What Is a Randomized Controlled Trial? Reading Hypnosis Research

When a headline says research shows hypnosis helps with a problem, the first useful question is what kind of research it was. Randomized controlled trials, or RCTs, are the most informative routine design for testing whether an intervention causes an effect. They are not flawless, and hypnosis exposes several of their weak points at once. This guide explains how RCTs work, where hypnosis trials are most vulnerable and what to check before you trust a result.

The core idea: chance decides who gets what

In an RCT, researchers recruit participants and assign each one by chance to the intervention or to a comparison condition. Everyone is measured on the same outcomes, and the groups are compared.

Chance matters because choice introduces bias. If clinicians picked who received hypnosis, they might pick people who seemed more motivated or less unwell, and any later difference could reflect that. Random assignment balances the groups on measured characteristics such as age and severity, and also on ones nobody thought to measure. Balance is not guaranteed in a small trial, but systematic bias in who receives what is removed.

Randomization only works if it cannot be subverted. Allocation concealment means the person enrolling a participant cannot know which group that participant will be assigned to. Cochrane’s risk-of-bias tool treats this as part of the first of its five bias domains, alongside deviations from the intended intervention, missing outcome data, outcome measurement and selective reporting [2].

Trial reports are expected to describe all of this. The CONSORT 2025 reporting guideline sets out a 30-item checklist and a diagram of participant flow, and adds a new section on open science such as registration and data sharing [1].

What to look for in a trial report

• Population. Who took part, how they were recruited and who was excluded. Results apply most directly to people like those studied.
• Intervention. How many sessions, live or recorded, individual or group, and delivered by whom.
• Comparison. What the other group received. This choice shapes what the trial can conclude; see Expectation, Placebo and Hypnosis: Why Comparison Groups Matter.
• Primary outcome. One main measure chosen in advance, not the best-looking result picked from many.
• Registration. A public record of the plan made before the trial began.
• Dropout and analysis. How many people left, from which groups, and whether the analysis counted everyone as randomized.
• Follow-up. Whether effects were measured only at the end of treatment or months later.

A worked example

The IMAGINE trial shows what good reporting makes visible. It randomized 354 adults with irritable bowel syndrome across 11 Dutch hospitals to six sessions of individual hypnotherapy, group hypnotherapy or educational supportive therapy, and it was registered [6]. Dropout was reported by arm: 15% in each hypnotherapy group and 20% in the control group.

The authors reported two analyses. In the intention-to-treat analysis, which counts everyone as randomized, adequate symptom relief at three months was 40.8% with individual hypnotherapy, 33.2% with group hypnotherapy and 16.7% with the control. In the per-protocol analysis, which counts only those who completed treatment as planned, the hypnotherapy figures rose to 49.9% and 42.7% [6]. Same trial, same patients, two different headline numbers. The intention-to-treat figure is the more conservative estimate of what happens when a treatment is offered.

Where hypnosis trials are most vulnerable

Blinding is rarely possible

In a drug trial, neither patients nor clinicians need to know who received the active pill. In a hypnosis trial, participants usually know whether they were hypnotized and the practitioner always does. The Cochrane Handbook notes that blinding “can be difficult or impossible in some contexts” and that it is particularly difficult for participant-reported outcomes, where knowing the assigned treatment can influence the assessment [2].

Most outcomes are self-reported

Pain ratings, anxiety scales and sleep diaries are the usual outcomes in hypnosis research. That matters. A combined analysis of 146 meta-analyses and 1,346 trials found that, for subjective outcomes, trials without blinding exaggerated effects (ratio of odds ratios 0.75), as did trials with inadequate or unclear allocation concealment (0.69). For objective outcomes, there was little evidence of either bias [3]. Hypnosis trials combine the two conditions under which bias is largest.

Small trials report bigger effects

Across 93 meta-analyses and 735 trials, treatment effects were on average 48% larger in trials with fewer than 50 patients than in trials with 1,000 or more [5]. Many hypnosis trials enrol a few dozen people, so a large effect from a small study deserves extra caution.

Outcomes can be selected after the fact

When researchers compared 102 trial protocols with the published papers, 62% of trials had at least one primary outcome that was changed, added or dropped. Statistically significant efficacy outcomes had 2.4 times the odds of being fully reported compared with non-significant ones [4]. Registration lets readers check what was planned against what was published.

Harms are under-measured

A review of hypnosis trials registered on ClinicalTrials.gov found no serious adverse events likely caused by hypnosis and a 0.47% rate of other adverse events, but several trials reported no adverse-event data at all [9]. When one insomnia trial asked participants directly after sessions, adverse events were reported by 37% to 52% of them, depending on session content. Most were mild, and 10% of participants discontinued [10]. Part of that gap probably reflects how closely researchers looked.

Reading the numbers

A statistically significant result is one unlikely to be explained by chance alone, given the study’s assumptions. It does not tell you whether the effect is large enough to matter. Look for the effect size, which describes how big the difference was, and the confidence interval, which shows how precisely the average was estimated.

Meta-analyses sometimes add a prediction interval, which estimates the range of effects a new study in a new setting might find. It is often the more honest number. A 2021 meta-analysis of 50 trials in 4,269 surgical patients found that hypnosis reduced emotional distress (g = 0.55) and pain (g = 0.37) compared with standard care or attention control. Yet the 95% prediction intervals included no effect for every outcome except surgical procedure time, because results varied so much between studies [7]. The average was positive; a given hospital’s programme might not be.

One trial is rarely the final word

Confidence grows when independent teams replicate a finding and when systematic reviews pool trials transparently. Reviews inherit the weaknesses of the trials they include, and they vary in quality themselves. An overview of 49 meta-analyses of hypnosis, covering 261 primary studies, rated only nine as having high methodological quality [8].

Evidence you are reading	What it can support
Single small RCT, self-reported outcome	A plausible effect worth testing again
Several RCTs, consistent results, some blinded assessment	A probable effect, size uncertain
High-quality meta-analysis with active or sham comparisons	A reasonably established effect for that population

A six-question checklist

1. Were participants randomized, with allocation concealed?
2. What did the comparison group receive?
3. Was the primary outcome registered in advance, and is it the one reported?
4. How many people dropped out, and was the analysis intention-to-treat?
5. How large was the effect, and how wide were the intervals?
6. Who took part, and do they resemble the people you care about?

Participants also differ in how strongly they respond to suggestion, which can shape a trial’s results; see What Is Hypnotizability?.

What an RCT does not tell you

An RCT estimates an average effect in a group. It cannot predict how one person will respond, it does not by itself explain how an intervention works and it does not show that an approach suits a particular person’s health needs.

Using research for a health decision. A trial result describes groups, not you. If you are considering hypnosis for a diagnosed condition, discuss it with the clinician treating you and use it alongside established care, not instead of it. See a doctor first about new, severe or worsening symptoms rather than waiting to see whether a self-help approach works.