Unscreened controls: While the use of unscreened controls is common in studies of traits with low population prevalence, unscreened controls are not ideal for traits with the high population prevalence of AD (Wellcome Trust Case Control Consortium, 2007). However, the most likely impact of this lack of screening is for a proportion of controls to be unrecognized and phenotypically misclassified cases. This will reduce the contrast between cases and controls, increase Type II error and reduce study power to detect effects. Importantly, the use of unscreened controls is not expected to increase Type I error and produce spurious positive results.
