What does the study design badge on your citations mean?

It names the kind of study a citation reports, and each badge carries one sentence saying what that design can and cannot establish. A randomised controlled trial can attribute a difference to the treatment because assignment was by chance. A case series shows that an outcome occurred but cannot establish an effect, because there is no comparison group. In vitro work shows something is biologically possible at the concentration tested. The badge is there so a reader can bound a result before reading it.

Why does the number of participants matter so much?

It is the single figure that teaches the most about a human study, and it is the one most often left out of a summary. An underpowered study that reports an effect has usually reported a large one, and large effects found in small samples tend to shrink when the work is repeated. Small studies are how a field responsibly begins, so the number is not a criticism. It is context that a reader needs and that an adjective cannot supply.

Is preclinical or in vitro evidence worth anything?

Yes, for what it is. Culture work establishes that a pathway exists and can be engaged, and animal work shows an effect is producible in a living organism. Neither establishes that a person would notice a benefit. Attempts to reproduce landmark preclinical findings have confirmed only a minority of them, so the gap between mechanism and clinical result is both real and measured. A preclinical citation supports a plausible mechanism, and presenting one as evidence of a clinical outcome changes the subject.

How do I check whether a paper has been retracted?

Open the PubMed record rather than the publisher page. Retractions, corrections and expressions of concern are carried on the record itself and appear at the top. A retraction does not always indicate misconduct, since papers are withdrawn for honest error and for misidentified reagents too, but it always means the finding can no longer be cited as support.

What is the difference between a systematic review and a narrative review?

A systematic review states its search method in advance, including databases, search terms and inclusion criteria, so a reader can see what was excluded and judge whether that matters. PRISMA 2020 is the reporting standard and a review following it will say so. A narrative review has no stated protocol, which makes the selection invisible. Both are legitimate publications and only one can be audited.

What should I do with a supplier reference list?

Run five quick checks on the first few entries. Confirm the identifier resolves to the paper you were told about, note the design, find the participant count for any human study, confirm the study used material comparable to what you are buying, and check the PubMed record for a retraction. Then ask what each citation is offered to prove and whether it could prove that even if everything in it is correct.

Clinic launch guide

Reading the Evidence Behind a Product Claim

Reviewed by ExaVeyra Sciences editorial team · Updated 20 August 2026

A supplier sends a list of papers. The list is long, the journals sound serious, and there is no time to read thirty abstracts before a purchasing decision. The good news is that most of what a citation can tell you is available in about a minute, and it comes from three things: what kind of study it was, how many people were in it, and whether the paper still says what it said when it was published.

This page explains those three checks and the vocabulary around them. Every citation on this site carries a design badge for exactly this reason, and each badge has one sentence attached saying what that design can and cannot establish. What follows is how to use them.

What the design badge is telling you

Study design is the first thing to read because it bounds everything else. A result cannot be stronger than the design that produced it, no matter how large the effect or how good the journal. The designs you will meet in this field fall into a small number of kinds.

Design	What it can establish	What it cannot
Randomised controlled trial	That a difference between groups is attributable to the treatment, because assignment was by chance.	That the result generalises beyond the people studied, or that a rare harm is absent.
Clinical study, single arm or open label	What happened to a group of people who received something.	Whether the treatment caused it, since nothing separates it from time, attention and expectation.
Case series	That an outcome occurred, which is genuinely useful for reporting harm and for raising a question.	Any effect at all, because there is no comparison group.
Preclinical, meaning animal work	That an effect is producible in a living organism.	That the same happens in a person. Most compounds that work in animals do not go on to work in humans.
In vitro, meaning cells in culture	That something is biologically possible.	That the concentration used is reachable in a body, or that a cell in a dish behaves like a cell in tissue.
Systematic review and meta-analysis	A summary of a literature searched to a stated protocol, so you can see what was included and why.	More than the studies it pools. A meta-analysis of small studies is not a large study.
Consensus statement and guideline	What a field agrees good practice looks like, or how work should be reported.	What a treatment does. These answer a different question entirely.
Narrative review	Orientation, from someone who knows the area.	Much as evidence, because there is no stated search and what was left out is not recoverable.

The number that teaches the most

For any study in people, find the participant count before anything else. It is the single figure that does the most teaching, and it is often the figure a summary omits.

Consider what happens when it is present. A registered exosome trial in acute ischaemic stroke enrolled five people. That is a real study, ethically run and worth doing, and knowing the number tells a practitioner more about the maturity of the field than any adjective could. Across the registered exosome trials we track, the median enrolment is thirty. A reader who knows that has learned the state of the evidence in one number.

Small studies are not worthless. They are how a field starts, and early work has to be small before it can responsibly be large. The problem is what small studies do to the results that get published from them: an underpowered study that finds an effect has usually found a large one, and large effects found in small samples shrink when anyone looks again. This is why a striking result from a handful of participants is the most common way to be misled by a paper that did nothing wrong.

Work that has not reached people yet

A great deal of the literature around regenerative material is preclinical or in vitro, and this is normal for a field at this stage. What matters is that the citation is read as what it is. A mechanism demonstrated in culture establishes that a pathway exists and can be engaged. It does not establish that engaging it produces a benefit a person would notice, and the gap between those two statements is where most overclaiming happens.

The size of that gap has been measured. Attempts to reproduce landmark preclinical findings have repeatedly confirmed a minority of them, and analyses of the resulting waste put the irreproducible share of United States preclinical spending at roughly half. The lesson for a purchasing decision is narrow and practical: a preclinical citation supports a plausible mechanism, and a supplier who presents one as evidence of a clinical result has changed the subject.

Reviews, and the difference a protocol makes

Two documents can both be called a review and differ completely in what they are worth. A systematic review states its search in advance: which databases, which terms, which inclusion and exclusion criteria, and what happened to every record found. Because the method is stated, a reader can see what was left out and decide whether the omission matters. PRISMA 2020 is the reporting standard for this, and a review that follows it will say so.

A narrative review has no such protocol. It is an expert writing about an area they know, which makes it good orientation and weak evidence, because the selection is invisible. Both kinds are legitimate publications. Only one of them lets you audit the reasoning.

The same logic applies to a trial report. CONSORT 2010 sets out what a randomised trial should disclose, including how the randomisation sequence was generated and concealed, and how many participants were lost along the way. Concealment in particular is not a formality: trials with inadequately concealed allocation have been shown to report systematically larger effects than trials that concealed it properly.

Registration, and the result that was promised

Most clinical trials in the United States are required to be registered on ClinicalTrials.gov, and the registration is public before the results are. That ordering is what makes it useful. A registration records what the investigators said they would measure, so a published paper reporting a different primary outcome can be compared against what was planned.

Two things are worth doing with a registry record. Search the identifier a supplier quotes and confirm it describes the study they say it does. Then check the recruitment status and the enrolment figure, which tell you whether the study finished and how large it actually became rather than how large it was intended to be. Registration also carries a reporting obligation under FDAAA 801 for applicable trials, so a completed study with no posted results is itself informative.

Whether the paper still says what it said

A citation is a snapshot of a document that can change after publication. Papers are corrected, flagged with expressions of concern, and retracted, and a reference list assembled two years ago will not know about any of it. PubMed carries these notices on the record itself, which makes the check quick: open the PubMed entry rather than the publisher page, and any retraction or correction appears at the top.

A retraction is not always misconduct. Papers are withdrawn for honest error, for a figure assembled wrongly, or because a reagent turned out to be misidentified. What a retraction always means is that the finding can no longer be cited as support. When our own pages carry a flagged paper, the flag is shown in the markup as well as in the prose, because a reader scanning a citation list should not have to take our word for which references are sound.

Five checks for any citation you are handed

None of these requires reading the paper, and together they take about a minute per reference. Run them on the first three citations a supplier sends, and the character of the whole list is usually clear.

Does the identifier resolve? Search the DOI or PubMed ID and confirm the title matches what you were told. An identifier that resolves to a different paper is the fastest signal there is, and it happens.
What was the design? Human, animal or culture, and with a comparison group or without. This bounds every other question.
How many people? For any human study, find the enrolled figure. If a summary does not state it, that omission is itself worth noticing.
Does the paper study the material you are buying? A result for one preparation does not transfer to another with a different source, isolation method or characterisation.
Is the record clean? Check the PubMed entry for a retraction, correction or expression of concern before relying on it.

One further question sits underneath all five. Ask what the citation is being offered to prove, and whether it could prove that even if everything in it is correct. A well-conducted in vitro study is still an in vitro study, and no amount of methodological quality turns it into evidence of a clinical outcome.

Reporting standards specific to this field

Extracellular vesicle work has its own reporting standard, and it is directly useful when qualifying a supplier. MISEV2023, published in the Journal of Extracellular Vesicles, is the third iteration of a consensus document setting out how vesicle preparations should be produced, separated, characterised and reported. It gives a purchasing clinic a neutral yardstick that does not depend on any vendor framing.

Its practical use is comparison. Particle counts obtained by different methods are not interchangeable, so a count quoted without the method named cannot be compared to anything. Marker panels establish that vesicles are present and enriched, and they do not establish potency. A supplier working to the standard will already state both, and one who cannot has told you something useful.

Where to go next

If the open question is what to require from a supplier in writing, our page on adding exosomes to a practice lists the documentation to ask for before a first order and what each item does. If the open question is how a claim may be described once a service line is running, the advertising and claims material covers substantiation. And if you want to see these badges in use, the peptide science pages carry a design badge and a participant count on every citation, which is the point of this page.

Sources

ClinicalTrials.gov, US National Library of Medicine trial registry (accessed 20 August 2026)
ClinicalTrials.gov, Results Reporting Requirements under FDAAA 801 (accessed 20 August 2026)
US National Library of Medicine, Errata, Retractions and Other Linked Citations in PubMed (accessed 20 August 2026)
US National Library of Medicine, PubMed User Guide (accessed 20 August 2026)
US National Library of Medicine, MEDLINE Overview (accessed 20 August 2026)
US Department of Health and Human Services, Office of Research Integrity (accessed 20 August 2026)

Scientific literature

Each source carries the kind of study it was and, where the study enrolled people, how many. Study design decides what a result can establish, so it is stated rather than left to be inferred. Each line also says what that source is carrying on this page.

Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ, 2021. doi:10.1136/bmj.n71 PMID:33782057Consensus statementLandmarkSource 1 supports: The reporting standard that separates a systematic review from a narrative one. Cited for what a review must disclose about its own search, which is the thing that lets a reader audit it.
Schulz KF, Altman DG, Moher D, for the CONSORT Group. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ, 2010. doi:10.1136/bmj.c332 PMID:20332509Consensus statementLandmarkSource 2 supports: What a randomised trial report should disclose, including how allocation was generated and concealed and how many participants were lost. Cited for the checklist a reader can hold a trial report against.
Ioannidis JPA. Why most published research findings are false. PLoS Medicine, 2005. doi:10.1371/journal.pmed.0020124 PMID:16060722Narrative reviewLandmarkSource 3 supports: The argument for why study size and design bound what a positive result means. Cited for the relationship between low power and inflated effect estimates, which is the reason the participant count is the first number to find.
Begley CG, Ellis LM. Drug development: raise standards for preclinical cancer research. Nature, 2012. doi:10.1038/483531a PMID:22460880Narrative reviewSource 4 supports: The reproducibility gap in preclinical work, from an attempt to confirm landmark findings. Cited for the size of the distance between a culture or animal result and a clinical one.
Freedman LP, Cockburn IM, Simcoe TS. The economics of reproducibility in preclinical research. PLoS Biology, 2015. doi:10.1371/journal.pbio.1002165 PMID:26057340Narrative reviewSource 5 supports: An estimate of the irreproducible share of United States preclinical research spending. Cited to give the reproducibility problem a scale rather than leaving it as an impression.
Welsh JA, Goberdhan DCI, O’Driscoll L, et al. Minimal information for studies of extracellular vesicles (MISEV2023): from basic to advanced approaches. Journal of Extracellular Vesicles, 2024. doi:10.1002/jev2.12404 PMID:38326288Consensus statementLandmarkSource 6 supports: The field-specific reporting standard for vesicle preparations. Cited here as the worked example of a standard a purchasing clinic can hold a supplier to without relying on the supplier’s own framing.

Apply for wholesale access Ask our team