Case Study 22.2 — CGRP and Migraine: The Counter-Case, and What Actually Differed
A case in what has to be true, beyond a good mechanism, for a neuropeptide program to produce a drug.
Why this case
Read alone, the NK1 story in Case Study 22.1 supports a cynical conclusion: mechanism is worthless, preclinical data is theater, and drug development is a lottery. That conclusion is wrong, and this case is the reason.
CGRP is a sensory neuropeptide, found in primary afferent neurons — frequently the same neurons that contain substance P — released on intense stimulation, acting at a GPCR, implicated in pain, and characterized in enormous detail over decades. On every dimension that describes what kind of scientific object it is, CGRP and substance P are the same kind of thing.
One produced nothing for pain. The other produced the first migraine-specific preventive drug class in history. The interesting question is not which was the better molecule. It is what a successful program had that a failed one did not.
The molecule
Calcitonin gene-related peptide is a 37-residue peptide, and its name records its origin. The calcitonin gene is processed two ways: in thyroid C cells the transcript is spliced to produce calcitonin, and in neurons the same gene is spliced differently to produce CGRP. One gene, two peptides, unrelated physiology — the same precursor economy that yields substance P and neurokinin A from a single tachykinin gene.
CGRP is among the most potent vasodilators known in human tissue. It is heavily expressed in trigeminal sensory neurons, which innervate the meninges and cranial blood vessels — precisely the territory that hurts during a migraine.
How the case was built: three kinds of human evidence
The migraine hypothesis assembled over roughly three decades, and it did something the substance P program never managed. It demonstrated the peptide's role directly in humans, in the disease of interest.
Observation
CGRP levels rise in cranial venous blood during spontaneous migraine attacks and normalize as attacks resolve. Effective acute treatment with triptans lowers CGRP alongside the headache.
What this establishes: a temporal association between the peptide and the clinical event. What it does not: direction. A molecule released during an event is not necessarily causing it.
Provocation
Intravenous infusion of CGRP into people with migraine triggers delayed headaches resembling the participant's own attacks, at a far higher rate than in people without migraine.
What this establishes: causality, in humans, in the target population. The intervention was the peptide itself; the outcome was the clinical event. What it does not: that CGRP is the only cause, or that every attack is CGRP-driven — not everyone infused develops an attack. Nor does it identify where in the pathway the peptide acts, and laboratory provocation is not identical to a spontaneous attack.
Blockade
Blocking CGRP or its receptor reduces migraine frequency and severity in randomized controlled trials across multiple agents and two distinct molecular modalities.
What this establishes: the therapeutic claim. What it does not: that every patient benefits, or that the class is superior to older preventives on every patient-important outcome.
Together, this is a complete causal chain. The peptide rises during the event; administering it reproduces the event; blocking it prevents the event. Nothing comparable exists for substance P and pain.
What the drugs are — and are not
Two families reached the clinic.
Monoclonal antibodies. Erenumab targets the CGRP receptor; fremanezumab, galcanezumab, and eptinezumab target the CGRP peptide. Given by injection or infusion at long intervals, approved for migraine prevention.
Small-molecule receptor antagonists — the "gepants." Ubrogepant, rimegepant, atogepant, and zavegepant, taken orally or nasally. Some are used acutely, some preventively, rimegepant both ways. An earlier generation was abandoned over liver toxicity signals, which is why the class took longer to arrive than the biology predicted.
Now read the names with Chapter 1 §1.8 in hand. Erenumab, fremanezumab, galcanezumab,
eptinezumab — all -mab. Monoclonal antibodies: roughly 150,000 daltons, produced in cell culture,
about thirty times the mass of a peptide. Not peptides. And the gepants are small molecules in the
size class of aspirin. Also not peptides.
The peptide is the target, not the drug.
This is the single most-misreported fact about the class. Headlines describing "a new peptide drug for migraine" have it backwards, and the error is not cosmetic: it changes what you should expect about manufacturing, cost, dosing interval, half-life, and delivery.
Three relationships between a drug and a peptide recur in this book, and none of them is "the drug is the peptide":
| Relationship | Example | The peptide's role |
|---|---|---|
| Blocked by an antibody or small molecule | CGRP therapies (Ch 22) | target |
| Used as a delivery address for a payload | radioligand therapy (Ch 27) | address |
| Raised indirectly by inhibiting its degradation | neprilysin inhibition (Ch 28) | effector |
📊 Evidence Rating
Claim: CGRP-targeting therapies (anti-CGRP and anti-receptor monoclonal antibodies, and small-molecule CGRP receptor antagonists) reduce migraine frequency in adults with episodic or chronic migraine. Rating: ✅ Strong clinical evidence Reason: Multiple adequately powered randomized placebo-controlled trials across several structurally distinct agents and two molecular modalities, consistent reductions in monthly migraine days, regulatory approval in major jurisdictions, and accumulating real-world use. What would change it: A long-term safety signal — CGRP is a vasodilator with cardiovascular and wound-healing roles, and the consequences of blocking it for years are still being characterized. The ✅ does not claim universal response or superiority to older preventives. (Rated as of 2026.)
The comparison, laid out
| Substance P / NK1 for pain and depression | CGRP for migraine | |
|---|---|---|
| Peptide type | sensory neuropeptide | sensory neuropeptide |
| Often found in | small-diameter primary afferents | small-diameter primary afferents (often the same cells) |
| Receptor | GPCR (NK1) | GPCR (CLR–RAMP1 complex) |
| Preclinical case | extensive, converging, decades deep | extensive, converging, decades deep |
| Indication targeted | chronic pain; major depression — heterogeneous, subjective endpoints | migraine — discrete entity, countable endpoint |
| Human causal evidence | none of provocation type | provocation study reproduced the clinical event |
| Required site of action | central | substantially peripheral |
| Modalities available | brain-penetrant small molecules only | antibodies and small molecules |
| Outcome | comprehensive clinical failure | approved therapeutic class |
Three rows are bolded because they are the three that differed. Mechanism is not among them.
The lesson, stated precisely
The wrong conclusion is "mechanism is useless." CGRP's success came directly out of decades of mechanistic work; the migraine drugs exist because somebody carefully characterized a peptide, its receptor, and its release pattern in trigeminal neurons. Mechanism was necessary.
The right conclusion is this:
Mechanism alone is a hypothesis. Mechanism plus a well-defined indication with a validated endpoint, plus direct human evidence that the target matters in that indication, plus an accessible target, is a different proposition — and it is the one that produces drugs.
Substance P had the first item and not the rest. CGRP had all four.
This is a checklist you can run yourself. When you read that some peptide "regulates inflammation" or "modulates the stress response" or "plays a key role in neuroprotection," you have been handed item one. Ask for the other three: what specific condition, in what population, measured how; what human evidence that this molecule matters there; and by what route does the drug reach the tissue?
Discussion questions
1. Substance P and CGRP are co-expressed in many of the same sensory neurons. Construct the strongest version of the argument that their divergent clinical fates are therefore surprising — then answer it using the three bolded rows of the comparison table. Which of the three do you find most persuasive, and why?
2. The provocation study is the piece of evidence §22.9 treats as decisive. Design, in outline, the substance P equivalent that was never run. What would you infuse, into whom, and what outcome would you measure? Then identify the reasons — practical, ethical, and conceptual — that such a study was not feasible, and say whether those reasons are themselves informative about the hypothesis.
3. "Migraine has a countable endpoint; depression does not." Interrogate that claim. Is the difference between monthly migraine days and a depression rating scale one of kind or one of degree? What would it take to give a heterogeneous condition a migraine-quality endpoint, and what is lost in the attempt?
4. CGRP therapies work largely from outside the blood-brain barrier. Does that make them a solution to Part IV's central obstacle, or a demonstration that the obstacle can sometimes be avoided? Argue for one reading, and then name a neuropeptide indication where the same strategy would be unavailable and explain why.
5. Every approved CGRP therapy is either an antibody or a small molecule; none is a peptide. Does this class belong in a book about peptides at all? Make the case that it does — and in doing so, articulate what "understanding peptides" means if it does not mean "understanding peptide drugs."
6. Apply the four-item checklist to a neuropeptide program currently in development — orexin agonists for narcolepsy (§22.6) are one option, or choose another from Part IV. Which items does it satisfy today? Which are outstanding? And based on that assessment alone, without consulting any efficacy data, what would you predict — and what would you want to see before betting either way?