BiostatisticsTier 2Medical Sciences concept

Blinding in clinical trials

Core concept

*Blinding (masking) = concealing allocation after randomisation*
- Distinct from allocation concealment, which operates at randomisation and prevents selection bias
- Allocation concealment is always possible; blinding often is not

Who is blinded	Bias prevented
Participant	Placebo response, differential dropout, co-intervention
Treating clinician	Performance bias - differential care, unblinded escalation
Outcome assessor	Detection / ascertainment bias
Data analyst / DSMB	Analytical bias, selective outcome reporting

PROBE design - Prospective, Randomised, Open, Blinded Endpoint
- Used where blinding therapy is impractical (devices, surgery, titrated antihypertensives, anticoagulants needing INR)
- Preserves assessor blinding, which is where most of the protection sits
Objective vs subjective outcomes
- All-cause mortality is nearly bias-proof - blinding matters least
- Pain, function, quality of life, "clinical improvement", imaging read, cause-specific death - blinding is essential
- Empirically, lack of blinding exaggerates effect estimates most for subjective outcomes
Sham/placebo integrity
- Match appearance, taste, schedule; matched-placebo arms
- Active placebo where side effects unblind (dry mouth with TCAs, bradycardia with beta-blockers, flushing with niacin)
- Test the blind - ask participants to guess allocation
Sham procedures: ORBITA (sham PCI in stable angina), sham renal denervation, vertebroplasty - each shrank or abolished an apparently large open-label effect
When blinding is impossible: surgery vs medicine, behavioural and exercise interventions, dietary trials, most device trials

Open-label trials with soft endpoints are the ones to distrust - look for a blinded adjudication committee
Unblinding in practice: emergency code-break procedures, and any drug with a signature effect or lab change (INR, HbA1c, potassium, eosinophil count)
Blinding does not fix attrition - report and analyse dropouts (ITT, sensitivity analyses for missing data)
DSMB sees unblinded data so the investigators do not; early stopping rules must be pre-specified

3 more sections, plus exam facts

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