Appendix F — Red Flags: Evaluating a Peptide Information Source

Most of what you will read about peptides is not written by anyone trying to deceive you. It is written by people who are enthusiastic, who read a study they did not have the training to evaluate, and who repeated it. This appendix is not a guide to spotting liars. Liars are easy. It is a guide to spotting the far more common thing: sincere, confident, well-produced content that is wrong.

That distinction governs how to use what follows. A red flag is a reason to check, not a verdict. Several of the flags below appear in perfectly sound writing — a source with a financial interest can still be accurate, and Chapter 42 was explicit that a conflict of interest is a reason to verify a claim rather than evidence that it is false. Treating any single flag as disqualifying will make you confidently wrong in a new direction, which is not an improvement.

The flags are grouped by what they tell you about, and roughly ordered within each group from most to least diagnostic.

F.1 Red flags about the claim

1. The claim has no population. "Peptide X repairs tendons." In whom — rats, athletes with a specific injury, healthy adults, the elderly? Chapter 5 established that a claim without a population is not evaluable, and Chapter 8's STEP 1 and STEP 2 are the standing demonstration: the same drug at the same dose produced ~−15% weight loss in adults without diabetes and ~−10% in adults with it. Population is not a detail.

2. The claim has no endpoint, or a vague one. "Improves recovery." "Supports cognitive function." "Optimizes metabolic health." None of these name a thing that could be measured. A claim with no endpoint cannot be tested, which means it can never be shown to be wrong — and a claim that cannot be wrong is not doing the work its confident phrasing implies.

3. The claim has no comparator. Compared to what? Placebo, standard care, nothing, or a previous personal best? Chapter 5's method requires this and Chapter 28 showed why it matters: sacubitril/valsartan's result carries the weight it does partly because the comparator was an active drug already known to work.

4. A surrogate endpoint is presented as an outcome. The compound raised IGF-1, improved a body-composition scan, lowered a biomarker, increased collagen synthesis in a dish. Chapter 16 is entirely about this gap and Chapter 28 supplied the cleanest example: nesiritide improved the hemodynamic measures it was approved on and did not reduce death or rehospitalization when a large outcome trial tested it. A moved number is a hypothesis about a better life, not evidence of one.

5. Relative risk is quoted without the absolute. "Reduced cardiovascular events by 20%." Chapter 10's SELECT numbers are the reference case: a 20% relative reduction, and an absolute change from roughly 8% to roughly 6.5% — about 1.5 percentage points, with a number-needed-to-treat around 65 to 70. All three statements describe the same result. The first sounds about ten times more impressive than the third, and a source that consistently reaches for it is telling you about its purpose.

6. The claim is unfalsifiable as stated. "It works at the cellular level." "Results vary by individual." "It's optimizing your body's own systems." Ask what observation would be inconsistent with the claim. If there is none, you are not looking at a claim.

7. False precision. A compound with no completed human trials described as producing "a 40% improvement in healing time." A number that specific requires a study that specific, and its absence is more informative than its value. Precision borrowed from a rodent experiment and stated as a human result is one of the most common failures in this literature.

F.2 Red flags about the evidence

8. Animal or in-vitro data described in human terms. The most common single error in peptide content. Chapter 17's BPC-157 literature is substantially rodent work, and rodent work is a legitimate reason to run a human trial — it is not a small version of one. Watch for the tell: the study is described accurately in one paragraph and then discussed in human terms for the rest of the article.

9. Volume of citations substituting for quality. "Over 100 published studies." Chapter 5's method asks what the best study is and how it was designed. A hundred small, uncontrolled, or preclinical papers do not aggregate into one good trial; they aggregate into a large literature with a known shape. Counting is not reading.

10. Citation laundering. A claim cites a review, which cites another review, which cites a paper that says something narrower. Follow one chain to its origin and you will often find a primary source making a modest claim about a cell line. The signature is a confident sentence whose citation, when opened, does not contain it.

11. A single study carrying the entire argument. Chapter 5 established that a lone positive trial is a weaker thing than coverage implies — particularly a small one, particularly with an unusually large effect, and particularly when it has not been replicated. The effect sizes that generate headlines are disproportionately the ones that shrink on replication.

12. Testimonials in place of data. Chapter 6's territory. The problem is not that people lie in testimonials; most do not. The problem is selection — you are seeing the reports that got made and shared, drawn from an unknown denominator, with no comparison group and no account of what would have happened anyway. Chapter 9's Phase 2 optimism and Chapter 10's interim-stopping discussion are the same mechanism operating on professionals with statisticians, which should indicate how well informal impressions do.

13. "No reported side effects." For an unregulated compound this is close to meaningless, and Chapter 19 explained why: there is no reporting pathway. Absence of reports where no one collects reports is not evidence of safety. It is silence, and silence has no direction.

14. A mechanism doing the work of an outcome. "It upregulates VEGF, which drives angiogenesis, which accelerates healing." Every link may be true and the conclusion may still be false. Chapter 22's substance P antagonists bound their target precisely as designed and did not relieve pain. Rating rule 3 exists because this argument is so persuasive and so unreliable.

15. Ongoing or announced trials presented as results. "Currently in clinical trials" is a statement about activity, not about findings. Chapter 36's six pipeline questions apply: what phase, what endpoint, what population, what comparator, who is saying it, and what is the base rate — which is that most compounds entering human trials do not finish.

F.3 Red flags about the source

16. The source sells the compound. Not disqualifying, and worth knowing. A vendor's educational content is marketing with citations. The useful version of this flag is narrower: does the source's evidence standard change between the compounds it sells and the compounds it does not? A site that is rigorous about a competitor's product and credulous about its own has told you what its rigor is for.

17. The source is a clinic that dispenses what it recommends. The same structure with a white coat, and Chapter 39 §39.8 covered it. A recommendation is not evidence about a molecule any more than a testimonial is, and the reader who has learned to discount one but not the other has learned half the lesson.

18. Credentials imported from an adjacent field. A physician is not thereby an expert in peptide pharmacology; a PhD in an unrelated discipline confers no special authority here. Genuine expertise is narrow, and a source that leans on a title rather than on an argument you can check is asking you to skip the checking.

19. The "they don't want you to know" frame. Chapter 38 §38.10 addressed this directly. For most gray-market peptides, no application was ever submitted, and none was submitted because no one funded the trials, because there is no patentable commercial position in a well-known short sequence. That is a structural fact about drug economics, not suppression. The suppression frame is attractive precisely because it explains an absence of evidence without producing any.

20. Confidence uncorrelated with the evidence base. The most reliable flag in this appendix. Compare how certain a source sounds about a well-studied drug with how certain it sounds about an unstudied one. A source that speaks about BPC-157 in the same register it uses for insulin is not calibrated, and a source that is not calibrated cannot be used even where it happens to be right — because you would have no way of telling which parts those were.

21. No acknowledged uncertainty anywhere. Related but distinct. A source that has never written "we don't know" about anything has either never encountered a hard question or is not reporting the encounters.

22. Nothing ever gets downgraded. Check whether the source has ever revised a position downward. Enthusiasm ratchets: new positive findings are reported, negative ones are quietly dropped. A source with a five-year archive and no retractions is not telling you it has been right.

F.4 Red flags about the product

Chapter 19 owns this territory and Chapter 34 owns what analysis can establish. Briefly:

23. "Research chemical" or "not for human consumption" labeling. Chapter 6 named it and Chapter 19 dissected it: a liability posture, not a regulatory category. The disclaimer transfers risk to you without changing what is in the vial.

24. "Third-party tested," unqualified. Rated ❌ as a claim form in both Chapters 19 and 34. Identity, purity, content, sterility, and endotoxin are separate tests, rarely all performed, and a certificate describes a sample rather than the vial in your hand.

25. "99% purity" offered as a content guarantee. Chapter 34 §34.4: purity by HPLC peak area and peptide content by mass are different quantities. Counterion, residual water, and salts can be a substantial fraction of what is in the vial.

26. "Pharmaceutical grade." Not a regulatory term. Rated as a claim form in Chapter 32.

27. Sterility asserted without endotoxin testing. Chapter 19's formulation: sterility asks whether anything is alive in the vial; endotoxin asks whether anything ever was. Endotoxin survives sterilization and is not removed by killing or filtering out microbes. Endotoxin testing is routinely not performed outside regulated manufacture.

F.5 The five green flags

Absence of red flags is weak evidence. Presence of these is stronger, and they are rare enough to be genuinely diagnostic.

1. The source states what would change its mind. The single best indicator that you are reading someone reasoning rather than someone arguing. It is field 12 of the dossier, applied by the writer to their own position.

2. The source rates claims, not molecules. Any source that says "this compound is ✅ for X in population Y and ❌ for Z" is using the distinction that makes the rest of the reasoning possible.

3. The source distinguishes untested from tested and failed. Rarer than it should be. The field knows far more about nesiritide, which failed a large outcome trial, than about BPC-157, which has never had one — and the second is the one people treat as more open.

4. The source reports negative findings about things it likes. Including its own prior claims.

5. The source's confidence tracks its evidence. ✅ language for well-supported claims, hedged language for weak ones, and explicit "we don't know" where nothing is known. This is calibration, and it is the whole thing. A calibrated source can be wrong on specifics and still be useful, because its uncertainty is informative. An uncalibrated source cannot be used even when correct.

F.6 Applying this to a source you already trust

The uncomfortable exercise, and the one worth doing.

Pick the source that convinced you a peptide was worth taking seriously — the video, the article, the thread, the person. Run this appendix against it. Not to debunk it; several of the flags here appear in careful writing. The question is not whether the source is good but whether the specific claim you took from it survives the checks.

Then run the checks on this book. Chapter 5 stated the standard and the book is bound by it. Where this book gives a rating without a population and an endpoint, where it quotes a relative risk without the absolute, where it sounds more certain than its evidence supports — those are errors, and you now have the tools to catch them. A textbook exempting itself from its own method would be demonstrating flag 20 on the largest possible scale.

Related: Chapter 5 (the method) · Chapter 6 (hype and selection effects) · Chapter 19 (the gray market) · Chapter 34 (what analysis establishes) · Chapter 38 (regulatory status) · Chapter 42 (incentives) · Appendix C (the dossier) · Appendix D (reading a trial) · Appendix H (worked evaluations)