Appendix A — The Master Peptide Evidence Table

Every evidence rating in this book, in one place, alphabetical by compound.

Chapter 37 groups the same ratings by therapeutic area, because Chapter 37 is making an argument and the grouping is part of it. This appendix is not making an argument. It is a lookup table, built for the reader who arrives holding a name — tesamorelin, TB-500, linaclotide — and wants the verdict in ten seconds. If you want the reasoning, Chapter 37 has it. If you want the method that produced the reasoning, Chapter 5 has that. This page has the answers and their chapter numbers.

A.1 How to use this table

The tables below reproduce 154 ratings — 140 attached to molecules or indications, 14 attached to claim forms. Ten of them are split ratings shown with a slash, and four are claims the book explicitly declines to rate. They are stated as the source chapters state them, with the same wording, the same qualifiers, and the same split verdicts. Nothing has been rounded off to make a compound look better or worse than its chapter found it.

All of them are dated to 2026.

Four cautions, and the first one is the one that will actually change how you use this page.

A rating attaches to a claim, not to a molecule. This is rating rule 6 and it is the reason the table is built one row per claim rather than one row per compound. Semaglutide occupies nine rows. They are not nine opinions about semaglutide; they are nine different questions, and the answers run from ✅ through ⚠️ to 🔬. Desmopressin occupies three, BPC-157 three, and oxytocin four — and in oxytocin's case one of those rows is among the most secure ratings in the book while another is ❌. So before you take a verdict away from here, check that the row's population and endpoint match the question you came in with. Does oxytocin work? has no answer. Does intravenous oxytocin in a supervised obstetric setting produce uterine contraction sufficient to induce labor? has a very firm one. If the row you are reading names a population you are not in, or an endpoint you do not care about, you have found a rating — just not yours.

These ratings are date-stamped, and they will age. Each is a statement about the evidence available as of 2026, not a permanent property of the compound. Some will move up: the ⚠️ and 🔬 rows are mostly waiting on trials that are running now. Some will move down, which happens more often than people expect — calcitonin's fracture claim and intranasal oxytocin for autism are both in this table as downgrades, ratings that were higher once and fell as better evidence arrived. If you are reading this some years after it was written, treat every row as a starting position and check what has been published since. The rows that tell you what would change them, in the source chapters, are the ones that tell you what to go look for.

This table is a snapshot. Chapter 5's method is the durable thing. The single most useful outcome of reading this book is not memorizing which compounds got which symbols. It is being able to generate the rating yourself for a compound that is not on this page — to ask what the claim is, in whom, measured how, against what, and on what evidence, and to notice when the answer is nobody has checked. A reader who has the method does not need the table. A reader who has only the table will be stranded the first time someone mentions a molecule invented after 2026, which will be soon.

Absence from this table means the book did not rate it — not that it is unrated in the world, and certainly not that it is safe or effective. This is not a comprehensive registry of peptides. It is the set of claims this book examined, chosen because they illustrate something, because they are widely made, or because a reader is likely to meet them. Thousands of compounds are sold that appear nowhere here. Their absence carries no information at all, and reading it as a quiet endorsement is the exact inversion of what the empty space means.

One last note on scope. Nothing here is medical advice, nothing here contains dosing or protocol information, and nothing here tells you where to obtain anything. A ✅ says that a claim is well supported by evidence. It does not say the drug is right for a person, which is a different question belonging to that person and their clinician.

A.2 The four tiers, in one screen

✅ Strong clinical evidence. Adequately powered, well-designed human trials — usually more than one, usually randomized and controlled — have tested this claim in this population and found the effect. Regulatory approval, guideline inclusion, and a characterized adverse effect profile typically accompany it. A ✅ is bounded by the claim it attaches to and travels nowhere else.

⚠️ Promising but preliminary. Real human data exists and does not settle the question. This tier covers several distinct situations: small or short trials, surrogate endpoints rather than outcomes, Phase 2 results awaiting Phase 3, literatures that disagree with themselves, and literatures that are genuinely hard for a reader to reach and assess. A ⚠️ can read like a ✅ — retatrutide's weight number is larger than tirzepatide's approved one — and the tier is the warning that the number is sitting on a weaker footing than its size suggests.

❌ Hype outpaces evidence. The claim, as it is usually stated, is not supported. Note what this does not mean: it is not a statement that the molecule does nothing, and it is not a moral verdict on anyone who has used it. ❌ describes the state of the evidence, not the state of the molecule. It also arrives in two very different flavors, which §A.3 is about.

🔬 Frontier — too soon to rate. Serious science is proceeding properly through the ordinary sequence and has not yet produced a clinical answer. 🔬 is not a polite ❌ and not a soft ⚠️; it is what an early, well-conducted field is supposed to look like. It carries no probability of success. Most 🔬 becomes ❌, and saying so is part of using the tier honestly.

The six rating rules, compactly:

1. A rating attaches to a claim — and a claim can be an argument or a form of words as easily as it can be a molecule.
2. ❌ describes the evidence, not the molecule — nor the person who tried it.
3. Never upgrade with mechanism. A plausible pathway is a reason to run a trial, not a substitute for one.
4. Never downgrade with distaste. A claim that is annoying, commercially motivated, or aesthetically silly still gets rated on its evidence.
5. Falsifiability first. A claim that no result could contradict is not weakly supported; it is unrateable, and saying so is the rating.
6. One molecule, many ratings. The most-violated rule in every source you will read, and the reason this table has 140 molecule rows for far fewer molecules.

A.3 The two kinds of ❌

There are 50 ❌ ratings below and they are not all the same kind of thing. The difference matters more than the difference between two adjacent tiers, and readers get it backwards with remarkable consistency.

❌ evidence ABSENT. No adequate human trial was ever run, so the claim is unsupported and untested. Nobody knows. The compound might work; the compound might be harmful; the compound might do nothing. What is established is only that the question has not been asked in a way that could answer it — and often that no human safety data exists either. BPC-157 is the standing example. Three rows below, three ❌s, and behind all of them a substantial and genuine body of rodent work with no completed randomized human trial to translate it. Most of Part III sits here.

❌ evidence PRESENT AND NEGATIVE. Proper trials were run, in the right population, with prespecified endpoints, and they answered no. Nesiritide is the clean cardiovascular case: it was approved on hemodynamic surrogates, a large randomized outcome trial of roughly 7,100 patients followed, and it found no meaningful effect on death or rehospitalization. NK1 receptor antagonists for depression and chronic pain is the clean neuropsychiatric case: multiple adequately powered trials, across several structurally distinct compounds and sponsors, failed to beat placebo — with human PET data confirming the drugs were occupying the target receptor at the doses tested. The experiment was done. The answer arrived.

The second is a far stronger state of knowledge, and readers routinely treat it as the weaker one. A molecule that failed a trial is better understood than a molecule that never entered one. We know something about nesiritide that we do not know about anything in Part III. But the market rewards the opposite: a compound with disappointing trials is hard to sell, and a compound with no trials at all is easy, because there is nothing to argue with. Chapter 21 puts it plainly and it is worth carrying out of this appendix — the absence of a negative result is not the same as the absence of a problem, and it is usually the absence of anyone looking.

In the master table, rows the source chapters explicitly identify as present-and-negative are marked ❌ present & negative. An unmarked ❌ should be read as the absent kind unless the chapter says otherwise.

A note on the slash. Ten ratings in the master table carry two symbols separated by a slash — ⚠️ / ❌, ✅ / ⚠️, ⚠️ / 🔬. These are not indecision, and they must not be collapsed to a single tier. A split marks a claim whose verdict changes with how strongly the claim is stated or in which population it is asserted: GHK-Cu is ⚠️ for the modest instrument-measured effect and ❌ for the strong marketing version; insulin analogs are ✅ for nocturnal and severe hypoglycemia in type 1 diabetes and ⚠️ for broader superiority claims in type 2. That is rating rule 1 doing visible work inside a single row — the rating is attached to the claim, and when the claim comes in two strengths, so does the rating. A few splits also record a downgrade over time, a verdict that was higher once and fell as better evidence arrived; calcitonin's fracture claim and intranasal oxytocin for autism are both of that kind, and their cells say so.

A.4 The master table

140 ratings, alphabetical by compound. Where a compound carries several ratings they occupy adjacent rows under the same name, which makes rule 6 visible without anyone having to argue for it. Look at semaglutide, at oxytocin, at desmopressin, at BPC-157, at growth hormone, at the NK1 antagonists: the spread inside a single name is the whole point.

A handful of rows are not molecules. The book rates label claims and inference forms inside its molecule chapters — "pharmaceutical grade," "third-party tested," "nobody has reported problems," "used in animals for years," "this peptide is a nootropic" — and those are filed here at their alphabetical positions rather than moved, because that is where the harvest puts them. Section A.5 holds a further fourteen ratings of this kind. And four rows are not ratings at all — they carry NOT RATED, and they are collected and explained in §A.6.

Columns: Compound | Claim (population and endpoint) | Rating | Ch

A

Aprepitant: see NK1 receptor antagonists under N. Argireline: see Acetyl hexapeptide-8 above. Abaloparatide and other approved molecules without a rating in this book appear in Appendix E.

B

The three BPC-157 rows are the book's reference case for §A.3's first kind of ❌: the animal literature is real and the human question is genuinely open.

C

Cagrilintide: see Amylin analogs, long-acting under A. Copper tripeptide-1: see GHK-Cu under G.

Read the CNP analog's endpoint exactly as written. The ✅ covers annualized growth velocity over one year — how fast a child grows during treatment — and nothing beyond it. Chapter 28 is explicit that final adult height and any health or functional outcome are separate claims resting on long-term follow-up that is still accumulating, and they do not inherit this row's symbol.

D

Degarelix: see GnRH antagonists under G. DL-phenylalanine: see Endorphin supplements, oral under E.

Three desmopressin rows, three ✅s, and they are not the same ✅. Same molecule, same receptor, same mechanism — different population, endpoint, evidence base, and durability. Central diabetes insipidus is a permanent physiological correction with an immediately measurable endpoint. Nocturnal enuresis is a symptomatic treatment whose rating is deliberately narrow: it covers reduction in wet nights during treatment, not cure, and not durable benefit after stopping, for which the evidence is weaker and relapse is well recognized. A source that told you "desmopressin: ✅" would have compressed that difference away, which is exactly what rule 6 exists to prevent and exactly what one row per claim is for.

E

Three of these four rows are rateable empirical claims about a population and an outcome. The fourth — inevitability — is a prediction about aggregate future human behavior, which is not that kind of claim at all, and Chapter 43 declines to rate it. See §A.6.

F

The Follistatin-344 row is marked present-and-negative by extension: the relevant human trials were run in disease populations, increased lean mass, and failed on function — a result that argues against the healthy-adult claim rather than merely leaving it unaddressed. For the disease-population claim itself, see Myostatin / follistatin pathway inhibitors under M.

G

GHRP-2 and GHRP-6: see Sermorelin, GHRP-2, GHRP-6 under S. Goserelin and triptorelin: see GnRH agonists above.

The two growth hormone rows are the cleanest demonstration in this table of why rule 6 exists. Same molecule, same axis, adjacent rows, opposite verdicts — and the entire difference is whether the population has a documented deficiency. Any source that gives you a single verdict on "growth hormone" has thrown away the only thing you needed to know.

I

Ibutamoren: see MK-677 under M. Ipamorelin: see CJC-1295 + ipamorelin under C. The "immune modulation" row is a rating of a form of words, not of a molecule: it specifies no immune arm, no direction, no endpoint, and no population, which makes it compatible with every possible result and supportable by none.

Icatibant's ✅ is for acute attacks of hereditary angioedema, a rare disease with a precisely identified mediator. Chapter 29 notes that the drug has been studied in other conditions involving bradykinin-mediated swelling with results that did not repeat this success — a ✅ earned against one precisely identified mediator does not travel to a different disease with a superficially similar mechanism.

K

L

Larazotide: see KPV and larazotide under K. Leuprolide: see GnRH agonists under G. The two leptin rows are the table's tidiest ✅/❌ pair, and the reason is the population: total deficiency responds to replacement, and common obesity — in which circulating leptin is already elevated — does not.

M

Matrixyl: see Palmitoyl pentapeptide-4 under P. Metreleptin: see Leptin under L.

N

Four of the nine rows in this letter are present-and-negative ❌s, which makes N the densest concentration of genuine knowledge in the table. Nesiritide, NK1 antagonists for depression, and biomarker-guided titration were all tested properly and all came back no. That is a far better epistemic position than the untested compounds elsewhere in this table, and it is why those three are almost never marketed while the untested ones are.

O

Octreotide: see Somatostatin analogs under S. The four oxytocin rows are worth reading as a block. One is among the most secure ratings in the book. One is a rejected characterization. One is a downgrade driven by larger and better trials. One covers three separate populations that did not assemble into an indication. A single verdict on "oxytocin" would be false four different ways at once.

P

PT-141: see Bremelanotide under B. PRRT: see Lutetium Lu 177 dotatate under L. P21: see Dihexa / P21 under D. PSMA radioligand therapy: see Lutetium Lu 177 vipivotide tetraxetan under L.

Chapter 41 treats the public-figure disclosure claim as a case study in how an "ought" statement gets mistaken for an empirical one. It carries no tier; §A.6 explains why, and why that refusal is not a fifth tier.

R

Retatrutide is the book's standing example of a ⚠️ that reads like a ✅. Its number is larger than tirzepatide's approved-drug number, which makes it feel stronger, and it sits on a much weaker evidentiary footing. See also Triple agonists under T, which rates the same compound class on the same endpoint from Chapter 36's vantage point.

S

Selank: see Semax / Selank above. Somatropin: see Growth hormone under G. Suvorexant: see Orexin receptor antagonists, dual under O.

Ten semaglutide rows, four tiers. This is the single best argument in the book for reading the claim column before the rating column. Weight loss, glycemic control, and cardiovascular risk in a defined population are ✅. Kidney disease in a defined population is ✅. Steatohepatitis and preserved ejection fraction heart failure are ⚠️ because their endpoints are surrogates or symptom scores. Alzheimer's is 🔬. And the compounded-equivalence claim — which is not a claim about the drug at all, but about a category of products — is ❌.

The cardiovascular claim appears twice, from Chapters 5 and 8, because Chapter 5 uses it to teach the rating format and Chapter 8 uses it to make the clinical case. Both rows carry the same restriction, and it is the restriction that matters: established cardiovascular disease. People without it were not studied, and neither ✅ extends to them.

T

Thymosin β4: see TB-500 above. Triptorelin: see GnRH agonists under G.

Set tesamorelin against the other growth-hormone-axis entries. It is the one compound in that neighborhood with a ✅, and the reason is not that the molecule is better — it is that somebody ran randomized placebo-controlled trials with an imaging-based primary endpoint in a named population and took the result to a regulator.

U

This rating attaches to the inference, not to any molecule — which is why the veterinary GnRH row under G is ✅. Veterinary evidence is not weak evidence; it is evidence about a different claim.

V

W

Z

Note the route in that claim. This ✅ is intrathecal-only and cannot be inherited by any oral, transdermal, or subcutaneous product invoking "cone snail peptide" technology.

A.5 Claim-form ratings

The book rates fourteen claim forms — sentences rather than substances. These are kept separate from the master table on purpose, so that a reader scanning for a compound does not have to filter them out.

They are also, entry for entry, the most reusable material in this appendix. A rating on tesamorelin helps you exactly once, when someone mentions tesamorelin. A rating on "it's approved in another country, so the evidence is there" helps you every time anyone says it about anything, which will be often, about compounds that do not exist yet. Learn these fourteen and you can dispose of a large fraction of the claims you will meet without researching the molecule at all — because the defect is in the sentence's construction, not in the compound's data.

Three points about this table that are easy to miss.

Not every claim form is ❌. Two are ⚠️ and one is ✅. That matters, because a reader who concludes that "rated claim forms are the bad ones" has learned a superstition rather than a method. Disclosing substance use to a treating clinician earns a ✅ on structural grounds — differential diagnosis, interaction checking, monitoring, and perioperative assessment each take a medication history as an input and produce a wrong output when the input is incomplete. Medical supervision earns ⚠️ rather than ✅ because its direction is sound and its magnitude is unquantified, and because it adds no evidence whatsoever about whether the supervised compound works.

Several closely related claim forms are rated inside the molecule chapters and therefore appear in §A.4 at their alphabetical positions rather than here: "pharmaceutical grade" and the "98% purity by HPLC" certificate (both under P), "third-party tested" as rated in Chapter 19 and "this peptide is a nootropic" (under T and N), "nobody has reported problems" (under N), "used in animals for years" (under U), and "immune modulation" (under I). Two of these — purity and third-party testing — are rated twice, in two different chapters, from two different angles. The duplication is not an error; it is two chapters finding the same sentence wanting for different reasons.

The fix for a bad claim form is never to upgrade it. It is to replace it with a specific claim that can be checked — name the compound, the population, the endpoint, and the comparator — and then rate that. Every one of the ❌ entries above becomes rateable, and might land anywhere on the scale, the moment somebody supplies the missing specifics.

A.6 NOT RATED

Four claims in this book carry no tier at all, and none of them is an oversight.

The last one is the clearest case and the book closes on it deliberately.

It is a values question, not an empirical one. The rating system evaluates claims of the form X does Y in population Z, which evidence can settle. This claim has empirical components — how much of the population-level effect is environmental, how effective each approach is — and those can be rated, and are, elsewhere in this book. But the load-bearing word is "primarily," and it encodes a judgment about how a society allocates attention and resources between two approaches that are not mutually exclusive. No trial resolves that. A study showing structural interventions work does not establish that pharmacological ones should be deprioritized, or the reverse.

Rating it ❌ would be a category error. It would say the evidence does not support it, when the correct statement is evidence is not the relevant instrument. Rating it ⚠️ or 🔬 would be worse: it would imply a study is pending that might settle the matter, and none is, because none could be designed.

A refusal to rate is not a fifth tier. It is a statement about the kind of claim you are holding, made before any tier applies. The four tiers describe how much is known about an empirical question. NOT RATED says the question is not empirical, which is a different axis entirely — and reading it as a low score, or as a diplomatic ⚠️, misses the whole point of writing it down.

Now notice where these four sit. All of them are in Part VIII — the part of the book that turns from molecules to society, and rates claims about language, media, sport, work, food systems, and stigma. Every molecule claim in Part I through Part VII got a tier. The refusals appear only, and all together, at the point where the questions stop being about what a compound does in a population and start being about what a society should do.

That is a rating system behaving correctly. An instrument with a domain and no sense of its own edges is worse than no instrument, because it returns confident output on inputs it cannot read. These four rows are the system saying this is where I stop — and saying it out loud, in the same table, in the same format, next to the compounds it can assess. The boundary is part of the tool.

Why the category exists in a reference table at all. Because the most common way a moral or political position gets past a careful reader is by wearing an empirical costume. "Public figures should disclose" is a values claim and is not rated. "Disclosure by public figures reduces harmful weight-control behavior in their audiences" is an empirical claim about a population and an endpoint, and it would be rated. The two sentences get used interchangeably in argument, and they are not interchangeable. When you meet a claim, the first question is not whether it is true. It is what kind of claim it is.

Knowing when your own method does not apply is part of the method.

A.7 The distribution, and what it actually shows

  154 RATINGS
  ------------------------------------------------------------
  ✅  strong clinical evidence ....................  54
  ⚠️  promising but preliminary ...................  26
  ❌  hype outpaces evidence ......................  50
  🔬  frontier — too soon to rate .................  10
  /   split ratings (two tiers, one claim) ........  10
      NOT RATED (outside the system's domain) .....   4
  ------------------------------------------------------------
  by kind:   140 molecule / indication      14 claim form

Read the first and third lines against each other before anything else. More claims in this book are well supported than unsupported — 54 ✅ against 50 ❌.

Take the consequence immediately: this is not a debunking book. Fifty-four claims examined here are well supported, and if you add the twenty-six ⚠️ entries — most of which describe real human data that has not yet settled a question — plus the supported halves of the ten splits, considerably more of what this book looked at turned out to be substantiated than turned out to be empty. Anyone who finishes a book about peptides believing that peptides are mostly nonsense has read a different book. The category contains insulin, oxytocin, desmopressin, icatibant, the GnRH agonists, the somatostatin analogs, ziconotide, the natriuretic-pathway drugs that worked, and the incretin drugs, and it has contained most of them for decades.

But the margin is narrow, and the near-tie has structure. Fifty-four against fifty is not the book concluding that the field is roughly a coin flip. It is two very different populations of claims that happen to be close to the same size.

Sort the ✅ rows by where they live and they cluster hard: the approved pharmacopeia and the metabolic chapters. Semaglutide and tirzepatide. Insulin and its analogs and the closed-loop systems. The oncology hormone-axis drugs and the radioligand therapies. The natriuretic-pathway drugs that worked. Teriparatide, desmopressin, glucagon, linaclotide, icatibant, ziconotide, colistin, daptomycin, botulinum toxin. Every one of them has a sponsor who paid for a trial program because there was an approval to win at the end of it.

Now sort the ❌ rows. They cluster just as hard, in a different place: compounds sold direct to consumers without a prescription. BPC-157. TB-500. The growth-hormone secretagogues sold for body composition. IGF-1 and follistatin for muscle. Melanotan II. Dihexa. The cosmetic peptide making a botulinum-toxin comparison. The supplements claiming to raise endogenous GLP-1. And clustered around them, the label claims — pharmaceutical grade, third-party tested, 98% purity, nobody has reported problems, used in animals for years — which exist because the products they describe were never put through anything that would generate a real credential.

That is not a fact about peptides. It is a fact about which peptides get studied, which is a fact about who pays for trials. A randomized controlled trial with a clinical endpoint costs a great deal of money and returns nothing to whoever funds it unless there is an exclusive product at the other end. A compound sold as a research chemical has no such product, no such exclusivity, and therefore no such trial — and it does not need one, because it is already selling. The incentive to generate evidence and the incentive to generate revenue point in the same direction for a drug company and in opposite directions for a gray-market supplier. The distribution above is what that asymmetry looks like after it has run for twenty years.

Three corollaries worth carrying.

First: the ❌ column is not a list of molecules that failed. Overwhelmingly it is a list of molecules that were never tested — the §A.3 distinction again, and the reason it leads this appendix. Some of those compounds may work. The claim that they do is unsupported today, and the reason it is unsupported is structural rather than scientific.

Second: the 🔬 column is small — ten entries — and it is where the honest uncertainty lives. GLP-1 agonists for addiction. Semaglutide in Alzheimer's. KPV and larazotide. Kisspeptin. Antimicrobial peptides as systemic antibiotics. Individualized neoantigen vaccines. De novo designed binders. Lean-mass preservation during weight loss. Peptide-drug conjugates as a platform. Whether widespread GLP-1 use changes what the packaged food industry sells. These are the questions where serious people are doing serious work and nobody yet knows. Most of them will end up ❌, and that is not a prediction about any particular one; it is the base rate. If this table is revised in a decade, the 🔬 rows are where most of the movement will have happened, in both directions.

Third: the four NOT RATED rows are not a rounding error, and they all sit in Part VIII. The system rated every molecule claim it met. It stopped at the questions about what a society should do — and stopping there, visibly, in the same table, is §A.6's point and the last thing this book asks you to take from its method.

A.8 Where to go next

Chapter 37 — the same ratings, grouped by therapeutic area, with the reasoning attached. If a row here surprised you, that is the chapter that explains it.

Chapter 5 — the method. How the tiers are assigned, what the six rules mean, and why a claim without a population and an endpoint cannot be rated at all. This is the durable part; the table above is the perishable part.

Appendix C — the dossier: how to assemble everything that is known about a single compound, in order, so that you can produce a rating for something this book never covered.

Appendix H — twenty worked evaluations, start to finish. If you want to see the method applied rather than described, that is where it happens.

Appendix F — red flags. The patterns that predict, before you have read a single study, that a claim is going to end up in the ❌ column.

And one last time, because it is the sentence this appendix exists to make unavoidable: the rating belongs to the claim, not to the molecule. Check the population. Check the endpoint. Then take the symbol.

Related: Chapter 5 (evaluating evidence) · Chapter 12 (compounding and access) · Chapter 17 (BPC-157) · Chapter 21 (oxytocin) · Chapter 28 (natriuretic peptides and surrogates) · Chapter 31 (veterinary inference) · Chapter 37 (the ratings by therapeutic area) · Chapter 38 (reading claims) · Chapter 44 (values questions and the limits of prediction) · Appendix C (the compound dossier) · Appendix E (approved peptide medicines) · Appendix F (red flags) · Appendix H (worked evaluations) · Appendix K (glossary)