The Biggest Pile of Papers Is Not Where the Obesity-Drug Race Is Won
A worked example from a ForIntel R&D Velocity Audit. In the GLP-1 obesity field, the sub-field with the most published papers has the smallest patent estate — a search-term artifact, not real invention. Reading R&D velocity on publication counts alone points you at the wrong bet.
If you wanted a quick read on where the obesity-drug race is heating up, the obvious move is to count papers. More publications means more research attention means more momentum — so whichever sub-field has the biggest pile of literature is where the action is. It is a fast, intuitive heuristic.
It is also how you end up backing the wrong horse.
When we ran a sample R&D Velocity Audit on the GLP-1 / incretin obesity field, the sub-field with by far the largest publication count turned out to have the smallest patent estate in the field. The paper pile was real. The invention behind it was not proportional. Reading the field on publications alone would have pointed a partnership or investment thesis at exactly the sub-field the deeper signals rank third.
Here is why, and what to read instead.
The gap that gives it away
The audit reads the field across three velocity signals — the scholarly publication record, the clinical-trial registry, and the patent record — because no single one tells the truth on its own. Line up the four candidate sub-fields on publications versus patents and one row jumps out:
| Sub-field | Publications | Patent families | The gap |
|---|---|---|---|
| Dual / multi-agonist | 2,548 | 235 | The field's combined-signal leader |
| Long-acting injectable | 2,635 | 198 | Deepest late-stage trial bench |
| GLP-1 + amylin | 345 | 70 | Trial-dense for its size — the watch flag |
| Oral small-molecule GLP-1 | 7,398 | 32 | Widest gap — the term-breadth signature |
Oral small-molecule GLP-1 has almost three times the literature of the next sub-field and roughly one-seventh the patents of the leader. That is not what a genuinely differentiated, fast-inventing sub-field looks like. It is what a broad search term looks like: the free-text query that defines "oral small-molecule GLP-1" over-captures adjacent oral-agent literature, inflating the paper count without a matching rise in protected invention.
The patent record settles what the publication count could only hint at. When a sub-field has the biggest literature and the smallest patent estate, the paper pile is measuring vocabulary, not velocity.
Where the race is actually won
Strip the artifact out and the ranking falls into place, and the three signals agree on the order — the strongest form of corroboration this kind of read produces:
- Dual and multi-agonists — the combined-signal leader. A large, recency-saturated literature (2,548 works), the deepest trial bench (123 distinct trials, 37 late-phase), and the most in-field patent families (235). The literature rise holds up across a second, independent literature index, so it is real recent dates, not a single-source back-catalog effect. The field's two incumbents anchor its sponsor list.
- Long-acting injectables — the late-stage leader. A comparable literature (2,635) and the most commercially advanced trial bench in the audit: 66 of its 100 trials sit at Phase III/IV. It trails on total breadth but leads on late-stage depth — a divergence worth naming rather than smoothing over.
- Oral small-molecule GLP-1 — real, but term-inflated. Accelerating, with 100 trials, but ranked on its trial bench and thin patent base, not its headline paper count.
- GLP-1 + amylin — the emerging watch flag. The smallest literature (345 works) but a dense trial set (90 trials) concentrated almost entirely in one developer.
The watch flag has a specific shape
The amylin-combination sub-field is worth a second look, because it is the mirror image of the oral small-molecule trap. Oral has too many papers for its invention; amylin has too few papers for its trials. Only 345 works, yet 90 distinct trials — and 22 of them sponsored by a single developer, AstraZeneca.
That divergence — thin literature, dense trials, one developer — is the textbook watch-flag shape. It marks a sub-field that is moving on registration intent ahead of broad research consensus, and whose fortunes rise or fall largely with one company's program rather than a field-wide bet. Its trial starts are steady rather than accelerating, so the flag rests on the concentration and the density-against-thin-literature gap, not on a rising trajectory. It is an emerging bet to monitor — partnership-worthy if the literature broadens and the sponsorship diversifies, premature to treat as settled field consensus today.
You cannot see either the oral trap or the amylin flag from publication counts. You need the trial registry and the patent record standing next to the literature.
The measurement lesson: your counting method can flip the answer
The patent leg carried its own cautionary tale. An intermediate, quicker patent pull — US titles only — read Eli Lilly ahead of Novo Nordisk on the field's IP. The gold-standard read reversed it.
Family-deduplicated across US, European and PCT filings, with holder names harmonized, Novo Nordisk leads roughly 2:1 (242 families to Lilly's 115), with Hanmi, the University of California, Roche, Sanofi and Zealand behind. The title-only version was wrong for a mechanical reason: Lilly tends to name its US filings with obvious molecule terms while Novo files globally under generic titles, so a raw US-title count over-credited Lilly and under-counted Novo. Deduplicate the same invention across jurisdictions and the true holder concentration appears — Novo leads both developer-ranked signals, trials and patents, with no cross-signal split.
The takeaway is the same one that recurs across every field we map: how you count is not a footnote — it decides who ranks first. Raw publication counts over-credit broad search terms. Raw US-title patent counts over-credit whoever labels their filings loudly. Both errors are invisible until you put the signals side by side and deduplicate honestly.
One caveat travels with the whole patent layer and is worth repeating: it measures filing activity, not commercial value. A single licensed, foundational asset can outweigh dozens of narrow filings. Weight holders by what you are actually valuing.
What to do with this
- Never rank a field on publication counts alone. The biggest paper pile is often the broadest search term. Cross it against trials and patents before you believe it.
- For a partnership or licensing thesis, the highest-confidence bet is the sub-field the signals agree on — here, dual and multi-agonists, anchored to the two incumbents, with a late-stage overlay from long-acting injectables.
- Treat single-developer trial density as a watch flag, not a leader. Amylin combinations are real and moving, but concentrated in one program.
- Insist on family-deduplicated, global patent counts. A US-title tally can hand you the wrong front-runner with total confidence.
None of this is a claim about which molecule is safer or more effective. That is a clinical question for domain experts. This is a claim about where research effort, trial capital and protected invention are actually concentrating — and about how easily the wrong measurement points you at the wrong sub-field.
See the full velocity read
The worked example above is drawn from a public sample of a ForIntel R&D Velocity Audit, which reads a field across three velocity signals and ranks its sub-fields into partnership, watch and caution calls. Publication counts are read as floors, trial records as registration intent, and the patent leg at gold standard. It reports velocity and field structure only — no scientific or clinical claim about any molecule.
- Read the sample: ForIntel R&D Velocity Audit — GLP-1 obesity field
- Commission one for your field: reach the ForIntel desk at forintel@foragentis.com.
