In a double-blind study neither the participants nor the investigators or assessors who care for or evaluate them know which treatment each participant receives. Single-blind masks participants only; triple-blind usually extends masking to the people analysing the data.
Free sandbox · No credit card · 21 CFR Part 11 aligned
Blinding (masking)
noun · clinical research
Withholding knowledge of the assigned treatment from one or more groups involved in a trial, such as participants, investigators, assessors or analysts, so that expectations cannot influence conduct, outcomes or analysis.
Related terms
In short
Explained
Blinding (also called masking, particularly in eye research) means keeping information about treatment assignment from people who could be influenced by it. Every group that knows the assignment is a possible source of bias: participants may report symptoms differently, clinicians may treat or assess differently, and analysts may make different analytic choices.
Because the terms are not applied uniformly, one trial's "double-blind" can mean something different from another's. The CONSORT reporting guideline encourages authors to describe who was blinded and how, rather than just using the label. That is the rule to follow when writing your own protocol and publications.
Everyone knows the assignment. This is normal for surgical trials, many behavioural interventions and some oncology designs where blinding would be impossible or unsafe. The risk is bias in subjective outcomes, which is why open-label trials lean on objective endpoints such as survival or a laboratory value, or use blinded outcome assessors.
Usually only the participant is masked. Sometimes it is the outcome assessor instead. It is simpler to run than double-blind but leaves investigators free to treat or judge participants differently depending on the arm.
Participants and the investigators and staff who administer or assess the treatment are all masked. This usually requires a placebo or comparator that is indistinguishable from the active treatment (same appearance, taste, schedule or device), and a pharmacy or unblinded team that holds the allocation. Double-blind, randomized, placebo-controlled designs are the traditional benchmark for confirmatory Phase 3 trials.
Most commonly adds the statisticians or an independent data monitoring committee that review data without knowing which arm is which (for example, arms coded as A and B). Some authors use "triple-blind" differently, for example when the investigator and the outcome assessor are different people, so define it in the protocol.
| Open label | Single-blind | Double-blind | Triple-blind | |
|---|---|---|---|---|
| Participant | knows | masked | masked | masked |
| Investigator / assessor | knows | knows | masked | masked |
| Data analyst / committee | knows | knows | knows | masked |
| Unblinded pharmacist / statistician | n/a | role may exist | holds allocation | holds allocation |
Running a blinded trial
Blinding is a set of procedures, not a checkbox on the protocol.
Name each group (participants, investigators, assessors, sponsor staff, analysts) and say what each can see. Explain how the placebo or comparator is matched.
Allocation should be generated and held by people or a system outside the blinded team, with allocation concealed until a participant is enrolled. See clinical trial randomization software and block randomization.
Pharmacists, supply managers or unblinded statisticians may need to see the arm. Everyone else should not. See blinded and unblinded role management.
Define who may break the blind for one participant (for example for a serious adverse event), how it is recorded, and who must be told. Unblinding one participant should not unblind the study.
Blinded data cleaning uses datasets that exclude the arm. Treatment codes are released to analysts only after database lock.
Some trials ask participants or staff to guess their assignment to test whether blinding held. Report who was blinded and any unblinding events.
Create blinded and unblinded roles in the free Capture sandbox and check what each one can see, with sample data and no credit card.
Data and software
Blinding fails more often through side channels than through deliberate unblinding. Common leaks include a data listing that shows the randomization arm, a lab report that reveals an active drug effect, an export that includes the allocation column, or a query that mentions the treatment. A blinded study needs the system to hide the arm from blinded roles at the data layer, not only on the screen.
Edit checks, queries and monitoring reports must also be reviewed from a blinded point of view. For example, a safety listing grouped by treatment arm is for the unblinded team or the data monitoring committee only. A blinded medical monitor should see events without the arm label. Exports and the SDTM datasets used for analysis should be generated after lock, with a documented process for when the treatment codes are merged. See CDISC for the dataset standards involved.
Related decisions that depend on the arm, such as IRT/IWRS drug assignment, are normally handled by a randomization function the blinded team cannot read. The pragmatic trial is a counter-example where open-label designs are accepted for good reasons.
In Capture
Capture supports randomization and blinded roles. Blinded roles do not receive treatment-arm values; this is enforced by masked database views and row-level security at the database layer, not just in the user interface. Pair that with role-based access, a field-level audit trail and data exports you control, and you can test who sees what in the free sandbox before the first participant is randomized.
Roles tested
3 of 3
Open items
1
A study in which both the participants and the investigators or assessors who deliver or evaluate the treatment do not know which treatment each participant receives. It reduces bias from expectation in behaviour, care and outcome assessment.
In a single-blind study only one group, usually the participants, is unaware of the assignment. In a double-blind study the investigators and assessors are also unaware.
Usually a double-blind study in which the data analysts or an oversight committee are also masked until analysis. Usage varies, so the protocol should state exactly who is blinded.
To prevent performance bias, where knowing the treatment changes behaviour or care, and detection bias, where it changes how outcomes are assessed. It works alongside randomization, which prevents selection bias.
Unblinding reveals the treatment assignment. It is allowed under predefined rules, for example in an emergency where the treating clinician needs to know. The procedure should limit disclosure to the participant concerned and record who unblinded and why.
No. Surgical, behavioural and some oncology trials often cannot be blinded. They can reduce bias with objective endpoints or blinded outcome assessors.
By limiting what each role can see and export. In Capture, blinded roles never receive treatment-arm values, enforced by masked database views and row-level security.
Keep exploring
Blinded and unblinded role management
Setting up roles.
Randomization software
Generating allocations.
Block randomization
Balanced allocation.
IRT / IWRS
Randomization and supply.
Database lock
When the blind is released.
Pragmatic clinical trial
Open-label by design.
Phases of clinical trials
Where blinding is used.
Test blinded roles in a free sandbox with every feature. No credit card.