Chapter 37 — Key Takeaways

The chapter in one paragraph

Every evidence rating in this book is collected here: 154 of them, 140 attached to a molecule-and-indication claim and 14 attached to a recurring claim form. Nothing is added, softened, or collapsed, and the chapter issues nothing of its own — every rating was decided in the chapter that owned the claim. It is a snapshot dated 2026 and it will age unevenly. Chapter 5, which taught you to rate a claim, matters more than Chapter 37, which lists the answers that method produced on one particular day.

The eight things to carry

1. A rating attaches to a claim, never to a molecule. A claim names a population and an endpoint. "Is semaglutide good?" has eight answers in this table across four tiers, and the question is the problem, not the evidence. Desmopressin makes the same point from the other side: three ratings, one hormone, one physiological action, two clinical presentations that look nothing alike.

2. There are two kinds of ❌ and they are close to opposite. Evidence absent means adequate trials were never run — the claim is unsupported and untested. Evidence present and negative means the trials ran and answered no. The second is a far stronger state of knowledge, and readers routinely treat it as the weaker one. BPC-157 for tendon healing is the first kind. NK1 receptor antagonists for depression and nesiritide for heart failure outcomes are the second.

3. Splits are findings, not hedges. Ten ratings carry two glyphs. Eight of the ten are ⚠️/❌, and almost all of those have the same shape: a modest, instrument-measured version of a claim is supportable and the strong marketing version of the same claim is not. Collapsing a split to one glyph either flatters the product or libels the science.

4. Four claims are NOT RATED, all of them in Part VIII. Three are values questions; one has a direction that two opposed mechanisms both plausibly determine. A refusal to rate is not a rating and there is no fifth tier. The clustering is what a rating system behaving correctly looks like: it has a domain, and it says so at the boundary rather than producing output anyway.

5. The distribution is 54 ✅ against 50 ❌ — this is not a debunking book. More claims turn out well supported than unsupported. Do not lean on the margin; four ratings is too small a gap to carry an argument. Lean on the structure underneath it: ✅ clusters in the approved pharmacopeia and the metabolic chapters, ❌ clusters where compounds are sold direct to consumers without a prescription. 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 (Chapter 38 §38.10).

6. ⚠️ is rare because it is expensive. Only 26 of 154. "Real human data that does not settle the question" requires that somebody ran a trial at all, and most consumer-market claims never get that far. ⚠️ is not a way station every compound passes through; it is a distinction earned by having been studied.

7. The table ages unevenly, and the pattern is predictable. ✅ on hard outcomes is close to permanent. ✅ resting on surrogates is the most fragile ✅ — nesiritide is the object lesson. ⚠️ resting on one phase 2 result is more likely to shrink than grow, because larger, longer, better-blinded studies regress early effects (Chapter 9). ❌ evidence-absent entries move most easily in either direction, precisely because nothing anchors them.

8. In a drift audit, the count is not the point — the direction is. Generous drift rates up on compounds you have a stake in; severe drift rates down on compounds whose presentation you dislike. Neither is more respectable, and the severe one is harder to catch because it feels like rigor and is right more often in a market full of nonsense. Being right for the wrong reason is still not knowing.

Numbers worth remembering

The two sentences that do the most work

On evidence: "We do not know, and the reason we do not know is that nobody paid to find out." That is the accurate reading of most evidence-absent ❌s in this table, and it is neither dismissal nor encouragement.

On yourself: the Step 5 dossier sentence — one line, dated, naming the pattern in your errors and what produced it. Very few people can write that sentence about themselves, and the ones who can are noticeably harder to sell things to.

What this chapter does not do

• It gives no doses, protocols, sources, or recommendations.
• It rates nothing new. Every entry is reproduced from Chapters 2–44.
• It says nothing about safety. A ✅ means efficacy established and safety characterized, not benign.
• It says nothing about you. Population-level evidence is an input to a clinical decision, not the decision.

Where this goes next

Chapter 38 answers the question §37.8 raises and defers: why the evidence is distributed the way it is, and who decides which peptides get studied. If you read only one follow-on section, read §38.10.

Chapter 39 turns the table into a conversation — its ✅-rated claim form is disclose everything to a treating clinician, which is the one recommendation this book makes.

Chapter 40 closes the dossier project that §37.10 has just measured.

Appendix A holds the full four-line callouts — claim, rating, why, and what would change it — for every rating summarized here. When a row in this table matters to you, that is where to go next.