*Geek Box: “Healthy User Bias” and Selection Bias
“Healthy User Bias” is a bias that can commonly occur in observational reserach, where the characteristics of people who volunteer for reseerch may reflect overall health consciousness, tendency to exercise more, smoke less, have a healthier overall diet, and other health-promoting behaviours. This can often be difficult to avoid, given that large observational studies rely on volunteers willing to participate, and be followed-up over a period of time.
Selection Bias occurs when the selected study population differs in characteristics from the target population, or a group of eligible participants not included in the study. Ideally, a study cohort would be selected at random from the population, however, practically this is impossible as people must agree to participate in a study.
These factors are important when we come to consider relative risk (elaborated on below). For example, while the cohort in the EPIC-Oxford studies were healthier than the average UK population, they still had characteristics that may influence cancer outcomes: 35% of men and 25% of women were ex-smokers, whereas if we contrast this to the Seventh Day Adventist cohorts, who largely abstain from smoking, that population generally reflect much more constant overall health-promoting behaviours.
These factors influence the relative risk in both cohorts, and the magnitude of effect in comparing different diet patterns against each other.