If you’ve ever watched a clinical trial stall before enrollment really gets going, you know how painful it is. Every week lost is expensive. And a surprising amount of that delay comes down to one basic issue: patients don’t make it through prescreening.
Not because they’re ineligible, often they are, but because they never get through the labs, imaging, or diagnostic steps required to confirm it. Life gets in the way. Travel, scheduling, cost, or just inertia. Meanwhile, the data that could explain who’s likely eligible, lab values, comorbidities, medication history, biomarkers, sits scattered across hospitals, labs, and registries that can’t legally share it.
The result is a leaky funnel, sites stretched thin, and a long wait before the first patient ever gets enrolled.
But the industry is starting to move in a different direction.
There’s growing momentum behind AI-driven prescreening models, tools that help Pharma estimate a patient’s likelihood of meeting eligibility criteria before a single test or appointment happens.
If you can predict who’s a strong candidate early, you can:
The potential upside is big. The roadblock is just as big: these models rely on sensitive data that can’t be moved, pooled, or exposed.
That’s exactly where privacy-enhancing technologies (PETs) come in.
Our upcoming white paper breaks this down in more detail, but the basic idea is simple:
PETs let you run the model where the data is, without moving or revealing any of it and without exposing the Pharma’s proprietary model either.
Duality’s platform uses two key technologies to make that possible:
Think of them as secure “rooms” inside modern hardware. Data and models go in, computation happens, and nothing inside the room can be viewed by cloud operators, admins, or anyone else.
Instead of shipping data to the model, the model is securely sent to each hospital or lab and runs locally right where the data already lives.
Together, this means:
And suddenly, predictive prescreening becomes practical, not theoretical.
Trial operations teams know the pain points all too well:
These aren’t issues you fix with more staff or more reminders. They’re structural. And PETs are finally giving Pharma a way to get ahead of them instead of reacting to them.
When you can run predictive models directly on distributed data privately, securely, and at scale you remove the friction that has slowed trial startup for decades.
One global pharma company is using a PET platform to run its biomarker prediction model across three very different data partners: a hospital network, a diagnostics consortium, and a national claims database.
No patient-level data moved.
No model IP exposed.
Each site running the encrypted model locally inside a TEE.
Only encrypted eligibility scores came back.
What changes?
It’s shifting prescreening from slow and manual to fast, predictive, and trusted.
In the next couple of years, PET-enabled predictive modeling is going to become part of standard trial startup. Not only for prescreening, but eventually for feasibility, protocol design, adaptive enrichment.
This isn’t a privacy compromise. It’s privacy used as an enabler.
And pharma who take advantage of it will activate trials faster, reach more patients in more regions, and lift a huge operational burden off their sites.
If you’re serious about cutting prescreening timelines without creating privacy headaches, let’s talk about how PETs can fit into your prescreening strategy.
