Why the People Who Left the Study Matter as Much as the Ones Who Stayed
Missing data and participant dropout happen when people leave a clinical trial before the study ends, and if the missing group differs from the group that stayed, the reported result can be skewed without anyone stating it outright. A headline result only reflects the participants whose data made it into the final analysis — not everyone who originally enrolled.
This guide walks through the four counts every trial report should show, the questions worth asking about anyone who left partway through, and a short checklist you can use the next time you read a study summary.
The Four Numbers a Trial Should Report
Randomized trials are expected to report participant flow from start to finish. The CONSORT reporting guidelines, used by medical journals to standardize how trial results are written up, call for a flow diagram that tracks every participant from enrollment through final analysis. Four counts matter most:
- Enrolled — everyone who agreed to join the study and was screened for eligibility.
- Randomized (assigned) — everyone who met the criteria and was placed into a study group, such as treatment or comparison.
- Completed — everyone who stayed in the study through the point where the main outcome was measured.
- Analyzed — everyone whose data was actually included when the researchers calculated the result.
If a study enrolled 400 people but only analyzed 310, that difference of 90 people is not a footnote. It’s a question the summary should answer: who left, when, and why. For background on how these groups are formed in the first place, see our guide to understanding clinical study design types.
Questions to Ask About Anyone Who Left the Study
How many participants were missing, and from which group?
A small amount of missing data does not automatically mean a result is unreliable, and a large amount does not automatically mean it is wrong. According to the Cochrane Handbook’s chapter on assessing bias in randomized trials, there is no fixed cutoff, like 5% or 20%, that reliably separates “safe to ignore” from “a real problem.” What matters more is whether the dropout rate was similar across groups. If one group lost far more participants than the other, that imbalance is itself a signal worth investigating.
Why did people leave?
The reason someone dropped out can matter more than the fact that they dropped out. The Cochrane guidance points out that reasons like “lack of efficacy” or “adverse experience” are often connected to how well the treatment was actually working for that person. When the reasons for leaving differ between the treatment and comparison groups, or plausibly relate to the outcome being measured, the risk that missing data has skewed the result goes up. A study that simply lists “lost to follow-up” for everyone, with no further detail, hasn’t given you enough to judge this. Understanding who was in each group to begin with also helps here — see study populations, dosing, and outcomes.
Could the missing outcomes have looked different from the ones that were recorded?
This is the core question behind bias from missing data. If the people who dropped out were doing notably better or worse than the people who stayed, leaving them out of the analysis can shift the reported result in a misleading direction. The Cochrane Handbook offers an illustrative example worth sitting with: in a group of 1,000 participants, 10% missing data might sound minor, but if that missing 10% had a much higher rate of the event being measured than the reported group did, the true result for the full group could be nearly double what the analyzed data alone suggested. The size of the gap that missing data can hide depends on the outcome and how the numbers are distributed, not just on the percentage missing.
How did the researchers handle the missing values?
Researchers have a few common approaches when outcome data is incomplete:
- Complete-case analysis — including only participants with full data, and excluding everyone else. This is the simplest approach, but it can introduce bias if the excluded participants differ systematically from the included ones.
- Intention-to-treat analysis — analyzing participants according to the group they were originally assigned to, generally the preferred approach for judging whether an intervention works, though it still requires assumptions when outcome data is missing.
- Imputation — estimating what the missing values likely would have been using statistical methods, rather than dropping those participants from the analysis.
None of these approaches fully solves the problem of missing data on its own. A trial summary that names its approach, and explains why, is giving you more to work with than one that doesn’t mention this step at all.
A Quick Checklist for Reading a Trial Summary
Keep this list nearby the next time a headline or article cites a specific trial:
- Does the summary state how many people were enrolled, randomized, completed the study, and were analyzed?
- Is the number of participants who dropped out similar between the treatment and comparison groups?
- Are the reasons for dropout explained, or just labeled “lost to follow-up”?
- Does the summary say whether the analysis used all randomized participants (intention-to-treat) or only those who completed the study?
- If data was missing, does the summary mention how the researchers accounted for it?
- Would the conclusion change much if the missing participants had the least favorable outcome instead of being left out?
When Dropout Is a Bigger Red Flag Than Usual
Pay closer attention when you see any of the following together in a trial summary:
- A noticeably higher dropout rate in one group compared to the other.
- Dropout reasons connected to side effects or lack of improvement.
- No mention of how missing data was handled in the analysis.
- A conclusion that relies on a small difference between groups, where a modest shift from missing participants could erase or reverse it.
None of these automatically invalidate a study. They tell you where to look more carefully, or where to treat a headline conclusion as provisional rather than settled — a distinction covered further in clinical study safety data.
What This Guide Doesn’t Cover
This guide focuses on missing outcome data within a single randomized trial. It does not cover how missing or unpublished studies affect a broader body of evidence across many trials, which is a related but separate issue. It also does not tell you whether any specific study’s result is valid — that depends on details this guide can help you look for, not conclusions this guide can hand you.
Medical Disclaimer
This article is for educational purposes only and does not provide medical advice, diagnosis, or treatment recommendations. It does not evaluate any specific product, supplement, or therapy. See our full Medical Disclaimer for more information.
By ClinicalStudyConnect.com Research Desk | Last updated September 2026