Chapter 11: Insulin: The Original Peptide Drug — 100 Years of Saving Lives

"Insulin is not a cure for diabetes; it is a treatment. It enables the diabetic to burn sufficient carbohydrates." — Frederick Banting, Nobel lecture, 1925

He was right, and a century later it is still true, and that sentence is the most honest thing anyone has said about this drug.

Overview

Everything in Part II so far has been measured against modern standards of evidence: randomized trials, pre-specified endpoints, confidence intervals, estimands, data monitoring committees.

Insulin had none of that, and it is the most certainly effective drug in this book.

In January 1922, a fourteen-year-old boy in a Toronto hospital was dying of type 1 diabetes. He weighed about 65 pounds. The only treatment available was a starvation diet that extended life by months at enormous cost. He received an injection of pancreatic extract. His blood glucose fell. He received better-purified extract, and he recovered, and he lived for another thirteen years.

No control group. No randomization. No p-value. No trial.

And nobody has ever seriously questioned whether insulin works, because when a treatment reverses a uniformly fatal condition in the person in front of you, the epistemology takes care of itself.

This chapter is about that — and about everything that followed, because the century after 1922 is where insulin gets genuinely complicated. The drug is not the hard part. The dosing is. Insulin's history since its discovery is a hundred-year attempt to solve one problem: how do you replace a hormone whose correct dose changes every hour, when getting it wrong in one direction is dangerous within minutes?

And it is the chapter where this book has to talk about money, because insulin is the clearest case in medicine of a drug that costs little to make and a great deal to buy.

In this chapter, you will learn to:

• Describe the discovery and explain why it needed no trial
• Describe insulin's structure and what the disulfide bonds do
• Explain what insulin does at the cell, and why hypoglycemia defines the therapy
• Distinguish type 1 from type 2, and explain why one drug treats them differently
• Explain what recombinant production changed in 1982
• Compare the analogs and say what problem each was built to solve
• Explain why insulin does not stop working, when so many peptide therapies do
• Describe the pricing controversy accurately

Learning Paths

💊 GLP-1 — §11.4 and §11.7 are the two sections that will most improve your understanding of the drugs in Chapters 8–10, by contrast. 🔬 Science — full read; §11.2 and §11.6 are the structural and engineering core. 🏥 Clinical — mostly review, but §11.9 is worth reading for how to discuss cost with a patient without either defending or attacking anyone.

11.1 Toronto, 1921

The essential facts, which are Tier 1 and worth stating plainly.

By 1920 it was known that the pancreas produced something that controlled blood sugar, that removing it caused diabetes, and that extracts of pancreas had been tried without reliable success. The obstacle was that the pancreas also produces digestive enzymes, which destroy the very peptide anyone was trying to extract — Chapter 1 §1.3, operating as a laboratory problem before it was a pharmacological one.

At the University of Toronto in 1921, Frederick Banting and Charles Best, working in J.J.R. Macleod's laboratory, produced extracts that lowered blood glucose in depancreatized dogs. James Collip, a biochemist, developed the purification that made the extract safe enough to give to a person.

In January 1922 it was given to Leonard Thompson. The first extract was insufficiently pure and produced a sterile abscess. Collip's improved preparation worked.

The Nobel Prize followed in 1923 — awarded to Banting and Macleod, a decision that caused lasting acrimony. Banting shared his share with Best; Macleod shared his with Collip.

And they sold the patent to the University of Toronto for one dollar. Banting's stated position was that a discovery of this kind should not be a source of profit. §11.9 returns to this, because the subsequent history is not simple and the story is frequently told in a way that is.

🔬 Read the Study — the trial that never happened

text FIGURE 11.1 — "When you do not need a control group" [real historical record] THE STUDY Not a study. A series of clinical administrations of pancreatic extract to patients with type 1 diabetes, Toronto, beginning January 1922. No randomization, no control arm, no blinding, no pre-specified endpoint. THE QUESTION Does pancreatic extract reverse the metabolic derangement of type 1 diabetes? WHAT IT SHOWS Unambiguously yes. Blood glucose fell, ketosis resolved, and patients who were dying recovered — reproducibly, in patient after patient, in a condition with essentially 100% short-term mortality. WHAT IT DOESN'T It establishes nothing about optimal dose, long-term complications, comparative effectiveness, or safety over decades. All of that took the following century. THE VERDICT ✅ — and the strongest ✅ in this book, obtained without any of the apparatus Chapter 5 describes. THE LESSON Randomization exists to detect effects that could be confused with natural history, expectation, or chance. When a condition is uniformly fatal and a treatment reverses it in front of you, none of those alternatives is available. THIS IS RARE. Most treatments are not like this, which is why Chapter 5 exists — and a field that reasoned from insulin to "we do not need trials" would be catastrophically wrong.

The general principle, stated carefully: the strength of evidence required scales with the plausibility of alternative explanations. For insulin in 1922 there were none. For essentially everything else in this book, there are several.

11.2 The molecule

Insulin is 51 amino acids in two chains — an A chain of 21 and a B chain of 30 — joined by two interchain disulfide bonds, with a third disulfide within the A chain. Molecular weight about 5,800 Da.

INSULIN — two chains, three disulfides

     A CHAIN (21 residues)
     ┌─────────────────────────────────────┐
     │  ┌──── intrachain disulfide ────┐   │
     │  │                              │   │
   N─┴──┴──────────────────────────────┴───┴─C
        │                    │
        │  interchain        │  interchain
        │  disulfide         │  disulfide
        │                    │
   N────┴────────────────────┴──────────────────────C
     B CHAIN (30 residues)

  Remove the disulfides and you have two inactive peptide chains.
  The bonds are not decoration; they ARE the molecule's structure.

Why this matters for three later chapters:

Manufacturing. Chapter 32 covers this in full, but note now: making insulin recombinantly is not simply a matter of expressing two peptides. The chains must fold and the correct disulfides must form. Getting that right was the central technical problem of the first recombinant drug.

Storage. Insulin's activity depends on its three-dimensional structure, which depends on those bonds and on the surrounding conditions. Heat, agitation, and freezing degrade it. This is why cold chain matters and why Chapter 4 §4.7 flagged aggregation as an immunological as well as a potency question.

And the hexamer. In storage at high concentration with zinc, insulin molecules assemble into hexamers — six molecules in a stable complex. Hexamers are inactive; they must dissociate into single molecules to bind the receptor.

That dissociation takes time, and it is the reason "rapid-acting insulin" required engineering rather than just a different label. §11.6.

11.3 What insulin does at the cell

Insulin's receptor is not a GPCR. It is a receptor tyrosine kinase — a different architecture from almost everything else in this book, and worth one paragraph because the difference matters.

INSULIN SIGNALING — a different receptor family

  insulin  ──▶  ┌──────────────────┐   the receptor is a dimer that spans the
                │ INSULIN RECEPTOR │   membrane; binding brings its two halves
  ══════════════╪══════════════════╪═══ together
                └────────┬─────────┘
                         ▼   the receptor phosphorylates ITSELF
                    autophosphorylation
                         ▼
                  intracellular signaling cascade
                         ▼
        ┌────────────────┴─────────────────┐
        ▼                                  ▼
  GLUT4 transporters move to          metabolic switches flip:
  the cell surface                    · glycogen synthesis ON
        ▼                             · fat storage ON
  GLUCOSE ENTERS muscle               · gluconeogenesis OFF
  and fat cells                       · lipolysis OFF
        ▼
  BLOOD GLUCOSE FALLS

  Note: this is a much more direct mechanism than the GPCR cascades of Chapter 2.
  Insulin does not merely inform the cell; it operates a set of switches.

The single most important consequence: insulin causes glucose to leave the bloodstream, whether or not there is enough of it there. It has no glucose-dependence, no built-in safety property. Chapter 7 §7.4 explained why GLP-1's amplification design largely prevents hypoglycemia; insulin has no such design because it is not an amplifier. It is the signal itself.

Which brings us to the defining problem of the therapy.

11.4 Hypoglycemia: the side effect that defines everything

Give too much insulin and blood glucose falls below what the brain requires.

The brain runs almost exclusively on glucose and cannot store it. Symptoms progress from sweating, tremor, and confusion to seizure, unconsciousness, and — if uncorrected — death. It can develop within minutes and it is the single most important constraint on insulin therapy.

Why it is unavoidable in principle, and this is Chapter 3 §3.6's counter-regulation argument in its sharpest form:

WHY INJECTED INSULIN CANNOT SELF-CORRECT

  NORMAL PHYSIOLOGY                    INJECTED INSULIN
  ─────────────────                    ────────────────
  glucose falls                        glucose falls
       ▼                                    ▼
  beta cells STOP releasing insulin    the injected dose is ALREADY IN
       +                               and cannot be withdrawn
  alpha cells release GLUCAGON              +
       ▼                               alpha cells release glucagon
  glucose rises back to set point           ▼
                                       counter-regulation operates against
                                       a FIXED input it cannot influence
                                            ▼
                                       glucose may keep falling

  The natural system has TWO levers. Injected insulin removes one of them
  and leaves the other pushing against a constant.

And it gets worse over time. In long-standing type 1 diabetes, the counter-regulatory response itself becomes impaired — glucagon release in response to hypoglycemia is blunted, and the adrenaline response that produces warning symptoms can fade. Hypoglycemia unawareness is the resulting condition, and it is dangerous precisely because the warning system has failed.

This is the reason for essentially every insulin innovation since 1922. The analogs, the pumps, the continuous monitors, the closed-loop systems: all of them are attempts to make the delivered dose track the actual need more closely, because the gap between them is where hypoglycemia lives.

🩺 Safety and Risk — the asymmetry that shapes the whole therapy

Too much insulin can kill you in an hour. Too little kills you over years.

High glucose damages small blood vessels, nerves, kidneys, and retinas — cumulatively, over decades. Acutely it can produce diabetic ketoacidosis, which is a medical emergency, but the ordinary course of under-treatment is slow damage.

Low glucose can cause seizure, coma, and death within an hour.

The asymmetry is why glycemic targets are not simply "as low as possible." A target that minimizes long-term complications will produce more hypoglycemia; a target that minimizes hypoglycemia will accept more long-term damage. Every insulin regimen is a negotiated position on that trade-off, and the right position differs by person — an adolescent with decades ahead and an eighty-year-old living alone are not making the same calculation.

This is one of the clearest cases in medicine where "optimal" is genuinely individual, and it is a useful thing to have understood before Chapter 39.

11.5 Type 1 and type 2: different diseases, one drug

Two conditions, both called diabetes, with substantially different biology.

That last row is the one this book cares about most, because it is Chapter 2 §2.8's rule in its cleanest form.

In type 1, insulin replaces an absent signal. There is no intact system being overridden. No counter-regulation is provoked by the therapy itself. And crucially — the therapy does not lose effectiveness over decades. A person with type 1 diabetes who has taken insulin for forty years has not developed tolerance to it.

In type 2, insulin is being given to overcome resistance to a hormone that is already present. Doses required are often much higher, and the underlying resistance is not corrected by the therapy.

Compare with everything in Chapters 8 through 10. GLP-1 receptor agonists override an intact appetite-regulating system, and weight returns when they stop. Insulin in type 1 replaces a missing one, and nothing returns because nothing was being suppressed.

Same rule, opposite outcomes, and it predicts both.

11.6 From pig pancreas to recombinant DNA

For sixty years, insulin came from animals — extracted from bovine and porcine pancreases collected at slaughterhouses.

It worked, because pig insulin differs from human insulin at one residue and cow insulin at three. Chapter 1 §1.7's point about specificity cuts both ways: close enough is sometimes close enough.

And it had real problems. Supply was tied to meat production. Purity was variable. And the sequence differences meant a meaningful rate of immune reactions and, over time, anti-insulin antibodies — Chapter 4 §4.7's immunogenicity, in its first large-scale clinical manifestation.

In 1982, recombinant human insulin became the first recombinant DNA drug approved anywhere.

The significance is hard to overstate and Chapter 32 covers the technique. What matters here:

Supply became unlimited and decoupled from agriculture. No more dependence on slaughterhouse volumes.

The sequence became exactly human. Immunogenicity fell substantially.

And the door opened to engineering. Once you are producing insulin by instructing an organism from a DNA sequence, you can change the sequence — which is what produced the analogs.

The analogs, and what each solves

Human insulin has a fundamental timing problem: injected under the skin, it forms hexamers that must dissociate before absorption, making onset too slow for a meal and duration awkward for background needs. Each analog is a solution to a specific part of that.

INSULIN ANALOGS — one engineering problem, several answers

  RAPID-ACTING (lispro, aspart, glulisine)
    PROBLEM: hexamers dissociate too slowly; onset lags the meal
    SOLUTION: substitutions that DISCOURAGE hexamer formation
              → faster dissociation → faster onset, shorter duration
    → covers a meal

  SHORT-ACTING (regular human insulin)
    the unmodified molecule; slower onset, longer tail

  INTERMEDIATE (NPH)
    PROBLEM: need background coverage
    SOLUTION: complexed with protamine to slow absorption
    → variable, with a pronounced peak — its main drawback

  LONG-ACTING (glargine, detemir, degludec)
    PROBLEM: need FLAT background coverage without a peak
    SOLUTIONS, three different ones:
      · glargine  — modified isoelectric point; soluble in the vial,
                    precipitates in subcutaneous tissue, dissolves slowly
      · detemir   — FATTY ACID for albumin binding  ← the Chapter 4 strategy
      · degludec  — forms long multihexamer chains at the injection site
                    that release slowly; ultra-long duration
    → covers background needs with less peak, and less nocturnal hypoglycemia

Notice detemir. A fatty acid attached for albumin binding — precisely the semaglutide strategy from Chapter 4 §4.5, applied to a different peptide for a different purpose. The toolkit is general, which is Chapter 33's subject.

📊 Evidence Rating — insulin analogs versus human insulin for hypoglycemia reduction

Claim: Insulin analogs reduce hypoglycemia compared with human insulin.

Rating: ✅ for nocturnal and severe hypoglycemia in type 1 diabetes; ⚠️ for broader claims of superiority in type 2 diabetes.

Why: The analogs' pharmacokinetic advantages are real and translate into reduced nocturnal hypoglycemia in type 1, which is well supported. In type 2 diabetes the incremental benefit over human insulin is smaller and has been genuinely contested — which matters because the cost difference is large.

What would change it: for the type 2 claim, adequately powered comparative trials with hypoglycemia and glycemic outcomes as pre-specified endpoints.

Why this rating is split: because "analogs are better" is a molecule-level claim, and the evidence differs by population. Rule 6, again.

11.6a The dosing problem, concretely

This book has described insulin's difficulty abstractly — "the correct dose changes hourly." It is worth making that concrete, because it explains the entire device industry and because it is the part of this therapy that people outside it consistently underestimate.

Consider what a person with type 1 diabetes has to compute before a meal.

INPUTS TO A SINGLE MEALTIME DOSE

  ① CURRENT GLUCOSE            and, with a sensor, its DIRECTION and RATE of change.
                               Falling at 3 mg/dL per minute is a different situation
                               from the same number rising.

  ② CARBOHYDRATE CONTENT       of food that has not been weighed, cooked by someone
                               else, of uncertain portion size. Estimation error here
                               propagates directly into dose error.

  ③ FAT AND PROTEIN CONTENT    which delay and prolong absorption. A high-fat meal
                               shifts the glucose rise hours later, so a dose timed
                               for carbohydrate alone can cause a low first and a
                               high afterward.

  ④ INSULIN STILL ACTIVE       from the previous dose. Rapid-acting insulin works for
     ("insulin on board")      several hours; stacking doses is a common cause of
                               severe hypoglycemia.

  ⑤ ACTIVITY, RECENT AND       exercise increases insulin sensitivity for many hours
     PLANNED                   afterward — sometimes overnight.

  ⑥ ILLNESS, STRESS, SLEEP,    all of which shift insulin requirements, sometimes
     HORMONAL CYCLE            substantially, and not always predictably.

  ⑦ SITE AND ABSORPTION        injection site, temperature, and local blood flow all
                               affect how fast the dose acts.

  ─────────────────────────────────────────────────────────────────────────────
  This calculation is performed several times a day, every day, for decades,
  by a person who is also doing something else with their life. Getting it
  badly wrong in one direction is dangerous within an hour.

Two observations follow, and both matter for how this book talks about the rest of Part II.

First, the therapeutic burden is enormous and largely invisible. Insulin is frequently described as a solved problem — the disease is manageable, the drug works, people live full lives. All true, and it omits that "manageable" means performing this computation indefinitely with no days off. When Chapter 39 discusses what patients want from clinicians, this is the context.

Second, it reframes what the devices are for. A continuous glucose monitor does not merely replace finger-stick testing; it supplies input ① with direction and rate, which is qualitatively new information. A closed-loop system does not merely automate delivery; it takes over a recurring calculation that a human being was performing under time pressure while distracted. The benefit is partly glycemic and partly cognitive, and trials that measure only time-in-range capture half of it.

💊 In the Clinic — why "just count carbs" is harder than it sounds

Carbohydrate counting is taught as a skill and it is a genuinely learnable one. It is also subject to irreducible error: restaurant portions are not standardized, packaged food labels have tolerances, and the same meal prepared twice is not the same meal.

Studies of carbohydrate estimation accuracy consistently find meaningful error rates even among experienced, well-educated patients. This is not a compliance failure. It is a measurement problem that no amount of motivation solves.

Which is why the framing matters clinically. A patient whose glucose is erratic may be doing everything correctly and hitting the limits of what the method can deliver. Treating variability as evidence of poor adherence is both common and frequently wrong, and it damages the relationship that the therapy depends on.

This is also the strongest practical argument for automated systems: they do not eliminate the estimation error, but they correct for its consequences continuously rather than waiting for the person to notice.

11.7 Pumps, sensors, and closed-loop systems

The most significant advance in insulin therapy since the analogs is not a molecule.

Continuous glucose monitors measure interstitial glucose every few minutes and display it continuously. This changed the therapy fundamentally: a person who previously had four glucose measurements a day now has a curve, with direction and rate of change.

Insulin pumps deliver continuous background insulin with user-directed meal doses, replacing injections.

Closed-loop systems connect them: the monitor's readings drive an algorithm that adjusts pump delivery automatically. These are sometimes called artificial pancreas systems, which overstates it — they modulate insulin delivery and do not replicate the pancreas's glucagon arm.

📊 Evidence Rating — closed-loop systems for glycemic control in type 1 diabetes

Claim: Automated insulin delivery systems improve glycemic control and reduce hypoglycemia in type 1 diabetes.

Rating: ✅ Strong clinical evidence.

Why: Randomized trials have demonstrated increased time in target glucose range and reduced hypoglycemia across age groups, supporting regulatory clearance of multiple systems.

What would change it: little for the core claim. Open questions concern which systems, in which populations, at what cost, and with what burden — not whether the approach works.

Note what this rating is about: a device-and-algorithm system, not a peptide. The most important recent progress in insulin therapy has been in delivery and control rather than in molecules — which is a theme Chapter 36 develops.

And what remains unsolved. Meals still generally require user input, because even rapid-acting insulin is slower than eating. Systems have no glucagon arm, so they can reduce insulin but cannot actively raise glucose. And the burden is real: devices, adhesives, alarms, failures, and cost.

"Smart insulin" — glucose-responsive insulin that self-regulates its activity according to surrounding glucose — would be the genuine solution, since it would restore glucose-dependence to a molecule that lacks it. It has been pursued for decades. As of this writing it remains investigational, and it is rated 🔬.

11.8 Why insulin never stops working

Worth its own short section, because it is the cleanest demonstration of a rule this book has been building since Chapter 2.

A person with type 1 diabetes who has used insulin for fifty years has not developed tolerance to it. The dose may change with weight, activity, illness, and age, but the drug has not stopped working.

Compare the pattern elsewhere in this book: GLP-1 agonists' nausea attenuates; some peptide therapies develop tachyphylaxis within weeks; growth hormone secretagogues face a desensitizing axis (Chapter 15).

The rule (Chapter 2 §2.8): expect tolerance when you override an intact regulated system; expect durability when you replace an absent one.

Type 1 diabetes is the purest replacement case in medicine. There is no beta cell to downregulate, no feedback loop measuring exogenous insulin and reducing endogenous output, and no set point being defended. The therapy is not fighting anything.

This single distinction predicts a great deal, and it is worth carrying into Part III — where several compounds are marketed as though they were replacements and are in fact overrides.

A caution about applying it, because the rule is easy to over-extend. "Replacement" does not mean "natural" and it does not mean "safe." Thyroid hormone replacement is a replacement and is dangerous in excess. Corticosteroid replacement is a replacement and suppresses the axis if the dose exceeds physiological need — which is Chapter 3 §3.4, and which shows that the categories can blur when a replacement is over-supplied.

The rule is about durability, not about virtue. It predicts whether an effect will persist and whether stopping produces rebound. It says nothing about whether the therapy is a good idea, and a compound being a genuine replacement is not an argument for taking it.

Type 1 diabetes is the clean case precisely because the deficiency is total: there is no residual production to suppress and no set point to defend. The further a situation departs from that, the less cleanly the rule applies — which is why Part III's compounds, most of which are supplied to people with no established deficiency at all, sit at the opposite end.

11.9 The pricing controversy, stated accurately

Insulin is the clearest example in medicine of a drug that costs little to manufacture and has been expensive to buy, and it deserves an accurate account rather than a satisfying one.

What is true:

Insulin was patented and sold for a dollar in 1923 with the explicit intention that it not be a source of profit. In the United States, list prices rose substantially over the 1990s–2010s, to a degree difficult to explain by manufacturing cost. People have rationed insulin for cost reasons, and there are documented deaths from doing so. Prices in the United States have been far above those in most other high-income countries for the same products.

What complicates the simple story:

The 1923 patent covered the 1923 product. Every insulin in current use is a later product with its own intellectual property — recombinant human insulin, then the analogs, each separately developed and patented. "They gave away the patent and then charged for it" describes different molecules, which does not excuse the pricing but does mean the story is not straightforwardly a betrayal.

List price is not net price. In the U.S. system, rebates flow between manufacturers, pharmacy benefit managers, and insurers. The list price a manufacturer sets is not what most payers pay — and it is very close to what an uninsured person pays. The people harmed most by high list prices are the ones with the least insurance, which is a structural feature rather than an accident.

Biosimilars have been slow. Chapter 32 explains why: a protein product's copy must be shown to be biologically comparable, which is much harder and more expensive than demonstrating that a small-molecule generic is chemically identical. That regulatory reality slowed competition, and the slowness is not purely a pricing decision.

And prices have moved. Policy interventions, cap programs, and competitive pressure have reduced out-of-pocket costs for many people in recent years. The situation as of this writing is better than it was and is not resolved, and any specific figure in this chapter would be wrong by the time you read it.

⚠️ Hype Check — both directions on insulin pricing

The simple outrage version: "They sold the patent for a dollar and now it costs hundreds." Emotionally powerful, and it conflates the 1923 product with modern analogs, ignores that list price is not net price, and omits the biosimilar regulatory story. Everything it points at is real; the causal account is wrong.

The industry-defense version: "Insulin is complex to manufacture and prices reflect that." Manufacturing a modern analog is genuinely non-trivial. It does not explain a severalfold international price difference for identical products.

The honest version: insulin's price is set by market structure, intellectual property on successive generations of product, an intermediary system that inflates list prices, and slow biosimilar competition — not by manufacturing cost. That combination produced real harm to a specific and predictable group: uninsured and underinsured people paying near list price for a drug they cannot stop taking.

Why this belongs in a peptide book: Chapter 12 asks the same questions about GLP-1 agonists, and insulin is the century-long natural experiment that shows what happens when a life-sustaining peptide becomes a pricing question. The GLP-1 drugs are earlier in the same story.

📋 Your Evidence Dossier

This chapter teaches Field 5 at historical depth: what a century of evidence looks like.

FIELD 5, AT DEPTH — evidence accumulated over time
  First evidence          what established it originally, and by what standard
  How the standard        what evidence would be required TODAY for the same claim
    has changed
  What accumulated        outcome data, comparative data, long-term safety
  What is STILL open      after all this time, what remains unresolved
  What the drug's         what does its history predict about newer drugs in
    history predicts      the same class or with the same architecture?

Worked demonstration — insulin

FIELD 5 (AT DEPTH) — INSULIN                          [worked demonstration]
  First evidence      Uncontrolled clinical administration, Toronto 1922. Reversal of a
                      uniformly fatal condition, reproducibly, in patient after patient.
                      No randomization, no control, no statistics — and no serious doubt.
  Standard today      A new drug for a fatal condition with no alternative could still be
                      approved on dramatic uncontrolled response, but would require formal
                      safety characterization, dose-finding, and post-marketing commitments.
                      The bar for a NON-fatal condition would be a randomized trial.
  Accumulated         A century of clinical use; the DCCT and comparable long-term studies
                      establishing that tighter glycemic control reduces microvascular
                      complications; comparative trials among formulations; extensive
                      safety characterization; and hard-won knowledge about hypoglycemia.
  STILL open          The optimal glycemic target for an individual; whether analogs justify
                      their cost in type 2 diabetes; how to solve postprandial timing; and
                      glucose-responsive "smart" insulin, still investigational after decades.
  What it predicts    (1) REPLACEMENT therapies are durable; overrides are not (Ch 2 §2.8).
                      (2) The hard problem is usually DELIVERY AND CONTROL, not the molecule.
                      (3) A life-sustaining peptide becomes a pricing question eventually,
                          and the mechanism is market structure rather than manufacturing.

Your task

For the best-evidenced compound in your dossier, complete Field 5 at depth, including the final line.

The "what it predicts" line is the transferable one. A compound's history is evidence about compounds like it — and insulin's history predicts three things that apply directly to the GLP-1 drugs, none of which are pharmacological.

Conclusion

Insulin was demonstrated to work in 1922 without a randomized trial, because a treatment that reverses a uniformly fatal condition in front of you leaves no alternative explanation standing. That is rare, and a field that reasoned from insulin to "we don't need trials" would be catastrophically wrong.

The molecule is 51 amino acids in two chains held by three disulfide bonds, and those bonds are the structure rather than an ornament — which made recombinant production a folding problem and makes storage a real constraint.

Its receptor is a tyrosine kinase, not a GPCR, and insulin operates switches rather than amplifying a signal. It has no glucose-dependence, which is why hypoglycemia is the defining risk: injected insulin cannot be withdrawn when glucose falls, so counter-regulation pushes against a fixed input. Every innovation since — analogs, pumps, monitors, closed loops — is an attempt to close the gap between delivered dose and actual need.

Type 1 and type 2 are different diseases, and insulin is a replacement in one and an override of resistance in the other. That distinction explains the chapter's most important observation: insulin never stops working. There is no intact system being overridden, so nothing adapts — which is the exact inverse of the weight regain in Chapter 8, and the same rule predicts both.

Recombinant production in 1982 — the first recombinant drug — decoupled supply from agriculture, eliminated the immunogenicity of animal sequences, and opened the door to the analogs, each of which solves a specific timing problem. One of them, detemir, uses the same albumin-binding fatty acid strategy as semaglutide.

And the dosing problem is unsolved. A century after the molecule was characterized, a person with type 1 diabetes still performs a seven-input calculation several times a day, with irreducible estimation error, and the best available systems automate the correction rather than eliminating the problem. The molecule was the easy part, which is a sentence worth carrying into every remaining chapter of this book.

And the pricing story is not the simple one. The 1923 dollar patent covered a different product; list price is not net price; biosimilar competition was slow for real regulatory reasons; and the harm fell predictably on people paying near list price for a drug they cannot stop taking. Chapter 12 asks the same questions about a newer class, and insulin is the century-long preview.

Key Terms

Type 1 diabetes — autoimmune destruction of pancreatic beta cells, producing absolute insulin deficiency; insulin therapy is required for survival.

Type 2 diabetes — insulin resistance with progressive beta cell dysfunction; insulin is one treatment option among several.

Insulin receptor — a receptor tyrosine kinase, not a GPCR; binding triggers autophosphorylation and a cascade that moves glucose transporters to the cell surface.

GLUT4 — the glucose transporter that relocates to the cell membrane in response to insulin, allowing glucose into muscle and fat cells.

Hypoglycemia — blood glucose below the level the brain requires; insulin therapy's defining risk, capable of causing seizure, coma, and death within an hour.

Hypoglycemia unawareness — loss of the warning symptoms of low blood glucose, common in long-standing type 1 diabetes and dangerous for that reason.

Diabetic ketoacidosis — the acute emergency of severe insulin deficiency.

Basal insulin — background insulin covering needs between meals and overnight.

Bolus insulin — insulin given to cover a meal.

Insulin analog — an engineered insulin with substitutions or modifications altering its absorption and duration.

Hexamer — the six-molecule complex insulin forms in storage with zinc; inactive, and must dissociate before the molecule can act.

Isoelectric point — the pH at which a molecule carries no net charge and is least soluble; manipulating it is the glargine strategy.

Recombinant DNA — the technique of inserting a gene into an organism to produce a protein; human insulin in 1982 was the first approved drug made this way.

Biosimilar — a copy of a biologic product demonstrated to be highly similar to the reference product; harder to establish than small-molecule generic equivalence.

Continuous glucose monitor — a device measuring interstitial glucose continuously.

Closed-loop system — an insulin pump whose delivery is driven automatically by continuous glucose monitor readings.

Smart insulin — investigational glucose-responsive insulin that would self-regulate its activity; rated 🔬.

Spaced Review

1. (Ch 5) Insulin was established without a randomized trial. State the general principle governing when randomization is necessary, and explain why insulin is a rare exception rather than a precedent.

2. (Ch 7) Explain why insulin causes hypoglycemia and GLP-1 receptor agonists largely do not. Name the property that differs.

3. (Ch 2, Ch 8) Insulin does not lose effectiveness over fifty years; GLP-1 agonists' weight effect reverses on discontinuation. Explain both with one rule.

4. (Ch 4) Insulin detemir carries a fatty acid for albumin binding. Name the other drug in this book using the same strategy and state what problem each was solving.

5. Explain the insulin pricing situation to someone who has heard "they sold the patent for a dollar" — accurately, without defending anyone and without overstating.