Open-label vs blinded clinical trials and when to trust each one

Some treatments cannot be hidden. You cannot give someone a dummy brain procedure, and it is hard to disguise whether a person is taking a daily blood thinner when they and their doctor are managing it together. Trials of treatments like these are often open-label, meaning everyone knows who is getting what.

That raises a fair question for anyone reading the results. If patients and doctors know the treatment, can the findings be trusted? The short answer is that it depends on what was measured and how. This guide explains the difference between open-label and blinded trials, the kinds of bias each design guards against, and how to judge an open-label result, using two 2026 trials as examples: LATE-MT, a stroke trial of clot removal up to 72 hours after symptoms began, and SINGLE-AF, a trial of blood thinners in atrial fibrillation.

This guide is for general information and is not medical advice. Please talk to your own doctor about any treatment decision.

The two designs in brief

In a blinded trial, some or all of the people involved do not know which treatment each participant receives. In a double-blind trial, neither participants nor the trial staff know. A placebo or a matching active drug keeps the groups looking the same. Our guide to placebo and blinding covers how that works.

In an open-label trial, the treatment is known to everyone. Participants are still randomised, so chance decides who gets the new treatment. The difference is what happens after randomisation, when knowledge of the treatment can affect behaviour and judgement.

There is a middle option, often written as PROBE: prospective, randomised, open, blinded end-point. Participants and doctors know the treatment, but the people who judge the outcomes do not. Many large heart and stroke trials use this design.

Why open-label trials happen

There are good reasons to run a trial open-label:

• The treatment is a procedure or surgery that cannot be faked ethically.
• A sham version would carry its own risks, such as anaesthesia or a needle into an artery.
• Treatments are given in very different ways, such as a tablet against an injection, and matching them would be impractical.
• The trial is comparing strategies, such as treating against not treating, where blinding would distort real-world practice.

Open-label trials also tend to be cheaper and simpler, which can make large trials possible in areas where industry funding is scarce.

What can go wrong without blinding

Knowing the treatment can bias a trial in several ways.

Performance bias. Doctors may look after patients in the new-treatment group a little differently, with closer monitoring or extra care. Patients may change their behaviour too.

Detection bias. Whoever assesses the outcome may score it differently if they know the treatment. This is the biggest worry when the outcome needs judgement, such as a disability rating or a symptom questionnaire.

Reporting bias. Participants who know they got the new treatment may report feeling better, or report side effects more readily.

Drop-out. People who learn they are in the control group may be more likely to leave the trial, which can unbalance the groups.

How much these matter depends heavily on the outcome being measured.

Hard outcomes and soft outcomes

Researchers often talk about hard and soft outcomes.

Hard outcomes are events that are difficult to misjudge, such as death, a confirmed stroke on a brain scan, or a heart attack confirmed by blood tests. Bias has less room to act.

Soft outcomes depend on judgement or self-report, such as pain scores, mood, quality of life, or how a person rates their own function.

Open-label trials are most convincing when their main outcome is hard and assessed by people who do not know the treatment. They are least convincing when the main outcome is a patient-reported score and nobody is blinded.

The LATE-MT example

LATE-MT tested mechanical thrombectomy, a procedure in which a catheter is threaded through an artery to pull out a blood clot, in people with a large-vessel ischaemic stroke who reached hospital 24 to 72 hours after symptoms began or were last known to be well. Brain scans had to show tissue that could still be saved. The trial ran at stroke centres in China and was published in the New England Journal of Medicine on 23 October 2026. We reported it in our LATE-MT article.

A sham thrombectomy would mean threading a catheter into the brain arteries for no benefit, which would not be ethical. So the trial was open-label, with blinded assessment of outcomes.

The main outcome was the modified Rankin scale at 90 days, a 0 to 6 measure of disability where 0 means no symptoms and 6 means death. It was assessed by staff who did not know which treatment each patient had received.

Among 336 patients analysed, 168 in each group:

• thrombectomy shifted Rankin scores towards better recovery, with a common odds ratio of 1.74 (95% confidence interval 1.19 to 2.55)
• about 49% of treated patients were independent at 90 days, against about 35% with standard care
• symptomatic brain bleeding occurred in 4.8% after thrombectomy and 0% with standard care
• deaths by 90 days were 6.0% after thrombectomy and 1.8% with standard care

Judging the LATE-MT result

The Rankin scale involves a structured interview, so it sits between hard and soft. Blinding the assessors was the key safeguard. A patient might still mention the procedure during the interview, which is one reason trialists train assessors and sometimes record interviews for checking.

Some parts of the result are close to bias-proof. Death is hard to misclassify. Brain bleeding was confirmed on scans. Those safety signals point against thrombectomy, which is a sign the trial was not tilted in favour of the new treatment.

The benefit on disability was large, and the confidence interval sits well clear of 1.0. Put together, LATE-MT gives fairly strong evidence of a trade-off: more people recovering independence, with more bleeding and more deaths. Confirmation from other trials, in other populations, will still matter before practice changes widely.

The SINGLE-AF example

SINGLE-AF compared a direct oral anticoagulant with no anticoagulant in 1,803 people with atrial fibrillation and an intermediate stroke risk. It was open-label. Participants knew whether they were taking a blood thinner. We cover it in our SINGLE-AF report.

The main outcome combined stroke, systemic embolism and cardiovascular death, all events with clear definitions. Events were few: 4 in the anticoagulant group and 13 in the comparison group. With so few events, even a small amount of bias in how strokes were picked up could matter, so independent adjudication of each event is important.

There is also a subtler issue. People who know they are not on a blood thinner might report new symptoms sooner, or doctors might scan them more readily, which could lead to more strokes being found in that group. Careful trials look for this kind of imbalance in how often patients were tested.

Our guide on how to read clinical trial results walks through SINGLE-AF's numbers in detail.

A checklist for open-label results

When you read an open-label trial, ask:

1. Could the treatment have been blinded, and if not, why not?
2. Is the main outcome hard or soft?
3. Were the people judging outcomes blinded?
4. Was each event confirmed by an independent committee using set definitions?
5. Did both groups get the same follow-up and tests?
6. Did more people drop out of one group than the other?
7. Do the safety results point the same way as the benefits, or against them?

A trial that scores well on these points can be as reliable as a blinded one for hard outcomes. A trial that scores poorly deserves more caution, especially if its main outcome is a symptom score.

Open-label extensions are different

You may also see the term open-label extension. This is a stage after a blinded trial ends, in which all participants can take the active drug. Extensions are useful for long-term safety data and for giving participants access to a promising treatment. They have no comparison group, though, so they cannot show on their own whether the drug works.

Why regulators accept open-label evidence

Regulators such as the FDA and the MHRA accept open-label trials when blinding is impractical and the trial has other safeguards. Many approvals in cancer, rare disease and gene therapy rest on open-label studies, sometimes with a single treatment arm. The UK approval of Casgevy for children aged five and over and the US accelerated approval of Voceryq for recurrent respiratory papillomatosis are examples of decisions based on open-label data. In such cases, regulators often ask for further evidence after approval.

Questions to ask if you are offered an open-label trial

• Will I know which treatment I am getting from the start?
• Who will assess how I am doing, and will they know my treatment?
• If I am in the comparison group, can I get the new treatment later if it works?
• What extra tests or visits will I have, and will they be the same in both groups?

Knowing the design helps you understand what the trial can and cannot prove. An open-label trial with blinded assessment of a clear outcome can give strong answers. The design is a starting point for judging a result, and the details of how outcomes were measured tell you the rest.

Related reading

• Late thrombectomy up to 72 hours, the LATE-MT results
• What it means when a clinical trial fails
• How new medicines get approved and reach patients