Mechanism is not proof

Every deep-dive on this site ends the same way: bullish on the science, sceptical on the page. It is not a hedge. It is a claim that two things people routinely merge are actually separate and graded on different evidence. “The science” is the mechanism — the receptor, the pathway, the plausible chain of cause and effect. “The page” is the benefit — the thing the molecule is sold to do in an actual person. A compound can have landmark-grade science and an empty page, and most of the peptide world lives in exactly that gap.

This piece is the map of the gap: the ladder the whole reference grades against, told as an argument rather than a standard. Learn to read the rung a claim actually sits on and the field reorganizes itself — the genuinely proven separates from the merely promising, and the sentence “it works” splits into the several very different things it can mean.

Evidence is not one thing. It comes in rungs, and each rung up is dramatically harder to reach, rarer, and more predictive of whether a drug helps a person. At the bottom sits a plausible mechanism — a receptor and a story — which is cheap, abundant, and only loosely tied to benefit. Above it, cell data, then animal efficacy: real results, but a mouse is not a person, and the graveyard of drugs that cured rodents and failed humans is enormous. Higher still, human surrogate markers — a number in the blood or on a scan moves. Near the top, the rung that actually matters: a controlled human outcome, the thing you wanted, measured against placebo. And at the summit, independent replication and approval — the result confirmed by people with no stake in it.

The trouble is that a claim’s confidence should be set by its highest rung of real evidence, and the market sets it by its most flattering sentence. A molecule with nothing but a receptor and a rodent study gets described in the language of the top rung. Reading well is mostly the discipline of asking, every time: what is the best evidence here, and which rung is it actually on?

The most seductive error on the ladder is mistaking a surrogate for an outcome — a marker that moves for the benefit you actually want. It is seductive because the marker is real and the graph looks like success. The definitive warning comes from outside the peptide world: in the CAST trial, drugs that suppressed the heart arrhythmias thought to cause sudden death — the surrogate moved beautifully — increased mortality (Echt et al., NEJM, 1991). The thing everyone was sure the marker predicted went the opposite way.

The peptide field’s own version is quieter but identical in shape. MK-677 raised IGF-1 and added lean mass on the scan for a full year — surrogates moving exactly as promised — while strength and physical function, the outcomes anyone actually wanted, did not improve, and glucose drifted up (Nass et al., Ann Intern Med, 2008). The same split shadows the whole field: GH secretagogues that raise IGF-1, incretins measured by craving scores, fragments judged by a marker in a mouse. A surrogate is a promissory note. The outcome is the only thing that pays it, and sometimes it bounces.

Two more failures live high on the ladder, where the evidence looks strong. The first is the cross-trial comparison. Even a solid controlled result cannot be ranked against a result from a different trial — different people, durations, doses and analyses mean the numbers aren’t on the same ruler. When amycretin’s early 24% weight loss gets lined up against other drugs’ phase-3 figures, the comparison is exactly the one the data cannot support. A trial tells you about its own population; it does not hand you a league table.

The second is the top rung itself: independent replication. A striking finding produced by a single group and never reproduced by anyone else is not proven — it is unaudited, which is a more honest word than true or false. Epitalon’s telomerase and longevity claims are the clean example: elegant, bold, and almost entirely confined to the network that produced them. And a rung below even that is the one the headlines keep failing — the difference between a trial that has enrolled and a trial that has reported. A protocol is not a result, and a committee vote is not a rule.

None of this is a counsel of dismissal, and that is the half that keeps the site bullish. Mechanism, cell data, animal efficacy — these are real science, frequently beautiful, and they are where discovery actually happens. A molecule sitting on a low rung is not a fraud; it is the frontier working as it should, a candidate that has earned attention and not yet earned a verdict. The sin is never possessing early-stage evidence. The sin is presenting it as if it were proof — dressing a rodent study in the language of a clinical outcome, selling a receptor as a result.

So the discipline is not scepticism for its own sake; it is calibration. Say which rung a claim sits on, plainly, and a reader can be genuinely excited about the biology while staying honest about the benefit. The standard the whole catalog runs on captures it in one line: downgrading a claim is free, but upgrading it requires a source.

Almost every confusing peptide claim resolves into a question about rungs. The molecule with a brilliant mechanism and no human data; the trial whose marker soared while its outcome sat still; the breakthrough nobody outside one lab has reproduced; the phase 3 that has started but not finished — each is a low rung wearing a high rung’s clothes. There is one question that cuts through all of them, and it is worth carrying to every page on this site and every claim off it: what is the best evidence, and which rung is it on? Answer that honestly and the rest follows.

• Science and benefit are graded separately. 'Bullish on the science, sceptical on the page' means the mechanism can be real while the proof of benefit is absent — different claims, different evidence.
• Evidence comes in rungs. Plausible mechanism → cell data → animal efficacy → human surrogate markers → controlled human outcome → independent replication. Each rung up is rarer and more predictive; confidence should track the highest real rung.
• The surrogate trap. A marker can move while the outcome doesn't follow or worsens — CAST (arrhythmia suppression, more deaths); MK-677 (IGF-1 and lean mass up, strength flat). A moving marker is a promissory note.
• The cross-trial and replication traps. You can't rank numbers across separate trials (amycretin), and a result no independent lab has reproduced is unaudited, not proven (epitalon). Enrolled is not reported.
• Low rungs are real science — just label them. Mechanism and animal data are where discovery lives; the error is selling them as proof. Grade the rung plainly: downgrade is free, upgrade needs a source.