Biases in Genetic Association of Coronary Heart Disease Events May Be Less Likely Than Suspected: Here Is When to Check for Them.

Abstract

Investigating lethal diseases like coronary heart disease (CHD) and major adverse events like myocardial infarction (MI) and death can sometimes seem a bit macabre. We are interested in understanding the events with the hope of preventing them; yet, to demonstrate effects, it is essential for a high rate of such unfortunate events to occur and to be observed. Fortunately, the increasing availability of big [event] data supports an unprecedented ability and power to explore genetic influences on primary and subsequent CHD events. Tempering the enthusiasm around this opportunity is the concern for biases that threaten the internal and external validity of such investigations. See Article by Hu and Schmidt et al In particular, selection and survival biases are of concern to the context of CHD events. When subjects are nonrandomly selected or tend to be systematically included based on the presence of a related risk profile, diagnosis, or event, this is deemed selection bias (also known as index event bias). Survival bias occurs when sampling is dependent on an individual’s likelihood of surviving an event in the first place, or when attrition or study closure prevents complete observation of time to censoring. These types of bias can systematically distort variance and lead to spurious outcomes. Selection bias has been a well-acknowledged concern in genetic association.1 …