Sample ForIntel R&D Velocity Audit — Where the Obesity-Drug Race Is Actually Accelerating

The verdict

The obesity-drug field is accelerating fastest in dual / multi-agonists and long-acting injectables; amylin combinations are the emerging watch flag — and two developers anchor it all.

This is an R&D-velocity audit of the GLP-1 / incretin obesity & metabolic field: which sub-fields and which developers are pulling ahead, ranked by the combined publication × trial × patent signal. Across all three velocity signals, the order is clear. Dual / multi-agonists lead — a large, recency-saturated literature (2,548 works) and the deepest trial bench (123 distinct trials, 37 late-phase), with the rise independently corroborated across two literature indexes. Long-acting injectables rank second on the strength of the most late-stage clinical work in the audit (66 of 100 trials at Phase III/IV) — the commercial-stage signal. Oral small-molecule GLP-1 shows the largest raw publication field (7,398) but the lead is partly a broad-search-term artifact, so it ranks third. GLP-1 + amylin is the emerging watch flag: a thin literature (345 works) but a dense trial set concentrated in a single developer (AstraZeneca, 22 trials). Two incumbents — Novo Nordisk and Eli Lilly — anchor the field's sponsor concentration. The patent record — the third velocity signal, now read at gold standard — corroborates this order (most in-field patent families in dual / multi-agonist at 235, fewest in oral at 32, confirming the term-breadth call) and resolves the holder picture: on a family-deduplicated global basis Novo Nordisk leads the IP ~2:1 over Eli Lilly (242 vs 115 families), so Novo leads both developer-ranked signals — trials and patents — with no cross-signal divergence, in a commercial-pharma-led field. The patent leg measures activity, not commercial value.

• Dual & multi-agonists are the field's combined-signal leader — large literature and the deepest trial bench. The dual / multi-agonist sub-field (the GLP-1/GIP and GLP-1/glucagon chemistries) carries a large, recency-saturated publication field (2,548 works on file) and the most total clinical-trial activity of any sub-field — 123 distinct registry records, 37 of them late-phase (III/IV). The two field incumbents (Novo Nordisk, Eli Lilly) lead its sponsor list. The acceleration is independently corroborated for recency: a separate biomedical-literature check on the same sub-field returned a capped recent-window sample, every item dated to the current year — so the rise is in genuine recent publication dates in a second index too, not a single-source artifact. It ranks first on combined signal.
• Long-acting injectables carry the most late-stage clinical momentum — the commercial-stage signal. The long-acting / next-gen injectable sub-field has a comparably large publication field (2,635 works) and, critically, the highest concentration of late-phase work in the audit — 66 of its 100 distinct trials are Phase III/IV. That is where the field's commercial-stage momentum sits, and its sponsor list is led by Eli Lilly and Sanofi. It ranks second on combined signal: slightly behind dual/multi-agonists on total trial breadth, but ahead on late-stage depth — a divergence the ranking names rather than smooths over.
• Oral small-molecule GLP-1 leads on raw literature volume — but the lead is partly a broad-term artifact. The oral small-molecule sub-field shows the largest raw publication field (7,398 works) and a fully recency-saturated recent sample, with 100 distinct trials (30 late-phase). But the counter-signal pass flags it: that 7,398 rests on the broadest free-text term of the four sub-fields, which over-captures adjacent oral-agent literature, so its publication lead is inflated relative to its true differentiated output. The patent record now confirms this directly: the oral sub-field is the field's largest publication sub-field but its smallest patent estate — just 32 families against its 7,398 publications — the widest publication-to-invention gap of any sub-field, exactly the term-breadth-not-invention signature. It ranks third — real and accelerating, but not as far ahead as the raw publication count suggests.
• GLP-1 + amylin is the emerging watch flag — small literature, but trial-dense and developer-concentrated. The amylin-combination sub-field has by far the smallest publication field (345 works), yet it is disproportionately trial-dense — 90 distinct registry records, 29 of them late-phase. And its trial activity is concentrated in a single emerging developer: AstraZeneca sponsors 22 of the sub-field's trials. That divergence — thin publications, dense and concentrated trials — is the classic watch-flag shape: trial-dense relative to a thin literature and concentrated in one developer, not yet broadly corroborated. Its trial starts are steady rather than accelerating (about level recently versus the prior multi-year window), so the flag rests on the density-against-thin-literature gap and the single-developer dependence, not on a rising trajectory. It ranks fourth, as the field's emerging bet rather than its established leader.
• The patent record — now read at gold standard — corroborates the sub-field ranking and resolves the holder question. The third velocity signal is now in hand at gold standard (full-text + classification, family-deduplicated, holder-name-harmonized) and it does two things. First it confirms the combined ranking: the dual / multi-agonist sub-field carries the most in-field patent families (235), and oral small-molecule by far the fewest (32) — so the patent leg disambiguates the oral term-breadth question harder than the publication gap alone. Second, the holder concentration removes an apparent twist rather than adding one: on a family-deduplicated global basis Novo Nordisk leads the IP ~2:1 over Eli Lilly (242 vs 115), with Hanmi (48), the University of California (44), Roche (36), Sanofi (34) and Zealand (33) behind — so Novo leads both developer-ranked signals with no cross-signal divergence. (The intermediate title-only pull had read Lilly ahead; that was a US-title first-applicant artifact, corrected by family-dedup across US/EP/WO and global coverage.) The field is commercial-pharma-led — the mirror image of an academic/state-led research field. The read is gold-standard (High confidence) with one caveat: it measures patent activity / volume, not commercial value — one licensed asset can outweigh dozens of filings.

In one line: across the three velocity signals — the scholarly publication record, the clinical-trial registry and the patent record — the GLP-1 obesity field's combined R&D velocity ranks dual / multi-agonists first (largest trial bench, corroborated literature rise, most patent families at 235), long-acting injectables second (deepest late-stage pipeline), oral small-molecule third (biggest raw literature, but term-inflated and smallest patent estate at 32), and GLP-1 + amylin fourth as the emerging watch flag (thin literature, dense and AstraZeneca-concentrated trials). Novo Nordisk leads the IP ~2:1 over Eli Lilly (242 vs 115 families), so Novo leads both developer-ranked signals, no divergence, in a commercial-pharma-led field. The patent leg measures activity, not commercial value.

How to read this audit. This is an R&D-velocity and field-structure read — it reports who is publishing, trialing and patenting what, and how fast, and ranks the sub-fields by that combined signal. It makes no scientific or clinical claim about any molecule's safety, efficacy or merit; that is the buyer's call with domain experts. A High confidence chip marks a signal observed across signals or independently corroborated; a Medium chip marks a signal that rests on one signal or a term-breadth caveat, stated inline. Publication counts are read as floors (the recent-window samples are capped and saturated by recent work) and trial records are registration-intent signals deduplicated by record. The patent-velocity leg is read at gold standard (Section 04) and folded into the ranking — full-text + classification, family-deduplicated, holder-name-harmonized (High confidence on the holder ranking), with one caveat stated inline: it measures patent activity / volume, not commercial value.

01 · Publication Velocity

Which sub-fields the literature is moving toward. (Confidence: High.)

The first velocity signal is the scholarly publication record — where the research output is accumulating, and how recent it is. Across the four candidate sub-fields the field sizes diverge sharply. For the three larger sub-fields, the most-recent sample we could pull was entirely saturated by recent-window (current-and-prior-year) work — meaning the in-window output exceeds the sample ceiling, so these counts are floors, not a complete census.

Figure — Scholarly-publication field size per candidate sub-field. The three larger sub-fields are recency-saturated (the recent-window sample hits the sample ceiling, so counts are floors). The oral small-molecule total rests on the broadest free-text term and is read with that caveat (Section 04).

Two reads matter. First, the publication record alone would put oral small-molecule on top — but that headline carries a caveat the counter-signal pass surfaces (Section 04): its field is measured on the broadest search term of the four, which over-captures adjacent oral-agent literature, so the raw lead overstates its differentiated output. Second, an independent biomedical-literature check on the dual / multi-agonist sub-field, run on a different index and restricted to the recent window, returned a full recent-window batch dated entirely to the current year — confirming that sub-field's acceleration is in genuine recent publication dates across two indexes, not a single-indexer back-catalog artifact. One limitation travels with the layer and is named, not buried: these are recent-window floors, and the per-period split is reconstructed from separate dated queries rather than a clean pre-grouped time series.

02 · Trial Velocity & Phase Mix

Where the field is moving into the clinic — and how late-stage. (Confidence: High.)

Publications show where the research is; the clinical-trial registry shows where the field is committing capital and moving into the clinic, and at what stage. Deduplicated on registry record within each sub-field, the trial counts and phase mix are:

Figure — Distinct trial records per sub-field, deduplicated on registry ID, split late-phase (III/IV) vs other. Dual / multi-agonist carries the most total trials; long-acting injectable carries the most late-phase work (66 of 100). Records are registration-intent signals; a single program may register more than one.

The divergence is the signal. Dual / multi-agonists have the broadest bench (123 trials, 37 late-phase). But long-acting injectables are the most commercially advanced: 66 of their 100 trials are Phase III/IV, the highest late-stage concentration in the audit, which is why they rank second on combined signal despite a slightly smaller total bench. GLP-1 + amylin is the standout for its size: with the smallest publication field, it still fields 90 trials (29 late-phase) — trial-dense relative to its literature, the first tell of an emerging sub-field. One arithmetic note so the numbers reconcile: the per-sub-field counts are deduplicated within each sub-field, so a multi-mechanism trial (e.g. a GLP-1/GIP/amylin tri-agonist) is counted in more than one sub-field — the sub-field counts therefore sum to more than the 340 distinct field-wide records and are deliberately not additive.

03 · Developer Concentration

Who is driving the field — and where an emerging player is massing. (Confidence: High.)

The last structural question is which organizations are driving the activity. Across the deduplicated trial set (340 distinct registry records field-wide), the raw sponsor ranking is Novo Nordisk (27 distinct trials), AstraZeneca (25), Eli Lilly (19) and Sanofi (12), with a long tail of biotech and academic sponsors behind them. But raw volume hides the structure: AstraZeneca's second-place 25 is almost entirely one sub-field (22 of its 25 trials sit in GLP-1 + amylin — the watch-flag concentration), so the field's two broad-based incumbents are Novo Nordisk and Eli Lilly, who anchor the dual/multi-agonist and long-acting-injectable sponsor lists, while AstraZeneca is a single-sub-field specialist rather than a broad field leader.

Figure — Leading developers by distinct sponsored trials across the deduplicated registry set. Two incumbents (Novo Nordisk, Eli Lilly) anchor the field; the third-ranked developer (AstraZeneca, 25 trials) is concentrated almost entirely in the amylin-combination sub-field — 22 of its 25 trials — the emerging-vendor watch flag. Sponsor naming is imperfectly normalized; one organization may appear under variant names.

The structural surprise sits in third place. AstraZeneca ranks among the most active sponsors in the field (25 distinct trials) — but almost all of that activity is in one sub-field: 22 of its 25 trials are in GLP-1 + amylin. That single-sub-field concentration is exactly why amylin combinations register as the audit's watch flag — a thin publication base but a dense, single-developer-driven trial program (its late-phase share, 29 of 90, sits at the field average — the watch-flag signal is the concentration, not the stage mix). For a buyer, the read is that the amylin opportunity is real and moving but concentrated: it rises or falls largely with one developer's program rather than a broad field consensus. One limitation is named: sponsor names are imperfectly normalized — a single organization can appear under variant legal entities, so the counts are decision-grade for the concentration shape, not for an exact per-entity tally.

04 · Patent Velocity & IP Concentration

The third signal — what the field has actually invented, and who owns it. (Confidence: High.)

The third velocity signal is the patent record — normally the leading commercial-R&D indicator, because it measures what the field has translated into defensible invention, not just published or registered. Read at gold standard (full-text title-plus-abstract + classification match, family-deduplicated so each invention counts once across jurisdictions), the in-field patent estate is large and recent: 4,597 distinct patent families match the GLP-1 / incretin field, of which 2,825 were filed since 2018 and 56% of those (1,580) in the last few years — the same recent acceleration the other two signals show.

Figure — Distinct in-field patent families by holder (family-deduplicated across jurisdictions; full-text + classification match; holder names variant-harmonized). Novo Nordisk leads the IP (242) ~2:1 over Eli Lilly (115) — the same direction as the trial signal, where Novo also led. The field is commercial-pharma-led. Counts are patent activity, not commercial value.

Two reads matter, and both sharpen calls the other signals could only flag. First, the patent leg resolves the oral small-molecule term-breadth question directly, and harder than the publication record alone could: oral small-molecule carries the field's largest publication count (7,398) but the smallest patent estate — just 32 families, by far the widest publication-to-invention gap of the four sub-fields. That is the signature of term breadth, not differentiated invention, and it is why oral ranks third on the combined signal despite its raw publication lead. (The amylin estate is genuinely modest here at 70 families; an earlier title-only pull had read it far higher, ~192, but that swept in older pramlintide and insulin-delivery patents outside the GLP-1-obesity scope — the family-deduplicated specific-token method excludes them.) Second, the holder concentration removes an apparent twist rather than adding one: Novo Nordisk leads the IP ~2:1 over Eli Lilly (242 vs 115 families), so Novo leads both developer-ranked signals with no cross-signal divergence. An intermediate title-only pull had read Lilly ahead (69 vs 26); that was a US-title first-applicant artifact — Lilly names US filings with molecule terms while Novo files globally under generic titles — corrected here by family-dedup across US/EP/WO and global coverage. The field is commercial-pharma-led, the mirror image of an academic/state-led research field. One caveat travels with the layer and is named: the read measures patent activity / volume, not commercial value — a single licensed asset can outweigh dozens of filings — so weight holders by what you are valuing; and claims/description full-text was not scanned (a deliberate cost scope; title + abstract is the standard analytics grain).

05 · The Combined-Signal Ranking

Putting the signals together — and stress-testing each call. (Confidence: High.)

Combining the three velocity signals — publication, trial, and patent — at the sub-field and organization-name level (the defensible join; individual-asset threading is a declared boundary), the ranking falls out cleanly, and each load-bearing call was stress-tested against its strongest opposing case before being named.

• 1 · Dual / multi-agonists — the combined-signal leader. Large, recency-saturated literature (2,548) + the most total trials (123, 37 late-phase) + the most in-field patent families (235) + the two incumbents on its sponsor list. Opposing case tested: could the literature rise be an indexing artifact? No — an independent biomedical-index check returned a full current-year batch, so the rise is in real recent dates across two indexes. Proceed.
• 2 · Long-acting injectables — the late-stage leader. Comparable literature (2,635) + the deepest late-phase bench (66 of 100 trials at III/IV) + 198 patent families. It trails dual/multi-agonists on total breadth but leads on commercial-stage depth — a divergence named, not smoothed. Proceed.
• 3 · Oral small-molecule GLP-1 — real, but term-inflated (now patent-confirmed). Largest raw literature (7,398) + 100 trials (30 late). Opposing case tested: the 7,398 rests on the broadest free-text term and over-captures adjacent oral-agent work. The patent record now settles it: oral is the field's largest publication sub-field but its smallest patent estate (just 32 families against 7,398 publications) — the widest publication-to-invention gap in the field. Proceed with caveat — accelerating, but ranked below its raw count, now on three signals not two.
• 4 · GLP-1 + amylin — the emerging watch flag. Smallest literature (345) but trial-dense (90, 29 late) and concentrated in one developer (AstraZeneca, 22); a modest patent estate (70). Opposing case tested: a single-developer-driven trial count can inflate on registration intent — so it is named as a concentrated, watch-flag bet, not an established leader. Proceed with caveat.

The ranking now rests on all three velocity signals — publication, trial, and patent — and they agree on the relative order, which is the strongest form of corroboration this instrument produces. The patent leg is the gold-standard one (full-text + classification, family-deduplicated), so its holder ranking and family counts are exact-grain, not floors; the caution that remains is on the other two signals — the publication samples are capped and the trial records are registration-intent. So the ranking is decision-grade for the relative order of the sub-fields, while the publication / trial magnitudes remain floors and patent counts are read as activity, not commercial value. The order is decision-grade; the magnitudes are bounded.

06 · What This Means For You

1 · Partnership bet — dual / multi-agonists, anchored to the two incumbents. This is the sub-field with the strongest combined velocity: the broadest trial bench, a large and independently-corroborated literature rise, the most in-field patent families, and the field's two leading developers (Novo Nordisk, Eli Lilly) already concentrated in it. For a BD / licensing or investment thesis, it is the highest-confidence, highest-leverage partnership target. Pair the partnership scan with a late-stage overlay from the long-acting-injectable sub-field, which carries the deepest Phase III/IV bench.

2 · Watch flag — GLP-1 + amylin, and specifically AstraZeneca's program. The amylin-combination sub-field is trial-dense relative to its thin literature and single-developer-concentrated (AstraZeneca sponsors 22 of the sub-field's trials), but not yet corroborated on publications, and its patent estate is modest (70 families). Its trial starts are about level recently versus the prior multi-year window — steady, not accelerating. That is the textbook watch-flag shape: narrow corroboration, single-developer dependence. Track it as an emerging bet to monitor — partnership-worthy if the literature broadens and the concentration diversifies, but premature to treat as an established field consensus today.

3 · Caution — do not over-weight the oral small-molecule headline (now patent-confirmed). Oral small-molecule GLP-1 shows the largest raw publication count, which makes it the easy headline pick — but that lead is partly a search-term-breadth artifact, and the patent record now confirms it: it is the field's largest publication sub-field but its smallest patent estate (32 families against 7,398 publications), the widest publication-to-invention gap in the field. Treat it as a genuinely accelerating but over-hyped-relative-to-raw-count sub-field: real, but rank it on its trial bench and its (thin) patent base, not on the headline publication number.

4 · Read the field as commercial-pharma-led, with Novo Nordisk ahead on both developer-ranked signals. The gold-standard patent layer settles the holder question the trial bench could only hint at: Novo Nordisk leads the IP ~2:1 over Eli Lilly (242 vs 115 families), so Novo leads both developer-ranked signals — trials and patents — with no cross-signal divergence. For a BD / licensing or competitive thesis, the two incumbents are not split by stage (an earlier title-only read suggested they were — that was a US-filing artifact); Novo is the consistent front-runner, with Lilly the clear second and a pharma tail (Hanmi, Roche, Sanofi, Zealand) plus university holders (California) behind. Weight holders by commercial value, not family count — the patent leg measures activity, and one licensed asset can outweigh dozens of filings.

Scope, Confidence & What a Deeper Engagement Adds

This R&D Velocity Audit reads three velocity signals — the scholarly publication record, the clinical-trial registry, and the patent record — all directly observed, with the lead sub-field's acceleration corroborated across two literature indexes and the sub-field order corroborated across all three signals. The boundaries below are named with the specific reason and the work that closes each. They are diligence boundaries, not findings, and are never presented as such. Above all, this audit reports R&D velocity and field structure only — it makes no scientific or clinical claim about any molecule.

• Patent velocity — now read at gold standard (Section 04); the residual is value-weighting, not coverage. The patent record is now in hand at gold standard (full-text title-plus-abstract + classification, family-deduplicated across jurisdictions, holder names variant-harmonized) and folded into the ranking as the third signal. Two residual boundaries remain, both proven-hard and neither a coverage gap: (a) the read measures patent activity / volume, not commercial value — a single licensed asset can outweigh dozens of filings, so holder rank is "who is filing", not "who owns the most valuable estate"; and (b) claims/description full-text was not scanned (a deliberate cost scope — title + abstract is the standard analytics grain). Closing them: a value-weighted overlay (citations, litigation, licensing, family geography) and, if needed, a claims-level pull — an IP-valuation engagement distinct from this velocity read.
• Velocity period-granularity — reconstructed, not pre-grouped. The publication record does not return counts pre-grouped by period and sub-field, so the per-period velocity was reconstructed from separate dated queries and the recent-window samples are capped and saturated — counts are floors, not a complete time series. The direction (which sub-fields are accelerating) is the load-bearing read and is unaffected; the exact slope is bounded. Closing it: uncapped, pre-grouped per-period pulls to quantify the precise velocity curve per sub-field.
• Cross-corpus threading — at the sub-field / organization level, not per-asset. The audit threads the signals at the sub-field and organization-name level (which developer leads which sub-field across publications, trials and patents) — the defensible join. It does not thread an individual publication to an individual trial to an individual patent via shared identifiers; that chain is thin and lossy and is declared a boundary rather than asserted as a false linkage. Closing it: an identifier-level threading pass to complete the publication→trial→patent chain per asset.
• Trial registry is registration-intent, deduplicated by record. The clinical-trial registry is registration-based — a registered trial is an intent signal, not a guaranteed-completed study — coverage of ex-US trials is uneven, and a single program may register more than one record. Counts here are deduplicated by registry record and read as pipeline-intent, not a completed-study or program census. Closing it: a program-level dedup and a status/results overlay to convert intent into completion.
• Structural, not clinical — the deliberate scope. The audit reports who is publishing, trialing and patenting what, and how fast. It does not adjudicate which molecule is safer or more efficacious, rank drugs by clinical outcome, or make any medical recommendation. Any scientific, clinical or regulatory interpretation is the buyer's, with domain experts. This is a deliberate tier-scope, not a gap. Closing it: a dedicated clinical / regulatory evidence engagement — a different instrument from this velocity read.

This is a three-signal R&D-velocity and field-structure read at the R&D Velocity Audit tier — publication, trial, and patent (the patent leg at gold standard: full-text + classification, family-deduplicated). The natural next step is a deeper engagement that sharpens it: (1) a value-weighted IP overlay — citations, litigation, licensing and family geography — to convert the family-count holder ranking into a commercial-value map; (2) uncapped, pre-grouped per-period pulls for the precise velocity curve per sub-field; and (3) an identifier-level cross-corpus threading pass to complete the publication-to-trial-to-patent chain per asset. To commission it, reach the ForIntel desk directly at forintel@foragentis.com.