Table 3.

Potential sources of bias in a clinical trial

Types of BiasPotential Effects of Bias
Intervention assignment bias, e.g., unconcealed allocation or unblindingKnowledge of assignment might lead to performance bias, e.g., extra attention to participants for one of the intervention groups than the other; unblinding of participants to interim study results might result in performance bias
Exclusion bias (exclusions after randomization)Nonrandom withdrawal in intervention groups might result in intervention groups becoming less comparable than at randomization
Performance biasIntervention groups might receive cointerventions differentially (e.g., greater counseling or follow-up for one group than the other)
Differential compliance biasSubjects in the two arms of the trial may have different levels of compliance with the assigned therapy; this might dilute the intervention effect; or subjects who cross over from one arm to the other might dilute the intervention effect
Follow-up biasMore subjects in one arm lost to follow-up than in the other arm; or differential follow-up between the intervention and control groups?
Measurement biasInaccuracy of assessment method especially if nonrandomly distributed between intervention groups
Detection biasDifferential rates of outcomes assessment between the intervention and control groups
Selective reporting of resultsConflict of interest because of industry sponsorship with selective reporting of favorable results or lack of reporting of unfavorable data