13 Research methods, statistics, critical review and evidence-based practice13.3 Experimental studies
13.3.1.5 Blinding and measurement bias
1 min read1 figure
Written by Resident Doctor Committee · Published
Blinding and allocation concealment
Allocation concealment keeps the upcoming assignment secret before and until enrolment, which guards against selection bias.
Blinding keeps secret which arm a participant is in after allocation, which guards against performance bias and measurement bias.
Blinding chiefly protects against measurement (information) bias, where knowing the allocation distorts how outcomes are recorded.
Randomisation and allocation concealment, not blinding, deal with selection bias and confounding.
Feature | Allocation concealment | Blinding |
|---|---|---|
Example of failure | A clinician delays a very unwell patient until the favoured arm comes round | An assessor who knows the arm scores ambiguous items in the expected direction |
Feasibility | Possible in almost any trial | Often impossible, as in psychotherapy trials |
Who to blind
Terms such as single, double and triple blind are used inconsistently, so a trial report should state exactly who was blinded.
Party | Bias if they know the allocation |
|---|---|
Participant | Reporting bias, differential attrition |
Treating clinician | Performance bias, through co-intervention, dose changes and rescue medication |
Outcome assessor | Observer or detection bias |
Endpoint adjudicator | Outcome misclassification |
Data analyst | Selective analysis bias |
Subjective outcomes and masked assessment
Subjective outcomes, such as rating scales, are the most open to measurement bias, while all-cause mortality needs the least judgement.
Apparently objective outcomes can still involve judgement: hospital admission depends partly on clinical decisions.
When participants and therapists cannot be blinded, outcome assessors should be separate from treating staff and kept unaware of allocation.
A double dummy design masks treatments of different forms: one arm gets active tablets with placebo injections, the other placebo tablets with active injections.
An active placebo mimics some adverse effects without the intended therapeutic action, so that side effects such as sedation do not reveal the allocation.