Case Study 2 — The Graveyard of Weight-Loss Drugs

Why the field was so pessimistic, and what that context does to semaglutide's result

Type: Real, public, historical · Tier 1 withdrawal facts, Tier 2 detail · Relevance: §8.5, §8.7 · Context for the whole of Part II

Background: a century of failure

Semaglutide's result looks impressive on its own. It looks considerably more impressive against the history it emerged from — and that history is also the reason serious pharmacologists were pessimistic about the entire category.

A compressed account of what came before. Details vary by jurisdiction and by drug; the pattern does not.

Amphetamines and derivatives, mid-twentieth century. They suppressed appetite effectively. They also produced dependence, cardiovascular effects, and psychiatric harm. Some remain available in restricted forms; most did not survive.

Fenfluramine-phentermine ("fen-phen"), 1990s. A widely used combination that produced substantial weight loss. It was withdrawn after an association with valvular heart disease was identified — damage to heart valves, in some cases requiring surgery. The mechanism was eventually traced to serotonin receptor activity on valve tissue. This is the case that most shaped professional caution about weight-loss drugs, and it is the reason later serotonergic agents faced intense scrutiny.

Sibutramine, 2000s. A serotonin-norepinephrine reuptake inhibitor with a real weight effect. It was withdrawn in many jurisdictions after a cardiovascular outcomes trial found increased cardiovascular events in a high-risk population — a trial that was, notably, only conducted because regulators required it.

Rimonabant, 2000s. A cannabinoid receptor antagonist that produced weight loss and metabolic improvements and looked, briefly, like the answer. It was withdrawn following serious psychiatric adverse effects, including depression and suicidality. It was never approved in the United States.

And several others approved with modest efficacy, restricted labels, or subsequent constraints.

The pattern

Read across these and a structure appears.

WHY THE OLD DRUGS FAILED

  MECHANISM                      WHAT WENT WRONG
  ────────────────────────────────────────────────────────────────────────
  central monoamine              the same neurotransmitter systems govern
  manipulation                   mood, cardiovascular tone, and reward.
  (amphetamines, sibutramine)    You cannot suppress appetite through them
                                 without touching the rest.

  serotonergic action            serotonin receptors are present on heart
  (fenfluramine)                 valve tissue. Peripheral receptor expression
                                 was not appreciated until valves were damaged.

  cannabinoid receptor           the endocannabinoid system is deeply involved
  antagonism (rimonabant)        in mood and reward. Blocking it centrally
                                 produced predictable psychiatric harm — in
                                 hindsight.

  ────────────────────────────────────────────────────────────────────────
  THE COMMON FEATURE: these drugs acted on broad central systems with many
  jobs. Appetite was one output among many, and the others came along.

And a second, methodological pattern: in several cases the harm was identified only when a cardiovascular outcomes trial was finally required. Efficacy was established on the surrogate; harm appeared when someone measured the hard endpoint. Chapter 5 §5.6, twice over.

What is different about the incretin approach

Four things, and it is worth being precise because "this time is different" is a sentence with a poor track record.

The target has a specific job. The incretin system's physiological function is to respond to a meal. It is not a broad neuromodulatory system with many other responsibilities. Acting on it is inherently more constrained than acting on serotonin.

Glucose-dependence. Chapter 7 §7.4. The mechanism has a built-in safety property that none of the earlier drugs had.

Peripheral origin with a neural relay. The signal originates in the gut. Even where central action occurs, it operates through circuitry evolved to receive this specific message rather than through general neurotransmitter manipulation.

And — decisively — the hard endpoint was measured, and it went the right way. SELECT is the category's answer to its own history. Previous drugs failed when someone finally ran the cardiovascular trial. This one was run and showed benefit.

⚠️ Hype Check — "this time is different"

The case for: all four points above are real. The mechanism is genuinely more targeted, the safety property is genuinely built in, and — most importantly — the outcome trial that killed previous drugs has been run and passed.

The case for caution, which this book takes seriously:

Fen-phen's valve problem was not predicted. Nobody anticipated serotonin receptors on heart valves. The harms that end drug classes are, characteristically, the ones nobody was looking for — and looking harder does not reliably find them, because you cannot look for what you have not imagined.

Exposure duration is short. SELECT ran about three years. These are drugs intended for decades of continuous use in millions of people. Chapter 3's cadaver growth hormone case had a decades-long incubation period. Three years of data is genuinely reassuring and is not thirty.

The population using them now differs from the trial populations. Off-label use at lower BMI, in younger people, in people without comorbidity, and without medical supervision, is a different exposure from the one that was studied.

And GLP-1 receptors are more widely distributed than anyone initially appreciated — heart, kidney, vasculature, immune cells, multiple brain regions. Broad receptor distribution is how unexpected effects happen, in both directions. It is the reason for Chapter 10's expanding benefits and it is also the reason to remain attentive.

The honest position: the mechanistic reasons for optimism are real and better than any previous weight-loss drug class had. The historical base rate for "this time is different" in this specific therapeutic area is poor. Both are true, and the resolution will come from continued surveillance rather than from argument.

What this case teaches

Context changes how impressive a result is. Semaglutide's cardiovascular benefit is a good result in isolation. Against a century in which the field's drugs repeatedly reduced weight and harmed people, it is a much bigger deal — and reading it without that context understates it.

The trial that mattered was the one regulators forced. Sibutramine's harm appeared in a cardiovascular outcomes trial conducted because of a regulatory requirement. That requirement, introduced in response to earlier problems, is why we know what we know about this class too. The regulatory apparatus that Chapter 38 examines is not merely an obstacle; in this therapeutic area it is the reason the evidence exists.

Surrogates failed repeatedly, in the same direction. Every one of these drugs improved weight. Weight is a good surrogate — better than most — and it was insufficient every time.

And "the mechanism is more targeted" is a reason for optimism, not a guarantee. Chapter 2's rule applies to good news as well as bad: mechanism is necessary and never sufficient, and a well-reasoned argument for why this class should be safer is exactly the kind of argument that preceded every previous failure.

Discussion questions

1. Each failed drug acted on a broad central system. Explain why that made harm more likely, and identify what the incretin approach does differently.

2. Fen-phen's valve damage was not predicted. What does that suggest about how much confidence to place in the absence of an anticipated harm? Distinguish "we looked and found nothing" from "nobody thought to look."

3. Sibutramine's harm appeared only in a required outcomes trial. Should such trials be required for all weight-loss drugs? What does requiring them cost, and who pays that cost?

4. SELECT provides about three years of data for a drug intended for decades of use. How much reassurance is three years? Compare to Chapter 3's cadaver growth hormone case, where the incubation period was decades.

5. The Hype Check states that the historical base rate for "this time is different" in weight-loss pharmacology is poor. Does citing a base rate constitute an argument against a specific drug with its own evidence? When is base-rate reasoning appropriate and when is it lazy?

6. The population using these drugs now differs from the trial populations — off-label, lower BMI, younger, less supervised. Name three specific ways that could produce a different safety picture than the trials found, and say which you would prioritize monitoring.