Key features of clinical trial design - blinding versus concealment

Two protections, two moments

Two different protections, at two different moments in a trial. This is the distinction the exam tests.

	Allocation concealment	Blinding (masking)
When	Before and at the moment of allocation	After allocation
Who is kept unaware	The person enrolling the participant	Participants, clinicians, outcome assessors, analysts
Protects against	Selection bias at randomisation	Performance and detection bias
Always possible?	Yes - in every randomised trial	No - impossible for surgery, physiotherapy, many devices

Allocation concealment prevents those enrolling participants from knowing or predicting the upcoming treatment assignment, protecting against selection bias at the point of randomisation
It is distinct from blinding, which conceals treatment assignment after allocation has occurred
An unblindable trial can still - and must still - be concealed

Empirical evidence: trials with inadequate or unclear allocation concealment exaggerate treatment effects by ~20-40% (odds ratios inflated by roughly 30%)
Inadequate blinding of outcome assessment inflates effects most for subjective outcomes; little effect on all-cause mortality
Concealment is reported inadequately in a large proportion of published trials - "randomised" alone tells you nothing about it

Alternate allocation, allocation by day of week, hospital record number, or date of birth - these are quasi-randomisation, not randomisation, and are trivially predictable
Unsealed or translucent envelopes, envelopes held on site, an unsecured allocation list
The enroller who can predict the next assignment can
- Delay or defer enrolling a sicker patient until the "right" arm comes up
- Steer patients toward or away from the trial
- -> systematic baseline imbalance - the randomisation is destroyed before the trial starts

Central randomisation - telephone, web-based or pharmacy-controlled service, remote from the enrolling site (the gold standard)
Sequentially numbered, opaque, sealed envelopes (SNOSE) - acceptable if rigorously handled
Identical pre-numbered containers prepared by an independent pharmacy

Single blind - participant
Double blind - participant + clinician (the term is used inconsistently; CONSORT asks who exactly was blinded, not for the label)
Triple blind - + outcome assessor or data analyst
PROBE design - Prospective Randomised Open, Blinded Endpoint: pragmatic when treatment cannot be masked, but a blinded committee adjudicates outcomes

Performance bias - differential co-intervention, differential adherence, differential follow-up intensity
Detection bias - a knowing assessor scores subjective outcomes differently
Attrition bias - differential dropout when participants learn their arm

"Randomised" tells you nothing about concealment. Look for the method: central randomisation, pharmacy-controlled, SNOSE
Ask who was blinded, and whether the blinding was successful (some trials formally test it by asking participants to guess)
Objective outcomes (death, dialysis, biochemical) are relatively robust to unblinding. Subjective outcomes (pain, function, quality of life, "clinical improvement") are not
A blinded endpoint committee substantially rescues an open-label trial - look for one
The placebo must be indistinguishable in appearance, taste and adverse effect profile
- A drug with a distinctive side effect (flushing with niacin, cough with ACE inhibitors, bradycardia with beta-blockers) unblinds itself - "active placebo" designs address this

Use a central, independent randomisation service - it is cheap and it solves concealment completely
Randomisation refinements
- Block randomisation - keeps arms balanced in size. Small fixed blocks are guessable at the end of each block; use variable block sizes and do not disclose them
- Stratified randomisation - by site and by strong prognostic factors, to prevent chance imbalance
- Minimisation - dynamic allocation balancing multiple covariates; include a random element to preserve unpredictability
Blind whoever can be blinded, and always blind the outcome assessor and the statistician even in an open trial
Pre-specify emergency unblinding procedures and report how often they were used
Report all of this per CONSORT items 8-11

The two words are constantly conflated in exam stems and in published papers - concealment is about allocation, blinding is about everything afterwards
A trial can be perfectly blinded yet have failed concealment (envelopes opened early), and vice versa
Baseline imbalance in a large trial is a red flag for failed concealment, not merely bad luck
Do not accept "double blind" as a synonym for methodological quality - ask who, and ask whether it held
Loss of blinding is more likely as the trial progresses and adverse effects accumulate; late outcomes are more vulnerable than early ones

9 more sections, plus exam facts

Premium unlocks every note across every specialty, and the full exam fact library behind it.