An outlier, an observation that lies an unexpected distance from other values, does
not in itself prove that an error or violation has occurred. Therefore, the primary
data analysis should include all such observations, including “outliers”. A sponsor
may perform some specific supplemental analyses on subsets of the data, if
clinically and scientifically justified, that may exclude outlying data points.
Excluding large amounts of data, regardless of the reasons for exclusion, will
seriously bias the results.
