Clinical trials are planned around measurements: a symptom score at six months, a blood test at a visit, a weight recorded at the end of a study. In real life, not every measurement is collected. Someone misses an appointment, moves, becomes ill or leaves the study.
Missing information doesn't automatically make a trial useless. But it can affect how confidently researchers interpret the results. The important questions are how much is missing, why it may be missing and what the analysis assumes about it.
A missing measurement isn't always a missing person
Imagine a fictional study with visits at three, six and twelve months. A participant misses the six-month appointment but returns at twelve months. One measurement is missing, while other information is available.
Another participant stops the assigned treatment but continues to attend follow-up visits. A third leaves the study and provides no later measurements. These are different situations, and a report should make clear what information was collected from each group of participants.
Reasons for missing data can be practical or medical. A participant may move, have a scheduling conflict, feel unwell or decide not to continue. Researchers may not always know the reason. That uncertainty is part of the problem they must address.
Why the missing information can change the picture
Suppose a weight-care trial reports the average result only for people who attended the final visit. If the people missing from that visit had different experiences from those who returned, the observed average might not represent everyone originally assigned to that group.
The difference might have little effect or it might matter considerably. You can't tell just by knowing that some data are missing. Researchers need to examine the amount and pattern of missingness, the available information and the assumptions behind their analysis.
Compare the groups as well. If much more follow-up is missing from one group, or the reasons differ, that may complicate the comparison. Similar percentages don't automatically remove every concern, but unequal follow-up is a useful detail to investigate.
How researchers can handle it
One approach analyzes only participants with the relevant outcome available. That can be straightforward to describe, but it may be biased if those people differ systematically from those whose outcomes are missing.
Other methods use the information collected to estimate results under stated assumptions. Some use statistical models for repeated measurements; others use multiple imputation, which represents uncertainty about missing values through several plausible datasets. None can recreate the exact unobserved measurement with certainty.
The method needs to fit the study question and the pattern of available information. A report should explain what it did instead of presenting the final number as though every measurement had been observed.
Assumptions deserve daylight
Every way of handling missing data relies on assumptions, even if they aren't obvious. For example, a method may assume that the available information is sufficient to account for differences related to missingness. That may or may not be plausible in a particular study.
Researchers can use sensitivity analyses to see whether conclusions change under different reasonable assumptions. If the result looks similar across several approaches, that can be reassuring about those scenarios. If it changes substantially, the uncertainty deserves more attention.
A sensitivity analysis isn't a guarantee that one assumption is true. It helps readers understand how much the conclusion depends on choices that can't be verified completely from missing observations.
Find the numbers in the report
Look for a participant flow diagram and the results section. How many people were assigned to each group? How many provided the main outcome at the planned time? How many stopped treatment but remained in follow-up? Were reasons for missing data described?
Then read the analysis explanation. A brief question can guide you: “What happened to participants whose final measurements weren't available?” If the answer is unclear, don't assume they were included in the same way as participants with complete records.
For a study with multiple outcomes, missingness may differ by measure. Someone might complete a questionnaire but miss a laboratory visit. Check the specific outcome behind the headline. In the STEP 4 semaglutide trial, for example, 770 of the 803 randomized participants provided a weight measurement at week 68. That count is useful context, but it doesn't by itself tell you how missing measurements were handled.
Put the limitation in context
A small amount of missing information isn't automatically harmless, and a larger amount doesn't always make a result impossible to interpret. The pattern, reasons, design and analysis all matter. Compare the study with other evidence and consider whether the missing data could plausibly change the conclusion in a way that matters.
If a trial influences a treatment decision, ask your clinician: “How complete was follow-up, and did the researchers test whether missing information changed the result?” You can also ask what is known about benefits and side effects from other studies.
The goal isn't to demand a perfect study. It's to see where the evidence is strong and where the unobserved part of the story leaves room for uncertainty.










