The shared pool: reading competitive saturation off the registry before you sign a site

You have a site you like. Strong institution, an investigator you have known for years, prior enrolment numbers that stack up on every question your feasibility survey thought to ask. On paper it is the easiest yes on your site-selection shortlist.

Now open ClinicalTrials.gov and search that investigator by name. In IgA nephropathy, the indication I have been digging through for this piece, three physicians on a plausible shortlist turn out to be tied to three separate Phase 2/3 trials, all currently recruiting, all from different sponsors, all fishing the same local patients. In two of those cases the physician appears under a second, differently-branded facility name on one of the trials, so a check that matched on site names would have waved him straight through.

That is the blind spot. Your feasibility questionnaire measured whether the site could run your trial. It never asked the one question that decides whether the site can fill it: who else is already recruiting from the same patients? Enrolment is a shared-pool problem, and the answer sits in a public registry the survey cannot reach.

Why the feasibility survey can't see the competition

Two structural blindnesses, both baked in.

The first is what the survey measures. A feasibility questionnaire is a capability instrument. It asks how many staff the site has, what equipment sits in the building, how many eligible patients pass through the clinic in a month, what the investigator has run before. Every one of those is a property of the site in isolation. None of them captures what the site is doing for everybody else. Capability and competition are different axes, and the form only has a column for one of them. That is precisely the kind of gap the modern biotech playbook is meant to close [1].

The second is who fills it in. A feasibility survey is a negotiation dressed as a measurement. The party completing the form is the party that wins the study if the answers look good, so the incentive runs one way. The ASCO Research Statement on site feasibility (Kurbegov et al., 2021) is blunt about the state of the practice, calling current assessment "costly, inefficient, unnecessarily burdensome, and resource intensive" and, overall, of "uncertain effectiveness" [2].

Even where nobody is gaming anything, the estimate skews optimistic for structural reasons. Hulstaert and colleagues (2024) put their finger on why, writing that "this overestimation is because investigators do not have full protocol information and limited time for a thorough trial feasibility assessment" [3]. Bogin (2022) is blunter about how sponsors ought to read the numbers, noting that "sponsors rightfully view some CRO's optimistic recruitment projections as a marketing strategy" [4].

Years ago I worked on an early-phase programme in a small indication where our lead site aced feasibility on every axis: a respected investigator, a dedicated coordinator, a clinic apparently full of exactly the right patients. It then enrolled a fraction of what it had promised. Nobody had lied to us. The site was simply running three other studies that drew from the same clinic, and we were fourth in the queue for the same people. The survey had no box for that.

So the Monday change is small and consequential. Stop treating the completed questionnaire as your primary enrolment signal. It is a starting bid, and the track record sits elsewhere, in public and free to read.

The shared pool: what saturation does to accrual

The eligible patients around a site sit over a shared aquifer. Every trial recruiting in that indication sinks another well into the same water table, and your feasibility survey only ever measures the pump: how fast this one site can draw, never the wells already sunk around it. Two sites can each look excellent alone and still come up dry together, because they were never drawing from separate supplies.

This is measurable, and the measurement is uncomfortable. Bennette and colleagues (2015) studied low-accruing NCI cooperative-group trials and found they had been launched into more crowded competitive settings: a median of 4.4 competing trials per 10,000 eligible patients per year, against 2.9 for the rest. Put as a risk, each additional competing trial per 10,000 eligible patients carried an odds ratio of 1.88 (95% CI 1.32–2.68) for low accrual [5]: nearly double the odds of under-enrolling, per rival in the pool. That is the number to hold in your head when a site tells you it can deliver.

And the pool is thin before you arrive. Tran and colleagues (2020), working from the AACT database, put the national ratio at 12.6 newly diagnosed patients for every open oncology trial slot, ranging by cancer type from 6.0 in brain and CNS to 24.7 in colorectal [6]. Those are the odds of the draw. Your indication may be kinder or crueller, but the shape holds: finite water, more wells arriving every quarter.

The operational consequence is unavoidable. Discount your enrolment assumptions by the competitive load on each site; do not accept the figure the survey volunteered at face value. An under-delivering site does not only cost you patients; it burns the runway you are measuring every other decision against [7].

You can check this today: three investigators, three sponsors, one pool

None of this is hypothetical. Everything below came out of a public ClinicalTrials.gov search run on 20 July 2026, in one indication, in a single afternoon.

In IgA nephropathy, a public registry search identified three investigators participating across three concurrently recruiting industry Phase 2/3 trials from three different sponsors: Novartis's iptacopan biopsy study (NCT06797518), Roche's sefaxersen study IMAGINATION (NCT05797610), and Takeda's mezagitamab study (NCT06963827). Across the three studies, the same investigators appear repeatedly in the same disease area and overlapping geographic catchments. Three sponsors, all recruiting through the same investigator network.

Here is the part a name-match would miss. Call it the facility-name blind spot. In the Takeda trial, two of those investigators appear under different facility names than they do elsewhere in the public registry. Match your shortlist on the letterhead and they look like different sites. Match on the investigator instead, and you reveal that what appear to be separate facilities are, in fact, linked through the same investigator and, by extension, the same underlying referral network.

The competition is not only from rivals. Novartis alone runs at least six simultaneously open IgAN protocols across three different assets iptacopan, zigakibart and atrasentan [8]. A single sponsor can create internal competition for the same investigator network before a competitor even enters the picture.

And the registry records the withdrawals too. Roche's IMAGINATION trial lists 204 sites; nine US locations are marked WITHDRAWN while the trial as a whole remains open to recruitment [8]. Sites selected, then pulled before they ever opened a chart. A feasibility survey captures a site's intentions; the registry captures what happened.

The Monday move: search by investigator, not by facility, across your entire competitive landscape before you sign anyone.

Note: ClinicalTrials.gov records participating investigators and recruiting facilities but does not disclose referral patterns, screening activity or enrolment volumes. Accordingly, the observations above are based on publicly listed investigator overlap and indicate the potential for recruitment competition; they do not demonstrate shared patients, referral pathways or actual enrolment.

The site-selection registry pass you can run this week

How do you check for competitive saturation before signing a site?

Here is a disciplined ClinicalTrials.gov read a lean team can do itself, without a CRO engagement. Six steps.

• 1. Map your competitive set by sponsor. Search by sponsor across every asset chasing your indication and list each protocol still RECRUITING or ACTIVE_NOT_RECRUITING. The Novartis example, six open IgAN protocols from one company, is why this comes first: you cannot judge one site's load until you can see the whole field drawing on it.
• 2. Search investigators, not facilities. Query by condition and geography, then match on the physician and their referral network rather than the name over the door. This is the only check that catches the facility-name feint. A site-name match sails straight past it.
• 3. Read the prior-enrolment cadence. Pull a shortlisted site's completed trials in the indication and work out patients per site per year from the registry's own numbers. 
• 4. Weight for start-up speed. Site activation time predicts enrolment success and varies enormously between sites. Goyal and colleagues (2021) found a median cardiovascular-trial site start-up of 255 days (IQR 177–350), while the top decile averaged around 106 days [10]. Ratnayake and colleagues (2025), across 315 studies at one NCI-designated centre, found that studies which hit their 70% accrual threshold had a median activation of 140.5 days against 187 days for those that fell short, with each extra activation day cutting the odds of study success by 0.5% [11]. A slow site is not a neutral site.
• 5. Sanity-check against a registry-only baseline. Structured registry fields, with no self-report at all, already predict enrolment better than nothing. A epidemiological model forecasting patients by site based on rich data is a genuinely powerful tool to not only identify potential sites but also sense check CRO predictions (InovaCS does this incredibly well). 
• 6. Learn from the graveyards. Pull the terminated trials in your indication and read their target against their actual enrolment. Two examples make the point: RemeGen's telitacicept study (NCT04905212) registered a per-protocol target of 30 and shows 15 enrolled [13]; EMD Serono/Merck KGaA's atacicept trial (NCT02808429) enrolled 16 across 18 sites, listed globally, before terminating [14]. This gives you a shortfall you can calculate straight from the trial's own target-versus-actual numbers and is signal a survey will never send you. (Treat a recruiting trial's enrolment as an estimate and a completed or terminated figure as actual; the registry does not label them, so that reading is an inference from status.)

The honest objection: every good site is busy

Every capable site is busy; the fast, experienced, in-demand sites are precisely the ones your own start-up-speed data tells you to want, which means a busy site is very often a signal of quality rather than a red flag, and competition is unavoidable in any hot indication anyway. You cannot out-manoeuvre an entire field, and if you strike off every saturated site on your list you will have no list left. So why not just trust the survey and the KOL relationship and get on with it?

Because the goal was never to avoid busy sites, which is impossible and often wrong. A busy site genuinely can signal capability and saturation at once, which is exactly why you want both readings rather than betting everything on one. What the registry check actually buys you is an honest count of the competing load instead of an inflated enrolment assumption, the option to place sites in catchments that do not overlap rather than stacking three trials on one pool, and a guard against unknowingly counting the same investigator twice behind two facility names. Carry that into your next CRO negotiation and you are arguing from a position you can defend, not a hope.

What the registry still won't tell you

There is plenty the registry still will not hand you: A flag telling you whether an enrolment count is an estimate or a final number, a single view of every trial competing for one site's patients that is assembled for you instead of by you. Until those exist, you build that picture yourself, out of public data, one investigator search at a time. It is unglamorous work, and it is also the difference between walking into clinical trial site selection with a track record in hand and walking in with a promise.

That assembly, the competitive landscape and the site history and the enrolment signal pulled into one strategic read, is the kind of registry intelligence InovaCS was built to compress. But the discipline matters more than any tool. Search the investigator, not the letterhead, and count the wells before you trust the pump.

References

[1] The biotech playbook is in desperate need of an update. Inovia Bio. https://blog.inovia.bio/inovia-bio-insights/updated-biotech-playbook

[2] Kurbegov D, et al. (2021). ASCO Research Statement on site-based feasibility assessment. PMID: 33405975. https://pubmed.ncbi.nlm.nih.gov/33405975/

[3] Hulstaert L, et al. (2024). "Enhancing site selection strategies in clinical trial recruitment using real-world data modeling." PLOS ONE. DOI: 10.1371/journal.pone.0300109. https://doi.org/10.1371/journal.pone.0300109

[4] Bogin V. (2022). Narrative review on clinical trial recruitment and site feasibility. PMID: 35198795. https://pubmed.ncbi.nlm.nih.gov/35198795/

[5] Bennette CS, et al. (2015). Journal of the National Cancer Institute. PMID: 26714555. https://pubmed.ncbi.nlm.nih.gov/26714555/

[6] Tran G, et al. (2020). JCO Clinical Cancer Informatics (built from the AACT database). PMID: 31977253. https://pubmed.ncbi.nlm.nih.gov/31977253/

[7] How Biotechs Can Maximise Runway. Inovia Bio. https://blog.inovia.bio/inovia-bio-insights/biotechs-maximise-runway

[8] IgA nephropathy competitive landscape, ClinicalTrials.gov, searched 20 July 2026: Novartis "iptacopan biopsy study" NCT06797518 (https://clinicaltrials.gov/study/NCT06797518); Roche sefaxersen "IMAGINATION" NCT05797610 (https://clinicaltrials.gov/study/NCT05797610); Takeda mezagitamab NCT06963827 (https://clinicaltrials.gov/study/NCT06963827).

[9] Mayo Clinic IgAN trial history, ClinicalTrials.gov: NCT00498368 (https://clinicaltrials.gov/study/NCT00498368); NCT02282930 (https://clinicaltrials.gov/study/NCT02282930).

[10] Goyal N, et al. (2021). Cardiovascular clinical-trial site start-up analysis. PMID: 34297072. https://pubmed.ncbi.nlm.nih.gov/34297072/

[11] Ratnayake S, et al. (2025). Study activation time and accrual success at an NCI-designated cancer centre. PMID: 41211192. https://pubmed.ncbi.nlm.nih.gov/41211192/

[12] Bieganek C, et al. (2022). Predicting clinical-trial enrolment from pre-trial ClinicalTrials.gov structured fields. PMID: 35202402. https://pubmed.ncbi.nlm.nih.gov/35202402/

[13] RemeGen. Telitacicept in IgA nephropathy (TERMINATED). ClinicalTrials.gov: NCT04905212. https://clinicaltrials.gov/study/NCT04905212

[14] EMD Serono / Merck KGaA. Atacicept in IgA nephropathy (TERMINATED). ClinicalTrials.gov: NCT02808429. https://clinicaltrials.gov/study/NCT02808429

[15] ICH E6(R3) Good Clinical Practice. Final; adopted 6 January 2025. International Council for Harmonisation. https://www.ich.org/page/efficacy-guidelines

[16] The blessing and curse of KOLs in biotech. Inovia Bio. https://blog.inovia.bio/inovia-bio-insights/kols-in-biotech