Chapter 19: The Gray Market — Research Peptides, Compounding Pharmacies, and the Risks of Self-Experimentation

"My chemists and I deeply regret the fatal results… I do not feel that there was any responsibility on our part." — Samuel Evans Massengill, on the deaths caused by Elixir Sulfanilamide (1937). His company had dissolved a good drug in a solvent nobody had tested. No law then required anyone to test it.

Overview

This chapter is about people who buy peptides outside the pharmaceutical supply chain. Statistically, some of the people reading this sentence are among them, and this chapter is not going to spend nine thousand words telling them they are foolish.

Here is the position, stated once at the top so that nothing later has to be read defensively: this is not an anti-peptide chapter. It is a pro-knowing-what-you-have chapter. The risk word in everything that follows is unregulated, and it is not a synonym for peptide. Insulin is a peptide. Semaglutide is a peptide. The growth hormone in Chapter 3 that transmitted a fatal neurological disease was entirely natural, extracted from human tissue, and killed people anyway. The category tells you nothing. What is actually in the preparation tells you a great deal.

Chapter 6 put four vials on a table — four identical white powders, one pharmaceutical, one genuine research-grade material, one of unverifiable origin, and one containing something other than its label — and showed that no amount of chemistry knowledge distinguishes them by inspection. This chapter opens them. It asks what can be wrong inside a vial, how those things go wrong, which of them anybody checks, and — the section that matters most — what a clinician adds that no quantity of self-directed reading can supply.

That last part is where most writing on this subject fails. It gestures at "consult your doctor" as a liability formality and never says what the doctor would do. So §19.7 says it, item by item: baselines, compound-specific monitoring, the difference between a dose that stopped working and a receptor that stopped listening, interaction checking, a differential diagnosis instead of an attribution, and somebody reachable at two in the morning.

In this chapter, you will learn to:

• Map the peptide supply landscape and say what each channel is accountable for
• Take apart the "not for human consumption" framing and name what it accomplishes
• Enumerate the six independent quality questions about any vial
• Explain endotoxin, and why sterility testing does not cover it
• Explain why mishandling a peptide changes its risk profile and not just its strength
• State honestly what testing programs have found — and what they cannot establish
• Read the compounded semaglutide episode as a quality story
• Say specifically what medical supervision provides
• Explain why "nobody has reported a problem" is nearly empty information here
• Separate the quality of a preparation from the strength of the evidence for a claim

Learning Paths

All five paths read this chapter in full. It is the safety spine of the book, and the chapter most likely to change what a reader actually does.

💊 GLP-1 — §19.5. The compounded semaglutide episode happened to a drug with excellent evidence, which is the point of §19.10. 🏋️ Performance — §19.3, §19.6, §19.9. The most unregulated material in circulation and the most confident advice about it. The anti-doping discussion is not optional if you compete. 🔬 Science — §19.3 and §19.4 are the analytical core, and set up Chapters 32 and 34. 💄 Cosmetic — lighter relevance, but §19.2's laundering diagram describes the cosmetic peptide market almost perfectly. 🏥 Clinical — §19.7 and §19.8. You will meet patients already doing this, and the most useful skill here is receiving that disclosure without producing the reaction that guarantees you never hear it again.

19.1 How the market is actually structured

People talk about "the peptide market" as though it were one thing with a fuzzy edge. It is at least five distinguishable channels that happen to share molecules, and the useful question about any of them is not is this legitimate? but who released this, against what specification, and what happens if it turns out to be wrong?

Start with a word that does quiet work in pharmaceutical manufacturing: release. A batch of drug substance is not a product when it comes off the line; it is a candidate. It becomes a product when a qualified person, whose name goes in a record, compares the batch's test results against a written specification and signs a decision that it conforms. That signature is the difference between material and medicine — a legal act with a document behind it, creating the thread a recall later pulls on.

FIVE CHANNELS — the molecule may be identical; the accountability is not

  CHANNEL                     WHO RELEASES IT       WHAT RECORDS EXIST    RECALL PATH?
  ────────────────────────────────────────────────────────────────────────────────────
  Licensed manufacturer       a qualified person    full batch record,    yes — lots are
  → pharmacy → patient        against a written     chain of custody,     traceable to
                              specification         stability data        distributors

  503B outsourcing facility   the facility, under   CGMP records,         yes, though
  → clinic or patient         CGMP requirements;    FDA-inspectable       narrower in
                              NOT an approval                             reach

  503A compounding pharmacy   the pharmacist,       the prescription,     limited; runs
  → one named patient         under state board     compounding log       through the
                              rules                                       state board

  "Research use only" vendor  nobody, for human     a certificate that    none exists
  → buyer                     use — by the          may describe a
                              vendor's own terms    different batch

  Peer-to-peer resale         nobody                none                  none

Read the right-hand column first. It changes fastest as you go down, and nobody thinks about it until something is wrong. Two features of this landscape surprise people.

The molecules often come from the same places. Peptide synthesis is a mature industrial process (Chapter 32). A relatively small number of contract manufacturers produce bulk peptide at scale; some supply pharmaceutical customers under quality agreements and audits, some supply research suppliers, and some supply both from the same building under entirely different quality systems. So "it comes from the same factory as the real stuff" is sometimes literally true and almost entirely uninformative — what differs is not where the chemistry happened but what was tested, what was documented, what was rejected, and who signed.

"Legitimate" is not one property. A 503B outsourcing facility operating under current good manufacturing practice, inspected by a regulator, is operating legally and to a real standard — and its product is still not FDA-approved and still has not been reviewed for safety or efficacy. Most confusion here comes from readers who assume legality, quality, and evidence are one axis. They are three, and §19.10 draws the picture.

The middle of the map is hardest. Between the licensed supply chain and the anonymous vendor sits a growing zone of wellness clinics, longevity practices, and telehealth prescribers; some are excellent, some are a supply chain with a stethoscope, and from a website they look identical. You cannot tell the channels apart by how they present, any more than you can tell the vials apart by looking.

🔍 Check Your Understanding

1. What does "release" mean in pharmaceutical manufacturing, and why does it matter more than where a molecule was synthesized?
2. A vendor states, truthfully, that its material comes from a facility that also supplies pharmaceutical customers. What has that told you, and what has it not told you?
3. Why is "legitimate" an unhelpfully compressed word here? Name the three axes it collapses.

19.2 "Not for human consumption": the legal fiction, dissected

Chapter 6 §6.4 introduced the phrase. Here we take it apart.

A large fraction of this market operates under a label stating that the contents are sold for research use only and are not for human consumption. Everyone involved knows what the material is for — the vendor, the buyer, and the forum where the buyer learned the vendor's name. The phrase persists anyway, and what it accomplishes is not nothing and not what people assume.

What the phrase does. It positions the product outside the regulatory definition of a drug. A drug, in most systems, is defined substantially by intended use — by what the seller claims or implies the product is for. Sell a compound while claiming it treats an injury and you have sold an unapproved drug. Sell the same compound as a laboratory reagent, with no therapeutic claim attached, and you have sold a chemical. The molecule is identical; the regulatory object is different, because the regulatory object was never the molecule. It was the claim. The phrase also does liability work — it is the sentence a seller points to afterward, and shaping that argument is most of what it is for.

What the phrase does not do. It does not change what is in the vial, or what happens in a body, and it confers no quality standard — quite the reverse, since research-use material is released against research specifications, which suit dissolving something in a beaker and are not the specifications used for anything injected into a person.

And here is where people invert it. Preserve Chapter 6's distinction between vial B and vial C: genuine research-grade material with an honest certificate of analysis is a real thing, made competently and sold accurately. It is not the same object as a pharmaceutical product, and the difference is not that it is fake. The difference is that nobody ever asked, tested, or certified whether it is fit to put in a human being. The most accurate sentence on that label is the one everybody has agreed to treat as a formality.

Epistemic laundering

No single participant in this market makes a therapeutic claim. Collectively, one is made anyway.

EPISTEMIC LAUNDERING — how a claim gets made that nobody made

  THE VENDOR        sells "a research reagent." Attaches no therapeutic claim.
       │            Defensible: it is a reagent, and they said so.
       ▼
  THE FORUM         members describe personal "experiences." Sell nothing.
       │            Defensible: people may describe their own lives.
       ▼
  THE INFLUENCER    describes "what people are doing," disclaims everything,
       │            treats nobody, prescribes nothing. Defensible: commentary.
       ▼
  THE CLINIC        advertises "optimization" in general terms, naming no
       │            compound. Defensible: a practice describing its services.
       ▼
  THE BUYER         now believes that a specific compound treats a specific
                    condition, at a specific dose, with a specific result.

  Nobody in that column made that claim. The claim IS the column.

That is the diagram to remember. Each element is individually defensible; the assembly is the claim. Nothing has to be a lie for the aggregate to be false. It is a distributed manufacture of confidence in which every participant can honestly say they did not manufacture it — the care is distributed and the confidence accumulates. Chapter 17 gave the BPC-157 case in full: an enormous popular literature, enthusiastic secondhand testimony, and a published human trial record that does not support the claims being made. Nobody wrote the claim. The column wrote it.

⚠️ Hype Check — "the 'research use only' label is just legal cover; the product is the same thing pharmacies sell"

The claim, in its usual form:

"Everyone knows what it's for. The disclaimer is a formality so the vendor doesn't get sued. The peptide inside is the same molecule the pharmacy dispenses — you're just cutting out the markup."

What's true in it. More than people expect. The disclaimer is doing legal rather than chemical work. Peptide synthesis is not exotic, and a competent research supplier can make a genuine, correctly sequenced peptide. Chapter 12's account of drug pricing is not a fantasy, and the resentment behind this claim is often earned.

Where it fails. The claim answers a question about the molecule; the label is a statement about the batch. A pharmaceutical product is a molecule plus a specification, plus testing against it, plus a documented sterile process, plus an endotoxin limit, plus stability data, plus a release decision with a name on it, plus a recall pathway if any of that turns out to have failed. Strip those away and the molecule may survive intact — but you have bought one component of a product and inherited responsibility for the rest. "Cutting out the markup" implies you know which markup you cut. You did not remove profit; you removed testing, documentation, and accountability, which happen to be the expensive parts.

Verdict: right that the disclaimer is legal theater, wrong about what it conceals. It is not concealing that the product is good. It is concealing that nobody checked.

19.3 What can actually be wrong with a vial

This is the technical core of the chapter, and the section to reread.

Six independent things can be wrong with any vial. They are independent in a specific and important sense: each requires a separate analytical method, has a separate acceptance criterion, and can pass or fail without regard to the others. A vial can contain exactly the right peptide at exactly the right amount and still be dangerous. A vial can be sterile and still cause a fever.

SIX INDEPENDENT QUESTIONS ABOUT ONE VIAL
  (an answer to any one of them tells you nothing about the other five)

  1  IDENTITY     Is the peptide in here the peptide on the label?
  2  PURITY       What else came along from the synthesis?
  3  CONTENT      How much actual peptide is present?
  4  STERILITY    Is anything alive in here?
  5  ENDOTOXIN    Was anything ever alive in here?
  6  RESIDUALS    What is left over from making and purifying it?

  Six questions, six methods, six costs. A document answering 1 and 2 has said
  nothing about 3 through 6 — and 1 and 2 are the two that unregulated material
  is most likely to have documentation for, because they are the two research
  customers care about.

Identity. A vial may contain a different peptide entirely, or — more insidiously — a closely related one: a fragment, a truncation, a different analog from the same family, a sequence differing at one or two positions. Chapter 1 already told you why "closely related" is not reassuring. Oxytocin and vasopressin are both nine residues and differ at two of them; one drives uterine contraction and milk ejection, the other water retention and vasoconstriction. Two residues, two physiologies. A molecule that is 90% the same is not 90% as good — it may be inert, or do something else entirely. Identity failures arise from honest mistakes as readily as from substitution, and both produce the same object: a white powder that looks correct.

Purity. Peptides are usually assembled one residue at a time on a solid support (Chapter 32). Each coupling step is efficient but not perfect, and when one fails on some fraction of the growing chains, the synthesis does not stop — those chains keep growing, now permanently missing a residue. The result is a deletion sequence: one residue short, chemically almost identical, similar in mass, behaving almost identically in the chromatography used to purify it. That "almost" is the whole problem. Related substances that closely resemble the product are the hardest impurities to remove and to see, and alongside deletions come truncations, incompletely deprotected species, oxidized residues, and stereochemical variants. A purity number — "98%" — is a claim produced by a particular method with a particular detection limit, and different methods disagree about the same sample. Related substances also matter beyond arithmetic: they are a likely source of an immune response, being close enough to the intended molecule to prime a response that then recognizes it too.

Content and potency. Two traps live here. The first is the difference between purity and peptide content, which almost nobody outside the field knows about. A lyophilized peptide is not pure peptide by weight; it is peptide plus bound water plus counterions left from purification. Purity says what fraction of the peptide-like material is the right molecule; content says what fraction of the total mass is peptide at all. Both are legitimate numbers, they are different numbers, and a product labeled by gross weight contains less peptide than the label implies unless content has been determined.

The second trap is what a content error means physiologically, and this is where Chapter 2 earns its keep. Peptide receptors amplify (§2.4): one bound molecule triggers thousands of downstream events, and concentration relates to effect along a curve with a threshold at one end and saturation at the other, not a straight line. A concentration error is therefore not a proportional error in effect. Being off by a factor of two might be undetectable if both values sit on a flat part of the curve, or the entire difference between nothing and a maximal response if they straddle the steep part. In an unregulated context you know neither the concentration nor where on the curve you are.

Sterility. Not a property material acquires by being dry, cold, or clean-looking. It is produced by a process — terminal sterilization of the sealed container, or aseptic processing under validated procedures — and verified by a specific test. Lyophilized powder is not sterile by virtue of being lyophilized; drying stops most organisms from growing and does not remove them. Contamination can enter during manufacture, filling, transport, storage — and afterward, by the user (§19.6).

Endotoxin. The failure mode most people have never heard of, and the one most worth learning. Endotoxins are lipopolysaccharide fragments of the outer cell wall of Gram-negative bacteria, with three properties that matter enormously. They survive sterilization — they are molecular debris, not organisms, and autoclaving kills bacteria while leaving their cell wall fragments behind, still biologically active. They are not removed by the filtration used to remove microbes, because sterilizing filters retain organisms and endotoxin molecules are orders of magnitude smaller. And injected, they produce a systemic inflammatory response — fever, chills, rigors, headache, malaise, and in serious exposures hypotension and worse.

Injectable pharmaceutical products carry an endotoxin limit, and conformance requires the bacterial endotoxins test: a specific assay, separate from the sterility test, run against a numeric limit derived from route and amount. That test is routinely not performed on research-grade material, because research use does not require it. A certificate showing excellent identity and purity may be entirely honest and entirely silent about endotoxin, and silence is the normal case.

🩺 Safety and Risk — the two questions people think are one question

The single most useful sentence in this chapter:

Sterility asks whether anything is alive in the vial. Endotoxin asks whether anything ever was.

Different tests, different methods, different limits, different consequences — and a product can pass either while failing the other. Sterile material with high endotoxin causes fever and inflammation with no infection to find. Non-sterile material with low endotoxin can cause an infection that develops over days.

Why this matters practically: an endotoxin reaction is among the failure modes least likely to be correctly attributed. It presents hours after exposure as fever, chills, aching, and general awfulness — indistinguishable from a viral illness. In a supervised setting it gets investigated, because someone knows what was administered and when, and "reaction to the preparation" is on the differential. Outside one it gets called a bad night. The person recovers, concludes nothing, and the batch stays in the refrigerator.

This is also why Chapter 1's reassurance — peptides break down into ordinary amino acids — does not cover this territory at all. It is a true statement about the peptide. Endotoxin is not the peptide. It is what was in the room while the peptide was being made.

Residuals. Peptide synthesis uses an inventory of organic solvents, coupling reagents, scavengers, and protecting groups; purification adds more chemistry, which is where the final salt's counterion usually comes from. Purification is designed to remove all of it to defined levels, and pharmaceutical manufacturing verifies the removal by residual-solvent testing against established limits. The gray market inherits the first half of that sentence and not the second. Add whatever was deliberately put in the vial — bulking agents, buffers, stabilizers — and whether those excipients are of a grade appropriate for injection, which is again a separate question.

The seventh problem: aggregation, which is about handling

The six questions concern how a vial was made. There is a seventh concerning what has happened to it since, and it is the one most likely to be created by an ordinary person doing something ordinary. Chapter 4 §4.7 introduced aggregation: under the wrong conditions, peptide molecules stop behaving as individual molecules and clump into higher-order assemblies. Temperature excursions, freeze-thaw cycles, agitation, light, inappropriate solvent chemistry, and time in solution all contribute. Pharmaceutical formulation development is substantially the science of preventing this, which is why approved peptide products come with storage requirements that read like fussiness and are not.

🧬 The Molecule — why aggregation is a risk problem, not a strength problem

The intuitive model of a mishandled vial is that it lost potency: some peptide broke down, so the effect is smaller. Under-dosing — wasteful, not dangerous.

That model is wrong in a specific way. Aggregated peptide is substantially more immunogenic than properly formulated peptide. An immune system that largely ignores a small soluble molecule responds much more vigorously to the same molecule presented as a repetitive, multivalent, particle-like assembly, because that arrangement resembles what immune surveillance evolved to detect. The mishandled vial has not become weaker. It has become a different kind of exposure.

Consequences run from irrelevant to serious. The common outcome is nothing detectable. A less common one is anti-drug antibodies: the body raises antibodies against the administered peptide, which may neutralize it, so the compound gradually stops working — and the natural response is to conclude that tolerance has developed and use more (§19.7 explains why that inference is a trap in more ways than one).

The rare and serious outcome is cross-reactivity. If the administered peptide closely resembles one your own body produces — and many compounds here do, by design — antibodies raised against the injected version can in principle recognize the endogenous one. That is a different category of harm from a drug not working. It is a drug having taught your immune system to interfere with your own physiology, and it does not necessarily resolve when you stop.

This is rare, and it is documented in therapeutic protein and peptide development, which is why formulation and storage are controlled so tightly for approved products. The point here is narrower and firm: mishandling changes the risk profile, not merely the potency — and a vial that spent three days in a warm delivery van looks exactly like one that did not.

The document that is supposed to settle all this

Certificates of analysis get a full chapter (Chapter 34), so this is only the frame you need here. A certificate of analysis is a report of tests performed on a sample — that is its logical form, and everything that goes wrong with it follows from that form. It may describe a different batch. It may have been produced by the supplier's supplier, describing bulk material subsequently divided, transported, stored, and relabeled by parties who ran no tests at all. It may be a real document for a real sample with no relationship to your vial. It may be a picture of a document.

A certificate is a claim about a sample. It is not a property of the vial in your hand. Hold that sentence; §19.4 builds a rating on it.

19.4 What testing programs have actually found

Independent parties — academic groups, journalists, regulators, testing services — have obtained samples of unregulated peptide products and analyzed them. It is worth being precise about what that literature supports, because it is cited in both directions and usually with invented numbers.

What has been reported, by category. Analyses of products from unregulated channels have found:

• Products containing no detectable amount of the labeled peptide.
• Products containing a different peptide than the one labeled, including related analogs and fragments.
• Substantial deviation between labeled and measured content, in both directions.
• Related impurities and unidentified peptide-like species consistent with incomplete purification.
• Unlabeled additional substances present in the preparation.
• Microbiological contamination, and, where it was measured at all, endotoxin above the limits that would apply to an injectable pharmaceutical product.

Separately, regulators have taken visible action: warning letters, import alerts, and public communications about unapproved peptide products — and, in §19.5's compounded semaglutide episode, statements describing adverse event reports the agency had received.

What that literature does not establish, which matters more than the list above. You will see confident percentages quoted for how many gray-market products "fail testing." Treat every one of them skeptically, including any you might derive from the list above, because:

• The samples are not random. Products get tested because somebody was suspicious, or because they were easy to obtain. Convenience samples cannot produce a population rate.
• The test panels differ. One survey checks identity and purity, another content; very few check sterility and endotoxin, the tests most likely to reveal the most dangerous failures. A "pass rate" across surveys with different panels is not a coherent quantity.
• There is no denominator, and results are batch- and time-specific. Analysis characterizes the samples analyzed, and nobody knows how many products are in circulation.

The honest summary — and this book would rather be honest than quotable — is: failures occur in every one of the six categories in §19.3, they are not rare enough to treat as freak events, and the rate is not knowable from available data. Anyone who gives you a number for that rate has either invented it or borrowed it from someone who did.

🔬 Read the Study — what an analytical survey of unregulated products can and cannot conclude

The design. Investigators obtain products through ordinary channels, blind and code the samples, submit them for analysis — mass spectrometry for identity, chromatography for purity and content, sometimes microbiological and endotoxin testing — and report per-sample results against the label.

What it is genuinely good at. Existence proofs. If a sample labeled as one peptide contains a different one, that is a finding requiring no statistics at all. The category list above is well supported in exactly that sense: these failures happen.

What it cannot do. Estimate prevalence. Generalize to a supplier, a country of origin, or a compound class. Or certify that untested samples from the same source are fine — the inference readers most want, and the one the design most clearly forbids.

The limitation nobody mentions. A survey reports what it measured. When one checks identity, purity, and content and finds most samples acceptable, the reasonable reading is "most samples contained approximately the right peptide," not "most samples were suitable for injection." Absence of a failure you did not test for is not a pass.

What would make it much stronger. Randomized sampling from a defined population; a pre-registered full panel including sterility and endotoxin; published methods and detection limits; repeated sampling over time. None of that exists at scale, and its absence is §19.8.

📊 Evidence Rating

Claim: "This product is third-party tested, which means it contains what the label says." Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Why: A certificate is a report on a sample that may or may not correspond to the vial in hand, testing scope varies enormously, and identity, purity, content, sterility, endotoxin, and residuals are six separate tests of which most programs run two or three. What would change it: Per-lot testing tied to the container by a verifiable identifier, by an independent laboratory whose methods and detection limits are published, with a full panel including sterility and endotoxin, and a documented chain of custody linking the tested sample to the distributed lot. That is approximately a description of pharmaceutical batch release — which is the point.

19.5 Compounding pharmacies: legitimate, marginal, and neither

Compounding is old, real, and necessary, and it deserves better than the reputation it has acquired by association. A compounding pharmacy prepares a medication for a patient whose need cannot be met by a commercial product: a child who cannot swallow a tablet, a patient allergic to a dye, a strength no manufacturer produces. That is a genuine clinical service, and it is not what this section warns about.

The United States framework, which Chapter 38 covers properly:

503A pharmacies compound for an identified individual patient pursuant to a valid prescription. They are regulated primarily by state boards of pharmacy, follow professional practice standards, and are exempt from certain federal requirements — including premarket approval and, significantly, current good manufacturing practice.

503B outsourcing facilities are a category created after a 2012 fungal meningitis outbreak traced to contaminated compounded injections, an event that killed dozens of people. A 503B registers with the FDA, must comply with CGMP, is subject to FDA inspection, and may produce batches without patient-specific prescriptions. A substantially higher standard than 503A.

And now the sentence neither category escapes: compounded products are not FDA-approved. They are not reviewed by any regulator for safety or efficacy before they are dispensed. Not the 503A preparation, not the 503B batch. Regulatory attention in compounding is directed at how the product is made — facility, process, contamination control — and not at whether it works or is safe for the purpose intended. A great deal of confusion here comes from patients assuming that a legally compounded product carries an efficacy judgment. It does not. It never did.

The compounded semaglutide episode, read as a quality story

Chapter 12 owns this episode's access-and-cost dimension and Chapter 6 owns the hype mechanism. Here it belongs because it is the clearest demonstration available of what compounding does and does not guarantee.

When a drug is in official shortage, restrictions that ordinarily prevent compounders from producing copies of commercially available products loosen. Semaglutide went into shortage, and a very large compounding market appeared spanning the full range — from 503B facilities operating under CGMP with documented sourcing to arrangements whose provenance was not determinable at all. When the shortage resolved, the permission narrowed again, with a great deal of product and a great many patients mid-course. Two quality concerns from that episode generalize far beyond semaglutide.

Salt forms. Regulators raised concerns about products containing semaglutide sodium and semaglutide acetate. These are not semaglutide base. A salt form is a different chemical entity with different properties, and the safety and efficacy data supporting the approved product were generated with the base. No principle transfers evidence automatically to a salt form; the transfer has to be demonstrated, and it had not been. Precise, technical, unglamorous — and exactly the kind of distinction that separates a product with data behind it from one that resembles it.

Delivery device substitution. The approved product is supplied in a pre-filled pen delivering fixed increments; a compounded version is typically supplied as a vial and a syringe. That substitution silently transfers the measurement task from an engineered device to the patient, and regulators reported receiving adverse event reports involving dosing errors, including errors of large magnitude. Note the structure: no molecule changed and no vendor lied. The failure mode was introduced by the format.

Generalize both. A compounded product is not the approved product with a different label. It may differ in substance, concentration, excipients, stability, storage, and administration — and each difference is a place where the approved product's evidence may fail to transfer.

💊 In the Clinic — what "compounded" means from the patient's chair

It can be entirely appropriate. A specific allergy, a formulation that does not exist commercially, a genuine shortage — good reasons, and a good clinician will name one.

It is not FDA-approved, and that is not a technicality. No regulator evaluated this preparation for whether it works or whether it is safe. Your clinician's judgment is doing that work. That may be entirely sufficient; it is worth knowing that it is what is happening.

The manufacturing standard varies by category. A 503B operates under CGMP and FDA inspection; a 503A under state board rules. Both can be excellent. They are not the same standard, and a patient is entitled to ask which one made their medicine.

A legal supply chain is not an evidentiary claim. The most transferable sentence in the section. That a clinic prescribes legally and a licensed pharmacy compounds legally establishes that the supply was lawful, and nothing about whether the compound treats the condition. Chapter 5's rating system applies to the claim regardless of how impeccable the paperwork is.

The questions worth asking are unglamorous: what is the substance and its exact form, who compounded it and under which category, what monitoring is planned, and what would cause you to stop.

19.6 The preparation problem, described without instruction

Most peptides are supplied as a lyophilized powder — freeze-dried, because a dry solid is far more stable than a solution. Before anything can be administered, that powder has to be dissolved.

In pharmaceutical manufacturing, if that step exists at all, it happens inside a controlled environment: classified air, validated procedures, qualified diluent, trained personnel, documented process, and a test at the end. In the gray market the entire step is transferred to a person at a kitchen table with no environmental control, no qualified diluent, no training, no documentation, and — the part that matters most — no way to detect a failure.

This book will not describe how any of it is done. That is not squeamishness: a description precise enough to be useful would be precise enough to be followed, and this chapter's whole argument is that following it is the problem. What follows is a taxonomy of what goes wrong, written so you can recognize the shape of the risk without acquiring a method.

Contamination introduced at the point of preparation. Every transfer between containers is an opportunity for organisms and particulates to enter something that will be injected. The pharmaceutical industry spends extraordinary sums engineering those opportunities out of existence, and the reason is the 2012 meningitis outbreak and the long list of similar events before it.

Diluent composition and stability in solution. Peptides are not indifferent to what they are dissolved in: pH, ionic strength, tonicity, and the presence or absence of a preservative all affect whether a peptide stays soluble, intact, and monomeric, and different peptides have different answers. In solution a peptide is also far less stable than as a dry solid. Approved products carry in-use periods set by actual stability studies; outside that framework the answers come from assumption, and there is no observable difference between a solution that is fine and one that is not. Relatedly, a container designed to be entered once and one formulated with a preservative to survive repeated entry are different products, and using one as the other is a category error with microbiological consequences.

The arithmetic problem. Converting a labeled mass into a delivered amount requires two numbers: how much peptide is present, and what volume it is in. §19.3 established that the first is uncertain here — and when an input is uncertain, arithmetic to three decimal places produces false precision, not accuracy. People trust the calculation because it is the part they can verify.

Running through all of it: none of these failures is visible. A contaminated solution looks like a clean one. An aggregated solution can look perfectly clear. Endotoxin has no appearance at all.

🩺 Safety and Risk — the injection-site problem, and why it is a clinical question

Introducing a non-sterile preparation into subcutaneous tissue produces a range of outcomes: localized redness and swelling; cellulitis; abscess formation requiring drainage; and, less commonly but seriously, spread beyond the local site.

The genuinely difficult part is that a sterile inflammatory reaction and an early infection can look similar at the beginning and require opposite responses. Not every red, sore site is infected — some are reactions to the preparation, to an excipient, or to aggregated material. Some are infected and will worsen without treatment. Distinguishing them involves time course, systemic signs, examination, and sometimes imaging or culture. It is a clinical judgment, not reliably made by the affected person, and definitively not made by photographs posted for strangers.

The pattern worth internalizing: the failure modes of unregulated preparation converge on presentations that require a clinician to sort out, arriving in a person who has structural reasons not to consult one. That combination turns a manageable event into a serious one, and it is precisely what §19.7 is about.

If something following an injection looks like a spreading infection, or comes with fever, it is a medical problem now and not an internet problem — and the thing to say to the clinician is what was actually taken. They have seen worse, and they cannot help without it.

19.7 What medical supervision actually adds

Most safety writing on this subject ends with "consult a healthcare professional," which functions as a disclaimer rather than an argument and persuades nobody. So here is the argument, made without a single moral claim: seven specific things a clinician does that no amount of reading, testing, or community advice supplies.

1. A baseline. Before-values, taken before anything starts, appropriate to the specific compound. Baselines matter because they are irreproducible after the fact. Six months in, an abnormal result is ambiguous — was this caused by the compound, or has it always been like this? With a baseline that question has an answer; without one it has an argument. Baselines are the cheapest, most valuable, and most commonly skipped element of the whole enterprise.

2. Monitoring matched to the specific compound. Different compounds put different parameters at risk, and the relevant list is not obvious from the marketing. Growth-hormone-axis compounds raise questions about glucose handling and insulin sensitivity, and produce a biochemical readout far more informative than how someone reports feeling. GLP-1 receptor agonists raise questions about gastrointestinal tolerance, hydration, specific red-flag presentations, and — with sustained weight loss — nutritional adequacy and lean mass. The point is that a compound-specific list exists, that it differs by compound, and that somebody has to know which one applies — including which findings are harmless and which mean stop today.

3. Dose adjustment based on response and on data. Response is not the same as feeling response. Clinical adjustment uses measured parameters, reported symptoms, and time, and it moves in both directions.

This is where Chapter 2 §2.8's trap lives, and it is the most valuable item in this section. When a compound's effect fades, the untrained interpretation is the dose has become insufficient, and the untrained response is to increase it. But receptor systems desensitize: sustained stimulation causes receptors to be internalized and downregulated, and a fading effect may be the system withdrawing from the conversation rather than the signal becoming too quiet. If so, increasing the dose is exactly wrong — it deepens the desensitization while raising exposure and everything that scales with it. The intuitive response to "it stopped working" is the opposite of the correct one, and telling the two situations apart requires knowledge that feeling does not provide.

4. Interaction checking against everything else — not just prescriptions, but over-the-counter medicines, supplements, other peptides, alcohol. Peptides interact less often than small molecules through the liver enzyme pathways people think of first, and constantly through physiology: anything slowing gastric emptying changes how oral medications taken alongside it are absorbed; anything lowering glucose stacks with other glucose-lowering agents. There is an epistemic version too — someone using several compounds at once has an attribution problem that cannot be solved retrospectively.

5. A differential diagnosis instead of an attribution. This is the reason the section exists.

When a new symptom appears, the untrained move is attribution: deciding whether it is "from the peptide" on the basis of prior belief. Enthusiasts attribute away; skeptics attribute toward. Both errors are dangerous, in opposite directions. Attribute a serious symptom to the compound and you may stop it, feel reassured, and leave an unrelated condition untreated. Attribute it away and you may continue an exposure that is causing harm.

A clinician instead builds a list of possible causes, ranks them by likelihood and by danger, and investigates. The compound is one hypothesis on the list, alongside everything that would have happened to a person of that age and history anyway. Concretely: new upper-abdominal pain in someone taking a GLP-1 receptor agonist could be an expected gastrointestinal effect, or gallstones — genuinely more common during rapid weight loss — or pancreatitis. Those require different responses, and no amount of reading distinguishes them from the inside.

6. Someone to call at two in the morning. A forum is not on call. A vendor's support inbox does not practice medicine. A search engine does not know your history. And the failure mode is not silence — it is that somebody answers immediately, confidently, for free, and wrongly.

7. Someone with the standing to say stop. A clinician can say this is not working, or this is harming you, and hold that position against a patient's investment in the decision. That is structurally unavailable elsewhere. A community organized around a practice has great difficulty producing the verdict that the practice should end — the members most likely to have reached it have already left, so the population you are consulting is filtered in precisely the direction that biases the answer.

The honest counter-argument

Not all supervision is real supervision. Some clinics take no meaningful history, order no baselines, monitor nothing, decline nothing, and function as a supply chain with a stethoscope attached. A reader told to "get a doctor" who already has one of those is entitled to say the advice has not helped. So the useful version is not find a prescriber. It is: notice whether the arrangement you have actually delivers the seven items above. Is a full history taken? Are there baseline labs, repeated? A monitoring plan specific to this compound? Has this practitioner ever declined a request? Is there a record you could hand to another clinician, a way to reach someone urgently, a stated criterion for stopping? A practitioner who has never said no is not exercising judgment, and a relationship with no exit criterion is not a treatment plan.

💊 In the Clinic — the two-in-the-morning test

The whole section reduced to one scenario. A person wakes at two in the morning with severe abdominal pain, twelve days into using something.

Supervised. They call the number they were given. Someone who knows what they are taking, when they started, what their baseline labs showed, and what else they are on decides whether this is an emergency department problem. If it is, they arrive with a history, and the clinician there starts from a real list instead of a mystery. The compound is on the record, so it is on the differential — ruled in or out rather than assumed either way.

Unsupervised. They search their symptoms and find a thread. Several strangers say this is normal and passes; one says stop immediately. Nobody knows their history, and the reassuring answer is more comfortable than the alarming one. If they do reach an emergency department, they may not mention what they have been taking, because they expect to be judged — and the clinician works a differential with a piece missing.

That last sentence is the one clinicians should sit with. The fear of being judged is itself a safety risk, and it is one the profession partly creates. A patient who expects a lecture withholds information, and withheld information turns a diagnosable event into a difficult one.

One boundary this chapter will not blur: nothing here suggests that testing a product, monitoring yourself, or reading carefully can substitute for clinical supervision. Analysis tells you what is in a vial. It cannot tell you what it will do in you, whether it should be continued, or what a new symptom means.

19.8 Adverse events with nowhere to go

Approved medicines exist inside a surveillance system. It is imperfect, underused, and slow, and still one of the most important safety structures in medicine: mandatory reporting by manufacturers; spontaneous reporting systems that accept reports from clinicians and patients (FAERS in the United States, the Yellow Card scheme in the United Kingdom, EudraVigilance in the European Union); statistical signal detection; the authority to change a label, restrict a use, or withdraw a product; and a recall system with the records to execute it.

The gray market has none of this. Not a weakened version — none. No mandatory reporting, because there is no manufacturer of record with a regulatory obligation. No spontaneous reporting, because the product does not officially exist as a medicine. No signal detection, because there are no accumulated reports. And no recall pathway — which deserves its own sentence: a recall requires knowing who has the product. Pharmaceutical distribution maintains lot-level records precisely so that when a batch fails it can be found; an anonymous transaction leaves no thread to pull. If a batch of gray-market material were found contaminated tomorrow, there would be no mechanism to tell the people holding it.

Why this makes "no reported problems" nearly empty

Chapter 10's second case study established an important asymmetry: observational data is strong for detecting harms and weak for establishing benefits. Rare, serious, temporally clustered events show up in large observational datasets even when trials were far too small to see them; benefits are the opposite, because confounding by indication and healthy-user effects make observational benefit estimates unreliable.

Now notice the missing precondition. That strength is entirely contingent on a collection system existing. Observational data does not accumulate by itself; it accumulates because clinicians file reports, databases ingest them, and someone runs the analyses. Remove the collection and you do not get weak evidence of safety. You get no evidence at all — and it looks exactly like a clean safety record. Absence of reported harm, in a system with no reporting pathway, is close to uninformative.

There is a second reason, and it is worse. The failure modes that matter most here are the ones least likely to generate an attributable event even if somebody were collecting:

• An endotoxin reaction presents as fever and malaise hours later. It looks like a virus.
• Wrong potency presents as "it didn't work for me," read as a non-response, not a defect.
• Immunogenicity presents as gradually diminishing effect, read as tolerance and answered with escalation (§19.7).
• Long-latency harm presents years later, in a different city, to a different clinician, with nothing connecting it to anything.

Attribution requires a hypothesis, and a hypothesis requires a record. Nobody diagnoses a cause they have no reason to consider.

The standing precedent

Chapter 3 told this story, and it is the most important historical case in the book for this section. Before recombinant production, human growth hormone was extracted from the pituitary glands of cadavers and given to children with growth hormone deficiency through national programs, over decades. Some batches were contaminated with the infectious agent that causes Creutzfeldt-Jakob disease, and recipients developed a fatal, untreatable neurodegenerative disease years to decades after treatment ended.

Now the part belonging to this chapter. That harm was detected only because the programs were documented. There were records: who received hormone, from which batch, when, and what happened afterward. The pattern was visible because the data existed to make it visible — and even then it took years, in a supervised, institutional program with named physicians and national registries.

In an undocumented market that harm would never have been attributed to anything. Rare neurological disease, decades later, scattered across geography and time, in people who had long since stopped seeing the clinician involved. No cluster, because no denominator, no registry, and no question anyone thought to ask — tragedies with no visible common cause. That case also involves a completely natural, human-derived, bioidentical substance, which is where §19.10 is going.

What a person can actually do

Report anyway. National adverse event systems accept reports from consumers, not only clinicians — FDA MedWatch and the UK's Yellow Card scheme both do. A report describing what was taken, from what kind of source, and what happened is a data point, and such points do occasionally aggregate into regulator action.

Tell your clinician the truth. Not because a book says to be virtuous, but because §19.7's entire argument collapses without it. A clinician working from a fictional history is one whose differential diagnosis is missing an entry. Chapter 39 is devoted to making that conversation go better.

📊 Evidence Rating

Claim: "Nobody has reported any problems with this source, so the product is safe." Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Why: No reporting pathway exists for unapproved products, so the absence of reports reflects the absence of collection rather than the absence of harm — and the failure modes that matter most here (endotoxin, wrong content, immunogenicity, long-latency harm) do not produce events anyone would attribute to a product even if a pathway existed. What would change it: A functioning surveillance system covering these products — routine lot-level identification, a channel for reporting events against a specific lot, enough reporting volume for signal detection, and published analyses. Short of that, a long-running prospective cohort of users with systematic follow-up would provide real information. Neither exists.

19.9 Legality: a structure, not an answer

This book cannot tell you whether a given peptide is legal where you live. It varies by country, by state or province, by compound, and over time, and any specific answer printed here would be wrong somewhere and out of date eventually. What is durable is the structure of the question, which most people get wrong by assuming there is one question.

SIX QUESTIONS THAT PEOPLE COMPRESS INTO ONE

  1  APPROVAL     Is this substance an approved medicine in this jurisdiction,
                  for any indication?
  2  SALE         May it be sold here, by whom, and in what form? (Selling a
                  substance and selling it *as a medicine* are different acts.)
  3  IMPORT       May it be brought in? Is there a personal-use exemption, and
                  what are its limits? Customs and health regulators may answer
                  differently from each other.
  4  POSSESSION   May an individual lawfully hold it? Some substances are
                  controlled; most peptides are not, but some are.
  5  PRESCRIBING  May a licensed practitioner prescribe it off-label, or arrange
                  compounding of it? Often the widest door in the structure.
  6  SPORT        Is it prohibited in competition? An entirely separate regime
                  that does not care what the other five say.

  A substance can be legal to possess, unlawful to sell as a medicine, importable
  in one country and seizable in the next, legal to prescribe, and banned in
  sport — all at once.

Three features cause most of the real-world confusion.

The questions are answered by different authorities. A national medicines regulator, a customs agency, a state pharmacy board, a professional licensing body, and an anti-doping organization each answer a different question, and none is obliged to agree with the others. "It's legal" usually means "one of these six answered yes," and the speaker rarely knows which.

Sport is a separate universe, and stricter than people expect. The World Anti-Doping Code's Prohibited List includes growth hormone, its releasing factors and secretagogues, and a range of peptide hormones and growth factors. It also contains a category — non-approved substances — that prohibits any pharmacological substance not currently approved by any governmental regulatory health authority for human therapeutic use. Read that carefully: it captures most research peptides by definition, without naming any of them. Anti-doping also operates on strict liability — the athlete is responsible for what is found in their body, regardless of intent or of what a label said. Combine that with §19.3's identity and purity failures and the arithmetic is stark: an unverified vial is a career risk existing independently of whether it is a health risk.

Statuses change. Substances get scheduled, import alerts get issued, shortage lists open and close — which is what drove §19.5's episode — and enforcement priorities shift. A forum answer from several years ago is not evidence about what is permitted now.

Verify locally, from a primary source, recently. Chapter 38 does the regulatory framework properly.

🔍 Check Your Understanding

1. A compound is not listed by name on the anti-doping Prohibited List. Why might it nonetheless be prohibited in competition?
2. Someone says "it's legal, my doctor prescribes it." Which of the six questions has that answered, and which four has it left open?
3. Why does strict liability interact so badly with the identity and purity failures in §19.3?

19.10 This chapter is not anti-peptide

Now the position, stated plainly, because it has been implicit for nine sections.

The risk word in this chapter is unregulated. It is not peptide.

Insulin is a peptide, millions inject it every day, and it is among the great medical achievements of the twentieth century. Semaglutide is a peptide with a trial program of extraordinary size behind it. Oxytocin is a peptide used in obstetric practice worldwide. A reader who finishes this chapter believing peptides are inherently dangerous has learned the wrong lesson.

Run the argument the other way, because it is more instructive. Chapter 3's cadaver growth hormone was entirely natural — human in origin, bioidentical in the most literal sense available, administered by physicians in national programs to children who genuinely needed it. And it transmitted a fatal, untreatable disease, because it carried something nobody knew to test for. Natural did not help. Bioidentical did not help. Good intentions did not help. What was missing was a test, and before that the knowledge that there was something to test for.

So here is the claim this chapter actually makes: what predicts the safety of a preparation is what is in it, how it was made, what could have contaminated it, and whether anyone checked. Not whether the molecule is synthetic or natural. Not whether the underlying compound is promising.

The two axes, and why arguments about this go nowhere

Most disagreements about peptide safety are two people using one word for two different things. The word is "safe," doing the work of two independent variables.

TWO AXES, NOT ONE
                              THE PREPARATION
                    known-good              unverified
                 ┌──────────────────────┬──────────────────────┐
  strong         │ approved medicine,   │ THE EVIDENCE DOES    │
  evidence       │ correctly dispensed. │ NOT TRAVEL WITH THE  │
  for the        │ The only quadrant    │ MOLECULE — trial data│
  CLAIM          │ where "safe and      │ describes the tested │
                 │ effective" is a      │ product, not         │
                 │ supported sentence.  │ whatever this is.    │
                 ├──────────────────────┼──────────────────────┤
  weak or        │ a well-made vial of  │ two independent      │
  absent         │ something unproven:  │ unknowns — and they  │
  evidence       │ wasted time, unknown │ do not add. They     │
                 │ benefit.             │ multiply.            │
                 └──────────────────────┴──────────────────────┘

  Most arguments about peptide safety are two people standing in different
  quadrants, using the same word, and both being locally correct.

The upper-right quadrant is where a great deal of GLP-1 discussion currently lives. Semaglutide's evidence is genuinely excellent (Chapter 8). That evidence was generated using a specific manufactured product, at specific doses, delivered by a specific device, in specific populations. The evidence attaches to that product, not to the molecular formula. A different substance, at an uncertain content, in a different format, is not covered by it — and saying so is not skepticism about semaglutide. It is the opposite: taking the evidence seriously enough to notice what it was about.

The lower-left is the quadrant people forget exists. A carefully made, correctly identified, sterile, endotoxin-tested vial of a compound with no human efficacy data is not dangerous in the way this chapter has described. It is unproven, which costs money, time, and opportunity — a different failure from the one in the right-hand column, and the axis Chapter 17's assessment of BPC-157 lives on.

⚠️ Hype Check — "regulation is about protecting profits, not people"

The claim: "The only reason these aren't approved is that nobody can patent them. The warnings are about market share, not safety."

What's true in it. A surprising amount, and pretending otherwise costs credibility. Development economics shape which compounds get studied, and unpatentable ones get studied less. Drug pricing in some markets is difficult to defend on any principled basis. The commercial interest in restricting compounded competition during and after a shortage was real. Chapter 12 covers those access failures without flinching, and they are failures.

Where it fails. The critique is about the system and is being used to license a conclusion about a vial. No company's profit is protected by a vial containing the wrong peptide, or by endotoxin, or by a content assay nobody ran. The six failures in §19.3 are physical properties of a container, unaffected by whatever you conclude about the industry. There is also a historical answer: the apparatus this claim treats as a commercial construction was built out of specific bodies — the 1937 elixir, thalidomide, the 2012 meningitis outbreak, and Chapter 3's cadaver growth hormone programs.

Verdict: the critique is frequently fair, and it transfers to policy arguments, not to the contents of a container. Both are true: the access failures are real, and the quality risks are real. Chapter 12 is about the first; this chapter is about the second.

📋 Your Evidence Dossier

This chapter fills Field 10's quality-risk line.

Field 10 asks what the practical risk is of obtaining and using a compound the way it is actually obtained and used. Most people, asked to write that line, produce a general warning: "gray-market products can be contaminated or mislabeled." True, useless, and identical for every compound — which is how you know it is doing no work. The version that works is specific, and it is a list of what you cannot answer about a particular preparation, not a list of general hazards.

FIELD 10 — QUALITY AND PREPARATION RISK

  Channel              approved product / 503B / 503A / research-use / unknown
  Identity             per-lot evidence this is the labeled peptide?
  Purity               per-lot evidence, by what method, at what limit?
  Content              is peptide content (not just purity) established for THIS lot?
  Sterility            made under sterile process, and tested?
  Endotoxin            tested? (assume not unless documented)
  Residuals            solvents and reagents tested against limits?
  Handling history     what happened to it between manufacture and you?
  Recall path          if this lot were found defective, how would you learn?
  Unanswered           which of the above you genuinely cannot answer

  Then one sentence: what would have to be true for those to become answerable?

A note on what this is not. It is not a checklist for evaluating a purchase, and not a list of questions to send anyone. In an unregulated channel most of these have no obtainable answer, and that unavailability is the finding. This is an inventory of your own uncertainty, not a buying guide.

Worked demonstration — two entries, deliberately opposite

FIELD 10 — SEMAGLUTIDE (prescribed, dispensed by a licensed pharmacy)
  Channel            approved product, licensed supply chain
  Identity/Purity/   all yes — released against specification, per-lot, by
    Content            registered methods; the device controls the amount given
  Sterility          yes — sterile process, tested, documented
  Endotoxin          yes — tested against a limit set for this route
  Residuals          yes — tested against established limits
  Handling history   cold chain maintained and documented to the pharmacy
  Recall path        yes — lot traceable to distributor; I would be contactable
  Unanswered         essentially none at the product level
  What would change  n/a. NOTE: none of this addresses whether the drug is right
                     for me — that is Field 6 (evidence) and Field 11 (monitoring).

FIELD 10 — BPC-157 (research-use channel)
  Channel            research-use vendor; provenance not determinable
  Identity           a certificate exists; it describes a sample, and I cannot
                     establish that the sample corresponds to this container
  Purity             a number is stated; method and limit of detection are not;
                     related substances not addressed
  Content            NOT stated. Purity is not peptide content. I do not know what
                     fraction of the mass in the vial is peptide
  Sterility          not stated, not claimed, and by the product's own stated
                     terms not required
  Endotoxin          not stated. Assume untested. A separate test from sterility
  Residuals          not stated. Purification is supposed to remove them; no
                     verification exists
  Handling history   unknown transport temperature, unknown storage duration,
                     unknown excursions. Aggregation state unknown
  Recall path        none. If this lot were defective nobody could tell me
  Unanswered         content, sterility, endotoxin, residuals, handling, recall —
                     six of nine, and they include every one that matters after
                     the vial is opened
  What would change  per-lot testing tied to this container by a verifiable
                     identifier, full panel, independent lab, published methods.
                     I have no way to obtain that.

Notice what the second entry is and is not. It is not a verdict that the vial is bad — it might be excellent material. What it establishes is that six of nine questions have no available answer, and that the unanswered ones are disproportionately those whose failure modes are invisible. That is what the word unregulated costs, made specific.

For each peptide in your dossier, write this line as it stands today, and date it. Then add one sentence naming which single unanswered question would most change your thinking if it were answered. For most readers it is content or endotoxin, and noticing that is worth more than the rest of the exercise.

Conclusion

The gray market is not one thing. It runs from outsourcing facilities under real manufacturing requirements to anonymous transactions with no records at all, and the molecules moving through it are sometimes identical. What differs is not chemistry. It is who tested what, against which specification, with whose name on the decision, and what happens if that turns out to have been wrong.

Six independent things can be wrong with a vial: identity, purity, content, sterility, endotoxin, and residuals. Documentation accompanying unregulated material typically addresses one or two, and the two most dangerous are the ones least often addressed, because research use does not require them. Sterility asks whether anything is alive in there. Endotoxin asks whether anything ever was. A seventh problem, aggregation, arrives after manufacture and changes the kind of risk rather than its magnitude.

Testing programs have found real failures in every one of those categories; they have not established a rate, and nobody who quotes you one has either. Compounding is legitimate, necessary, and not an approval — the semaglutide episode showed how much range a legal supply chain can contain. And no surveillance system watches any of it, which is why "nobody has reported a problem" is nearly empty here, and why Chapter 3's cadaver growth hormone remains the standing precedent.

But the conclusion is not that peptides are dangerous. It is that unverified preparations are, and that the two axes — how good the evidence is for a claim, and how good the preparation is — are independent. Collapse them and you will argue past everyone.

What supervision adds is not a formality: baselines that cannot be recovered later, monitoring matched to the compound, the judgment to tell a receptor that has stopped listening from a dose that has become too small, interaction checking, a differential diagnosis instead of an attribution, somebody reachable at two in the morning, and somebody with the standing to say stop.

Chapter 20 opens Part IV. But if you take one operational thing from this chapter, make it the smallest one: whatever you are doing, tell your clinician what you are actually taking. The worst outcome here is not a bad vial. It is a bad vial in a person whose doctor does not know it exists.

Key Terms

Gray market — trade in substances that is not clearly illegal but operates outside the regulated pharmaceutical supply chain, typically under a "research use only" framing.

Research use only (RUO) — a designation placing a substance outside the regulatory definition of a drug by attaching no therapeutic claim. It describes what was not done to the material.

Epistemic laundering — the distributed construction of a therapeutic claim by parties who each individually make no such claim. Each element is defensible; the assembly is the claim.

Certificate of analysis (CoA) — a report of tests performed on a sample. A claim about a sample, not a property of a container. Chapter 34.

Batch release — the documented decision by a qualified person that a batch conforms to its written specification and may be supplied. The act that creates traceability and a recall path.

Related substances — impurities structurally similar to the intended peptide, arising during synthesis; the hardest kind to remove and to detect.

Deletion sequence — a peptide missing one residue because a coupling step failed while the chain continued growing. Nearly identical to the target in mass and chromatographic behavior.

Peptide content — the fraction of a product's total mass that is actually peptide, as distinct from bound water and counterions. Different from purity, and often not stated.

Sterility — absence of viable organisms, achieved by a validated process and verified by a specific test. Not conferred by dryness.

Endotoxin — lipopolysaccharide fragments of Gram-negative bacterial cell walls. They survive sterilization, pass through sterilizing filters, and cause fever and inflammatory responses when injected. A separate concern from sterility.

Bacterial endotoxins test — the assay demonstrating that a product meets a numeric endotoxin limit set for its route and amount. Routinely not performed on research-grade material.

Excipient — an inactive formulation ingredient: bulking agent, buffer, stabilizer, preservative. Its grade and suitability for injection are separate questions from the identity of the active.

Residual solvent — solvent remaining from synthesis or purification, controlled to defined limits in pharmaceutical products and typically unverified elsewhere.

Aggregation — clumping of peptide molecules into higher-order assemblies under adverse conditions of temperature, agitation, light, concentration, or solvent. Increases immunogenicity.

Immunogenicity — the propensity of a substance to provoke an immune response. Increased by aggregation and by closely related impurities.

Anti-drug antibody (ADA) — an antibody raised against an administered therapeutic. May neutralize it, and in rare cases may cross-react with the corresponding endogenous molecule.

503A compounding pharmacy — prepares medications for individually identified patients under valid prescriptions, regulated primarily by state boards, exempt from CGMP and premarket approval.

503B outsourcing facility — registers with the FDA, complies with CGMP, is FDA-inspected, and may compound in batches without patient-specific prescriptions. Still not an approval.

Salt form — a compound paired with a counterion (for example sodium or acetate). A different chemical entity from the base, and evidence does not automatically transfer between them.

Pharmacovigilance — the systematic collection, analysis, and action on safety information about medicines in use. Requires a reporting pathway to exist.

Adverse event report — a submitted account of a suspected harm associated with a medicine. FDA MedWatch and the UK Yellow Card scheme accept reports directly from patients.

Strict liability (anti-doping) — the principle that an athlete is responsible for any prohibited substance found in their body, regardless of intent or of what a label claimed.

Spaced Review

1. Chapter 6's four identical vials included a pharmaceutical product (A) and genuine research material with an honest certificate (B). Using §19.3, name the specific questions that distinguish them — and explain why "vial B is real peptide" narrows the gap less than people assume.

2. Chapter 17 concluded that the human evidence for BPC-157's popular claims does not support them, and Chapter 5's rules say a ❌ describes the evidence, not the molecule. Suppose a well-conducted randomized trial next year produced a clear positive result for one indication. Which of this chapter's concerns would that resolve, and which would be entirely untouched?

3. Using §19.7, name the situation in which "it's working less well than it was" is the most dangerous thing a person could report to a forum, and say what the correct next step would be.

4. Chapter 6 argued that the "research use only" framing shifts liability without changing the product, and Chapter 5 requires every rating to be falsifiable. Write a 📊 Evidence Rating for the claim "a vial accompanied by a certificate of analysis contains what the certificate says."

5. Explain to someone with no scientific background, in four sentences, why "it's the same molecule the pharmacy sells" is not the reassurance they think it is — without using the words endotoxin, specification, or pharmacovigilance.