Every competitive-strategy team I have sat with watches its rivals through the same narrow window: the press release, the conference readout, the analyst note that lands three weeks after the news stopped being news (the kind a competitive-intelligence vendor will happily sell you back at a premium). It feels like intelligence. Most of it is theatre: a rival's communications team, or a banker paid to hold a view, telling you what they would like you to believe.
The most honest document a competitor produces is one they almost never point you towards.
It is the clinical trial protocol they filed with a regulator.
A press release is a brochure. A registered protocol is closer to a tax return: filed with an authority, costly to misstate, written by nobody to impress you. It is a commitment, and commitments reveal what slogans hide. Economists call this "revealed preference": you learn what someone actually values from what they do under constraint, not from what they announce. A rival's registered endpoint, comparator and eligibility criteria are their revealed preferences, sitting free on ClinicalTrials.gov, and most teams never read them as such.
Years ago I sat in a competitive-strategy session where the entire read on a rival's lead asset came from a conference poster and a two-line banker's note. Sharp people. Good instincts. Not one of us had opened the actual protocol. We were reverse-engineering a strategy from its shadow, while the document that would have answered half our questions sat public, free and three clicks away. When I later went looking for write-ups of this exact practice, I found almost nothing named. That absence is part of why it is worth writing down.
So stop treating ClinicalTrials.gov as a recruitment tool. It is already the most candid competitive intelligence your rivals will ever file, and almost nobody reads it that way.
Take one crowded indication: obesity, and the incretin-class drugs chasing it. Six sponsors, six current Phase 3 programmes, all searchable on 20 July 2026. Line their registered pivotal designs up next to each other and something appears that no single press release would ever hand you.
| Trial | Sponsor | Registered primary endpoint | Comparator | Notable eligibility feature |
|---|---|---|---|---|
| SURMOUNT-MMO (NCT05556512) [1] | Eli Lilly | Time to first MACE-plus-heart-failure composite event, up to 5 years | Placebo | Cardiovascular risk enrichment by design |
| ATTAIN-1 (NCT05869903) [2] | Eli Lilly | % change in body weight, Week 72 | Placebo | BMI ≥30, or ≥27 with a comorbidity; excludes T1D/T2D |
| NCT06131437 [3] | Novo Nordisk | Non-inferiority then superiority vs. tirzepatide, % body-weight change, Week 84 | Active: tirzepatide 15 mg (no placebo arm) | BMI ≥30; excludes T1D/T2D and HbA1c ≥6.5% |
| REDEFINE 1 (NCT05567796) [4] | Novo Nordisk | Co-primary: % body-weight change and ≥5% responder rate vs. placebo, Week 68 | Placebo, plus internal active arms | Overweight or obesity, BMI ≥27/30 |
| VANQUISH-1 (NCT07104500) [5] | Viking Therapeutics | % change in body weight, Week 78 | Placebo | BMI ≥30, or ≥27 with a comorbidity; excludes T1D/T2D |
| MARITIME-1 (NCT06858839) [6] | Amgen | % change in body weight, Week 72 | Placebo | Same BMI/comorbidity framing, plus an explicit exclusion for unstable major depression, severe psychiatric disorder and lifetime suicide-attempt history |
Read down the endpoint column first. Five of the six sponsors have powered their headline trial on the same surrogate: percentage change in body weight, read somewhere between Week 68 and Week 84. It is the fast route to a label. One sponsor breaks the pattern. Eli Lilly is running that surrogate trial (ATTAIN-1) and, in parallel, a five-year cardiovascular-outcomes trial (SURMOUNT-MMO) whose primary endpoint is time to a first major-adverse-cardiovascular-event-plus-heart-failure composite. Two entirely different evidentiary bets, same company, two different molecules. One buys a label. The other buys a claim a weight-loss-only rival cannot answer.
Now the comparator column. Read only Novo Nordisk's core pivotal (REDEFINE 1, a conventional placebo-controlled design with internal active arms) and you would file the company under "business as usual." You would be wrong. In parallel, Novo has run a separate trial (NCT06131437) with no placebo arm at all, testing its own candidate for non-inferiority and then superiority against Lilly's tirzepatide at 15 mg. The market leader's approved product, named in the design. That trial reads as COMPLETED, with 809 participants (the actual enrolment reported in the registry record). Read one protocol and you misread the strategy. Read the set, and the real posture shows. Novo is doing two things at once: shoring up its own filing, and picking a direct, named fight with the market leader.
The discipline is simple, and almost nobody applies it. Do not read the one trial you heard about at a conference. Pull the rival's whole registered programme ( and read the designs against each other.
Free download
The TPP Template
Target and minimally-acceptable profiles side by side, every claim linked to the evidence that has to support it.
Get the template →Once you have the set in front of you, three fields carry most of the signal, and here is how to read each for what it actually tells you about the strategy.
Endpoint: the bar they believe they will be held to. A primary endpoint is a public statement about the evidence a sponsor thinks it needs to win. A surrogate like weight change at 72 weeks is a bet on speed: label, market, worry about the rest later. A five-year cardiovascular-outcomes endpoint is a costlier bet on a durability-and-differentiation claim a weight-loss-only competitor cannot make. Read a rival reaching past the surrogate everyone else uses as the higher bar they have chosen to clear, and ask what it lets them claim that you cannot. This is minimum viable evidence read in reverse: the evidence package a competitor decided was worth the wait.
Comparator: confidence and commercial ambition, read with care. The control arm is where a sponsor tells you who they think they can beat. ICH E10 (the ICH's final guideline on choice of control group, adopted in 2000) puts it plainly: "The choice of control group is always a critical decision in designing a clinical trial. That choice affects the inferences that can be drawn from the trial... the acceptability of the results by regulatory authorities, and many other features of the study." [7] A placebo arm asks the modest question; an active head-to-head against the market leader, as in Novo's design, asks a far bolder one and puts real regulatory and commercial risk on the table. But resist the reflex that active comparator equals confidence. The same guideline warns that "there are circumstances in which a finding of non-inferiority cannot be interpreted as evidence of efficacy." [7] A non-inferiority design can be defensive as easily as bullish. Read the comparator with that sophistication, the way you would read any control-arm choice a regulator will interrogate, not as automatic swagger.
Eligibility: the population they are claiming, and the weakness they may be designing around. Eligibility text is the least-read, most revealing part of a protocol. Set three near-identical obesity pivotals side by side (Lilly's ATTAIN-1, Viking's VANQUISH-1 and Amgen's MARITIME-1) and one carries a criterion the other two do not. MARITIME-1 explicitly excludes patients with unstable major depressive disorder (MDD) or other severe psychiatric disorder, or a lifetime history of suicide attempt. That is a checkable, named difference in registered text: same disease, same calendar window. That is consistent with a sponsor managing a known sensitivity around neuropsychiatric safety. It is not proof of one — and here I have to be careful, because a single eligibility line does not license reading Amgen's mind. Treat it as a hypothesis to pressure-test, not a conclusion to bank. Keep the Pembrolizumab story in view: a biomarker-defined population, read at the time as a narrowing, turned out to be the whole strategy.
The same discipline, turned inside out, is where the real positioning value sits. What a protocol declines to test is as revealing as what it does, and it is the most concrete source of differentiation you will find.
A single trial won't show you this. You need the shape of a sponsor's whole registered programme. Viking Therapeutics has nine trials on the registry. Read across all nine and the absences line up: no cardiovascular-outcomes trial, no dedicated heart-failure study, no obstructive-sleep-apnoea trial, no paediatric programme. Lilly, Novo Nordisk and Amgen each field at least one registered trial in every one of those adjacent claim areas. That is real, checkable whitespace in one competitor's public pipeline.
That said, one discipline keeps this honest. A gap is evidence of an opportunity. It is not proof of a rival's private reasoning. A programme can be missing a cardiovascular trial because a sponsor made a deliberate strategic call, because a smaller company is sequencing its spend, or through plain institutional inertia, and the registry alone cannot tell you which. Published work on restrictive eligibility bears this out: across myeloma, triple-negative breast cancer and gynaecologic cancer, criteria stay narrow for years despite guidance pushing the other way, and nobody has cleanly separated deliberate strategy from unexamined default. [13] So read a gap as a lead, not a verdict.
So what do you do with a lead? Then act on it. Map every rival's registered claim areas against the disease's full patient population, and against each other. The empty cells are your differentiation shortlist, pulled straight off a public database. That is what a target product profile built on evidence looks like before anyone opens PowerPoint.
A protocol is a snapshot, not a strategy. Sponsors change endpoints all the time. A registered design can be a placeholder, a CRO's house template, or an eligibility list copied from the last programme without much thought. Read too much into a bureaucratic filing, the argument runs, and you will confidently infer intent that was never there.
It is a fair warning, and the fix is discipline, not retreat. You read directionally, never as certainty. Design and eligibility choices correlate with regulatory strategy and risk profile in aggregate, [11][12] which is enough to frame an inference as "consistent with" a bet and not enough to license a confident single-company mind-read. The protection is triangulation: never hang a read on one field of one trial. Pull the whole set, as with Novo above, and let the pattern carry the weight a single protocol cannot.
You might expect the next move to be watching a rival's protocol evolve, amendment by amendment: endpoints swapped, eligibility loosened, arms dropped, each edit a tell. The data exists in principle. ClinicalTrials.gov keeps a Record History. In practice you cannot audit it by hand. The history sits behind a view that resists a simple lookup, and the researchers who study amendments built bespoke software to do it, because manual inspection "is unfeasible for large sets of clinical trials." [10] Amendments are also common and rarely explained. In commercial trials they are now the norm: one analysis of 950 industry protocols found the share carrying at least one amendment rose to 76%. [8] Yet in a study of 1,746 trials, 55% changed a primary outcome at some point while only 1% of those reported it in the resulting publication. [9] So do not build a read on amendment diffs you cannot check. Read what is stable, filed and public.
At Inovia we run this same read at pipeline scale inside InoviaCS, because doing it by hand across a whole competitive set instead of one trial would eat a week nobody has. But the manual version above costs nothing. The registry hands your competitors' filed commitments to anyone willing to read them closely. Most teams never open the file.
Get the monthly digest
The 5 things evidence leads need to know each month: regulatory moves, RWE developments and what they mean in practice. No pitch, one email a month.