How to read a clinical trial result: a reader checklist

A trial result in a headline is a claim until you have checked a few basic things. This checklist sets out seven questions to ask before repeating a result. Each question has a short reason and a source.

This guide is about reading evidence. It is not medical advice, and it does not tell anyone whether to start, stop or change a treatment. Decisions about your own care belong with you and a qualified health professional.

Background

The CONSORT statement describes itself as an "evidence-based, minimum set of recommendations for reporting the results of randomised trials" (CONSORT-SPIRIT site). The 2025 version sets out its recommendations in a 30-item checklist and a flow diagram. The flow diagram displays the progress of all participants through the trial (CONSORT-SPIRIT site). The EQUATOR Network page gives the scope as "Reporting of randomised trials", with clinical trials and experimental studies as the study designs (EQUATOR Network).

The checks

1. Was the trial registered before the first participant was enrolled? Why: ICMJE recommends that editors require registration "at or before the time of first patient consent for enrollment". It also says that retrospective registration, for example at the time of manuscript submission, "meets none of these purposes". Exceptions are possible, but ICMJE says they should be rare (ICMJE trial registration).

2. What was the comparison group? Why: ICMJE's definition of a clinical trial covers studies "with or without concurrent comparison or control groups", so not every trial has one (ICMJE trial registration). A result can only be judged once you know what, if anything, it was compared against.

3. How many people were in each group, and how many were analysed? Why: the CONSORT flow diagram displays the progress of all participants through the trial, so the numbers at each stage can be followed from start to finish (CONSORT-SPIRIT site).

4. Do the reported outcomes match the registered ones? Why: ICMJE says the purpose of registration includes preventing "selective reporting of research outcomes", and that it is the authors' responsibility to explain any discrepancies between results reported in registries and in journal publications (ICMJE trial registration). The registry entry is the place to make the comparison.

5. Are the results given as absolute risks as well as relative ones? Why: the Cochrane Handbook gives an example where a risk ratio of 0.75 could correspond to a clinically important reduction in events "from 80% to 60%", or to a "small, less clinically important reduction from 4% to 3%" (Cochrane Handbook, chapter 6). The same chapter says a risk difference is "straightforward to interpret", but that its clinical importance "may depend on the underlying risk of events in the population" (Cochrane Handbook, chapter 6).

6. Who paid for the trial, and did the sponsor have a role? Why: ICMJE says articles should be published with statements declaring sources of support, including sponsor names with an explanation of any role those sources had. This covers study design, the collection, analysis and interpretation of data, writing the report, and any restrictions on submitting the report for publication (ICMJE conflicts of interest).

7. Did the authors have access to the study data? Why: ICMJE says articles should also declare "whether the authors had access to the study data", with an explanation of the nature and extent of that access (ICMJE conflicts of interest).

Using Vrulum labels

We use the labels set out on the Vrulum methodology page (Vrulum methodology). That page says company announcements are treated as reported claims until someone else confirms them. "Reported, unverified" means "Stated by a company or team but not independently confirmed", so a trial result that a company or team has announced but no one has independently confirmed carries that label. "Demonstrated" means "Shown in a published experiment, test or shipping product". "Proposed" means "A design or plan that has not yet been built or tested", so a trial design that has not yet been run carries that label. The same page says every AI-drafted record goes through an automated check against its sources and then human editorial review before publication.

Gaps in this note

The full CONSORT 2025 checklist and its explanation document were not opened for this review, so their item-by-item wording for sample size, primary outcomes and follow-up was not checked. Check 3 rests on the flow diagram description only. Before publishing any article based on this checklist, read the full CONSORT 2025 checklist from the CONSORT-SPIRIT site.