Stefano Garzon

Research · Oct 23, 2026

Blinding: single, double, triple and blinded outcome assessment

Concept

Blinding means keeping the identity of the intervention hidden from people involved in the trial. It protects the phase after allocation: symptom reporting, concomitant treatment, data collection and outcome assessment. A large sample size does not reduce bias.

Terminology is inconsistent. Devereaux and colleagues asked 91 internists who was blinded in “single”, “double” and “triple-blind” trials. They obtained 10, 17 and 15 different interpretations; the most common interpretation of double-blind was given by 38%. A paper needs to state who was blinded: participants, care providers, data collectors, outcome assessors, analysts.

Open trial

Participant and investigators know the allocation. This is the rule in surgery, devices and lifestyle changes. The risk lies in symptom reporting, compensatory treatment and differential dropout. In the European Coronary Surgery Study, smoking was equal between groups at entry, but early in follow-up there were fewer smokers in the surgical group, possibly from more advice by surgeons.

Single-blind

Only the participant is unaware of allocation. The investigator can still influence concomitant treatment, data collection and assessment. SPORTIF III and V compared ximelagatran with warfarin. SPORTIF III, single-blind with blinded event assessment, had a risk ratio of 0.71 (95% CI 0.48 to 1.07). SPORTIF V, double-blind, had 1.38 (95% CI 0.91 to 2.10). In SYMPLICITY HTN-3, renal denervation, which had shown large reductions in open studies, did not differ from a sham procedure: a difference of −2.39 mmHg in office systolic pressure (95% CI −6.89 to 2.12).

Double-blind

Participants and the team that follows, collects and assesses are unaware of allocation. This requires drugs identical in appearance, taste, odor and packaging, individual codes per participant and a code-break procedure for emergencies. When two active drugs differ in form, a double-dummy is used: each participant takes one active drug and the placebo of the other. Dose-adjusted treatments such as warfarin can be blinded with sham adjustments in the placebo group.

Double-blinding does not prevent imbalance in concomitant treatment. That use should be reported by group.

Triple-blind

The monitoring committee also receives data as groups A and B. The theoretical advantage is more objective assessment. The drawback is limiting the committee’s safety role, which depends on the direction of trends. In RALES, the committee started blinded; gynecomastia and hyperkalemia, which would reveal the group, were labeled X and Y. The committee should be able to request unblinding at any time.

Blinded outcome assessment and PROBE

When participant and staff cannot be blinded, those who classify the outcome can still be, usually a central committee. The PROBE design (prospective randomized open, blinded end-point) combines open treatment with blinded adjudication. Hansson and colleagues presented it as cheaper and closer to practice. Protection depends on the outcome. Objective outcomes such as death are little subject to assessment bias. Outcomes that depend on clinical decisions (hospitalization, revascularization) or patient report remain exposed, because the event occurs or is recorded by people who know the allocation.

Loss of blinding

Characteristic effects reveal the group: tremor with beta-agonists, dark tongue with bismuth, colored urine with rifampin, raised INR with warfarin. In the vitamin C trial for the common cold, participants identified the capsules by taste. Vitamin C had no benefit among those unaware of their group and appeared beneficial among those who knew or guessed.

Schulz and colleagues found odds ratios exaggerated by 17% in trials that were not double-blind. Fergusson and colleagues reviewed 191 placebo-controlled trials; 15 reported assessing the success of blinding, and in 9 it was imperfect. Asking at the end of the trial mixes blinding failure with guesses based on treatment response. CONSORT 2010 dropped the recommendation to report this test and asks who was blinded and how.

Common errors when reading a paper

What to check when reading a trial

  1. Which people were blinded, named one by one.
  2. How similarity between interventions was achieved (identical placebo, double-dummy, sham procedure).
  3. Whether outcomes were adjudicated by a blinded committee and whether they are objective.
  4. How many code breaks occurred and why.
  5. Concomitant treatment and dropout by group.

Related topics: allocation concealment, composite outcomes.

References