01
1. Freeze the exact wording.
Write the claim exactly as stated. “Supports recovery” is too vague. Recovery of what tissue, in which population, measured how, compared with what, and over what period?
Do not improve the claim for the person making it. The verdict belongs to the wording actually used.
02
2. Resolve molecule and product identity.
List aliases, sequence, modifications, salt form, formulation, route, and product name. Check whether the cited paper studied the same evidence object.
A shared nickname is not enough.
03
3. Find the closest primary record.
Start with the current drug label or FDA review for status and safety. Use a posted trial result or full publication for outcomes. Check ClinicalTrials.gov to distinguish a registered plan from completed results.
Reviews can map a field, but the central claim should link to the original record whenever possible.
04
4. Extract the scope.
- Population and important exclusions
- Exact product, route, and comparator
- Primary endpoint and time horizon
- Magnitude and uncertainty of the effect
- Adverse events, withdrawals, and missing data
- Sponsor role and conflicts
05
5. Write a verdict that keeps the boundary.
Use “supported with scope” when the exact outcome is supported for the studied context. Use “not established” when reliable evidence is missing or indirect. Use “misleading” when wording erases a material distinction.
A good verdict acknowledges positive evidence and states what it cannot establish in the same breath.
06 · Claim meaning
Evaluate what a reasonable reader will understand
A claim includes more than its literal sentence. Images, headings, testimonials, before-and-after stories, omissions, and placement can create an implied message. The FTC calls this the net impression. “Clinically studied” beside a product can imply that the product itself produced the advertised result even when the cited paper examined only an ingredient or a different formulation.
Freeze the strongest reasonable interpretation before searching. “Supports recovery” might imply faster return to activity, less pain, tissue healing, or a biomarker change. Those are different claims. If the promotion leaves the outcome vague while surrounding it with athletic imagery, record both the exact words and the implied patient result.
A disclaimer cannot reliably cure a contradictory headline. “Results may vary” does not turn an unsupported efficacy claim into a supported one, and “for research only” does not neutralize consumer instructions or human testimonials. Verification should address the overall message, not reward carefully engineered ambiguity.
07 · Identity
Resolve aliases before opening the literature
Create an identity record with the formal name, development code, sequence or fragment, modifications, salt form, formulation, route, and product name. Search common aliases, but do not merge them until a reliable source establishes that they refer to the same object. Online peptide names frequently collapse a family, fragment, and finished product into one term.
Then compare the marketed item with the intervention in the cited study. Was the material supplied by a sponsor under a clinical protocol? Was it an approved product, compounded preparation, laboratory reagent, or undefined bulk substance? If product identity is unverified, the verdict must preserve that uncertainty even when the molecule has strong evidence.
Use FDA application records and labels for approved products, ClinicalTrials.gov identifiers for studies, and full publications for intervention descriptions. A database synonym can help discovery, but it does not prove equivalence between commercial items.
08 · Source order
Start with the record closest to the question
For U.S. approval, indication, and current label questions, start with FDA rather than a clinic page or news story. For a registered trial, use ClinicalTrials.gov and check whether results are posted. For a claimed outcome, find the full trial report or posted results. For anti-doping status, use the current World Anti-Doping Agency list and relevant sport rules.
A press release can document what a sponsor announced. It cannot replace the full methods and tables. A review can map the field and point to primary papers. It should not be the only support for a precise effect when the original trial is available. Search results, snippets, AI summaries, and social posts are discovery aids—not evidence records.
Record access date because labels, trial statuses, safety communications, and prohibited lists can change. A claim audit without a date can silently become stale.
09 · Trial design
Read the methods before the abstract conclusion
Extract eligibility criteria, setting, sample size, randomization, masking, comparator, intervention, duration, primary outcome, and analysis population. These details show what question the trial was designed to answer. A result in adults with a specific disease cannot automatically support prevention in healthy adolescents or a different route.
Check participant flow. How many were randomized, received the intervention, completed follow-up, and were included in each analysis? Differential withdrawal can bias a result. Look for protocol changes, deviations, missing data, and whether the analysis followed the prespecified plan.
Compare the publication with the trial registry. Registration before results can reduce selective outcome reporting, but a registered record is not proof of completion or success. ClinicalTrials.gov results, when posted, include participant flow, baseline characteristics, outcomes, and adverse events that can reveal details absent from a headline.
10 · Results
Separate statistical detection from meaningful benefit
A p-value does not tell the size, importance, or certainty of an effect. Extract the absolute and relative differences, confidence intervals, baseline risk, units, and time point. A statistically significant biomarker shift can be too small to matter clinically, and a nonsignificant result can still be compatible with meaningful benefit or harm when the interval is wide.
Group averages do not predict every individual response. Responder thresholds can be informative, but check whether they were defined in advance and whether missing participants were handled consistently. Subgroup findings are less reliable when many comparisons were explored after the data were seen.
Match the endpoint to the advertising. Weight change, liver fat on imaging, hormone concentration, pain score, return to sport, hospitalization, and survival answer different questions. FDA distinguishes clinical outcomes from surrogate endpoints; a surrogate requires evidence that it predicts the clinical benefit in the context being claimed.
11 · Safety
“Well tolerated” is a starting phrase, not a safety analysis
Extract common adverse events, serious events, withdrawals due to events, laboratory abnormalities, vital-sign changes, and deaths by group. Note the duration and number of exposed participants. A small or short study can identify frequent problems but cannot reliably rule out rare or delayed harms.
Check who was excluded. Trials may omit pregnancy, organ impairment, interacting medications, older adults, adolescents, or people with complex disease. A favorable safety summary in a selected population does not establish safety for everyone. Product-quality risk is a separate layer: trial monitoring cannot validate an unrelated marketed item.
Avoid treating absence of evidence as evidence of absence. “No serious events occurred in 40 participants over 12 weeks” is accurate. “The peptide is safe” exceeds the data.
12 · Total evidence
One favorable paper does not erase the surrounding record
Search for replications, larger trials, null findings, withdrawals, regulatory reviews, corrections, and retractions. Multiple papers from the same dataset or research group are not independent confirmations. Ten mechanistic studies cannot substitute for a controlled patient-outcome trial when the claim is clinical.
Funding and author conflicts do not automatically invalidate research. They should prompt attention to protocol, comparator, analysis choices, reporting completeness, and independent replication. Peer review also varies in rigor; publication is not a universal quality seal.
FTC guidance emphasizes the totality of reliable evidence and notes that numerous flawed studies do not add up to competent and reliable support. A verdict should explain the strongest evidence on both sides and why one source carries more weight.
13 · Verdict
Write a conclusion another reader can audit
Lead with the exact claim and a scoped label. “Supported with scope” means reliable evidence supports the named outcome for the studied product, population, comparator, and duration. “Not established” means direct reliable evidence is missing, insufficient, or too indirect. “Misleading” means the wording materially erases a distinction such as molecule versus product or registered versus successful.
Follow the label with the evidence basis, the most important limitation, and direct source links. State when the newest result is sponsor-reported rather than peer-reviewed. Include the review date and what evidence would change the conclusion.
The goal is not to win an argument. It is to produce a reproducible path from wording to source to boundary. A reader should be able to inspect the same records and understand why the verdict stops where it does.
- Exact claim and implied message
- Resolved molecule and product identity
- Primary regulatory and study records
- Population, comparator, endpoint, and duration
- Effect size, uncertainty, attrition, and safety
- Totality of evidence and independent replication
- Dated verdict with a clear evidence limit
14 · Worked audits
Apply the workflow to two very different peptide claims
Claim one: “RETA is approved and causes 24% weight loss.” Identity resolution shows that RETA is informal shorthand for retatrutide, an investigational molecule. FDA and Lilly sources establish that it is not currently an approved product. The 24.2% figure comes from the least-squares mean change in one higher-maintenance group of a 48-week Phase 2 obesity trial, not every participant or every Phase 3 population.
The verdict is misleading. A scoped replacement would say: “In a randomized Phase 2 trial of adults with obesity or overweight plus a weight-related condition, the highest retatrutide maintenance group had a 24.2% estimated mean weight reduction at 48 weeks; retatrutide remains investigational.” Product identity and long-term safety questions remain separate.
Claim two: “BPC-157 heals tendons in people.” Identity is more uncertain because marketed items may not establish sequence, formulation, or quality. The frequently cited evidence is largely preclinical and includes animal injury models. Those studies can support mechanistic or model-specific repair hypotheses. They do not directly establish pain, function, return to activity, or reinjury outcomes in a controlled human trial.
The verdict is not established. That label does not mean researchers proved the molecule has no biological effect. It means the advertised human outcome exceeds the current reliable evidence. FDA’s compounding safety materials add product-quality and safety uncertainty but should not be misrepresented as a controlled efficacy trial.
Notice that the two audits fail for different reasons. Retatrutide has substantial controlled human outcomes but no approval and no evidence bridge to internet products. BPC-157’s promoted human healing claim lacks the required controlled human outcomes in the first place. A single “unapproved peptide” label would conceal the most educational difference.
A repeatable audit makes those distinctions visible. It prevents the optimistic error of treating plausibility as proof and the skeptical error of treating all unapproved molecules as scientifically identical.
Store the finished audit as a dated record with direct URLs and the claim’s original wording. If a new trial, label, safety communication, or approval appears, revise the verdict and preserve the history. Transparent corrections are evidence of a functioning editorial process, not a weakness.
Sources