Chapter 23: Nootropic and Neuroprotective Peptides — Semax, Selank, Cerebrolysin, and the Cognitive Claims

"If a man will begin with certainties, he shall end in doubts; but if he will be content to begin with doubts, he shall end in certainties." — Francis Bacon, The Advancement of Learning (1605)

Overview

Every chapter so far has asked you to compare a claim against the evidence for it. This chapter asks something harder: what to do when the evidence exists, is real, and you cannot read it.

Semax and Selank are registered medicines. Not in the United States and not in the European Union — in Russia, where they were developed and where they have been prescribed for decades. There is a literature behind that registration, peer-reviewed within its own system. Much of it is in Russian, a good deal is not indexed in the databases most English-speaking readers search, and a great deal predates the trial-registration and reporting conventions Chapter 6 taught you to look for — as does a great deal of everyone's older literature.

You have two obvious moves available, and both are failures.

The first is to dismiss the whole body of work because it is unfamiliar. This feels rigorous. It is the rule-4 error — downgrading on distaste rather than on evidence — wearing a lab coat, and it requires believing that findings become false when they are hard for you to reach.

The second is to accept the work because it exists and is voluminous. This feels open-minded. It abandons the standard you have applied to every Western compound in this book, and lets "there are hundreds of papers" substitute for "here is what the trials measured, in whom, against what."

The correct move is the uncomfortable third one: apply the same questions everywhere, and be explicit about which ones you could not answer and why. That is the chapter's actual subject, and the skill outlives peptides entirely.

The rest of the chapter earns that lesson twice. Cerebrolysin gives you a literature you can read in English, in quantity, with randomized trials and meta-analyses — and it still does not settle. Dihexa and P21 give the opposite case: striking animal data, no completed human trials, and a gray market running ahead of both. And cognitive endpoints turn out to be the easiest endpoints in this book to fool yourself with, for five specific and learnable reasons.

In this chapter, you will learn to:

• Take a "nootropic" claim apart and find the endpoint, the instrument, and the population hiding inside it — or establish that none are there
• Say accurately what Semax, Selank, and Cerebrolysin are, and where each one is and is not approved
• Evaluate a literature produced under a different regulatory and publishing culture without dismissing it and without deferring to it
• Read "mixed results" as the informative finding it usually is
• Name the five failure modes that make cognitive results untrustworthy, and spot each in the wild
• Explain why subjective enhancement is the worst possible endpoint for weak blinding
• Describe, in specifics, the trial that would change your mind

Learning Paths

All five paths should read §23.3, §23.7, and §23.8. They are the most transferable sections in Part IV, and none of them is really about peptides.

💊 GLP-1 — §23.7 and §23.8 are yours. Cognitive side effects and "brain fog" claims attach themselves to metabolic drugs constantly, and the instruments used to measure them are the ones described here. 🏋️ Performance — read in full. Same market, same vendors, and the same reasoning failures you met in Chapters 17 and 19, in a domain where the outcome is softer than "my shoulder feels better." 🔬 Science — read in full, slowly, and treat §23.3 as a methods chapter. §23.2's shared design motif is the cleanest illustration of Chapter 4's stabilization logic in the book. 💄 Cosmetic — §23.6 is short and worth your time; "is it really a peptide?" is the identical question you will face over INCI names in Chapter 30. Skim §23.4 and §23.5. 🏥 Clinical — §23.2, §23.4, and §23.5 are the practical ones. Patients arrive having imported these, and the useful conversation requires knowing what is registered where, and being able to say "the evidence I can assess is limited" without saying "that country's science is bad."

23.1 What a "nootropic" claim actually asserts

Start with the word, because the word is where the trouble begins.

Nootropic was coined in the early 1970s by Corneliu Giurgea, the Romanian-born chemist who developed piracetam, from Greek roots meaning roughly mind-turning. He did not use it loosely: a nootropic should enhance learning and memory, protect learned behavior against disruption, protect the brain against injury, improve higher cortical function, and do all of it without the sedation, stimulation, or toxicity of conventional psychoactive drugs.

Popular usage has since retained exactly one of those criteria — makes you smarter — and that one was the least precisely specified of the five. Everything else, including the demands for neuroprotection and for an unusually clean safety profile, has quietly fallen off.

So when a product, a forum post, or a clinic's website calls something a nootropic, what is being asserted? You cannot tell, and neither can they. The word is compatible with at least seven distinct claims, and they are not variations on a theme. They are different endpoints, requiring different instruments, populations, durations, and trials.

Read down that last column and notice something: a compound could plausibly do well on one row and nothing on any other. A drug that increases arousal will look excellent on a vigilance task administered to tired people at four in the afternoon and look like nothing on a delayed-recall test in rested undergraduates. That is not a contradiction. It is two different claims, one supported and one not, and a rating system that assigns one verdict per molecule cannot represent it.

You have met this exact structural failure twice already. Chapter 10 took apart "anti-inflammatory," a word that can mean suppression of a cytokine, reduction of a swelling, relief of a symptom, or alteration of a disease course — four claims of wildly different evidentiary standing, sharing one adjective. Chapter 18 did the same for "immune modulation," which can mean up, down, or sideways, and which is unfalsifiable precisely because it does not say which. "Nootropic" is the third instance of the same trick, in a third domain.

The pattern is worth naming because you will keep meeting it: a claim that names a direction but not an endpoint cannot be wrong, and a claim that cannot be wrong is not a claim. It is a mood.

📊 Evidence Rating

Claim: "This peptide is a nootropic" — offered without further specification. Rating: ❌ Hype outpaces evidence Reason: The statement names no endpoint, no measuring instrument, and no population, so there is no result that could confirm or refute it; it is not a weakly supported claim but an unevaluable one. What would change it: Restating it as a specific claim — compound X improves delayed recall on a validated word-list task in healthy adults aged 50–70 over eight weeks — at which point it becomes rateable, and might land anywhere from ❌ to ✅ depending on what has been done. (Rated as of this writing; see §23.9 for the chapter's dated rating table.)

Notice that the ❌ is not aimed at a molecule. It is aimed at a sentence form — rule 1 doing real work. Several compounds in this chapter would earn better ratings for specific, well-posed claims, and marketing worse than they deserve is part of what keeps them from being taken seriously.

🔍 Check Your Understanding

1. A product page says a peptide "supports cognitive function." Which of the seven rows in the table above is being claimed? What follows from the fact that you cannot tell?
2. Why is "improves attention in sleep-deprived adults" a better claim than "improves cognition," even though it promises less?
3. Chapter 10's "anti-inflammatory" and Chapter 18's "immune modulation" share a structural flaw with "nootropic." State the flaw in one sentence without using any of the three words.

23.2 Semax and Selank: what they are, and where they are medicines

Two compounds anchor this chapter, and they are best understood together, because they are two executions of the same design idea.

Semax is a synthetic peptide related to a fragment of adrenocorticotropic hormone (ACTH). ACTH is best known as the pituitary hormone driving cortisol release from the adrenal cortex, but a middle portion of the molecule — the ACTH(4–10) region — has long been studied for effects on attention, learning, and memory that appear separable from the hormonal function. Semax is built from that behaviorally active region rather than the hormonal one. It is a heptapeptide, and it is not a fragment as found in nature: a short Pro-Gly-Pro tail has been attached to the C-terminal end.

Selank is a synthetic analog related to tuftsin, an endogenous tetrapeptide from the heavy chain of immunoglobulin G, best known for immune-related activity — which is why tuftsin also appears briefly in Chapter 18. Selank is likewise a heptapeptide, constructed the same way: the tuftsin sequence with the same Pro-Gly-Pro tail appended.

TWO PEPTIDES, ONE DESIGN IDEA

  SEMAX     Met - Glu - His - Phe   +   Pro - Gly - Pro
            └── ACTH(4–7)-like ──┘      └─ stabilizing tail ─┘
            the part that carries         added; not present in
            the behavioral activity       the natural fragment

  SELANK    Thr - Lys - Pro - Arg   +   Pro - Gly - Pro
            └───── tuftsin ─────┘       └─ stabilizing tail ─┘
            an endogenous immune-         the same three residues,
            active tetrapeptide           the same job

  Two different active fragments. One shared engineering solution.

That shared tail is not decoration. Recall §1.2: proline is the residue that breaks regular structure, because its side chain loops back onto the backbone nitrogen. A run of prolines is rigid, awkward, and — the relevant part — a poor substrate for many of the peptidases that would otherwise chew a short peptide apart from its ends. Adding Pro-Gly-Pro to the C-terminus is a half-life intervention of exactly the kind Chapter 4 catalogs, and a strikingly economical one: three residues, no fatty acid, no PEG, no unnatural amino acids. Take a natural fragment with reported central activity; staple a protease-resistant tail on the end so it survives long enough to matter. That is real medicinal chemistry and deserves to be recognized as such.

Both compounds are registered medicines in Russia, where they were developed, and neither is approved in the United States or the European Union. Semax has been used there principally in neurological contexts, including cerebrovascular indications; Selank principally as an anxiolytic. Both are typically given intranasally, as drops or spray.

That last fact is not a footnote. It puts these compounds inside the argument of Chapters 21 and 22. Intranasal administration for a peptide intended to act in the brain is an attempt at the hardest delivery problem in this book: a molecule this size does not cross the blood-brain barrier in useful quantity from the bloodstream, and the nose-to-brain route is proposed as a partial bypass. Chapter 22 laid out what is established and what is not — that some transport along olfactory and trigeminal pathways is demonstrable, that the delivered fraction is generally small and hard to measure in living humans, and that "intranasal" in a protocol covers an enormous range of actual deposition depending on device, volume, and head position. Every one of those uncertainties applies here. A behavioral effect after intranasal dosing does not establish that the peptide reached the brain; a null does not establish that it would not have worked if it had.

Proposed mechanisms include effects on BDNF (brain-derived neurotrophic factor) and on monoaminergic systems — serotonergic and dopaminergic signaling in particular. Both are plausible, both have supporting preclinical work, and both are exactly what rule 3 forbids you to convert into an upgrade. Mechanism explains how an effect could occur; it is not evidence that it does. If you find yourself more confident about a clinical claim because the BDNF story is satisfying, you have made the error this book exists to prevent — the same one that kept "oxytocin is the love hormone" alive a decade past its expiry date (Chapter 21).

🧬 The Molecule — why Pro-Gly-Pro, and what it tells you about intent

Look again at the tail. Pro-Gly-Pro is three residues, and each one is chosen.

Proline is conformationally restricted and resists exopeptidase attack, which is why proline-rich stretches turn up so often at the ends of short bioactive peptides. Glycine, with its single-hydrogen side chain, is the most flexible residue there is (§1.2), and it sits between the two prolines like a hinge in an otherwise stiff segment.

Why does this matter for reading the literature? Because it tells you what the designers were worried about. Nobody adds a protease-resistant tail unless they expect rapid degradation — an admission, built into the molecule itself, that the parent fragment has a very short life in vivo.

Notice also what the design does not solve. A stabilizing tail addresses degradation. It does not address absorption, distribution, or the blood-brain barrier. Chapter 4's toolkit has separate tools for each, and solving one is not solving the others.

💊 In the Clinic — what "registered in one country, not another" actually means

Patients and readers routinely collapse two different things: is this a real medicine? and is this approved where I live? They are separate questions with separate answers, and getting the relationship right matters for this whole chapter.

A national registration is a real event. A regulator reviewed a dossier and permitted marketing for stated indications. That is more than a gray-market compound has (Chapter 19), and treating it as equivalent to nothing is simply inaccurate.

But registrations are not interchangeable across agencies or eras. Different systems have differed, at different times, in what evidence they required and how much of the review they published. The FDA and the EMA publish substantial assessment documents for products they approve; that transparency is useful to a reader, and its absence elsewhere is a limit on what you can inspect, not a statement about the quality of the work reviewed.

And non-approval in the U.S. or E.U. is usually not a rejection. This is the point most often missed. In the large majority of cases, a compound is unapproved in a given country because no sponsor ever filed there — a commercial decision about market size, patent life, and the cost of running new trials to that agency's specifications, made by people who were not evaluating the science at all. "The FDA has not approved it" and "the FDA looked and said no" are entirely different facts, and only the second one is evidence about the compound.

For a clinician the practical consequences are unglamorous: a patient who obtained one of these abroad or online is using a product with no U.S. or E.U. label, no pharmacovigilance pathway you can access, no assured identity or purity (Chapter 34), and no interaction data in the references on your desk. None of that requires insulting the underlying research. It requires saying what is true: I have no way to verify what is in this, and no framework here for monitoring it.

23.3 Reading evidence from a different regulatory culture

This is the section the chapter exists for. Everything else in it is a worked example.

Here is the situation, stated without spin. There is a substantial body of research on Semax, Selank, and related compounds. It was produced by trained scientists at established institutions. It was peer-reviewed within its own system, and it was sufficient to support registration by a national drug regulator. Much of it is published in Russian. A significant fraction sits in journals not indexed in the databases most English-speaking readers search, or is indexed only by title and abstract. Some of it predates the conventions on trial registration and structured reporting Chapter 6 taught you to look for.

You are now asked what to think. Two easy answers present themselves, and both fail.

The parochial failure

The first: I have not seen it in a journal I recognize, therefore it does not count.

This is common and indefensible, for a reason worth stating flatly. The accessibility of a finding to you is not a property of the finding. A trial does not become less true because its report is in a language you do not read, or because a commercial indexing service made a business decision about which journals to include. If you would not accept "I couldn't find it, so it's false" from someone dismissing a Western paper they failed to locate, you cannot accept it from yourself.

There is also a self-diagnostic to run. Rule 4 says: never downgrade with distaste. That rule was written with gray-market performance compounds in mind, where the temptation is to rate down because the marketing is obnoxious. It applies with equal force to unfamiliarity. A rating that drops because the research came from somewhere you associate with lower standards, rather than because a specific question about a specific study came back unfavorably, is a rule-4 violation in a more respectable costume. The test is mechanical: can I name the study-quality problem, or am I gesturing at a country? If the second, you have done no work — and note the asymmetry that makes this failure so tempting: parochial dismissal never costs you anything visible. You will never be shown the trial you refused to look for.

The credulous failure

The second: there are decades of clinical use and hundreds of publications, and Western regulators simply have not caught up.

This fails for reasons you know from earlier chapters, none of them specific to any country.

Volume is not weight. Chapter 6 established this in a Western context: a hundred small, non-randomized, non-blinded studies do not sum to one adequate trial. They sum to a hundred results you cannot interpret. The arithmetic is the same everywhere.

Duration of use is not evidence of efficacy. It is weak evidence about gross safety at customary doses, which is not nothing. But bloodletting had a long track record too. Long use tells you a practice persisted; persistence has many causes and efficacy is only one.

"They don't want you to know" is not a finding. The suppression frame is unfalsifiable, and it does exactly the same job as the parochial dismissal pointed the other way: it substitutes a story about institutions for a statement about data. It is also empirically weak — pharmaceutical companies license compounds from anywhere on earth when data supports the investment. The mundane explanation for non-approval, that nobody filed because filing means new trials to a new agency's specification at considerable cost with an unclear patent position, is duller and better supported.

The symmetric point, which is the important one

Now the part that makes this section honest rather than merely balanced.

Western literatures have their own well-documented, well-quantified problems. Not hypothetical ones. Measured ones, in this book:

• Publication bias. Positive results are published more often, faster, and more visibly. Not a suspicion — one of the most replicated findings in meta-research, and Chapter 6 showed you the funnel plots.
• Industry funding effects. Trials funded by a product's manufacturer are systematically more likely to report results favorable to that product. Again, measured, repeatedly.
• Selective outcome reporting. The gap between what protocols pre-specify and what papers report is large enough that trial registries were created specifically to expose it (§5.10).
• Replication difficulty. Chapter 21 walked you through the intranasal oxytocin literature — almost entirely in English, in indexed journals, from well-funded Western labs, producing a decade of confident, widely publicized findings, many of which did not survive larger and better-controlled replication attempts. That is not an aberration to be embarrassed about. It is what happens when small studies, soft endpoints, and an appealing story combine, and nothing about that combination is culturally specific.

Put those next to each other and the conclusion is uncomfortable but simple. The standard is not "Western good, other bad." It is the same set of questions, applied everywhere, plus an explicit accounting of when you could not apply them.

THE SAME SIX QUESTIONS — APPLIED TO EVERY TRIAL, EVERYWHERE
                                          |  Semax /  |  Oxytocin  | Cerebrolysin
  (from Chapters 5 and 6)                 |  Selank   |  (Ch 21)   |   (§23.4)
  ────────────────────────────────────────┼───────────┼────────────┼─────────────
  1  Randomized?                          |     ?     |    yes     |    yes
  2  Blinded — and did blinding hold?     |     ?     |  yes / ?   |  yes / ?
  3  Primary endpoint pre-specified?      |     ?     |   often no |   varies
  4  Adequately powered for that endpoint?|     ?     |  often no  |   varies
  5  Population defined and relevant?     |     ?     |   varies   |   varies
  6  Results reported regardless of       |     ?     |   no —     |   contested
     outcome, and registered in advance?  |           | pub. bias  |

  "?" IS A LEGITIMATE ANSWER. It is not a zero and it is not a pass.
  It means: I could not determine this, and here is why.
  Recording "?" honestly is the entire skill of this section.

That table is the practical heart of the chapter. It does not put the Russian literature in a special box. It puts the same six questions to three literatures and lets the answers fall where they fall — including a column of question marks, and a "no" in the most accessible column. The oxytocin column is not better because it is easier to read. It is easier to read, and it also has known problems. Two different facts, kept separate.

What you can check even when you cannot read the paper

The "?" is not the end of the process but the start of a smaller one. Even without full text in a language you read, several things remain checkable, and doing them is what separates honest uncertainty from lazy uncertainty.

WHEN YOU CANNOT READ THE PRIMARY LITERATURE

  YOU CAN USUALLY STILL CHECK          |  YOU CANNOT CHECK (say so)
  ─────────────────────────────────────┼──────────────────────────────────────
  Translated titles and abstracts      |  Whether the reported analysis was
  (often indexed even when full text   |  the pre-specified one
  is not)                              |
  Whether any English-language          |  Whether blinding held
  systematic review has assessed and   |
  included these studies               |  What the excluded participants did
  Whether independent groups OUTSIDE   |  Whether adverse events were
  the originating system have tried    |  collected systematically
  to replicate — and what happened     |
  Whether a regulator that publishes   |  Whether other trials were run and
  assessment reports has reviewed it   |  not reported
  Whether the compound appears in any  |
  public trial registry, anywhere      |  The actual numbers, if only an
                                       |  abstract is available
  Whether reviews describe the trials  |
  as small/short/single-center         |

Two of those rows deserve emphasis.

Independent replication outside the originating system is the single most informative thing to look for, and it requires no judgment about anyone's competence. Findings that generalize tend to travel. When a compound has been available for decades and independent groups elsewhere have not produced confirmatory trials, that is a real observation — but be careful what it licenses. It is consistent with the effect being absent. It is also consistent with nobody having tried, because there is no patent position and therefore no funder. Chapter 19 made this point about off-patent compounds generally: an absence of trials is often a fact about money rather than about biology. Note the absence; do not convert it into a negative result.

English-language systematic reviews are more useful here than people expect, because good ones search multiple-language sources and report what they found and could not obtain. A review that says "we identified n studies, of which k were available only in abstract" has done a large part of your accessibility work and told you the size of the gap.

🔬 Read the Study — the six questions when all you have is an abstract

Suppose you locate a translated abstract reporting that a peptide improved cognitive recovery in patients after an ischemic stroke. You cannot obtain the full text. What can you actually extract?

From the population sentence: who was enrolled, and how sick were they? "Patients in the acute period following ischemic stroke" is a population in which spontaneous recovery is the dominant signal. Anyone improves. That does not make a trial uninformative — it makes the control group load-bearing, and it means an uncontrolled design in this population tells you approximately nothing.

From the design sentence: does "randomized" appear? "Double-blind"? "Placebo-controlled"? Their absence in an abstract is weak evidence of absence in the study, because abstracts are compressed and conventions differ — but their presence is real information, and it is checkable.

From the numbers, if any: how many participants per arm? Thirty per arm can only detect a large effect. If a small sample yields a significant result on a soft endpoint, either the effect is large, or you are looking at one of the outcomes that happened to move (§5.10). You cannot tell which from an abstract, and you should not pretend otherwise.

From the endpoint sentence: is one primary outcome named, or is the abstract a list of things that improved? A list is a warning, in every language.

What you cannot extract: whether the analysis was pre-specified, whether blinding held, how many participants were lost, whether adverse events were systematically collected, and whether other trials by the same group went unpublished. Five of the most decision-relevant facts about any trial, and an abstract contains none of them.

The output of this exercise is not a verdict. It is a structured statement of what remains unknown. That statement is a legitimate research product. Write it down and move on; do not let the discomfort of an unresolved entry push you into resolving it with a guess in either direction.

The rule to take away

Here is the compressed version, and it generalizes far beyond this chapter:

Judge evidence by its design and reporting, never by its origin. When you cannot judge because you cannot access it, record that as an access limitation, not as a result. And apply the same rule when the inaccessible literature is one you would have liked to believe.

That last clause is the one people skip. The discipline is only real if it costs you something in both directions.

📊 Evidence Rating

Claim: Semax or Selank improves cognitive performance (Semax) or reduces anxiety (Selank) in adults, evaluated in Western evidentiary terms. Rating: ⚠️ Promising but preliminary Reason: These are registered medicines in their home jurisdiction with a real supporting literature, most of which is difficult for the typical reader of this book to obtain and assess, and which has not been replicated in independently conducted trials meeting the standard applied to every other compound rated here. What would change it: Independently conducted, pre-registered, adequately powered randomized trials with a single pre-specified cognitive or anxiety endpoint, reported in full and in an accessible venue — or, on the other side, well-conducted replication attempts that come back null. (Rated as of this writing.)

Read this rating carefully, because it is easy to misread. The ⚠️ describes the state of the evidence available to the reader, not a judgment that the underlying research is poor. Those are different statements and this book will not let them blur. A ⚠️ awarded for inaccessibility is a confession about the evaluator's position as much as a description of the evidence; a ⚠️ awarded for small, short, unblinded trials is a description of the studies. Both are ⚠️, and if you cannot say which kind you are looking at, you do not actually understand the rating you just gave.

23.4 Cerebrolysin and the stroke and dementia trials

Now a case with the opposite shape, and it is instructive precisely because the accessibility excuse is unavailable.

Cerebrolysin is a preparation of low-molecular-weight peptides and free amino acids derived from porcine brain tissue. It has been used for decades in a number of countries — across parts of Europe, Asia, and elsewhere — for acute ischemic stroke, traumatic brain injury, vascular dementia, and Alzheimer's disease. It is not approved in the United States. It is given by injection or infusion, not intranasally, which sidesteps the delivery uncertainty of §23.2 and replaces it with a different one.

Before the evidence, notice the identity problem, because it shapes everything downstream.

Every compound rated in this book so far has had an answer to Field 1 of your dossier: a name, a length, a sequence, a molecular weight. Cerebrolysin does not, in the same way. It is a mixture produced by enzymatic breakdown of biological tissue — a defined process yielding a characterized product, but not a single molecule with a sequence you can write down. This is not a scandal; several legitimate medicines are biologically derived mixtures, and heparin was the standard example for a century. But it has three consequences:

1. "What is it?" and "what is in this batch?" are different questions, so batch consistency becomes an evidentiary issue in its own right rather than an assumed background condition.
2. Mechanism claims are harder to pin down. Neurotrophic-like activity attributed to a mixture is a hypothesis about an ensemble, and rule 3 applies with extra force: "brain-derived peptides support neuronal survival" is exactly the satisfying story that should not move a rating by itself.
3. Cross-study comparisons are less secure, because you cannot be certain that what was tested in one decade is identical to what was tested in another.

Now the literature, which is genuinely large and genuinely unresolved.

There are randomized controlled trials, including multicenter ones, in acute ischemic stroke and in dementia. There are meta-analyses. There are Cochrane systematic reviews. And the results are mixed in a way that has not resolved with more data. Some trials and pooled analyses report benefit, particularly on some functional and cognitive scales and in some subgroups. Others, including large trials and systematic reviews applying strict methodological criteria, do not find convincing benefit on primary outcomes such as death or dependency, and reviewers have raised questions about adverse events and about the quality and consistency of the underlying studies. Methodological criticism has been persistent, and heterogeneity is substantial: different regimens, durations, outcome scales, populations, regions, eras.

That is an honest summary, and it is meant to be unsatisfying.

What "mixed results" usually means

Here a reader can improve on the average commentator. "Mixed results" is often treated as a null state — the question is open, so believe what you like. It is not a null state. It has content.

When randomized evidence stays mixed after many trials and many years, the possibilities are these, roughly in order of how often they turn out to be the answer:

The effect is small. Small effects produce exactly this pattern: significance that flickers with sample size, meta-analyses that land near the line and move when one study is added or excluded, and subgroup findings that do not replicate. A small effect may still be worth having — for a devastating condition with few options, a small real benefit matters — but it is a different clinical proposition from what marketing usually describes.

The effect is conditional. It exists in some population, at some point in the disease course, and not others — entirely plausible in stroke, where timing is everything. But conditional effects have to be demonstrated in a trial designed prospectively around that condition, not inferred afterward from which subgroups moved.

The effect is absent, and the positive results reflect the ordinary machinery of bias. Smaller trials, weaker blinding, publication bias, and outcome selection reliably manufacture a scatter of positive results around a true zero. Chapter 6 showed you this with funnel plots; the tell is a systematic relationship between study size and effect size.

The trials are not measuring the same thing. With a biologically derived mixture and a range of outcome scales, this is a live possibility here in a way it would not be for a single-molecule drug with one validated endpoint.

What the pattern almost never means is it works well and everyone has failed to notice. Large, reliable treatment effects are not shy. They show up in the first adequately powered trial and they survive skeptical reanalysis. Twenty years of disagreement is itself a measurement, and what it measures is an upper bound on effect size.

📊 Evidence Rating

Claim: Cerebrolysin improves functional or cognitive outcomes in acute ischemic stroke, or in vascular dementia and Alzheimer's disease. Rating: ⚠️ Promising but preliminary Reason: Randomized trials and multiple meta-analyses and systematic reviews exist and are accessible, and they disagree; positive findings coexist with null findings, substantial heterogeneity, and sustained methodological criticism, and the product is not approved in the United States. What would change it: A large, pre-registered, adequately powered independent trial in a defined population, with one pre-specified primary functional endpoint and a defined treatment window, reported in full — upgrading on a clear positive, downgrading on a clear null. Consistency in an updated systematic review that resolves the current heterogeneity would move it in either direction. (Rated as of this writing.)

This is the chapter's cleanest ⚠️, and worth contrasting with the Semax and Selank rating in §23.3. Both are ⚠️ for different reasons: one because the evidence is hard to reach, the other because it is easy to reach and does not agree with itself. Rule 6 says one molecule can carry many ratings; this pair shows one symbol can carry many meanings, which is why the four-line format requires a reason line. A rating without its reason is a number without units.

23.5 Dihexa, P21, and the compounds with only preclinical data

Two more compounds circulate in the same conversations, and their evidentiary situation is completely different. Being clear about how it differs is the entire lesson.

Dihexa is a synthetic analog of angiotensin IV — a fragment of the angiotensin family better known for blood-pressure regulation but with recognized central actions too. Dihexa has been reported, in animal models, to have remarkably potent effects on synaptogenesis, the formation of new synaptic connections, with hepatocyte growth factor and its receptor proposed as the mechanism. The preclinical reports include improved learning and memory measures in animal models of impairment.

P21 is a peptide derived from a neurotrophic factor, developed as a small, more drug-like surrogate for a full-length neurotrophic protein — a rational strategy, since the natural proteins are far too large and too poorly distributed to work as drugs. It has been reported in animal models to enhance neurogenesis and to affect markers relevant to neurodegenerative disease.

Neither compound has completed human trials. Both are sold on the gray market.

Now the Chapter 17 posture, which this book takes toward every preclinical-only compound and which requires holding two things at once without letting either cancel the other.

Take the preclinical work seriously. These are not made-up findings. The synaptogenesis work in particular describes effects at strikingly low concentrations, and if it holds up it is scientifically interesting on its own terms. Dismissing animal data as worthless is its own error — animal data is how nearly every drug you have ever taken began.

State the human situation accurately. There are no completed human efficacy trials. There is no established human dose-response, no human safety database, no human pharmacokinetics to speak of. Every question that matters for a person deciding whether to take something is unanswered, and "unanswered" here does not mean "probably fine."

And explain that both are true. The most common failure in this space is treating those two paragraphs as if one must be a rhetorical concession to the other. Interesting preclinical work and absent human evidence are simultaneously, unremarkably true of hundreds of compounds, most of which will never become drugs — not because the animal work was fraudulent, but because attrition from promising animal result to approved medicine is brutal for reasons that are themselves informative. Effects vanish across species. Doses reachable in a mouse are unreachable in a human. Toxicity appears in the longer exposures human use requires. A mechanism that looks clean in a healthy young animal behaves differently in an aging brain with comorbid disease.

That attrition is not a bug in the system. It is the system detecting things that were not true of humans, and it is exactly the detection step that a gray-market purchaser skips.

🩺 Safety and Risk — being the first-in-human, without the protections

When a compound moves into human testing through the normal route, a set of protections attaches to that transition: dose escalation starting far below the projected active dose, intensive monitoring, pre-defined stopping rules, an ethics committee that reviewed the plan, and a manufacturing standard with documented identity and purity.

A person who obtains a preclinical-only compound from an online vendor is doing the first-in-human exposure with none of that. That is the actual comparison — not "medicine versus supplement," but "phase 1 trial versus phase 1 trial without the safeguards, without the monitoring, and without anyone recording the result." Three hazards deserve naming.

Identity and purity are unverified. Chapter 34 covers the analytical findings in this market, and the summary is that label and vial correspond less reliably than buyers assume. For a compound with no human data, you are taking an unknown quantity of an unknown substance — two layers of unknown, not one.

"More synaptogenesis" is not obviously a quantity to maximize. The intuition that more connections must be better follows from nothing. The brain regulates synapse formation and elimination tightly, and both directions are necessary. A potent, sustained, unregulated push on a trophic pathway is a plausible way to cause a problem, and nobody has the human data to say whether it does.

No human data means no known dose-response, so the concept of a reasonable amount does not exist for these compounds. Not "is hard to find" — does not exist.

As always: this is a book, not a clinician. Decisions belong in a conversation with someone who knows your history and can order tests. Chapter 39 is about making that conversation productive.

📊 Evidence Rating

Claim: Dihexa or P21 improves cognition in humans, healthy or impaired. Rating: ❌ Hype outpaces evidence Reason: No completed human trials exist for either compound; the entire evidence base is preclinical, and marketing claims describe human outcomes that have never been measured. What would change it: A completed and published phase 1 study establishing human safety and pharmacokinetics, followed by a randomized controlled trial with a pre-specified cognitive endpoint in a defined population. A single such trial would move either compound to ⚠️ or 🔬 immediately — and the interesting preclinical work is a reason to want that trial run, not a substitute for it. (Rated as of this writing.)

Remember rule 2: the ❌ describes the evidence, not the molecule's potential. Dihexa and P21 may yet turn out to do something in humans. The ❌ says only that today, nobody knows, and the confident claims being made are therefore untethered. If the trials get run, this rating should change, and the point of writing down "what would change it" is so that you notice when it does.

23.6 Noopept — not a peptide, and why the label matters

A short section about a compound that keeps company with the ones above.

Noopept appears in every discussion of this category, is frequently described as a "peptide nootropic," and is not a peptide drug in the sense this book uses. It is a dipeptide-derived compound — a small synthetic molecule built around a modified proline-glycine core with additional groups attached, including an ester. That puts it at roughly 320 daltons, which on Chapter 1's size spectrum (§1.6) is squarely small-molecule territory, nowhere near the 1,000-to-5,000 dalton band where peptide drugs live.

The chemistry is not a technicality. It changes every practical property that matters:

Look at that table and notice the irony. The properties that make Noopept practically interesting are precisely the properties that peptides do not have. It can be swallowed. It is small enough to be a plausible central agent. Calling it a peptide does not merely misclassify it — it attributes to it the exact set of constraints it escaped.

So why does the label persist? Section 1.5 gave the answer in a different context. Calling something a peptide borrows the credibility of insulin and semaglutide — two of the most successful molecules in modern medicine — and connects any compound the word is applied to, however tenuously, to their reputation. That transfer is rhetorical, and it is free.

Structurally, a chemical category has been recruited to do evidentiary work. It cannot do that work. It could not do it for BPC-157 in Chapter 17 and it cannot do it here. A category is not a finding. When a claim leans on what kind of molecule something is, ask what the classification is supposed to license; almost always the answer is nothing.

This book issues no rating on Noopept's efficacy, and the omission is deliberate. It is out of category scope, and rating it here would quietly ratify the misclassification this section exists to correct. The reclassification is not a verdict either way. It means only that the compound belongs in a different file, judged by the standards appropriate to small-molecule central agents — which are the same six questions in a different setting.

⚠️ Hype Check — "it's been clinically used for thirty years"

The claim, in its usual form:

"This isn't some new research chemical. It's been a prescription medicine in Russia for decades, with millions of patient-years of use. That's more real-world evidence than most FDA-approved drugs have."

What's true in it. Long registered use is a real fact and not nothing. It is meaningful evidence about gross safety at customary doses — catastrophic, common harms tend to surface with that much exposure — and it reflects a regulatory review that actually happened. Refusing to acknowledge that is the parochial error of §23.3.

Where it fails. Three places.

First, duration of use is not a measure of efficacy. No amount of uncontrolled clinical experience answers the question a randomized trial answers, because the comparison the trial makes — against what would have happened otherwise — is exactly the comparison clinical experience cannot make. Chapter 6 made this argument about a Western drug class, and it is not weaker applied here.

Second, "millions of patient-years" is a number nobody has. Where prescription and outcome data are not published in a form outsiders can inspect, an exposure figure is an estimate offered by someone with an interest in its size. Ask where it came from; usually it came from the sentence.

Third, notice what the argument is used to skip. It avoids specifying an endpoint. Used for thirty years for what, measured how, compared to what? Its power is its vagueness (§23.1).

Verdict: the claim contains a legitimate safety-adjacent observation and converts it into an efficacy conclusion it cannot support. Accept the first part. Decline the conversion. And do not answer it by disparaging where the use occurred — answer it by asking what was measured.

23.7 Cognitive endpoints are unusually easy to fool

This is the most transferable section in the chapter, and if you retain one thing from Part IV it should be this. Cognitive outcomes are not merely soft. They are actively deceptive in five specific, well-characterized ways, each of which will manufacture an apparent effect from nothing if the design permits it.

Richard Feynman's line from Cargo Cult Science (1974) is the standing warning: "The first principle is that you must not fool yourself — and you are the easiest person to fool." He was talking about experimental physics. It applies harder here, because in physics your apparatus does not want the experiment to work.

1. Practice effects

People get better at cognitive tests by taking them. Not a little better — measurably, reliably better, on nearly every instrument in common use, and the improvement can persist for weeks or months. Participants learn the task structure, develop strategies, grow comfortable with the testing environment, and stop being anxious about being tested.

The consequence is immediate: an uncontrolled before-and-after cognitive study will show improvement, whatever you gave the participants, including nothing. Not sometimes. As the default. A before/after design here does not produce weak evidence; it produces approximately no evidence, because the expected result under the null hypothesis is the same as under the hypothesis being tested.

Partial fixes exist and their limits matter. Alternate or parallel test forms reduce item-specific learning but not learning of the task itself. Practice sessions before baseline push participants past the steepest part of the curve. Some instruments resist better than others. But the only complete fix is a control group that practices exactly as much as the treatment group — which is to say the control group is not a formality here. It is the measurement.

2. Expectancy

Subjective clarity, focus, and drive are among the most expectation-sensitive outcomes anyone measures. Someone who paid for a compound, waited for it to arrive, read about what it does, and took it expecting to notice something will quite often notice something. That is not weakness or dishonesty; it is how attention and self-report work, and it operates on everyone including researchers. It also compounds with practice effects rather than merely adding to them: a motivated participant tries harder on the second test, and trying harder improves scores.

3. Instrument multiplicity

Dozens of validated cognitive tests exist, most produce several sub-scores, and many are administered at multiple timepoints. Multiply those out and a modest-looking study can contain fifty or a hundred comparisons. Under those conditions something will move — not a flaw in the compound or the participants but arithmetic. The resulting paper can be entirely honest, in the sense that every number in it is real, while conveying something false, because the reader is shown the comparisons that moved and not the denominator they were drawn from.

This is the §5.10 red flag, and cognitive research is the easiest place to commit it. The tell is a results section that reads as a list: significant improvements were observed in X and Y, with a trend toward Z. A pre-specified primary endpoint produces a different sentence: the primary endpoint was A; here is what happened to A. One structure has a denominator. The other does not.

4. Baseline dependence

Effects are frequently much larger in impaired, stressed, or sleep-deprived populations than in healthy rested ones. This is among the most consistent patterns in cognitive pharmacology, and the explanation is straightforward: there is more room to move. Restoring a degraded function is a different task from improving an intact one, and there is no reason one should predict the other.

The practical consequences are large and routinely ignored:

• A positive result in patients with cognitive impairment does not support a claim about healthy adults.
• A positive result in sleep-deprived participants does not support a claim about rested ones.
• A null result in healthy young adults does not refute a claim about an impaired population.

All three inferences are made constantly, in both directions, and all three are invalid. The third cuts against skeptics: a well-conducted null in undergraduates is a real finding about undergraduates, and it does not close the question in patients.

A related phenomenon is stranger. For several classes of cognitive agent, effects follow an inverted-U, so individuals starting at a higher baseline may perform worse under a compound that helps lower-baseline individuals. An average effect of zero can conceal two real, opposite effects — a reason to care about variability and not just means.

5. State-dependence

Related to baseline dependence but distinct, and the most overlooked of the five. The same person is a different experimental subject at different times. An effect on a rested, well-fed, motivated participant at ten in the morning may be absent or reversed in that same participant at four in the afternoon on poor sleep and no food. Caffeine status alone moves many cognitive measures more than a study drug does.

This is why a study that does not control time of day, sleep, caffeine, and testing order carries noise that can swamp any real signal — and why individual self-experimentation, in which none of these are controlled and all of them vary, is close to uninterpretable no matter how carefully the records are kept.

WHAT AN OBSERVED BEFORE/AFTER COGNITIVE CHANGE IS ACTUALLY MADE OF

  Observed change  =   practice effect          (always positive, often large)
                   +   expectancy               (positive if you expect benefit)
                   +   regression to the mean   (positive if you started at a low point)
                   +   state differences        (sleep, caffeine, time of day, mood)
                   +   measurement noise        (can be large on a single administration)
                   +   TRUE DRUG EFFECT         (unknown, possibly zero)

  A control group subtracts the first five terms — because the control group
  gets all of them too. That is the entire logic of a control group,
  and in this domain the first five terms are big enough to hide the sixth
  completely.

Read the diagram again with the plus signs in mind. Every term above the last is systematically positive in a typical enhancement study. They do not cancel; they accumulate in one direction. That is why "I tested myself before and after and improved" is not weak evidence of an effect — it is the expected outcome of measuring yourself twice.

And notice the regression term, which matters here because of when people start. Nobody buys a cognitive enhancer during their sharpest week. People start when they feel foggy, tired, and worried about it — at a low point in a naturally fluctuating measure, the precise circumstance under which the next measurement will be higher for purely statistical reasons.

🔍 Check Your Understanding

1. A study gives twenty participants a compound for four weeks and tests them at the start and end on a battery of memory tasks. Scores improve significantly. Which of the six terms in the diagram above can you rule out as the explanation?
2. Why does a positive result in adults with mild cognitive impairment fail to support a claim about healthy thirty-year-olds? Name the specific principle.
3. A trial reports improvements on three of eleven cognitive measures. What single piece of information would tell you whether that is impressive or expected?
4. Explain why an alternate test form fixes part of the practice-effect problem but not all of it.

23.8 The placebo problem in subjective enhancement

Everything in §23.7 gets worse when the endpoint is how you feel you are thinking.

For an objective endpoint — HbA1c, body weight, a wound closed or not — placebo response exists but is bounded by physiology. For a subjective endpoint the report is the measurement, and expectation feeds into it with nothing in between. Placebo responses on subjective central endpoints are large, and in some literatures they have grown over decades as expectations have risen.

So blinding here is not one desirable design feature among several. It is load-bearing. If the blind fails, the study measures expectation.

Why the blind fails

Chapter 5 §5.5 made the general point: a compound with noticeable physical effects is functionally unblinded. Participants notice, form a belief about which arm they are in, and that belief acts on exactly the endpoint being measured. The ways it breaks here are specific and mundane:

• Route sensations. An intranasal preparation produces sting, taste at the back of the throat, drip, and smell that an inert vehicle may or may not reproduce. Participants compare notes with themselves across doses even when they cannot compare with each other.
• Perceptible arousal or calm. Anything that shifts alertness or anxiety announces itself — a particular problem for anxiolytic claims, where the sensation is adjacent to the endpoint.
• Somatic side effects. Headache, flushing, nausea, or disturbed sleep in the active arm and not the placebo arm is an unblinding signal, and because side effects are collected and reported, the information is often sitting in the paper.
• Recruitment context. People who volunteered for a study of a cognitive enhancer arrive with strong expectations about what they should feel, which lowers the threshold for reading any sensation as confirmation.

What good practice looks like

Two design features distinguish serious work here, and their absence is informative.

The blinding-integrity check. At the end of the study, ask every participant which arm they believe they were in, and report the result. If they guess correctly far more often than chance, the blind did not hold and the investigators know it. This costs one question and almost nobody does it. When you see it reported, the study has told you something about its authors independent of its result.

An active comparator. The strongest available design uses a comparator producing similar sensations without the hypothesized mechanism — sometimes called an active placebo. If the active compound stings the nose, the comparator should sting the nose. That is what separates "something is happening" from "the specific proposed thing is happening."

Very few studies in this space do either. That is the honest state of the field, and it applies to Western cognitive-enhancement research at least as much as to anything else in this chapter.

The deeper problem, which better blinding does not solve

For a disease endpoint, the placebo arm is a comparison condition. For a performance endpoint it may be an active intervention in its own right. Believing you have taken something that sharpens focus genuinely increases effort, arousal, and persistence — and those genuinely improve performance on cognitive tasks. The placebo arm is not inert; it is a competing intervention working through motivation.

That raises the bar: to beat placebo on a performance endpoint, a compound must outperform a real psychological effect rather than nothing. And it clarifies what you are buying. If an effect exists only when you know you took the compound, the effect is real in you — you did perform better, and that is not an illusion. But it is not a property of the molecule, and the molecule is what is being sold, at a price, with risk attached.

⚠️ Hype Check — "you can feel it working within twenty minutes"

The claim, in its usual form:

"I don't need a study. I took it and within twenty minutes the fog lifted and I could focus. You can feel this one — it's not subtle."

What's true in it. The person is almost certainly reporting their experience accurately. They did feel different. Immediacy is not automatically suspicious either: plenty of real central agents act within minutes, and dismissing rapid onset as implausible would be wrong.

Where it fails. Rapid, noticeable onset is, if anything, the condition under which a self-report is least informative, for three reasons that stack. It maximizes expectancy — a sensation arriving on schedule confirms the expectation that produced the attention that noticed it. It guarantees functional unblinding in any trial of the same compound, so the literature that would settle the question is harder to produce. And it says nothing about the endpoint: feeling focused and being more accurate are separable, and a person can feel sharper and test worse.

There is also a chronology problem specific to peptides. A trophic mechanism — new BDNF expression, synaptic remodeling — is not a twenty-minute process. If a compound's proposed explanation is structural change in the brain, an effect felt in twenty minutes is evidence against that explanation being the cause of what was felt, not for it. Notice how rarely that is pointed out by people who cite both the mechanism and the immediacy in the same paragraph.

Verdict: the experience is real and the inference is not licensed. "I felt it" is a report about a person, not a finding about a molecule, and the faster and stronger the feeling, the more carefully that distinction has to be held.

23.9 Ratings, and what a convincing trial would look like

The chapter's ratings

Each rating below attaches to a claim, with a population and an endpoint, per rule 1. Each is dated. Each states what would change it, per rule 5.

Three observations, because the table does more work than it looks like.

The two ⚠️ ratings mean different things — one access-limited, one conflicting-evidence. Record only the symbol and you have discarded the distinction that matters most, and you will not know what news would resolve it.

One ❌ is about compounds and one is about a sentence. Rule 1 says ratings attach to claims, and the fourth row is the purest demonstration: no molecule is named, and the claim still fails, because a claim that specifies nothing cannot be supported by anything.

Nothing here is ✅ or 🔬. No peptide in this chapter has multiple adequately powered, consistent, independently replicated human trials behind a specific cognitive claim. And 🔬 is reserved for work proceeding properly through early development; these compounds are mostly either long-registered elsewhere or sold without development at all.

What a convincing trial would look like

Vague skepticism is cheap. If you cannot say what would change your mind, you are not evaluating, you are declining. So, specifically, here is the study that would move any rating in this chapter.

THE TRIAL THAT WOULD SETTLE IT

  PRE-REGISTERED     protocol, primary endpoint, and analysis plan posted in a public
                     registry BEFORE enrollment. This is what makes §5.10 checkable.

  RANDOMIZED         with concealed allocation, so nobody choosing participants knows
                     what the next assignment will be.

  DEFINED POPULATION named in advance and homogeneous enough to interpret: healthy
                     adults 25–40, OR adults with mild cognitive impairment, OR
                     patients 24–72 hours after ischemic stroke. Not "adults."
                     (§23.7: baseline dependence makes this non-negotiable.)

  ONE PRE-SPECIFIED  a SINGLE primary cognitive endpoint on a validated instrument.
  PRIMARY ENDPOINT   Everything else is secondary and labeled exploratory.

  ACTIVE OR WELL-    a comparator producing similar sensations, plus a reported
  MATCHED COMPARATOR blinding-integrity check at the end. (§23.8)

  ADEQUATE DURATION  matched to the proposed mechanism. A trophic mechanism needs
                     weeks; an acute mechanism needs an acute design. Mismatch
                     between mechanism and duration is a design error.

  ADEQUATELY POWERED sample size justified in advance for the primary endpoint and
                     a pre-stated effect size worth detecting.

  REPORTED REGARDLESS results published whether positive, null, or negative — the
  OF OUTCOME          single most important item on this list, and the one most
                      often missing everywhere in the world.

Read that list back as a diagnostic rather than a wish. Every item is checkable in a paper or a registry entry, usually in under ten minutes. Together they are the six questions of §23.3 plus the domain-specific additions §23.7 and §23.8 forced on us: the population clause exists because of baseline dependence, the single-endpoint clause because of instrument multiplicity, the active-comparator clause because of the placebo problem.

And notice that this standard is not a Western standard. Nothing in that box is culturally specific. Every item answers a way that humans — all humans, in every research system — fool themselves. A trial run anywhere that meets it would move these ratings; a trial run anywhere that misses it would not.

📋 Your Evidence Dossier

This chapter fills Field 5 — Human Evidence — for the hardest case: when the evidence exists in another language, another indexing system, or another regulatory culture.

Field 5 is where you record what has actually been tested in people. Until now you have filled it by summarizing trials you could read. This chapter forces the question of what to write when you cannot read them — and the answer is emphatically not a blank field, and equally not "extensively studied" on someone else's authority.

The rule

Never let "I could not read it" silently become either "it does not count" or "it must be good." Record the access limitation as its own line, with a date and a reason.

An access limitation is a fact about your dossier, not about the compound. It belongs in the file, it should be visible, and it should be revisitable — because unlike most of what you write down, it can change without any new science being done. A translation appears. A systematic review searches non-English sources. A registry entry is posted. The unassessable entry becomes assessable, and you will only notice if you flagged it.

Field 5 with an access flag — worked demonstration

FIELD 5 — HUMAN EVIDENCE: SEMAX                        [worked demonstration]

  CLAIM BEING EVIDENCED
    Improves cognitive performance / supports recovery in cerebrovascular disease
    (note: TWO claims, different populations, different endpoints — split them
     before going further, per §23.1)

  REGULATORY RECORD I CAN INSPECT
    Registered medicine in Russia. Not approved in U.S. or E.U.
    Public assessment documentation: not available to me.
    Reason for non-approval elsewhere: no evidence of a filing and rejection —
    most likely no filing. NOT a negative regulatory finding.

  STUDIES I CAN READ IN FULL AND ASSESS
    [list them; for most readers this line is short or empty — write that down]

  STUDIES I KNOW EXIST BUT CANNOT ASSESS
    Reported literature in Russian, some not indexed in the databases I searched.
    Barrier type:  ☑ language   ☑ indexing   ☑ full text unavailable   ☐ paywall
    I have NOT concluded anything about these studies' quality. I have not read them.

  INDEPENDENT REPLICATION OUTSIDE THE ORIGINATING SYSTEM
    None that I could locate.
    Careful reading: consistent with no effect; ALSO consistent with no funder,
    since there is no patent position to motivate one (Ch 19). Not a null result.

  THE SIX QUESTIONS (Ch 5, Ch 6)
    randomized? ? | blinded/held? ? | endpoint pre-specified? ? |
    powered? ? | population defined? ? | reported regardless of outcome? ?

  MY RATING          ⚠️  — ACCESS-LIMITED
                     This ⚠️ describes what I can evaluate, not what was done.
  DATE               [today's date]

  WHAT WOULD CLEAR THE ACCESS FLAG
    - a systematic review that searched non-English sources and reports what it
      found and could not obtain
    - full-text translations of the principal trials
    - registry entries I can inspect
  WHAT WOULD CHANGE THE RATING ITSELF
    - an independently conducted, pre-registered RCT with one pre-specified
      endpoint — in either direction

Do this now

Step 1. Find every Field 5 entry you wrote from something you did not read yourself — a secondary source, a review's characterization, a vendor's citation list. Mark each. This is uncomfortable and it is the point.

Step 2. For any compound with a literature you cannot fully access, rewrite Field 5 in the format above, with the barrier type and the flag stated explicitly.

Step 3. Add one line: "If I am wrong about this, the most likely reason is ___." For an access-limited entry the honest completion is usually there is evidence I have not seen, in a direction I cannot predict. Write it anyway. Chapter 40 will ask whether your access-limited entries drifted in a consistent direction — and if they did, you have learned something about yourself rather than about peptides.

Conclusion

This chapter asked you to do something the previous twenty-two did not: to evaluate evidence you cannot fully reach, without pretending either that it is worthless or that it is sufficient.

The compounds are almost a pretext. Semax and Selank are rationally designed peptides — an ACTH-related fragment and a tuftsin analog, each stabilized with the same Pro-Gly-Pro tail, each registered where it was developed, neither approved in the U.S. or E.U., both usually given intranasally with all the uncertainty Chapter 22 attached to that route. Cerebrolysin is a brain-derived peptide mixture with an accessible, substantial, genuinely unresolved trial literature. Dihexa and P21 have interesting preclinical data and no completed human trials. Noopept is not a peptide, and noticing that is worth more than any opinion about whether it works.

But the durable content is the method.

Apply the same questions everywhere. Design and reporting decide a study's weight. Origin never does. A finding does not weaken because it is inconvenient to obtain, and it does not strengthen because obtaining it took effort.

Record what you could not check. A "?" honestly entered beats a confident answer manufactured to fill a blank. Both failure modes — parochial dismissal and credulous acceptance — are ways of avoiding the discomfort of an unresolved entry.

Distrust cognitive endpoints specifically. Practice effects, expectancy, instrument multiplicity, baseline dependence, and state-dependence stack in one direction and will produce an apparent benefit from an inert substance in any uncontrolled design. When the endpoint is a feeling about your own mind, blinding is the study, and almost nobody checks whether it held.

And notice what the standard in §23.9 is made of. Not one item is a Western convention. Every item counters a way that people — anywhere, in any system, with any degree of integrity — fool themselves. That is why the same box applies to a trial in Moscow, in Vienna, in Boston, and in a forum thread, and why "where did this come from?" is never the question worth asking. "How was it measured, in whom, against what, and was it reported either way?" is.

Chapter 24 turns from cognition to sleep and circadian peptides, where the endpoints are mercifully more objective — and where you will find that "more" is doing a lot of work.

Key Terms

Nootropic — a term coined in the early 1970s for compounds proposed to enhance learning and memory, protect the brain, and improve higher cortical function without conventional psychoactive toxicity. Current usage retains only "improves cognition," and as a claim form it names no endpoint, instrument, or population.

Semax — a synthetic heptapeptide related to a behaviorally active fragment of ACTH, extended with a C-terminal Pro-Gly-Pro tail; a registered medicine in Russia, not approved in the U.S. or E.U., typically given intranasally.

Selank — a synthetic heptapeptide analog of tuftsin carrying the same Pro-Gly-Pro tail; studied and registered in Russia for anxiolytic use, not approved in the U.S. or E.U., typically given intranasally.

Tuftsin — an endogenous tetrapeptide from the heavy chain of immunoglobulin G, best known for immune-related activity; the parent sequence of Selank.

ACTH fragment — a middle portion of adrenocorticotropic hormone, notably the ACTH(4–10) region, studied for behavioral effects apparently separable from the hormone's cortisol-releasing function.

Pro-Gly-Pro extension — the three-residue C-terminal tail on both Semax and Selank; the prolines resist peptidase attack and the glycine supplies flexibility. An economical half-life intervention of the kind cataloged in Chapter 4.

BDNF (brain-derived neurotrophic factor) — a growth factor supporting neuronal survival and synaptic plasticity, often proposed as a mechanism here. A proposed mechanism never justifies an upgrade (rule 3).

Cerebrolysin — a preparation of low-molecular-weight peptides and free amino acids derived from porcine brain tissue, used in several countries for stroke, traumatic brain injury, and dementia; not approved in the United States. A mixture, not a single defined molecule.

Neurotrophic — supporting the growth, survival, or maintenance of neurons; a hypothesis about how an effect might occur, not evidence that it does.

Dihexa — a synthetic angiotensin IV analog reported to have potent effects on synaptogenesis in animal models. No completed human trials.

P21 — a peptide derived from a neurotrophic factor, reported in animal models to affect neurogenesis and markers of neurodegeneration. No completed human trials.

Noopept — a dipeptide-derived compound of roughly 320 daltons: orally active, not protease-cleared in the same way, small enough to be a plausible central agent. A small molecule by every practical criterion, marketed alongside peptides.

Practice effect — the reliable improvement in cognitive test scores from having taken the test before; the reason uncontrolled before/after designs here are close to uninformative.

Expectancy effect — the influence of a participant's beliefs about a treatment on the outcome measured; especially powerful when the outcome is self-reported.

Instrument multiplicity — many tests, sub-scores, and timepoints producing many comparisons; reporting the ones that moved is the §5.10 red flag in its most tempting form.

Baseline dependence — larger effects in impaired, stressed, or sleep-deprived populations than in healthy rested ones, so a result in one population does not transfer to another.

State-dependence — variation in an individual's measured cognition with sleep, time of day, caffeine, food, and mood, which can exceed any drug effect under study.

Active comparator (active placebo) — a control reproducing the sensations of the active treatment without its hypothesized mechanism; the strongest defense against functional unblinding.

Blinding integrity check — asking participants which arm they believe they were in, and reporting the result. Cheap, rare, highly informative.

Pre-specified primary endpoint — the single outcome, named before enrollment, on which a trial stands or falls. It supplies the denominator a list of positive findings lacks.

Gray literature — research outside conventionally indexed journals: reports, theses, conference material. Hard to access is not the same as low quality.

Indexing — inclusion of a journal in the databases readers search; a commercial and editorial decision that affects what you can find and nothing about whether a finding is true.

Publication bias — the tendency for positive results to be published more often, faster, and more visibly than null ones. Documented in every literature examined for it, including Western ones.

Spaced Review

1. (Ch 22 + Ch 23) Semax and Selank are typically given intranasally. Using Chapter 22's account of nose-to-brain delivery, explain why a positive behavioral result after intranasal dosing does not by itself establish that the peptide reached the brain — and why a null does not establish that it failed to.

2. (Ch 5 + Ch 23) A trial reports that participants correctly guessed which arm they were in 84% of the time. Which section of Chapter 5 does this bear on, what has that number demonstrated, and how should it change your reading of the study's primary result?

3. (Ch 6 + Ch 23) Someone argues that a compound registered abroad for thirty years has "more real-world evidence than most approved drugs." Using Chapter 6's evidence hierarchy, identify what the argument gets right, what it gets wrong, and the one question that would move the conversation forward. Do not use the words "Western" or "foreign" in your answer.

4. (Ch 23) Write a Field 5 entry for Selank in five lines: claim, accessible evidence, access barrier, rating with its type, and what would clear the flag. Then state in one sentence what would have to happen for your rating to move without any new study being run.

5. (Ch 5 + Ch 23) A friend says a nootropic peptide "definitely works — same memory test before starting and after four weeks, score up 20%." Using the six-term diagram in §23.7, walk through every term that could account for that change before the compound is reached. Then give the smallest modification to their self-experiment that would make the result more informative — and say honestly how much more informative it would be.