Appendix E — Approved Peptide Medicines by Indication

Chapter 29 made a claim that sounds like a rhetorical flourish and is not one: most readers of this book already take, or have taken, a peptide medicine, and did not know it was one. Insulin is a peptide. So is the drug that induced a great many of the labors that produced a great many of the people reading this. So is the antibiotic in the tube of ointment in the bathroom cabinet, and so is the drug the anesthesiologist reached for when the blood pressure would not come back up.

This appendix is that claim's evidence. It is a directory of peptide medicines that have been approved somewhere, organized by what they are approved for, with a short account of what each molecule is. The tables are the skeleton. The prose between them is the appendix, because the directory's real subject is not the drugs but the shape of the list they make.

E.1 How to use this directory, and four cautions

Read the tables, then read what surrounds them. A list of approved drugs is a peculiar kind of document: it looks like a catalog and functions like an argument. The argument has two parts.

The first is about narrowness. Look at the third column of every table that follows. Acromegaly. Hepatorenal syndrome. Somatostatin-receptor-positive neuroendocrine tumors. Hypercalcemia. Adrenal function testing. Short bowel syndrome. Central precocious puberty. These are not marketing categories. They are precisely bounded clinical situations, often bounded further by the population in which the drug was studied, and each one represents a trial program that had to be run, reported, and reviewed before that phrase could appear on a label.

Now hold that against Part III. There you will find compounds credited, by the same source in the same paragraph, with healing tendons, repairing gut lining, improving mood, sharpening cognition, tightening skin, and accelerating recovery from essentially everything. Not one entry in this appendix makes a claim of that shape. Breadth of claim correlates inversely with strength of evidence, and this directory is the cleanest demonstration of it in the book — not because anyone argued the point, but because the constraint of having to prove something forces the claim to become small and specific. When a compound's advertised range of benefit expands, that is information. It is usually information about the evidence, and rarely good.

The second part of the argument is about axes. Chapter 38 §38.10 insisted that approval and evidence are separate dimensions, and this appendix keeps them visibly separate. The single most important entry here for understanding that is nesiritide (§E.7), which is approved, which is on this list, and whose clinical outcome claim this book rates ❌. Chapter 28 walked through why: it was approved on the strength of hemodynamic surrogates, and a large randomized outcome trial run afterward found no benefit on death or rehospitalization. Nothing was hidden and nobody cheated. The surrogate moved and the outcome did not.

So: a drug's presence in this directory is not a rating. It is a statement that a regulator somewhere concluded, at some point, that the evidence for a specified use met its standard. Appendix A carries this book's ratings; this appendix carries approval status. They are different columns in different tables and they disagree more often than the word "approved" leads people to expect.

                          EVIDENCE FOR THE CLAIM
                weak   <--------------------------->   strong

  APPROVED      nesiritide (outcome claim)  ❌         insulin
                calcitonin (fracture claim) ❌         GLP-1 receptor agonists
                                                       leuprolide
                                                       oxytocin
  ----------------------------------------------------------------------------
  NOT           most of Part III                       rare, and worth asking
  APPROVED                                             hard questions about
                                                       when you meet it

The upper-left quadrant is the one people forget exists. It is not large, and it is not a scandal — it is what happens when a decision has to be made before all the evidence arrives. But its existence is why "it's an approved drug" and "it works" are two sentences and not one.

Four cautions before the tables.

Approval is jurisdiction-specific, and this directory does not track jurisdiction. Regulators reach different conclusions, at different times, on different evidence packages, sometimes for the same molecule and the same indication. An entry here means the drug is approved somewhere for the named use. It does not tell you it is approved where you live, and one entry — thymosin alpha-1 (§E.10) — is included precisely because it is approved in some places and not others. Chapter 38 §38.4 is the section on what to make of "approved elsewhere," and its answer is that this is neither a credential nor a refutation, only a fact requiring interpretation.

Approval is indication-specific. This is the caution people most reliably discard. A drug is never simply "approved." It is approved for a use, in a population, by a route, on the basis of trials that studied that use in that population by that route. Every row below should be read as approved for the indication named, full stop. The distance between "approved" and "approved for X" is where a great deal of the marketing in Part III lives, because a compound that is approved for a narrow condition in a defined population can be described, truthfully and uselessly, as "an approved peptide."

This is not a complete list, and does not try to be. Peptide medicines number in the dozens to low hundreds depending on how you draw the boundary — and §E.11 shows that the boundary is genuinely hard to draw. Entries here were chosen because they appear elsewhere in this book, because they illustrate something, or because a reader is likely to have encountered them. Absence from this appendix means nothing at all.

Inclusion is not endorsement. Nothing here is a recommendation, nothing here contains dosing information, and route of administration is given only because it is chemically informative — it is the delivery problem of Chapter 4 showing up as a column in a table. If you want to know whether a drug is right for a person, that question belongs to that person and their clinician, and this book has never pretended otherwise.

E.2 Diabetes, obesity, and metabolic

Recombinant human insulin was approved in 1982, the first approved recombinant DNA drug of any kind. That is worth sitting with. The biotechnology industry's founding product was a peptide hormone, made because the alternative — extraction from animal pancreas — was a supply chain nobody could scale and a purity problem nobody could fully solve. Everything downstream in this appendix inherits from that decision: defined sequence, defined process, defined product. It is also the reason the world's most widely used peptide medicine is one that almost nobody calls a peptide. Ask a hundred people whether they have taken a peptide drug and then ask the same hundred whether anyone in their family uses insulin, and you will get two very different numbers about the same fact.

The GLP-1 rows repay slow reading, because they are the clearest available illustration of the indication-specificity caution. Exenatide was approved in 2005, the first GLP-1 receptor agonist — a molecule from Gila monster venom, which is Chapter 35's point about where leads come from, made concrete. Liraglutide was approved in 2010 for type 2 diabetes and in 2014 for weight management. Semaglutide was approved in 2017 for type 2 diabetes, in 2019 as an oral formulation, in 2021 for weight management, and in 2024 for cardiovascular risk reduction in a defined population. Tirzepatide was approved in 2022 for type 2 diabetes and in 2023 for weight management. Those are not one approval each. They are separate submissions resting on separate trial programs with separate endpoints, and the gaps between the dates are the years the trials took. A molecule does not accumulate indications by being good; it accumulates them by being studied, one question at a time, at enormous cost. This is what the honest version of "works for many things" looks like, and it looks nothing like a supplement label.

The oral semaglutide entry deserves a flag of its own, because it is Chapter 4's delivery problem appearing as a footnote in a table. A peptide taken by mouth meets stomach acid, then pancreatic proteases, then a gut wall built to admit small molecules and exclude large ones. Making an oral GLP-1 agonist required co-formulating the peptide with an absorption enhancer, and the fact that this counted as a distinct development program with its own approval — rather than a formulation tweak filed as an afterthought — tells you how hard the problem is. Note also what the oral entry is not evidence for. It does not establish that peptides in general survive the gut; it establishes that one heavily engineered peptide, in one specific formulation, does so well enough to be useful. Any claim that a peptide is "orally active" needs to say which peptide and in what formulation, because those are the two variables carrying all the weight.

Finally, glucagon and pramlintide are here as a reminder that this therapeutic area is not a single story. Glucagon reverses severe hypoglycemia — an emergency indication, narrow as a knife. Pramlintide is approved as an adjunct, a word that appears on labels precisely when the evidence supports a drug's use alongside something else rather than in place of it. Labels are written by people arguing about single words. Read them that way.

E.3 Bone and calcium

Teriparatide is a fragment of a human hormone. Say that sentence out loud, because it is exactly the sentence used to sell a dozen unapproved compounds in Part III — it's just a piece of something your body already makes. Teriparatide shows that the objection to those compounds was never that they are fragments. Fragments can be superb drugs; the first thirty-four residues of parathyroid hormone carry the whole receptor-activating business end, and the rest of the molecule turns out to be optional for that purpose. The objection to the Part III compounds is that nobody has run the trials. Teriparatide's fragment status is a chemistry fact. Its approval is an evidence fact. The marketing conflates them and the conflation is the trick.

The pharmacology here contains one of the most elegant facts in endocrinology and it is worth a sentence even in a directory. Parathyroid hormone signaling, sustained continuously, pulls calcium out of bone — chronic elevation is a bone-losing state. The same signal delivered intermittently builds bone. The receptor is identical; what differs is the pattern of exposure over time. The approved anabolic bone drugs exist because someone noticed that the direction of an effect can depend on its time course, which is a general lesson worth carrying: a molecule does not have an effect, a molecule plus an exposure pattern has an effect. Claims that name only the molecule have left out half the variables.

Calcitonin is the entry to study hardest, and it is here mainly because it is uncomfortable. It is approved, it has been in clinical use for a long time, and Chapter 29 rates its fracture-risk claim ❌ as of this writing — a downgrade over time, not an initial misjudgment. As trials accumulated and as long-term safety data were reviewed, the fracture-reduction case weakened rather than strengthened, and clinical practice moved away from it for osteoporosis while its use in hypercalcemia persisted. This is the shape of a hard truth about drug evidence: approval is a decision made at a moment, and the evidence keeps arriving afterward. Sometimes it arrives and confirms. Sometimes it arrives and does not. A directory that listed only drugs whose evidence had strengthened over time would be a comfortable document and a dishonest one.

E.4 Pituitary, growth, and water balance

Recombinant human growth hormone became available in 1985, and the year matters for a reason that has nothing to do with growth. 1985 is also the year that cadaver-derived growth hormone was withdrawn, following cases of iatrogenic Creutzfeldt-Jakob disease in people who had received it. Before recombinant production existed, the only source of human growth hormone was human pituitary glands, collected at autopsy and pooled. Pooling human tissue from many donors distributes any contaminant across the whole batch, and the contaminant in this case was a prion — an infectious agent that no purification step in use at the time removed and no test in use at the time detected. The people affected were mostly children, and the incubation period meant that the harm was invisible for years after the exposure. This is the strongest safety argument in the entire book for defined sequence and controlled process, and it is a great deal more persuasive than any abstract statement about quality standards. When Chapter 34 argues that testing a sample cannot substitute for knowing how the material was made, this is the historical case underneath.

Desmopressin looks, at first glance, like a counterexample to this appendix's narrowness thesis. It has three approved indications and they seem to have nothing in common: a rare pituitary disorder, a common pediatric condition, and a set of bleeding disorders. But look at how that breadth was acquired. Each indication rests on its own evidence, has its own defined population, and appears in its own section of the label. The molecule accumulated indications one trial at a time, the same way semaglutide did, and the fact that the three uses are mechanistically related — they all follow from a single receptor's effects on water handling and on the release of a clotting factor — did not excuse anyone from studying them separately. Breadth of approval and breadth of claim are not the same thing. The first is expensive and slow and shows up as a list of separately evidenced uses. The second is free and shows up as an adjective.

Two rows here are not treatments at all. Cosyntropin and secretin are diagnostic agents: peptides given to a person in order to provoke a measurable response and find out whether an organ is working. Cosyntropin is a fragment of ACTH, and its use is to ask the adrenal gland a question. Secretin asks the pancreas one. Sermorelin's principal remaining role is similar. These entries are a useful corrective to the assumption that "peptide medicine" means "peptide that treats something." Some peptides are reagents used in people — precise, purpose-built questions in molecular form — and they are approved on evidence that they answer the question reliably, which is a different kind of endpoint entirely. Note also how narrow those indications are, and then consider that the growth hormone secretagogues discussed in Part III are marketed with claims about body composition, sleep, recovery, and aging, while the two approved GHRH analogs in this table are approved for one condition and a diagnostic use. Same axis. Same class of molecule. Radically different breadth. That contrast is the appendix in miniature.

E.5 Somatostatin analogs

Octreotide is the design exemplar of this book and Chapter 33 treated it as such. Native somatostatin is fourteen residues with a half-life measured in a couple of minutes — too short to be a drug in any practical sense, which is the same wall GLP-1 hit. The solution was not to find a better natural molecule. It was to cut somatostatin down to the residues that mattered, close the resulting fragment into a ring so that exopeptidases have no free terminus to attack, and swap in D-amino acids at the positions where a protease would otherwise cleave. What comes out is shorter than the natural hormone, is not a sequence any ribosome could produce, and works for hours instead of minutes. Appendix B §B.7 covers the chemistry; this table shows what the chemistry bought.

What the table cannot show is why there are three drugs here instead of one. They are not interchangeable, and the reason is receptor subtypes. Somatostatin acts at a family of related receptors, and different tissues express them in different proportions. An analog that hits one subtype hard and the others weakly will do one thing; an analog with broader coverage across the family will do another. Pasireotide's approval in Cushing's disease and octreotide's in carcinoid syndrome are downstream of exactly this — the same peptide scaffold, tuned differently, ends up approved for different diseases. This is why "somatostatin analog" is a chemical description and not a clinical claim. Whenever you meet a compound described by its family rather than by its receptor profile and its evidence, you are being told the least informative true thing available.

There is one more thing hiding in this section, and it is the bridge to §E.6. These tumors are targeted with somatostatin analogs because they express somatostatin receptors in abundance. That same fact — the receptor is there, in quantity, on the cells you care about and not on most others — can be used for something other than signaling. It can be used as an address. Hold that thought for two paragraphs.

E.6 Oncology and reproductive

Look at the suffixes. -relin versus -relix is agonist versus antagonist at the same hormonal axis, one letter apart. Appendix I explains the nomenclature system that produced this, and it is not an accident or a naming collision — the stems were assigned deliberately, and the single changed letter is doing real work. But the consequence for a patient is not subtle at all. A GnRH agonist initially stimulates the axis before continued exposure desensitizes and shuts it down, which means the first days of treatment produce a surge of exactly the hormone the treatment is meant to suppress. In hormone-sensitive prostate cancer, that surge is the "flare," and it can transiently worsen symptoms. A GnRH antagonist blocks the receptor from the outset, so there is no initial stimulation and no flare. Degarelix's approval language names this explicitly — "without initial flare" is not marketing garnish, it is the clinical difference the molecule was built to deliver.

Which means that a reader who cannot tell -relin from -relix cannot tell a drug that causes an initial hormonal surge from a drug that does not. That is a large clinical difference resting on one letter, and it is the best argument in this book for learning the nomenclature conventions rather than skimming past them. The naming system is not decoration. It is compressed pharmacology, and Appendix I is the decompression algorithm.

The two lutetium entries do something the rest of this appendix does not, and it is worth understanding because it is where a great deal of the field is heading. In every other row here, the peptide is the drug: it binds a receptor and the binding produces the therapeutic effect. In radioligand therapy, the peptide binds a receptor and the binding produces proximity. The therapy is the radionuclide it carries, and the peptide's job is to deliver that payload to cells expressing the target and, crucially, not to cells that do not. Lutetium-177 dotatate was approved in 2018, using precisely the receptor abundance that §E.5's analogs exploit for signaling — the same address, a different letter delivered to it. Lutetium-177 PSMA-617 applies the same logic to a different target in prostate cancer, and it sits at the class boundary §E.11 is about: its targeting element is better described as a peptidomimetic ligand than as a peptide proper. Note what the indications demand. Not "prostate cancer" but "PSMA-positive advanced prostate cancer"; not "neuroendocrine tumors" but "somatostatin-receptor-positive" ones. The label requires the target to be demonstrated in the individual patient before the drug is given. That is about as narrow as an indication gets, and it is what precision looks like when it is real rather than claimed.

E.7 Cardiovascular

Nesiritide is the entry this appendix most needs, and it is here to be uncomfortable. It is recombinant human BNP — the actual hormone, correct sequence, produced by the heart itself under pressure. Chapter 28 told the story properly and the short version is this: it was approved on hemodynamic surrogates. It lowered filling pressures and improved measured breathlessness in the short term, and those effects were real. A large randomized outcome trial was subsequently run to ask the question the surrogates could not answer — does it keep people alive and out of the hospital — and it found no benefit on death or rehospitalization. This book therefore rates nesiritide's outcome claim ❌, and nesiritide appears in this directory, and both of those statements are correct at the same time. That is not a contradiction to be resolved. It is the two-axis structure of §E.1 made concrete in a single row.

Three lessons come out of that row and all three are general. First: a surrogate endpoint is a hypothesis about a mechanism, not a measurement of a benefit. Pressures fell, which was the mechanism working exactly as designed, and the mechanism working did not translate into people doing better. Second: the trial that mattered was run after approval, which is common and not sinister — outcome trials are large, slow, and expensive, and requiring one before every approval would keep useful drugs from people who need them. But it does mean the evidence base under an approved drug can change, and sometimes it changes downward. Third, and most useful for reading anything else: "it's the actual human hormone" is not an argument. Nesiritide was the actual human hormone, identical to what the body makes, and that fact did not make it work. Several compounds this book rates ❌ are exact fragments of human proteins, and their exactness is offered as though it were evidence. It is not evidence of anything except sequence.

The other two rows are quieter and make Chapter 35's point about lead discovery. Bivalirudin descends from the anticoagulant peptide a medicinal leech uses to keep its meal flowing; eptifibatide descends from a venom peptide that a viper uses to disrupt its prey's clotting. Organisms have spent a very long time evolving molecules that interfere precisely with mammalian physiology, and medicinal chemistry has been mining that library for decades. Both drugs are given intravenously in acute settings, both are approved for tightly bounded uses, and neither is described by anyone as supporting general cardiovascular health. The venom origin story that makes a compound sound exotic in a marketing paragraph is, in the approved formulary, completely routine — and it is always attached to a narrow indication and a trial program.

E.8 Infectious disease

This section is the counterexample to Chapter 25, and it must be read carefully, because the obvious reading of it is wrong. Chapter 25 rated antimicrobial peptides ⚠️/🔬 as systemic antibiotics, and a reader arriving here could reasonably conclude that the rating was simply mistaken, since half a dozen peptide-derived antibiotics sit in this table and get used in hospitals every day. That conclusion does not follow, and the reason is a distinction the rating depends on.

Chapter 25's subject is the host-defense antimicrobial peptides of the innate immune system — the defensins, the cathelicidins, the cationic amphipathic peptides your own tissues secrete — evaluated as a source of new systemic antibiotics. That specific proposition has repeatedly struggled in development: the molecules that kill bacteria beautifully in a dish, by disrupting the negatively charged bacterial membrane, tend to encounter serum proteins, host cell membranes, and rapid proteolysis when they enter a body, and the therapeutic window between "kills the pathogen" and "damages the host" narrows in ways that are hard to engineer around. The ⚠️/🔬 rating is about that development pathway, and it remains accurate.

The drugs in this table came from somewhere else entirely. Daptomycin, the polymyxins, bacitracin, vancomycin, and the echinocandins are microbial natural products — mostly non-ribosomally synthesized compounds that soil bacteria and fungi evolved to attack each other, found by screening programs rather than designed from human immunology. They are structurally strange by peptide standards: cyclic, lipidated, glycosylated, full of non-standard residues, and shaped by millions of years of chemical warfare that had already solved the stability problem before any chemist looked at them. So the precise statement is this: peptide-derived antibiotics are an established, approved, daily-use part of medicine, and the host-defense peptides of the innate immune system have not yet become one. Both halves are true, and collapsing them into "peptide antibiotics don't work" or "peptide antibiotics work great" loses the entire content. This is the sharpest correction in this appendix of a misreading a reader is likely to make, which is why it gets three paragraphs instead of one.

Two rows carry additional lessons. Daptomycin's label excludes pneumonia — not from caution, but because pulmonary surfactant inactivates the molecule, so it does not work there. That is indication narrowness with a mechanistic explanation attached, and it is the most literal possible demonstration that an approved antibiotic is approved for specified infections rather than for infection. And enfuvirtide, approved in 2003, is a 36-residue peptide that blocks HIV from fusing with a host cell. It works. It also requires subcutaneous injection, because a 36-residue peptide is not going to survive the gut, and injection-site reactions are the well-known consequence of delivering that much peptide under the skin. Enfuvirtide's role in therapy has always been constrained less by whether it works than by what taking it involves. That is Chapter 4's delivery problem determining a drug's clinical position rather than its pharmacology — which happens far more often than mechanism-focused accounts of medicine admit.

E.9 Neurology, pain, and immunology

Ziconotide was approved in 2004, and it is given intrathecally — delivered into the cerebrospinal fluid via an implanted pump and catheter — because it does not cross the blood-brain barrier. There is no clever formulation here, no absorption enhancer, no prodrug. The molecule cannot get where it needs to go, so medicine performs surgery on the patient to move the delivery point to the other side of the barrier. Chapters 4 and 35 both use this drug, and it deserves both mentions: it is simultaneously a triumph of natural product pharmacology — a cone snail peptide that blocks a specific neuronal calcium channel with a selectivity synthetic chemistry struggles to match — and a complete capitulation on delivery. When people say that peptides have a delivery problem, ziconotide is what the problem looks like when the drug is good enough to be worth solving it surgically. Note, too, what this implies about any claim that some other peptide "crosses the blood-brain barrier." Ziconotide's developers had every commercial reason to want an easier route and could not find one. Extraordinary claims about central nervous system penetration require the kind of evidence that building a pump into someone's abdomen represents the alternative to.

Cyclosporine is the opposite pole and Chapter 33 used it as such. It is a cyclic peptide, it contains D-amino acids, it is heavily N-methylated, and it is orally bioavailable — a peptide you can swallow that reaches systemic circulation and suppresses the immune system well enough to have made organ transplantation routine. Everything Appendix B §B.7 says about what D-substitution and N-methylation buy is visible in this one entry: D-residues that mammalian proteases cannot process, N-methylated backbone nitrogens that remove hydrogen bond donors and let the molecule fold into a membrane-permeable conformation, and a ring with no free termini for exopeptidases to grip. It is also a fungal natural product, which is to say that nobody designed it — chemistry found it and then worked out afterward why it behaved so unusually. Cyclosporine proves that oral peptide drugs are possible. It does not prove they are easy, and the decades between it and oral semaglutide are the measure of how hard.

Botulinum toxin gets its full treatment in §E.11, but its approval history belongs here because it is this appendix's thesis in a single row. It was first approved in 1989 for strabismus and blepharospasm — two narrow ophthalmic conditions. In 2002 it was approved for glabellar lines, the cosmetic indication that made it a household word. In 2010 it was approved for chronic migraine prophylaxis. Three approvals across three decades, in three unrelated clinical areas, each requiring its own evidence. The most famous "peptide" in the world — and it is not a peptide — did not become broadly used because someone claimed it was broadly useful. It became broadly used one narrowly specified, separately evidenced indication at a time. That is the difference between a drug's indications expanding and a claim's scope expanding, and it is the whole argument of §E.1 compressed into three dates.

E.10 Gastrointestinal and rare disease

Linaclotide inverts Chapter 4 completely, and this is the most interesting fact in the section. Every other peptide in this appendix fights absorption: the molecule must survive the gut, or the subcutaneous depot, or the bloodstream, and reach a receptor somewhere distant. Linaclotide's receptor is on the inside surface of the intestine. It does not need to be absorbed. It needs to not be absorbed — to stay in the lumen, act locally on the cells lining the gut, and go no further. Poor systemic absorption is the design goal rather than the problem, and the whole delivery obstacle that shapes the rest of this appendix simply does not apply. Plecanatide works the same way on the same target. The lesson generalizes beyond these two drugs: the delivery problem is not a property of peptides, it is a property of the relationship between where a molecule enters and where its target sits. Change the target's location and the problem dissolves.

The two hereditary angioedema entries show what precise mechanism buys you in a rare disease. Hereditary angioedema attacks are driven by uncontrolled bradykinin generation, and the two drugs intervene at different points on that pathway — icatibant blocks the receptor bradykinin acts on; ecallantide inhibits the enzyme that generates it. Same disease, same pathway, two different molecular strategies, both approved. Ecallantide is also a useful boundary case for §E.11, since at around sixty residues it is comfortably past the forty-amino-acid line Chapter 38 §38.2 uses to separate peptides from proteins — a recombinant protein that everyone files under peptide therapeutics because that is the field it grew up in. Teduglutide belongs to the same engineering lineage as the GLP-1 drugs in §E.2: GLP-2 is a gut hormone with the same DPP-4 vulnerability, and stabilizing it against that enzyme is what turned a hormone into a drug for short bowel syndrome. One enzyme, two hormone families, two therapeutic areas. Chapter 33 was not exaggerating about how much of peptide drug design is the same trick applied repeatedly.

Bremelanotide was approved in 2019, and its indication is a good closing example of how narrow these things actually get: not "low libido," but hypoactive sexual desire disorder, in premenopausal women, as a defined diagnosis. The population is specified as tightly as the condition. And thymosin alpha-1 is included as the deliberate exception, because it is approved in some jurisdictions and not others — which is the §E.1 jurisdiction caution stated as a row rather than a warning. Chapter 18 rates it ⚠️, and Chapter 38 §38.4 explains the general principle: "approved elsewhere" is neither a credential nor a refutation. Regulators differ in what evidence they require, what they will accept from which sources, and how they weigh benefit against uncertainty for a given condition. A drug approved in one country and not another has not been vindicated by the first regulator or condemned by the second. It has been assessed twice, by different standards, possibly on different data. The correct response to "approved elsewhere" is to go and look at the evidence package, not to treat the approval itself as the answer.

E.11 Things on this list that are not peptides

Several entries above are in this appendix and are not, strictly, peptides. That is not sloppiness. It is the most useful section here, because readers conflate these constantly, and because understanding why each one is borderline teaches the category better than any definition does.

Botulinum toxin is a large bacterial protein — thousands of residues, a complex multi-domain structure with enzymatic activity — and it is not a short peptide by any measure. It is in the popular imagination as the archetypal "peptide injection," and it appears in Chapter 30 alongside cosmetic peptides because that is the company it keeps in clinics and in conversation, not because it belongs to the same molecular class. If you want a single fact to fix the difference: botulinum toxin is an enzyme that cleaves proteins inside nerve terminals. The cosmetic peptides of Chapter 30 are short signaling or fragment molecules applied to skin. They share a marketing category and essentially nothing else.

Sacubitril/valsartan is a small molecule combination and is arguably the most important drug in Chapter 28 — which is a chapter about the natriuretic peptide system. It earns that place without being a peptide at all, because it works on the peptide system rather than as part of it: it inhibits the enzyme that degrades natriuretic peptides, raising their levels by slowing their destruction rather than by supplying more. Set it beside nesiritide (§E.7) and the contrast is instructive. Nesiritide supplied the peptide directly and did not improve outcomes; a small molecule that raises endogenous peptide levels did. Peptide biology and peptide drugs are different subjects, and the best drug acting on a peptide pathway need not be a peptide.

Captopril is a small molecule descended from a venom peptide lead — Chapter 35's peptidomimetic story in one word. The starting point was a peptide from a pit viper venom that inhibited angiotensin-converting enzyme; the destination was an orally active small molecule that kept the binding interaction and discarded the peptide backbone. This is what "peptidomimetic" means in practice: not a peptide made more stable, but a non-peptide that reproduces the pharmacophore — the specific arrangement of chemical features that the target recognizes — in a chemistry that survives the gut. Captopril is the field's proof that a peptide lead can be a route to a drug that is not a peptide at all, and that this is sometimes the better outcome.

Vancomycin is a glycopeptide — a peptide core, non-ribosomally assembled, with sugars attached and unusual crosslinks that lock it into a rigid shape. Is it a peptide? Structurally, part of it plainly is. Behaviorally, it has almost nothing in common with the linear signaling peptides that occupy most of this book. The echinocandins — caspofungin, micafungin, anidulafungin — are lipopeptides: cyclic peptide cores with fatty acid tails, likewise borderline, likewise built by enzymes rather than ribosomes. Daptomycin (§E.8) is in the same architectural family. These compounds occupy a region of chemical space that the word "peptide" reaches into without covering.

Glatiramer acetate is the strangest entry in the entire directory and the best test of what "peptide" means. It is a random copolymer of four amino acids — glutamate, lysine, alanine, and tyrosine — polymerized in fixed ratios rather than assembled in a defined order. There is no sequence. There is no single molecule. A vial contains an enormous population of different chains of different lengths with different residue orders, characterized statistically rather than structurally, and it is approved and used in multiple sclerosis. Everything Chapter 1 asserted about a peptide being a sequence rather than a substance runs directly into this drug, which is a substance and not a sequence. It is a genuine anomaly, and the honest position is to say so rather than to stretch the definition until it fits. It also raises a hard manufacturing question worth carrying: what does "the same product" mean for a material defined by ratios and a process rather than by a structure? The answer, in practice, is that the process is the definition — which is Chapter 34's argument arriving from an unexpected direction.

Antibody drugs — everything ending in -mab — are not peptides. They are large, folded, glycosylated proteins of many hundreds of residues, and they belong to a different manufacturing world, a different regulatory pathway, and a different set of design problems. Chapter 38 §38.2 draws the working line at roughly forty amino acids: below it, "peptide" is the natural word; above it, "protein." That line is a convention rather than a law, and it does not fall in the same place for everyone. Appendix I's stem table is the practical tool here — the -mab stem tells you unambiguously that you are looking at an antibody and not a peptide, which is more than most people get from the name.

Here is the teaching point, and it is the reason this section exists. The boundary of the category "peptide" is genuinely fuzzy. Size is a spectrum, not a threshold. Cyclization, glycosylation, lipidation, and non-standard residues each pull a molecule away from the simple picture. Some compounds are built by ribosomes and some by dedicated enzyme assembly lines and some by chemists, and the origin changes what the molecule can be. A reader who can explain why each entry above is borderline — which axis it departs on, and how far — understands the class considerably better than a reader who can recite a definition and apply it confidently to everything. Definitions are useful right up to the edge of a category, which is exactly where the interesting molecules live.

E.12 Closing: what the list shows

Four observations, in order of how much they change how you read the rest of the book.

One: the indications are relentlessly narrow. Go back through the third column. Hepatorenal syndrome. Central precocious puberty. Somatostatin-receptor-positive neuroendocrine tumors. Adrenal function testing. HIV-associated lipodystrophy. Hypoactive sexual desire disorder in premenopausal women. Acute attacks of hereditary angioedema. Every one is a bounded clinical situation, most are bounded further by population, and several require a target to be demonstrated in the individual patient before treatment begins. Not one entry claims general benefit for anything. Now recall the compounds of Part III, where a single molecule is routinely credited with helping tendons, gut, brain, and skin at once. Breadth of claim correlates inversely with strength of evidence, and this directory is where you can see it most cleanly, because the mechanism producing the correlation is visible: proving something requires specifying it, and specifying it makes it small. A claim that has never had to be specified is free to be as large as its author likes.

Two: the oldest drugs on this list are the best evidenced. Insulin, from 1982 in its recombinant form and much older in its extracted one. Oxytocin. The GnRH agonists. Vancomycin and the polymyxins. Cyclosporine. These have decades of use across enormous populations, and what is known about them is known from more than a trial program — it is known from the accumulated experience of millions of people taking them under every kind of circumstance, including the circumstances no trial protocol would allow. Time is an evidence-generating process. It is the one input to certainty that no amount of money or enthusiasm can accelerate, which is precisely why novelty is such a poor recommendation for a drug and such a good one for a product. When a compound's principal selling point is that it is new, notice that you are being offered the absence of the thing that matters most.

Three: look at the routes. They are overwhelmingly injectable.

  INJECTED (subcutaneous, intramuscular, intravenous, or implanted depot)
    insulin and analogs · exenatide · liraglutide · dulaglutide · semaglutide
    tirzepatide · pramlintide · glucagon · teriparatide · abaloparatide
    calcitonin · somatropin · tesamorelin · sermorelin · desmopressin
    vasopressin · terlipressin · oxytocin · carbetocin · cosyntropin · secretin
    octreotide · lanreotide · pasireotide · leuprolide · goserelin · triptorelin
    degarelix · cetrorelix · ganirelix · lutetium-177 dotatate
    lutetium-177 PSMA-617 · nesiritide · bivalirudin · eptifibatide · daptomycin
    polymyxin B · colistin · enfuvirtide · vancomycin · caspofungin · micafungin
    anidulafungin · glatiramer acetate · teduglutide · icatibant · ecallantide
    bremelanotide · thymosin alpha-1 · botulinum toxin (injected into tissue)

  ORAL
    cyclosporine ................ systemic; cyclic, D-residues, N-methylated
    semaglutide ................. one formulation, with an absorption enhancer
    linaclotide · plecanatide ... act in the gut lumen; absorption not wanted

  TOPICAL ....... bacitracin
  INTRATHECAL ... ziconotide          (pump and catheter, past the barrier)
  NASAL ......... desmopressin · calcitonin (some formulations)

That is Chapter 4's delivery problem rendered as a table. The injectable column is not a commercial preference or a conservatism of habit; it is the direct consequence of what the gut does to a chain of amino acids. And look at what it takes to escape it. Cyclosporine escaped by being an extraordinary natural product that chemistry did not design. Semaglutide escaped in one formulation built specifically for the purpose. Linaclotide and plecanatide escaped by not needing to go anywhere. Ziconotide did not escape at all — it required a surgically implanted pump. Four routes out of the injection column, and each one is either a rare accident of nature, a major engineering program, a target that happens to sit in the lumen, or an operation. Set that against the casual claim that some unapproved compound is effective taken orally or applied to skin, and the burden of proof such a claim carries becomes obvious.

Four, and this is the one to take with you: if you scanned this list and found something you have taken, you have just learned this book's central lesson experientially rather than by argument. Perhaps it was insulin, or oxytocin during a birth, or a tube of bacitracin, or an antibiotic during a hospital stay, or one of the GLP-1 drugs, or the injection that fixed a migraine or a frown line. Whatever it was, you took it without the word "peptide" entering the transaction. It was tested. It was manufactured to a specification. It was prescribed for a named condition, and it did — or did not do — a specific thing that someone had measured in advance.

That is what the category actually contains: ordinary, well-evidenced, precisely indicated medicine, sitting quietly in pharmacies and hospital formularies for decades. The peculiar modern usage in which "peptides" names a shadow category of unapproved compounds sold on breadth of promise is a recent and local distortion of a word with a long, boring, honorable history. The class was never exotic. The evidence-free corner of it is what's exotic, and it is a small corner of a large and unglamorous field.

A reader who finishes this appendix should be able to do one thing they could not do before: hold "approved" and "well evidenced" apart in their head, and notice which one a given sentence is actually asserting. Nesiritide is approved. Calcitonin is approved. Both have a central claim this book rates ❌. Meanwhile Part III's compounds are neither, and are described in language that borrows the confidence of the first column without having earned it. The word doing the work in any drug claim is almost never "peptide." It is whichever word tells you what was measured, in whom, against what. Find that word first. If it isn't there, you have learned something too.

Related: Chapter 4 (delivery) · Chapter 25 (antimicrobial peptides) · Chapter 28 (natriuretic peptides and surrogates) · Chapter 29 (approved peptide medicines) · Chapter 33 (engineering) · Chapter 35 (leads and peptidomimetics) · Chapter 38 (reading claims) · Appendix A (evidence ratings) · Appendix B (amino acid reference) · Appendix I (nomenclature) · Appendix K (glossary)