Table 3.

Bias in cohort studies

Type of Bias	Description of Bias
Selection bias	A systematic error in creating intervention groups, causing them to differ with respect to measured or unmeasured baseline characteristics, and ultimately prognosis
Adjustment for causal intermediates	Adjusting for variables on the causal pathway between treatment and outcome can result in biased estimation of both the total effect of treatment and the direct effect that is not mediated through the adjustment variables
Immortal person-time bias	Occur whenever information assessed during follow-up is used to determine a patient’s inclusion or exclusion in the study or treatment group assignmentFor example, when assessing a new drug vs. an old comparator drug, some cohort studies first identify all patients receiving the new drug to maximize the size of this group, and then identify patients receiving the old comparator drug who never receive the new drug, beginning follow-up at the initiation of the relevant treatment for each group [Figure 4]. Patients who survived ‘immortal time’ on an old drug were switched to the new drug, and selectively excluded from the comparator group, making the old drug appear worse
Depletion of susceptibles or “survivorship bias”	In the Nurses’ Health study, prevalent users of HRT were followed for outcomes and compared with nonusers. Because the HRT group included many patients who had been on treatment for several years, it effectively excluded cardiovascular events occurring shortly after therapy initiation, leaving a cohort of hormone users that were less susceptible to the outcome
Reverse causation	When an apparent association between treatment and outcome is because outcome status influences treatment choice, rather than treatment impacting the outcome.

HRT=Hormone therapy