Missing data means information a diagnostic device study was supposed to collect but didn’t. Withdrawn or excluded participants are people who were enrolled but never made it into the final results. When either happens without explanation, a study’s reported accuracy may not reflect everyone who started it.
What Do “Missing Data” and “Withdrawn Participants” Actually Mean?
A diagnostic device study usually starts by enrolling a set number of people. By the time it reports final accuracy numbers, that number often shrinks. People can leave early, be excluded by the research team, or produce a test result that doesn’t fit neatly into “positive” or “negative.”
Three terms show up often in this kind of research:
- Missing data — information that was supposed to be collected but never was, for any reason.
- Withdrawn or excluded participants — people enrolled at the start but removed from the final analysis, by their own choice or the research team’s decision.
- Ambiguous or indeterminate results — test outcomes that are not clearly positive or negative, including results described as unclear, incomplete, or “in between.”
None of these are automatically a problem. Some dropout and some unclear results happen in almost every real-world study. The issue is whether the study tells you how many people this happened to, why, and how those cases were handled in the final numbers.
Myth vs. Reality: How People Read These Studies
It’s easy to skim past this part of a study. Here is how a few common assumptions compare with what careful reporting actually requires.
- Myth: A high reported accuracy percentage applies to everyone who was originally enrolled.
Reality: That number usually reflects only the people whose results made it into the final analysis. If a meaningful number of people were excluded, the reported accuracy may not represent everyone who started the study. - Myth: Unclear or unreadable test results are too rare to need their own reporting.
Reality: FDA guidance treats ambiguous results as their own category that should be counted and reported, not folded quietly into either the positive or negative group. - Myth: If a study doesn’t mention withdrawals or exclusions, none happened.
Reality: Silence on this point could mean there were none, or it could mean the reporting is incomplete. A well-reported study states the number clearly rather than leaving readers to guess. - Myth: How missing data is handled is a technical detail that doesn’t change what the study means.
Reality: The handling choice can change what the reported accuracy actually represents, which is why FDA guidance recommends this be planned before data collection even starts.
What Does FDA Guidance Say About Missing Data and Withdrawn Participants?
The FDA has published statistical guidance for reporting results of studies evaluating diagnostic tests. Three points from that guidance are especially useful when reading a study yourself.
A complete accounting of participants. The guidance recommends a study report the number of subjects planned to be tested, the number actually tested, the number used in the final analysis, and the number omitted from the final analysis. In plain terms: the study should show the full path from “everyone we signed up” to “everyone whose results we actually used,” with the gap explained.
Ambiguous results need their own count. The guidance defines ambiguous results broadly: intermediate, inconclusive, incomplete, uninterpretable, unsatisfactory, unavailable, in a “gray zone,” or otherwise different from a clear positive or negative. It recommends reporting the number of these results specifically.
Discarding unclear results is discouraged. Simply throwing out equivocal results before calculating accuracy is not considered good practice. The guidance instead recommends reporting performance measures more than one way — once counting ambiguous results as positive, once as negative — so a reader can see how much that choice would change the outcome.
What This Guidance Doesn’t Tell You
It’s worth being upfront about the limits here. FDA’s training materials on pivotal device study design confirm that a plan for handling missing data should be written into a study’s statistical analysis plan before data collection starts. But those materials don’t specify an acceptable dropout rate, a required method for handling missing data, or a fixed list of valid exclusion reasons.
That means there’s no single number separating an acceptable amount of missing data from a concerning amount. The underlying principle still holds: a study should tell you what happened to every enrolled participant, rather than leaving you to assume nothing was left out.
What Should I Check When I Read a Diagnostic Device Study?
You don’t need a statistics background to use this. When reading a summary of a diagnostic device study, look for answers to these questions:
- Does the study state how many people were originally enrolled?
- Does it state how many people’s results were used in the final accuracy calculation?
- If those two numbers differ, does the study explain why?
- Does the study mention ambiguous, unclear, or indeterminate results at all?
- If ambiguous results happened, does it say how many, and how they were counted in the final numbers?
- Does the study describe its plan for missing data, or only show the final, “clean” results?
If a study answers most of these clearly, that’s a sign of more complete reporting. If it answers none of them, that’s not proof the study is wrong — but it means you’re missing information you’d need to judge how far the reported accuracy really extends. A practical rule: if a study excludes a noticeable share of ambiguous or incomplete results without explaining how, treat the headline accuracy number as provisional rather than final, and look for a fuller or peer-reviewed version of the same study before relying on it.
A Worksheet for Your Own Notes
If you want to work through a specific study, fill in these lines in your own words as you read it:
- Number of people enrolled at the start: ______
- Number of people whose results were used in the final analysis: ______
- Reason given for the difference between those two numbers, if any: ______
- Number of ambiguous or unclear results reported, if any: ______
- How those ambiguous results were counted in the final accuracy numbers: ______
- One question this study leaves unanswered for me: ______
Keeping notes like this for two or three studies on the same type of test can also help you see which one reports this information more clearly.
What This Guide Does Not Do
This guide is meant to help you read and think about study reporting. It does not diagnose any condition, evaluate any specific product, or tell you whether a particular device or test result is accurate for your situation. If you are dealing with a health concern that feels urgent, contact local emergency services or a qualified healthcare provider rather than relying on this article.
If you’re new here, our Start Here page explains what Connected Diagnostics Evidence covers and doesn’t cover. Our How We Research page explains how we select and verify the sources we cite, including the FDA guidance referenced above.
Sources and Independence
This article draws on FDA’s statistical guidance for reporting diagnostic test study results and FDA’s CDRH Learn training materials on pivotal device study design. Connected Diagnostics Evidence is an independent educational publication. It is not affiliated with, and does not continue the operations, research, staff, or products of, any company that previously operated this domain.
Educational information only. This article does not provide medical advice, diagnosis, or treatment recommendations. Consult a qualified healthcare provider for guidance specific to your situation.
By Connected Diagnostics Evidence Editorial Team. Last updated September 2026.
Leave a Reply