COMPLIANCE BIAS:

When some patients do not adhere to the assigned medication in experimental studies, such as randomized controlled trials, the efficacy of the medication is confounded by issues of patient compliance¹ (the study groups become non-comparable). This bias is an issue because the pre-defined dosing strategy for the study is not adhered to, and thus a causal dose-response relationship cannot be determined without post-randomization adjustments for confounding by compliance. Studies which involve minimal patient supervision during treatment (such as pills patients take at home etc.) are especially under threat from this bias. Strategies to verify patient compliance (e.g. pill counting etc.) are essential for these studies. Also see: Performance Bias, Confounding, and Information Bias.

Reference:

1. Sackett DL. Bias in analytic research. J Chronic Dis. 1979;32 (1-2):51-63. (Link to Reference)