Chapter 41: Ozempic Enters the Language

"The limits of my language mean the limits of my world." — Ludwig Wittgenstein, Tractatus Logico-Philosophicus (1922), §5.6

Overview

Somewhere in the last few years, a brand name stopped being a brand name.

It happened the way these things always happen: quietly, by ordinary people needing a short word for a long idea, and without anyone deciding. The long idea was glucagon-like peptide-1 receptor agonist, which is eleven syllables and a hyphen, or incretin mimetic, which is worse because it sounds like a term you should already know. The short word was three syllables, easy to say, and belonged to a single product from a single manufacturer approved for a single indication.

The short word won. It now stands for a molecule, for several other molecules, for an entire drug class, for pharmacological weight loss in general, for a cultural moment, and — in the construction "the Ozempic of X" — for the abstract idea of an intervention that works far better than anyone expected. That is a remarkable career for a word, and this chapter is about it.

It is not a chapter about semaglutide. Chapter 8 did that, and Chapter 12 did access and coverage. This chapter is about what happened to the language, and about consequences that are not merely linguistic. When one brand name absorbs a whole category, patients lose the ability to say what they are taking, prescribers field requests for the wrong product, and coverage conversations become confused in ways that cost people money and time.

Then it gets harder. A real clinical observation — that substantial weight loss by any means reduces facial fat — acquired a brand-name label, and that label became a way of alleging, about a stranger, that their appearance is the product of a drug they have never discussed with you. This chapter takes that seriously, and it takes seriously the question underneath it: what do you owe a stranger whose body you are speculating about?

One warning about method, and it applies to the whole of Part VIII. Nothing here relaxes the standards of the previous forty chapters. Claims about culture have populations, endpoints, and evidence bases too — usually thin ones — and they get rated with the same discipline. And no private individual is named anywhere in this chapter. Not to criticize and not to praise. The reason for that rule is one of the things the chapter is about, so breaking it would refute the argument.

In this chapter, you will learn to:

• Explain how a brand name genericizes, and name the rungs of that ladder
• Describe four specific ways genericization damages medical communication
• Trace how a descriptive clinical term became an accusation, and identify the structural move
• State both sides of the disclosure question at full strength, and see why the rating system declines it
• Show that speculation about individuals is unfalsifiable, and use that as the decisive objection
• Explain why a joke transmits a social fact that no evidence supports
• Give the culture credit where it has earned it, which is in several real places
• Identify the harms that fall on people who need these drugs medically
• Apply five practices that keep your own language precise without becoming insufferable

Learning Paths

All five paths read this chapter. It is short on pharmacology and long on the environment in which every future peptide claim will reach you.

💊 GLP-1 — this is your chapter. §41.2 is the one to read twice; it is the difference between a productive appointment and a confused one. §41.8 explains why strangers keep asking you to justify a prescription. 🏋️ Performance — §41.5 and §41.6 transfer directly. The speculation economy and the joke-as-vector operate identically in gym culture, with a different vocabulary and the same structure. 🔬 Science — §41.3 is a compact case study in what happens when a clinical observation escapes the clinic, and §41.5 is an applied falsifiability argument. Both are Chapter 5's machinery in a new domain. 💄 Cosmetic — §41.3 and §41.7 are the sections that matter most, and Chapter 30's skepticism about appearance claims applies here in reverse: the claim is now about someone else's face. 🏥 Clinical — §41.2 and §41.8 are the practical ones. The medication-history problem in §41.2 is already showing up at pre-procedure screening, and §41.8 names a burden your patients are carrying that has nothing to do with the drug.

41.1 How a brand name became a common noun

Words get borrowed. It is what words are for.

When a new thing arrives and the language has no convenient handle for it, people reach for the nearest available handle, and the nearest available handle is very often a brand. This is old and ordinary. You have almost certainly asked someone to xerox a document on a machine made by a competitor, put a band-aid on a cut using an own-brand adhesive strip, or looked something up by googling it. Linguists call the phenomenon genericization; trademark lawyers, who have a different stake in it, call the endpoint genericide, because a mark that becomes the common name for its category can lose legal protection entirely. Aspirin, escalator, thermos, cellophane, trampoline, and heroin were all brand names once. All of them lost.

So the process is not new. What is worth attention is why this particular word and how far it traveled, because the answers explain the damage.

Why this word. Three forces converged.

The first is that the alternatives are unusable. The class name is glucagon-like peptide-1 receptor agonist. Its acronym, GLP-1 RA, is not a word; it is a serial number. The generic name, semaglutide, is four syllables of unfamiliar phonology that most people have not seen written down before they have to say it aloud, and which — as Chapter 1 §1.8 explained — is built from a naming stem system that tells you a great deal if you know the system and nothing at all if you do not. Against that, a short trade name with a hard consonant in the middle is not competing on merit. It is competing on sayability, and it wins by a distance.

The second is timing. The brand that entered the language is the one that was widely prescribed and widely discussed first, during the window when the general public formed its first impression of the whole category. Genericization does not select the best-studied member of a class, or the most-used, or the one approved for the use people are actually talking about. It selects the most famous one at the moment the category becomes interesting. That is a fact about attention, not about medicine.

The third is that the word had somewhere to go. English is generous with productive constructions, and this brand name proved unusually good at slotting into them: as a modifier, as a suffix, as the second half of an "the X of Y" analogy. A word that can be bent is a word that spreads.

How far it traveled. Further than any of the classic examples, and faster. Here is the ladder.

HOW A BRAND NAME CLIMBS OUT OF ITS OWN CATEGORY

  RUNG 5   METAPHOR       "the Ozempic of higher education"
              ▲           the word has left medicine entirely; it now means
              │           "an intervention that works embarrassingly well"

  RUNG 4   PHENOMENON     "the Ozempic era," "the Ozempic economy"
              ▲           names a social moment, a market, a mood
              │

  RUNG 3   CATEGORY       "she's on Ozempic" — said of someone taking a
              ▲           different molecule made by a different company
              │           the brand now stands for every drug in the class,
              │           and loosely for pharmacological weight loss as such

  RUNG 2   SHORTHAND      "an Ozempic-type drug," "Ozempic and things like it"
              ▲           still recognizably a brand, used as a pointer
              │

  RUNG 1   BRAND          one molecule, one manufacturer, one route,
                          one approved indication

  Each rung upward adds reach and subtracts precision.
  By rung 3 the word no longer identifies a molecule.
  By rung 5 it no longer identifies a drug.

Most genericized brands stop at rung 3. Kleenex means any facial tissue; it does not name an era. This one climbed the whole ladder in a handful of years, which is unusual and is itself a piece of evidence about how much cultural weight the drugs were carrying.

A necessary caution. This book will not give you a number for any of this. You will see confident figures quoted for search volume, prescription counts, and the share of some population using these drugs; some are real and some are extrapolations presented as measurements, and this chapter cannot sort them for you. What it can do honestly is describe the shape — a brand name climbing a ladder it did not have to climb — and say plainly that the shape is not quantified here. If you meet a number, the Chapter 5 question applies: who counted, what exactly, over what population?

🧬 The Molecule — one molecule, three names, and the confusion that follows

Chapter 1 §1.8 introduced the three naming systems, and Chapter 8 applied them. Here is the part that matters for this chapter, stated as compactly as possible.

Generic name — assigned under international convention, one per molecule, worldwide. Semaglutide is semaglutide in every country and in every product that contains it.

Brand name — commercial property, one per product, and a product is a molecule plus a formulation plus a route plus an approved indication. One molecule can therefore have several brand names, and they are not interchangeable, because the products are not identical.

The consequence is the thing people find surprising: Ozempic and Wegovy contain the same active molecule and are not the same product. They differ in the doses supplied, in the device, and — most consequentially — in what they are approved to treat. A sentence like "Ozempic and Wegovy are different drugs" is wrong at the level of chemistry and right at the level of everything that determines what happens to you at a pharmacy counter. Both halves of that are true at once, which is exactly why the shorthand collapses.

The same structure applies to tirzepatide, sold under two brand names with two indication families. And approved indications change: labels are revised, new indications are added, and a statement about what a product is approved for has a date attached to it. Where this chapter describes a label, treat it as a snapshot and check the current one.

41.2 The genericization problem, and why it is a medical problem rather than a pedantic one

It would be easy to treat all of this as a language column — mildly interesting, ultimately harmless, the sort of thing that annoys people who care about words. That would be a mistake. Genericization here has a specific and traceable path to harm, and it runs through the fact that the word has been asked to carry more than any word can.

WHAT ONE WORD IS BEING ASKED TO CARRY

  MOLECULE          BRAND        ROUTE        APPROVED INDICATION FAMILY
  ─────────────────────────────────────────────────────────────────────
  semaglutide       Ozempic      injection    type 2 diabetes
  semaglutide       Wegovy       injection    weight management
  semaglutide       Rybelsus     oral         type 2 diabetes
  tirzepatide       Mounjaro     injection    type 2 diabetes
  tirzepatide       Zepbound     injection    weight management

  Two molecules. Five brands. Two routes. Two indication families.
  Different supplied dose ranges. Different coverage rules. Different labels.
  In general conversation, all of it is one word.

  And the same word is ALSO doing duty for:
    · compounded preparations, which are not any of the products above
    · gray-market vials of unverified identity and purity (Ch 19, Ch 34)
    · other, older drugs in the same class
    · "weight-loss drugs" as an undifferentiated idea
    · a social phenomenon that includes people taking nothing at all

Five products in the table. At least four additional referents underneath it. One word. Here is what breaks.

First: patients cannot say what they are taking. This is the most immediate harm and the least appreciated. A person who says "I'm on Ozempic" may be taking any of the five products above, a compounded preparation, or something bought outside the medical system entirely. The molecule matters, the route matters, and whether it is the diabetes product or the weight-management product matters. A clinician who takes the word at face value and writes it into the medication list has recorded something that may be false in three separate fields.

This is not hypothetical fussiness. Medication histories drive real decisions: pre-procedure fasting guidance, endoscopy and imaging preparation, interactions with drugs whose absorption is affected by delayed gastric emptying, decisions around pregnancy, and the correct attribution of side effects. If a patient reports a symptom and the record names the wrong molecule, the safety signal lands in the wrong place.

Second: prescribers field requests for the wrong product. A patient who has learned one word will ask for that word. If what they want is weight management and the word they know is the diabetes brand, then either the clinician spends appointment time on translation or a request goes in that does not match the intent. The results are predictable: coverage denials, off-label designations that trigger review, prior-authorization loops, and a patient who believes they were refused a drug when in fact they asked for a different product than the one they wanted.

Third: coverage conversations become confused. Chapter 12 laid out how payers decide — by product and indication, not by molecule. A formulary has no entry for "semaglutide, generally." When patient, prescriber, and payer each use one word to mean three different things, the conversation cannot converge, and people conclude that coverage decisions are arbitrary when what they have encountered is a naming collision.

Fourth, and largest: the brand name became a container for things that are not the drug. During the period of constrained supply, preparations that were not any approved product circulated widely under the same word, as did vials sold through channels with no oversight of identity, concentration, or sterility — the exact failure modes Chapters 19 and 34 document. And the failure is asymmetric in a nasty way: the approved products carry the safety and efficacy record, and the word lends that record to everything else it touches.

💊 In the Clinic — the question that has to be asked twice

Clinicians in several specialties have converged on a small habit, worth knowing about from the patient side because it is not an interrogation and it helps to answer it precisely.

The question is some version of: "You said Ozempic — can you tell me the exact name on the box, and whether it's a pen or a tablet?"

It is doing four jobs at once: establishing the molecule, the product, the route, and — by implication — the indication it was prescribed under. Asked once, it often returns the generic word again. Asked a second time, gently, it usually returns the box.

The most useful thing you can do as a patient is make the question unnecessary: keep a photograph of the label on your phone, or write the product name, the molecule, and the prescribing indication somewhere you can find them. Chapter 39 is about making clinical conversations useful, and this is one of the cheapest possible contributions to that.

No part of this callout is dosing guidance. It is a record-keeping habit, and it belongs to you.

Notice the structure of all four harms. None of them is caused by the drug. All of them are caused by the word. And all of them are, in principle, fixable by a change in habit that costs nothing — which is why §41.9 exists.

📊 Evidence Rating

Claim: Genericization of a brand name — one brand name coming to stand for an entire drug class — produces measurable harm to patient-prescriber communication and to medication-history accuracy.

Rating: ⚠️ Promising but preliminary (as of this writing)

Reason: The mechanism is clear and the individual failure modes are documented in other genericized drug classes — botulinum toxin products and epinephrine auto-injectors are the standard examples, and in both cases naming confusion has been formally recognized as a source of administration and dosing error. For this class specifically, the effect is described by clinicians and plausible on its face, but it has not been quantified at scale: there is no published estimate of how often a medication history for these drugs records the wrong product, or what fraction of coverage disputes trace to a naming collision. Mechanistically clear, under-measured.

What would change it: A study of medication-history accuracy in patients taking incretin-based therapies — comparing patient-reported product against dispensing records — with a reported error rate and a breakdown by error type. An error rate near zero would move this to ❌. A substantial and clinically consequential error rate, replicated, would move it to ✅. The study is entirely feasible and, to this book's knowledge, has not been done.

Pause on what just happened. That was a claim about society — about how people talk and what follows — and it received exactly the treatment a pharmacology claim receives: a population, an endpoint, a state of evidence, and a named falsifier. The discipline does not get suspended because the subject changed.

41.3 "Ozempic face": how a clinical observation became an insult

Chapter 8 §8.8 owns the clinical facts, and they are not in dispute. Compressed to three sentences: substantial weight reduction reduces fat throughout the body, including the discrete fat compartments of the face; the face is unusually sensitive to this because volume there is doing structural work, so its loss reads as hollowing, sagging, and apparent aging; and this happens after weight loss by any mechanism — dieting, bariatric surgery, illness, or a drug. It is a consequence of the weight change, not of the pharmacology that produced it. Rate and magnitude matter; mechanism does not.

That is the science. This chapter owns what happened to the term.

Stage one: description. A clinician noticed a presentation, needed a shorthand for it, and coined one. This is entirely ordinary and entirely legitimate. Medicine runs on descriptive shorthand — the literature is full of terms named after the thing that produced them — and a clinician who says "you may see some facial volume loss, what people are calling the face thing" is communicating efficiently with a patient about a real and predictable effect. At this stage the term has a referent, a context, and a purpose.

Stage two: amplification. The term was catchy, and catchy terms get picked up: it is short, it pairs a familiar brand with a body part, and it fits in a headline. Nothing sinister happened here either. But the term left the room where it had a context. In the clinic it described a patient in front of you whose history you knew. In a headline it described nobody in particular, which meant it could describe anybody.

Stage three: generalization. Once a term describes anybody, people start applying it. And here is where the failure occurs: applying it required a diagnostic inference from a photograph. The inference is: this person's face looks hollow, hollowing is associated with rapid weight loss, rapid weight loss is associated with these drugs, therefore this person is taking one. Each step in that chain is a probabilistic association being run as if it were an identification, and the chain terminates in a confident claim about a specific person's medical treatment.

Stage four: accusation. The final step took no effort at all, because the inference already carried an implied charge. To say that someone has "the face" is not merely to describe their appearance. It is to assert that their appearance is the result of a drug, and — nearly always — that they have not said so. It alleges a concealment. That is why the term stings in a way that "they've lost weight" does not, and why it is deployed most often in contexts where the point is to diminish rather than to describe.

Stage five: productization. The loop closes when clinics begin marketing treatments for the thing the term names. At that point the term has become a market category, which gives a commercial interest in its continued circulation. A word that sells something does not go away quietly.

Stage six: the productive suffix. The construction generalized. Attach the brand name to almost any body part or attribute and you get a coinage that will be understood immediately. Read that productivity correctly: the suffix spread because it is socially useful, not because each new coinage names a distinct clinical entity. Some of what it has been attached to corresponds to real consequences of rapid weight loss. Some does not. The grammar does not distinguish between them, and neither does the reader.

Now look at the whole sequence at once, because the structure is the lesson:

A real clinical observation was converted into a tool for public speculation about individuals.

The conversion is nearly invisible, and that is what makes it effective. The term keeps its clinical-sounding surface the entire way down. It sounds like a diagnosis. It functions as a guess. It lands as an accusation. Nothing in the phrase itself tells you which of those three things is happening in any given use, so the authority of the first attaches to the second and the third for free.

Three uses, then, and they are not the same act:

1. A clinician describing a presentation in a patient they are treating. Legitimate. There is an examination, a history, a known intervention, and a purpose.
2. A person forming a private hypothesis about a stranger. Not a diagnosis. No examination, no history, no reference standard, and — as §41.5 will show — no possible way to find out.
3. A person asserting that hypothesis publicly about a named individual. An accusation, which imposes a cost on someone who did not consent to the conversation and cannot exit it.

🔬 Read the Study — the study that has not been run

Suppose someone insists that facial appearance really does identify use. That is a testable claim, and medicine has a well-developed machinery for testing exactly this kind of claim: the diagnostic accuracy study. Its reporting standards are formalized — the STARD statement (Standards for Reporting of Diagnostic Accuracy Studies) has been the reference since 2003 and was updated in 2015 — and they tell you precisely what such a study would need.

The index test: the thing being evaluated. Here, a person looking at a photograph and judging whether the subject is using a GLP-1 receptor agonist.

The reference standard: the independent method that establishes the truth. Here, verified prescription or dispensing records, or confirmed disclosure. This is the step that kills the project outside a research setting. You cannot validate a visual test without knowing who is actually taking the drug — which requires exactly the disclosure that speculation exists to substitute for.

The sample: subjects representative of the population where the test would be used, with users and non-users at realistic proportions — not a curated set of dramatic before-and-after pairs.

Blinding: raters must not know the reference standard result. Prespecified thresholds, and reported sensitivity, specificity, and predictive values with confidence intervals.

To this book's knowledge, no such study exists for visual identification of GLP-1 receptor agonist use. If one is published, this chapter's rating changes accordingly, in whichever direction the numbers point. That is not a rhetorical flourish. It is the falsifier, stated in advance, exactly as Chapter 5 requires.

There is one more thing to say, and it is arithmetic rather than ethics. Even a good visual test would perform badly in the setting where it is actually used, because of base rates. Suppose — purely as an illustration, and these numbers are invented for the demonstration and correspond to no data — that a rater could correctly flag 80% of users and correctly clear 80% of non-users. Now apply that test to a room in which 10% of people are using. Out of a hundred people, ten are users and the test flags eight of them; ninety are not users and the test wrongly flags eighteen. Twenty-six flags, eight of them right. Most of the accusations are wrong even when the test is good, and the reason is not the test's quality but the rarity of what it is looking for. This is the same structure as screening for any uncommon condition, and it is why screening tests come with confirmation steps. Public speculation has no confirmation step.

📊 Evidence Rating

Claim: GLP-1 receptor agonist use can be identified in an individual by looking at their face.

Rating: ❌ Hype outpaces evidence (as of this writing)

Reason: Facial volume loss follows substantial weight reduction regardless of mechanism, and has many other causes — aging, dental changes, chronic illness, dehydration, endurance training, prior cosmetic procedures, medications unrelated to this class, and ordinary variation in how a face photographs under a given lens and light. There is no published diagnostic accuracy study of visual identification: no reported sensitivity, no specificity, no reference standard, no representative sample. And even under generous assumed accuracy, low base rates guarantee that most positive judgments in ordinary settings are false. This is a testable claim, and on the evidence available it fails.

What would change it: A STARD-compliant diagnostic accuracy study, in a representative sample, with verified prescription records as the reference standard, reporting sensitivity and specificity with confidence intervals. Sustained accuracy high enough to overcome plausible base rates would move this rating. Nothing short of that should.

Note carefully what this rating does and does not say. It does not say the underlying physiology is fake — Chapter 8 established that it is real. It says that the inferential claim, the one that runs backward from a face to a prescription, is unsupported. Rule 6 of the rating system is doing work here: one phenomenon, two claims, two ratings. "Substantial weight loss causes facial volume loss" is well supported. "Facial volume loss identifies drug use" is not. These are different claims and they are routinely stated as if they were one.

41.4 Disclosure, non-disclosure, and what a stranger owes you

Now the hardest section, and the one where the temptation to be glib is strongest.

The question: do people in public life have an obligation to disclose that they are using these drugs? It is a real question, asked in good faith by thoughtful people, and it deserves both sides presented at full strength rather than one side presented and the other characterized. So here they are.

The case for a disclosure obligation

Norms are set by visible examples. This is the strongest argument and it does not depend on any claim about individual virtue. When a highly visible transformation is presented without explanation, audiences infer a cause, and the cause they infer is usually effort. People then pursue that benchmark by the means they think produced it — which, when the actual means was pharmacological, can mean severe restriction and the cycle of failure and self-blame that Chapter 8's biology exists to interrupt. The harm is not to the person who stayed silent. It is to strangers who calibrated against a false reading.

It reinforces exactly the frame this book spent chapters dismantling. If effective pharmacological treatment is invisible while visible transformation is attributed to willpower, then the public evidence base — the thing most people actually reason from — systematically overrepresents willpower as a mechanism. That props up the moral model of body weight that the biology contradicts.

Commercial entanglement changes the calculus, and this is the sharpest version of the argument. If a person is selling a program, a supplement, a training method, or an aesthetic, and their own results are the implied proof of the product, then silence about a different cause is functioning as a claim about the product. A testimonial does not stop being a testimonial because it is delivered by implication rather than in words.

Audiences make decisions with this information. Money, time, and health behavior get allocated on the basis of what people believe worked. Information asymmetry between a person with an audience and the audience itself is the standard reason disclosure norms exist in advertising, medicine, and finance.

The case against any obligation

Medical privacy is a default, not a reward. It is not granted to people who behave well and withdrawn from people who become interesting. Every argument for an obligation has to explain why this medical information becomes public property on the achievement of visibility, and there is no clean answer. The information at issue is a prescription, among the most protected categories of personal information in every developed legal system.

The demand is applied selectively, and the pattern is not random. It falls hardest on women, and on people whose bodies have already been treated as public commentary material. It is made about weight and very rarely about the many other medical interventions with visible aesthetic consequences that nobody thinks require an announcement. An obligation enforced this unevenly is not an obligation; it is a license, issued selectively, to demand an answer from some people and not others.

Disclosure invites the abuse it is supposed to prevent. The observed reward for disclosing is frequently a fresh round of commentary — about the decision, the body, the timing, whether it counts. If the predictable consequence of complying is harassment, then the obligation is an obligation to accept harassment.

There is no stopping rule. Grant that a drug requires disclosure. Then what about a surgical procedure? A thyroid condition? A chef? A trainer? A schedule with four free hours in it? An inherited metabolic profile? Every one of those affects the achievability of a visible outcome, and most affect it more than a drug does. The line at "pharmacological" is intuitive and it is not principled — once drawn, it has to be defended against every adjacent case, and it cannot be.

And the plain version, which is not a small argument merely because it is short: nobody owes an explanation for their own body. Not to you, not to an audience, not to a comment section. The default is not silence-pending-justification. The default is that it is theirs.

Where this book lands

It declines to impose a duty. The arguments on both sides are serious, they engage each other honestly, and this is not a question a peptide textbook is positioned to settle. A reader who concludes there is an obligation has not made an error, and neither has a reader who concludes there is not.

But there is a distinction the book can draw, because it falls squarely in its lane:

Saying nothing and saying something false are not the same act.

A person who has not discussed their medical treatment has exercised a right every argument above concedes they have. There is nothing to evaluate, because no claim was made. Silence is not a proposition.

A person who actively attributes their results to a specific cause that was not the cause has done something categorically different. They have made a claim about how the world works — that this regimen produces this result — and claims are what this book is about. That is no longer a privacy question but an accuracy question, subject to the same evaluation as any other efficacy claim: what population, what endpoint, what evidence, and what would change our mind. A commercial interest strengthens the case for scrutiny in the usual way, which Chapter 42 takes up in detail.

That distinction lets you hold a firm position without demanding anything of anyone. You need not know what any individual is taking. You need only decline to accept an unevidenced efficacy claim — from anybody, about anything — which is the position this book has held since Chapter 5.

📊 Evidence Rating

Claim: Public figures have an ethical obligation to disclose their use of GLP-1 receptor agonists.

Rating: NOT RATED — this is a values question, not an empirical one.

Reason: The rating system measures the state of evidence supporting a claim about the world. It requires a population, an endpoint, and a study design that could in principle settle the matter. "Ought" claims have none of these. There is no trial that could show an obligation exists, because obligations are not the kind of thing a trial detects. Rating this ❌ would be a category error — it would mean "the evidence does not support it," when the correct statement is "evidence is not the relevant instrument."

Why this entry is here anyway: because the most common way an ethical claim gets smuggled past a careful reader is by wearing an empirical costume. Watch for the substitution. "Public figures should disclose" is a values claim and is not rated. "Disclosure by public figures reduces harmful weight-control behavior in their audiences" is an empirical claim, is rateable, would currently rate 🔬 for want of any study designed to test it, and is a different sentence entirely. The two get used interchangeably in argument, and they are not interchangeable. When you meet a claim, the first question is not whether it is true. It is what kind of claim it is.

41.5 The speculation economy

Speculating about who is using is content. That sentence explains most of what follows.

It is very cheap to produce: no reporting, no expertise, no access, no verification, and the raw material is a photograph that already exists. It is highly engaging, because it is a puzzle, and puzzles invite participation. It is socially safe, because it can be delivered as a question or an observation rather than an assertion, which leaves a retreat if challenged. And it has no correction mechanism, because nothing ever resolves.

Now the economics, which are the point of the section title. The cost of the speculation is not borne by the speculator. It is borne by the subject, who did not choose to participate, cannot respond without amplifying, and cannot exit. That is a textbook externality: a small benefit to many producers, a concentrated cost on one person, and nothing transferring one to the other. Systems built that way reliably overproduce.

You can make the moral argument from there, and many people have. This book has a different one to add, and it is the one that belongs in a book about evidence.

The speculation is not a claim

Here is the test Chapter 5 taught you: name what would change your mind. If you cannot, you are not holding a belief about the world. You are holding a stance.

Apply it.

WHY THE SPECULATION CANNOT BE WRONG

  OBSERVATION                          HOW IT GETS READ
  ────────────────────────────────────────────────────────────────
  Rapid weight loss                →   "obviously on it"
  Gradual weight loss              →   "on it, just being subtle about it"
  No weight change                 →   "on it, maintaining"
  Weight regain                    →   "came off it"
  Gaunt face                       →   "there's the face"
  Full face                        →   "fillers, to cover the face"
  Denies use                       →   "well, of course they'd deny it"
  Confirms use                     →   "knew it"
  Says nothing at all              →   "silence is an answer"

  Nine observations. Nine different states of the world. One conclusion.

  No possible observation counts AGAINST the hypothesis. Which, by the
  standard this book has used since Chapter 5, means there is no
  hypothesis here. There is a stance with a vocabulary attached.

Read that table twice, because its structure should be familiar. It is the same structure as "the peptide is working, your body just hasn't shown it yet" — the claim that absorbs every result. It is the same structure as "the trial failed because the dose was wrong, the population was wrong, and the endpoint was wrong" offered after the fact with no prespecification. Chapter 5 flagged that pattern as the single most reliable signal that you are looking at something other than a claim, and the pattern does not become respectable because the subject changed from a molecule to a person.

This matters for a practical reason. It is the strongest available objection, and it is an epistemic objection rather than a moral one. A reader who believes that public figures forfeit some privacy, who finds the whole thing harmless, who thinks the moral argument is precious — that reader still has to concede this one, because it does not ask them to share any value. It only asks them to apply the same standard they would apply to a supplement advertisement. You do not know. There is no observation available to you that would tell you. And a confident assertion built on no possible observation is the exact epistemic vice this book exists to correct.

Consistency is the force of the argument. It would be strange to spend forty chapters refusing unfalsifiable claims about molecules and then accept them about people.

The hedges do not rescue it

Three retreats show up predictably, and none of them works.

"I'm not saying they are, I'm just saying it's likely." Probabilistic claims require inputs: a base rate for use in the relevant population, and a likelihood ratio for the observation. Neither exists. You cannot perform a Bayesian update with two numbers you invented, and adding "likely" to a guess does not convert it into an estimate.

"Everyone knows." Common knowledge describes how widely a belief is held, which is silent on whether it is true. Chapter 6 covered how beliefs achieve wide circulation without passing through a verification step; this is that mechanism at full speed.

"It's obvious from the pictures." Addressed in §41.3. Absent a validated index test and a reference standard, this is a statement about the confidence of the viewer, not the state of the subject.

🔍 Check Your Understanding

1. A person says: "I'm not accusing anyone — I'm just noting a pattern." What would that pattern have to consist of, methodologically, before it supported an inference about any individual?
2. Why does a denial function as confirming evidence in this system, and what does that tell you about the system?
3. Give one example, from earlier in this book, of a pharmacological claim that has the same structure as the table above — every observation confirms, none disconfirms.
4. The epistemic objection in this section is offered as stronger than the moral one. Stronger in what specific sense? (Hint: it is not about which is more important.)

41.6 Jokes as an information vector

Chapter 6 built the compression ladder: a finding starts as a trial report full of qualifiers and descends through guideline, press release, headline, and post, losing a little more of what made it meaningful at every rung. Compressions are where qualifiers die. That was the chapter's thesis and it holds.

This chapter adds one rung below the bottom, and it behaves differently from all the others.

THE COMPRESSION LADDER — Chapter 6's version, with one rung added below

  TRIAL REPORT     population, endpoint, effect size, confidence interval,
       │           dropouts, adverse events, funding, prespecification
       ↓
  GUIDELINE        population, endpoint, strength of recommendation
       │
       ↓
  PRESS RELEASE    effect size, selected endpoint, no interval
       │
       ↓
  NEWS HEADLINE    direction of effect
       │
       ↓
  SOCIAL POST      claim plus affect
       │
       ↓
  JOKE             premise only — and the premise is never stated,
                   because the joke depends on the audience supplying it

  Every rung down loses qualifiers. The joke is the only rung that also
  loses the CLAIM. Nothing is asserted, so nothing can be checked — and
  yet something has been transmitted, and transmitted further and faster
  than any rung above it.

Four properties make jokes an unusual carrier, and they compound.

A joke cannot carry a population or an endpoint. It has a word budget of roughly a sentence, and the entire budget goes to the setup and the turn. There is no room for "in adults with a BMI above a threshold, over sixty-eight weeks, on a co-primary endpoint of percent weight change." A joke that tried would not be a joke. This is not a failure of the comedian; it is the form.

Jokes are repeated, not evaluated. This is the crucial asymmetry. When you meet a claim, the available responses include agreement, disagreement, and a request for evidence. When you meet a joke, the available responses are laughing, not laughing, and passing it on. Evaluating a joke is a social violation — the response to "is that actually true?" is "it's a joke" — which means the joke format immunizes its own payload. Whatever premise rode in on the joke enters the audience without ever being submitted for inspection.

A joke requires the audience to already hold the premise. This is what makes the form so efficient. Jokes work by shared assumption: the listener supplies the missing piece for the turn to land. So a joke does two things at once — it tests whether the premise is shared, and when the laugh comes it confirms to everyone present that it is. Laughter is a public receipt that a premise is common ground. No assertion, no argument, no evidence, and a room full of people who now know that everyone else assumes it too.

And so a joke establishes a social fact. This is the thing a straight claim cannot do. A joke about a drug does not tell you what fraction of people use it; it tells you — or rather, makes true — that the drug is common enough to be a reference, that using it is a recognizable and namable thing, and that there is an agreed attitude toward it. That social fact is entirely independent of any evidence about prevalence. A drug can become "the thing everybody is on" as shared understanding while the underlying numbers remain unknown to everyone in the room, including whoever made the joke.

Social facts have consequences. A belief that everyone is doing something changes what people ask their doctors for, what they assume about strangers, and how much shame attaches to a decision. It also changes what people conclude about themselves: if the premise is that the outcome is universally available in a pen, then not having the outcome becomes a choice again, and we are back in the moral model by a new route.

There is a second payload, and it usually rides along unnoticed. Most jokes in this territory carry not just this is common but an attitude — that it is a shortcut, that it is a little shameful, that the person using it opted out of something. That is a moral claim, delivered without argument, immunized by the format, and confirmed by laughter: a remarkably efficient delivery system for a position nobody had to defend.

⚠️ Hype Check — "everyone is on it"

The claim, in its usual forms:

"Half of Hollywood is on it." "Everyone at that party was on it." "You can't trust any before-and-after anymore — everyone's on it."

What's true in it. Uptake has been substantial and rapid by any reasonable reading; these are among the most-prescribed drugs of their moment, and there are populations and settings where use is genuinely common. Something real is being pointed at.

Where it fails. Three places.

First, "everyone" is not a quantity. It is an intensifier. It resists correction because it was never precise enough to be wrong, which is the §41.5 problem in miniature.

Second, the claim's evidence is almost always the social fact rather than any measurement. People believe it is everywhere because they have heard it referenced everywhere, and references are generated by the belief. That loop can run indefinitely without contacting a prescription figure.

Third, the claim does work its speaker rarely intends. "Everyone is on it" licenses individual speculation ("so probably them too"), flattens the difference between approved products and everything else circulating under the same name (§41.2), and feeds the framing that §41.8 shows falls hardest on people using these drugs for the indications they were approved for.

Verdict: the claim mistakes cultural salience for prevalence. Those are different measurements, and only one of them is being made. A true statement about prevalence comes from dispensing data with a stated population and period. Salience is measured by how often you hear about something — which is exactly what this chapter is about.

None of this is an argument against jokes. Comedy about medicine is ancient, it is frequently the first way a taboo subject becomes discussable at all, and §41.7 gives that its due. People are allowed to be funny about a drug. The error is not making the joke; the error is receiving a joke as information. Those are separable, and separating them is a skill.

41.7 What the culture gets right

A chapter that only criticized would be unfair, and — more to the point for a book like this one — it would be wrong. The public attention produced real goods. Several of them are goods that decades of patient, careful, evidence-based communication had failed to produce, which is a genuinely uncomfortable finding and should be stated plainly rather than buried.

It made a stigmatized condition discussable. For most of living memory, obesity was discussed in one of two registers: as a moral failing, or not at all. In a remarkably short period it became an ordinary topic of ordinary conversation — in workplaces, at dinner tables, on daytime television. That change is real and large, and whatever else the culture did, it did that.

It taught a very large population that body weight has biological determinants. This is the one that matters most. The idea that appetite is a regulated signal — that hunger is generated by identifiable molecules acting at identifiable receptors, that the system defends a weight, and that two people doing the same thing can get different results because their signaling differs — sat in the research literature and a few specialist clinics for decades. A great many people now understand it roughly and correctly. The mechanism of that transfer was not a public health campaign. It was a drug that visibly worked, and the question of why it worked. Chapter 8 is the answer; the culture supplied the question to a bigger audience than any textbook will.

It surfaced access and cost as political questions. Chapter 12 covers the substance — list versus net prices, formulary exclusions, coverage tied to indication rather than molecule, the gap between a drug existing and a person obtaining it. That machinery used to be understood mainly by people who worked inside it. It is now argued about by people who have never used the word "formulary" in their lives, and that argument is legitimate: these are distributive questions and they belong in public.

It generated pressure that appears to have contributed to coverage changes. Note the hedging; it is doing real work. Coverage decisions moved, and public attention was one input among several, alongside accumulating outcome data, payer negotiation, and manufacturer pricing decisions. Attributing the change to attention specifically is not something this book can do cleanly.

And, in a small irony, it created the occasion to fix its own confusion. The brand-name collapse of §41.2 is a genuine problem — and also the reason a large number of people now know that a molecule and a brand are different things, and that approval is granted for an indication rather than a substance. That is real pharmacological literacy, produced by the very confusion this chapter has been complaining about. Both things are true.

📊 Evidence Rating

Claim: Public attention to GLP-1 receptor agonists has increased appropriate access to these drugs for people who meet approved indications.

Rating: ⚠️ Promising but preliminary (as of this writing)

Reason: Plausible, partly supported by the observed direction of coverage changes and by a large increase in clinical conversations that would not otherwise have happened — but genuinely confounded. The same attention drove demand that contributed to supply constraints, which reduced access for other patients (§41.8), and drove a parallel unregulated market whose products are not the approved drugs at all. "Appropriate access" is also under-defined: access for whom, by what criterion, measured how? Net effect is not cleanly quantified, and it is entirely possible that it is positive for one population and negative for another simultaneously.

What would change it: Population-level analyses of initiation and continuation rates stratified by indication, compared against modeled counterfactuals, over a period long enough to span the supply constraint. Evidence that initiation among people meeting approved criteria rose faster than among people who do not, and that continuation was not disrupted by shortage, would move this toward ✅. The reverse pattern would move it toward ❌.

That rating is genuinely uncertain, and the uncertainty is not a hedge. Two of this chapter's own sections point in opposite directions — §41.7 lists the goods and §41.8 lists the harms, and both lists are accurate. A rating that resolved that tension by picking a side would be misrepresenting what is known.

41.8 The second-order effects on people who need these drugs

Everything in this section is a harm that flows from the cultural framing rather than from the pharmacology. That distinction is the point. None of it would show up in a trial, because trials study molecules and these effects are produced by words.

Shortage. Demand rose faster than manufacturing capacity, and for a period supply of certain products was constrained. Chapter 12 covers the mechanics — injector-device capacity, fill-finish constraints, and the long lead times of biologic-scale manufacturing. The point here is the distributional one: when supply is constrained, the people affected include those using the drug for the indication it was approved for, and reports of people with type 2 diabetes unable to fill established prescriptions were widespread during that period. Shortage causes are always multiple, so it would be wrong to lay it entirely at the door of demand outside approved indications — and equally wrong to pretend that culturally generated demand was not one of the inputs.

Coverage arguments got harder, not easier. This is the most consequential and the least visible. Payers respond not only to evidence but to how a treatment is categorized. A drug understood as treatment for a chronic disease with documented outcomes occupies one position in that argument; a drug understood as a cosmetic convenience occupies another, and the second is far easier to exclude, cap, or bury in criteria. Every cultural framing of these drugs as a vanity product strengthens the hand of anyone arguing they should not be covered — including for people whose need is unambiguous by any clinical standard. That is a real cost, imposed by a framing, on people who had no part in creating it.

The justification tax. People using these drugs for approved indications report being asked to explain themselves — at the pharmacy counter, at work, at family gatherings, in the comments under photographs posted for other reasons. Somebody managing type 2 diabetes should not have to conduct a defense of their treatment at a dinner table, and yet the collapse of a whole category into one word means anyone taking any of it is presumed to be taking it for the reason the culture finds most interesting. That is a social cost, paid in the currency of ordinary daily life.

The collapse runs further than most people realize. When a class of injectable medicine is absorbed into one brand name, the boundaries of the class stop being visible, and injectable medicines that predate this one by decades are now sometimes read by onlookers as members of it. A pen device on a restaurant table used to be invisible. Increasingly it is not.

The gray-market pull. When cultural demand exceeds what supply and coverage can meet, some of the excess routes around the system — standard behavior for any constrained market. The destination is precisely the environment Chapters 19 and 34 document: products of unverified identity, concentration, and sterility, sold under a brand name that belongs to none of them. The cultural demand created a market the culture has no way to inspect, and the brand name that generates the demand also supplies the borrowed credibility. That is how a word becomes a safety problem.

And the accusation runs both ways. A person who loses weight without any drug is now, in some settings, presumed to have used one, and finds themselves defending a negative. A person who used one is accused of a shortcut. Both accusations trace to the same habit, and the habit does not exempt anybody. There is no version of a body that is currently safe from a diagnostic guess, which is a strange thing to have to write and is, as far as this book can tell, accurate.

🩺 Safety and Risk — what the naming collapse does to a safety system

Drug safety monitoring depends on attribution. When someone experiences an adverse event, the system's ability to detect a real signal depends on correctly linking that event to a specific product, at a specific dose, from a specific manufacturer, in a person with a specific indication. That is how a genuine signal is distinguished from background noise, and it is how a problem with one product is distinguished from a problem with a class.

The naming collapse degrades every link in that chain.

A patient who reports "Ozempic" may be reporting any of five approved products, an oral rather than an injectable, a diabetes rather than a weight-management product — or a compounded or gray-market preparation that is not any of them. An event attributed to an approved product but caused by an unverified vial contaminates the record for the approved product. An event attributed to one manufacturer's molecule but caused by another's is a misattributed signal.

The direction of the error is not random. Because the famous brand absorbs the ambiguity, it absorbs adverse events that do not belong to it and lends its safety record to products that have not earned it. Both errors run toward the brand.

The correction is unglamorous and entirely within reach: say the molecule, say the product, say the route. If you are reporting a side effect to a clinician, to a pharmacist, or to a regulatory reporting system, that specificity is the single most valuable thing you can contribute. And as always, decisions about any of this belong with a clinician who knows your history, not with a book.

41.9 What a reader can do

Modest and concrete. None of this requires you to correct anybody, and one item explicitly asks you not to.

FIVE PRACTICES

  1. SAY THE MOLECULE          semaglutide, tirzepatide, liraglutide
                               one extra syllable; removes a whole error class

  2. SEPARATE DRUG FROM        "semaglutide, at the weight-management
     INDICATION                 product's label, for obesity" is a different
                                sentence from a brand name on its own

  3. DECLINE TO SPECULATE      not because it is rude — because it is not a
     ABOUT INDIVIDUALS          claim, and you have no way to be right

  4. NOTICE THE JOKE           ask: if this were stated flatly, as an
     DOING A CLAIM'S WORK       assertion, would I accept it?

  5. ASK HOW WE WOULD KNOW     the Chapter 5 question, applied to cultural
                               assertions exactly as to pharmacological ones

Each deserves a note.

Use generic names. The highest-yield habit in the chapter, and it costs almost nothing. Say semaglutide, tirzepatide, liraglutide; if you need the class, say GLP-1 receptor agonist. The stems from Chapter 1 §1.8 make this easier than it looks — once -glutide means something to you, the generic names stop being a wall of syllables and start being informative. The payoff is immediate: a medication history that says the right thing, and a coverage conversation conducted in the payer's actual categories.

Distinguish the drug from the indication. "Is Ozempic covered?" is not answerable. "Is this product covered for this indication under this plan?" is. That distinction settles a startling number of arguments, most of which turn out to be two people using one word for two different products approved for two different things.

Decline to speculate about individuals. Note the grounds. Not squeamishness — the grounds are epistemic: per §41.5, no observation available to you would settle it, so the belief you would be forming is not a belief about the world. Declining to hold beliefs you cannot in principle check is not a moral posture. It is the method, applied consistently.

Notice when a joke is doing a claim's work. One question: if this were stated flatly, as an assertion, would I accept it? Usually the flat version is something you would want evidence for and would not get. That does not mean the joke was bad. It means what you absorbed was a premise rather than a finding, and you can now hold it at the appropriate distance.

Ask how we would know. Chapter 5's question does not stop applying at the edge of the laboratory. "Everyone is on it" — how would we know? "It's changing the food industry" — how would we know, and has anybody measured it? Chapter 44 takes that seriously and finds the honest answer is mostly "not yet." The question is not hostile. It is the same one you would ask about a supplement, and the argument of this whole part is that cultural claims get no exemption.

And one thing not to do: do not become the person who audits everyone's drug names at parties. The goal is accuracy in your own reasoning and precision where precision changes an outcome — with a clinician, a pharmacist, an insurer, or someone who asked you a real question. Correcting a stranger's word choice in casual conversation achieves nothing except to make precision look unpleasant, and precision has enough enemies.

🔍 Check Your Understanding

1. Someone tells you their sister is "on Ozempic for weight loss." Name three distinct things you still do not know, and state which of them would matter to a clinician.
2. You see a claim that "these drugs have transformed the restaurant industry." Write the version of that claim that could be rated: give it a population, an endpoint, and a source that could in principle supply the data.
3. Why does this section ground the refusal to speculate in epistemics rather than in courtesy? What does that buy you in an argument with someone who does not share your moral intuitions?

📋 Your Evidence Dossier

Your dossier is finished. This is the field test.

By Chapter 40 you had complete entries on five to ten compounds — identity, mechanism, evidence, rating, falsifier, and the rest. Every one of those entries was built from the literature and from sources you chose deliberately. Part VIII asks the opposite question: what does the world say about your compound when you are not choosing the sources?

The exercise

Pick one entry. Choose the one that has the most cultural presence, which for most readers will be a GLP-1 receptor agonist, but any entry with public visibility works — a cosmetic peptide and a performance peptide both generate plenty of material, and the exercise is arguably sharper on a compound with less evidence behind it.

Find four descriptions of it, one from each of these:

1. An advertisement — from a telehealth service, a clinic, a compounding pharmacy, or a supplement seller. Anything where somebody is paying for your attention.
2. A video — short-form or long-form, from a creator rather than an institution.
3. A comment thread — under any of the above, or on a forum. Take the whole thread's consensus, not one comment.
4. A news article — from a general outlet, not a trade or scientific publication.

Now open Fields 5 and 6 of your own entry — the fields where you recorded what the evidence actually shows and the verdict you drew from it, with its falsifier (Appendix C has the blank workbook). For each of the four sources, record the departure.

CULTURAL DRIFT LOG — one compound, four sources

  COMPOUND: ______________________

  SOURCE          WHAT IT CLAIMS      DEPARTURE FROM     DIRECTION    WHAT IS
                                      FIELDS 5 / 6                    MISSING
  ─────────────────────────────────────────────────────────────────────────────
  Advertisement

  Video

  Comment thread

  News article

  DIRECTION is the column that teaches. Mark each one:
    ↑  INFLATES   — claims more certainty, more effect, or more breadth
    ↓  DEFLATES   — dismisses evidence that exists, or overstates risk
    ↔  DISTORTS   — roughly the right confidence, wrong population or endpoint
    =  ACCURATE   — and note it when it happens, because it does

  WHAT IS MISSING is usually more informative than what is claimed.
  Check specifically for: the population, the endpoint, the comparator,
  the duration, the harms, and who paid.

What to look for

The direction is not always up, which is why the log has four arrows rather than two. Advertisements inflate, reliably. Comment threads frequently deflate — dismissing well-supported findings, or holding an approved drug to a standard no drug could meet. News articles most often distort: broadly the right confidence level, attached to the wrong population. And every so often a source is simply accurate, worth recording precisely because you did not expect it.

Then ask the harder question. Which source came closest to your Field 5? If it was not the one you expected, that is the finding. And if none came close — if all four departed in different directions — notice that a person without your dossier would have had to choose among them with no way to tell which was which. That person is who you were before Chapter 1.

Optional extension

Do the log for a compound whose Field 6 rating is ❌ or 🔬. Strong evidence gets distorted in predictable ways; thin evidence gets distorted in far more interesting ones, because nothing anchors any of the descriptions. Watching four sources drift freely away from an empty evidence base is the clearest demonstration available of why the dossier was worth building.

Conclusion

A brand name climbed out of its own category, and it took some things with it.

It took the distinction between a molecule and a product — the distinction that determines what you can be prescribed and what will be covered. It took the difference between five approved products and an unregulated market that borrows their name. It took a real clinical observation about facial volume and converted it into an instrument for guessing about strangers, invisibly, because the term never stopped sounding clinical. And it took a set of moral positions about who deserves what and delivered them through jokes, where they could not be examined.

It also did good. It made a stigmatized condition ordinary to discuss. It moved a piece of real biology — that appetite is a regulated signal, not a referendum on character — into general understanding faster than any deliberate effort had managed in decades. It put access and cost into public argument, where distributive questions belong. Both columns are accurate, and a chapter that reported only one would be doing to the culture exactly what the culture did to the science.

The method survives all of it, which is the point of Part VIII. Cultural claims have populations and endpoints, and they can be rated. That you can identify use from a face is testable and fails, ❌, on grounds you can state in one sentence about base rates. That attention improved access is ⚠️, genuinely uncertain and entangled with a shortage. That genericization causes clinical harm is ⚠️, mechanistically clear and unmeasured. And that anyone owes you a disclosure is not rated at all, because it is not that kind of claim — and knowing which kind of claim you are holding is the most portable skill in this chapter.

The strongest thing here is also the simplest. Speculation about an individual's drug use is not merely unkind. It is unfalsifiable, and by this book's standard that means it is not a claim at all. Denials confirm it, admissions confirm it, silence confirms it, and every possible state of a person's body confirms it. You have spent forty chapters learning to recognize that shape in claims about molecules. It is the same shape.

Chapter 42 follows the money through the pipeline this chapter has been describing — who is paid at each step between a video and a prescription, and how platform incentives determine which claims get made at all. Chapter 6 explained how information degrades in transit. Chapter 42 explains who profits from the degradation.

Key Terms

Genericization — the process by which a brand name becomes the common noun for an entire product category. Aspirin, escalator, and thermos completed it.

Genericide — the trademark-law endpoint of genericization, in which a mark becomes the common name for its category and loses legal protection.

Brand name — the commercial name of a specific product: a molecule, in a formulation, by a route, approved for an indication. One molecule may carry several, and the products are not interchangeable.

Generic name — the single internationally assigned name for a molecule, identical across countries and across every product containing it (Chapter 1 §1.8).

Indication — the condition a product is approved to treat. Approval and coverage attach to a product-plus-indication pair, never to a molecule alone.

Off-label use — prescribing a product for a condition other than its approved indication. Legal in most jurisdictions, common, and a frequent source of coverage disputes.

Facial volume loss — reduction of fat in the facial compartments following substantial weight loss, producing hollowing and apparent aging. Occurs after weight loss by any mechanism (Chapter 8 §8.8).

Index test — in a diagnostic accuracy study, the test being evaluated. Visual inspection of a face would be the index test in §41.3's study.

Reference standard — the independent method establishing the truth against which an index test is judged. Its absence is what makes visual identification of drug use untestable outside research.

Diagnostic accuracy — how well a test identifies the presence or absence of a condition, reported as sensitivity, specificity, and predictive values. Never one number, never meaningful without a stated population.

Base rate — the underlying frequency of a condition in the population tested. Low base rates guarantee that most positive results from an imperfect test are false.

Unfalsifiable claim — an assertion for which no possible observation would count against it. By Chapter 5's standard, not a claim about the world.

Social fact — something true because a group treats it as true, independent of any measurement. "This drug is everywhere" can be a social fact while prevalence remains unknown to everyone asserting it.

Compression — reduction of a claim to a shorter form, losing qualifiers at each step. Chapter 6's central concept; a joke is its terminal case.

Disclosure norm — a social expectation that certain information be volunteered. §41.4 presents both sides and declines to impose one.

Misattribution — assigning a result to a cause that did not produce it. Silence makes no claim; misattribution makes a false one.

Second-order effect — a consequence that follows from the response to a thing rather than from the thing itself. §41.8's harms flow from how the drugs are talked about, not from what they do.

Formulary — the list of products a payer covers, with criteria attached. Organized by product and indication, which is why brand-name confusion has financial consequences (Chapter 12).

Spaced Review

1. (Ch 41 + Ch 8) A friend says: "You can always tell — they get that hollow look in the face." Using Chapter 8's physiology and this chapter's §41.3, separate the part of that sentence that is supported from the part that is not, and state what evidence would be required to support the second part.

2. (Ch 41 + Ch 12) A patient with type 2 diabetes is told their prescription cannot be filled, and concludes that their insurer has stopped covering "Ozempic." Name at least three distinct things that could actually be going on, and explain which of them the brand-name shorthand is preventing them from articulating.

3. (Ch 41 + Ch 6) Chapter 6 showed that qualifiers die at each step of compression. §41.6 argues that a joke does something additional that no other rung of the ladder does. What is it, and why does the joke format make its payload unusually difficult to evaluate?

4. (Ch 41) Rewrite the claim "public figures should disclose their use of these drugs" as an empirical claim that this book's rating system could evaluate. State the population, the endpoint, and what study would settle it. Then explain what your rewrite lost from the original.

5. (Ch 41 + Ch 5) The table in §41.5 lists nine observations that all get read as confirming the same hypothesis. Find one claim from Part III of this book with the same structure, write out its own version of that table, and state the falsifier that its proponents will not supply.