Reviewed: 25 October 2026
A trial headline may mention everyone who enrolled, while its main result uses a smaller group. Where did the other participants go? This is a useful question when reading cannabinoid research, not an automatic accusation that a study is unreliable. The aim is to understand whose outcomes were measured and how the analysis handled information that was unavailable.
Start with the participant flow
The CONSORT 2025 reporting guidance calls for clear reporting of participants through a randomised trial. Look for a flow diagram showing allocation, follow-up and analysis. Counts and explanations help readers distinguish people who did not receive the assigned intervention from people whose outcome measurements were missing.
These are different events. Someone can stop an intervention but still attend follow-up and provide outcome data. Someone else can continue an intervention but miss an assessment. A report that only says “completed the trial” may leave that distinction unclear. Check the diagram against the methods and results rather than assuming every departure means the same thing.
Before interpreting departures, establish the planned outcome and assessment date. Our guide to checking a clinical trial's registered plan explains where to look for that context.
Missing outcomes are not simply unsuccessful outcomes
According to the Cochrane Handbook's guidance on risk of bias, the concern is whether missingness could depend on the true outcome. People whose symptoms worsen might be harder to follow up; alternatively, people who feel better might stop attending. These are possibilities to investigate, not assumptions about any particular trial.
The direction of distortion cannot be read from a missing-data count alone. Nor is there one percentage that guarantees an unbiased result in every setting. Reasons, differences between groups and the frequency of the outcome all matter.
Read beyond “intention to treat”
Intention-to-treat principles involve analysing participants in their originally assigned groups and including all randomised participants. But the label is used inconsistently, as Cochrane notes. It does not make unmeasured outcomes reappear. Read exactly who was included, who was excluded and what assumptions were used to handle unavailable measurements. An analysis limited to people with observed outcomes still needs scrutiny for missing-data bias.
Check the outcome and time point separately
A study is not necessarily equally informative about every question it asks. Cochrane's RoB 2 explanation emphasises assessing a specific result. An early assessment can have little missing information, while a later assessment of the same outcome has much more.
Keep the outcome, measurement method and follow-up time together when recording a finding. Avoid carrying confidence in one result over to every other result in the paper. This complements our explanation of certainty in systematic reviews, where conclusions also belong to defined questions rather than broad labels.
A practical reading checklist
- Write down the number randomised to each group.
- Find how many contributed data to the particular result being discussed.
- Separate stopping an intervention from missing an outcome assessment.
- Look for reasons for missing data, not just a total.
- Read the analysis definition rather than relying on its label.
- Check whether the discussion acknowledges what remains uncertain.
A careful summary might say: “The reported result concerns this outcome at this time point; some measurements were unavailable, and the implications depend on why they were missing and how they were handled.” That is more informative than declaring either that dropouts invalidate everything or that randomisation removes every concern. Our guide to correlation and causation in health reporting offers further background on keeping conclusions proportionate.
This is an educational guide to research interpretation, not a recommendation for a cannabinoid product or treatment. Personal healthcare decisions should be discussed with a qualified healthcare professional.