared to answer/discuss the following questions: 1) What methods have you used to identify outliers in clinical trials? 2) How have you evaluated the influence of outliers in clinical trials?