Resilience Selection: A Grave Potential Bias in Clinical Trials

Physiological and psychological resilience has important implications for health, disease and treatment. Resilience is shown to boost treatment compliance as well as response and thereby reduce mortality. We consider the possibility that individuals having lower resilience are more likely to discontinue treatment in response to side effects of a drug. In randomized control trials (RCT) if a considerable proportion of individuals discontinue from the treatment group because of side effects, the average resilience in the remaining treatment group would be greater. As a result, the frequency or severity of adverse outcomes in the treatment group will be smaller than the control even when the drug has no effect. This bias is more likely to be serious for drugs with more frequent and/or serious side effects, but following intention to treat (ITT) protocols with some additional precautions can help in avoiding it. We suggest testable predictions of the resilience selection bias hypothesis along with ways to quantify and correct for the bias in RCTs. Attempts to detect, measure and correct for the resilience selection bias should be considered necessary for realistic evaluation of drug action in a clinical trial. Retrospective studies are more sensitive to RS bias than RCTs and need to be interpreted carefully.