Research
Peer-reviewed, preprint, or conference abstract
Three things that all look like studies when a product page links them. One has been checked by outside experts, one has not, and one is a summary that more than half the time disagrees with the paper it becomes.
Three tiers behind one word
A product page says a study. The link goes to something that looks like a study, has authors, has numbers, and sits on a scientific-looking site.
It could be any of three things. A paper that went through peer review at a journal. A preprint, posted publicly with no outside review. Or a conference abstract, which is a few hundred words summarizing work presented at a meeting.
They differ in how much checking they survived and in how likely their numbers are to survive contact with a full paper. Both differences have been measured.
What peer review actually is, and is not
Peer review is a journal sending a manuscript to outside experts who look for weaknesses in its assumptions, methods and conclusions. The editors decide whether the authors have answered those concerns before publishing.
Federal advertising guidance describes what that buys. The Federal Trade Commission says a rigorous, unbiased peer review process, like that required by established and reputable scientific journals, provides some level of assurance that the research meets accepted norms in the field.
Some level of assurance is a careful phrase, and the same guidance immediately limits it. The mere fact that a study is published is not a guarantee of quality, and not proof that a product is effective for the advertised benefit.
It goes further. The rigor of peer review varies widely from journal to journal, with some journals accepting studies based on little more than payment of a publication fee.
So peer review is a filter of uncertain strength, not a seal. A peer-reviewed paper in an obscure journal is not automatically better than a careful preprint. It is simply the one tier where somebody outside the author group was asked to object.
A preprint is the paper before anyone checked it
The National Library of Medicine defines the category plainly. Preprints are complete and public drafts of scientific documents, not yet certified by peer review.
The library is explicit that this is the whole distinction. Certification by peer review is the key difference between a preprint and an accepted manuscript or published article.
It also tells readers what to expect. Any aspect of the research, including the results and the conclusions, may change as a paper moves toward publication.
Preprints exist for good reasons. They circulate findings quickly and invite the criticism that improves them. A preprint is early, not fake.
The reading rule is that a preprint is a claim its authors are making, with nothing yet standing behind it except their own care.
Being in PubMed no longer means peer-reviewed
This is the part that catches careful readers, because the old shortcut has stopped working.
Under the National Library of Medicine's preprint pilot, preprints from research funded by the National Institutes of Health are made available through PubMed Central and, by extension, PubMed. The second phase of that pilot covers preprints acknowledging direct NIH support or having an NIH-affiliated author, posted from the start of 2023 onward.
A PubMed record is therefore no longer proof that anything was reviewed. The database labels preprints, and a reader has to look at the label.
That is worth knowing before treating a PubMed link on a sales page as a credential. It is a locator, not a verdict.
Most conference abstracts never become papers
A conference abstract is the weakest of the three, and the numbers behind that are unusually good.
A methodology review pooled 425 separate reports covering 307,028 abstracts presented at scientific meetings. It asked how many were later published in full.
The answer was 37.3 percent, with a confidence interval of 35.3 to 39.3 percent. That was significantly lower than the 44.5 percent the same review had found eleven years earlier.
Allowing a full decade helps but does not close the gap. Survival analysis estimated that 46.4 percent of all abstracts would be published in full within ten years, rising to 68.7 percent for randomized controlled trials and 44.9 percent for everything else.
Roughly six in ten meeting abstracts never become a paper anyone can check.
And the ones that do often say something different
Non-publication would matter less if the abstract were reliable on its own. It frequently is not.
One study followed 513 abstracts of randomized controlled trials presented at an ophthalmology meeting. Only 230 of them, 44.8 percent, reached publication at all.
Among the 86 pairs where the abstract and the eventual paper reported the same main outcome at the same time point, 47 pairs disagreed. That is 54.7 percent.
Most of those disagreements were about size. Seven of them were about direction, meaning the estimated effect pointed the other way in the published paper than it had in the abstract.
The authors concluded that more than half the pairs showed some discordance, which calls into question the dependability of conference abstracts.
Why the survivors lean positive
The abstracts that make it into journals are not a random sample of the abstracts presented, and the direction of the bias is measurable.
In the pooled methodology review, abstracts reporting a positive result were more likely to be published in full, with a risk ratio of 1.31. Defining positive more narrowly, as a result favoring the experimental treatment, still produced a risk ratio of 1.17.
That is publication bias described from the inside. The literature that exists on a question is tilted toward the answer people liked, and the tilt has been counted.
It is the strongest argument for reading a body of evidence rather than a single study, however good that study looks.
Why a regulator will not judge a study from its abstract
The Federal Trade Commission states the limit directly. It cannot evaluate the quality of a study from an abstract or an informal summary.
Its reason is practical. A study's write-up should contain enough detail to assess what actually took place, and the agency evaluates research partly on how closely it followed its protocol and how well the report explains any deviations.
An abstract has no room for any of that. It cannot show who dropped out, how the analysis was chosen, or what else was measured and not reported.
The guidance also raises the bar where review is missing. Research that has not been through a rigorous peer review process gets greater scrutiny, and the agency will often require the marketer to hand over underlying documents and raw data.
How to place a citation in about a minute
Look at where it lives. A journal name and a volume and page number mean a published paper. A preprint server or a PubMed record labeled as a preprint means no outside review yet.
Look at the length. If the entire item is a few hundred words with no methods section, it is an abstract or a summary, whatever it is called.
Look for a meeting name and a year. Those identify a conference abstract, and they are the citations most likely to have no paper behind them.
If it is an abstract, check whether the full paper exists. A result that has sat at abstract stage for years is in the majority that never gets published.
And weigh a single positive result against the tilt. Positive abstracts get published more often, so the published record on any question is friendlier to the treatment than the underlying research was.
Key takeaways
- Peer review means outside experts were asked to object; it is a filter of uncertain strength, not a seal of quality.
- A preprint is a complete public draft not yet certified by peer review, and its results and conclusions may still change.
- A PubMed record no longer proves peer review, because the National Library of Medicine now carries NIH-funded preprints there.
- Pooled across 307,028 abstracts, only 37.3 percent were ever published in full.
- Where abstracts and their eventual papers could be compared, 54.7 percent of pairs disagreed on the main outcome, and seven disagreed on direction.
- Positive abstracts are published more often, with a risk ratio of 1.31, so the published record leans friendlier than the underlying research.
- Federal advertising guidance states that a study cannot be judged from an abstract or an informal summary.
Frequently asked questions
What is the difference between a preprint and a published paper?
Peer review, and nothing else definitionally. The National Library of Medicine describes preprints as complete and public drafts of scientific documents, not yet certified by peer review. It says that certification is the key distinction between a preprint and an accepted manuscript or published article. It also warns that any aspect of the research, including the results and conclusions, may change on the way to publication.
Does appearing in PubMed mean a study was peer-reviewed?
Not any more. Under the National Library of Medicine's preprint pilot, preprints from research funded by the National Institutes of Health are made available through PubMed Central and, by extension, PubMed. The second phase covers preprints that acknowledge direct NIH support or have an NIH-affiliated author and were posted from the start of 2023. Records are labeled, so the label is what to read rather than the database name.
How many conference abstracts turn into real papers?
A methodology review pooling 425 reports and 307,028 abstracts found an overall full-publication proportion of 37.3 percent, with a confidence interval of 35.3 to 39.3 percent. Survival analysis put the ten-year figure at 46.4 percent overall, 68.7 percent for randomized controlled trials and 44.9 percent for other studies. Most meeting abstracts never become a paper anyone can check.
If an abstract does become a paper, do the numbers hold?
Often not. Among 513 abstracts of randomized trials presented at one ophthalmology meeting, 230 reached publication. Of the 86 pairs reporting the same main outcome at the same time point, 47 disagreed, which is 54.7 percent. Forty of those disagreements were about magnitude and seven were about direction, meaning the effect pointed the opposite way once the full paper appeared.
Is a peer-reviewed paper automatically trustworthy?
No, and the Federal Trade Commission says so in its guidance for health product marketers. Publication is not a guarantee of quality, nor proof that a product works for the advertised benefit. The rigor of peer review also varies widely between journals, with some accepting studies on little more than payment of a publication fee. Peer review is one filter of uncertain strength, not a seal.
Why will a regulator not accept an abstract as substantiation?
Because there is not enough in one to judge. The Federal Trade Commission states it cannot evaluate the quality of a study from an abstract or an informal summary, and that a write-up needs enough detail to assess what actually took place. The agency weighs how closely a study followed its protocol and how well the report explains deviations, and an abstract shows neither.
Sources
Each document below is named as it names itself, with the date printed on that document rather than the day it was read.
- NIH Preprint Pilot, including the definition of a preprint — National Library of Medicine, PubMed Central, April 2026
- Full publication of results initially presented in abstracts — Cochrane Database of Systematic Reviews, issue 11, article MR000005 (PubMed identifier 30480762), November 2018
- Dependability of results in conference abstracts of randomized controlled trials in ophthalmology and author financial conflicts of interest as a factor associated with full publication — Trials, volume 17, article 213 (PubMed identifier 27113767), April 2016
- Health Products Compliance Guidance — U.S. Federal Trade Commission, December 2022