Using unassessed population controls carries the possibility of misclassifying true cases as a controls, which is a key potential threat to the validity of using population controls and would reduce power to detect true associations.(Ho and Lange, 2010; Wellcome Trust Case Control, 2007) A recent analysis of nationally representative data estimates that 2.6% of the U.S. population 13 years of age or older has a lifetime history of injecting drugs.(Lansky et al., 2014) Thus, we expect the rate of misclassification of true cases (injection drug use in the past 30 days, plus use of an illicit drug 10+ times in the past 30 days) as population controls to not exceed 2.6%. This level of misclassification seems unlikely to affect study results but to the extent it does, it would increase risk of false negative findings and reduce risk of false positive findings. Additionally, our replication data set includes assessed controls (see Replication Study Cohort and Analyses), which helps to minimize any potential bias from using population controls affecting the conclusions of this study.
