An AI-generated image illustrating real-time trial recruitment at the moment of diagnosis.
In statistics and machine learning, there’s a tradeoff between precision and recall: tune a system to be more selective and you inevitably miss more true positives. Cancer clinical trials have stumbled into a similar trap. As oncology has shifted from tumor-type-centered studies to gene-directed approaches, eligibility criteria have grown exquisitely precise. Instead of enrolling patients with triple-negative breast cancer (TNBC), for example, a trial may now look for specific biomarkers within that group that indicate how aggressively the tumor is growing or how the patient’s immune system is responding. Scientifically sound, but recruitment can suffer when a marker appears in a sliver of a subset.
That broader challenge was front and center at Proscia’s Access25 event in late October. The company showcased Aperture, a platform designed to surface trial-eligible patients at the moment of diagnosis—rather than months later when data pipelines finally refresh.
“Aperture surfaces clinical signals in real time at the point of diagnosis—not months later.” —Nathan Buchbinder, co-founder
Participation in cancer treatment trials sits at roughly 7%, a rate that hasn’t changed substantially over time despite decades of initiatives. Meanwhile, the clock is unforgiving: the median time-to-treatment initiation in 2020 was 26 days. Many EHR datasets refresh monthly with 30-day recency, and Medicare research files often land 4.5–5.5 months after the quarter closes. Even when comprehensive tumor sequencing is ordered at diagnosis, turnaround times are typically 8–11 days for FoundationOne CDx and 10–14 days for Tempus xT. By the time lagging data catch up, many patients have already started therapy and become ineligible.
A neoadjuvant TNBC example
Proscia’s demo walked through a neoadjuvant TNBC scenario—an indication with tight eligibility windows and stringent biomarker requirements. A sponsor could configure Aperture to watch for newly diagnosed TNBC cases in patients with operable, early-stage disease who have not yet received systemic therapy and whose tumors show specific biomarkers indicating rapid growth and a significant immune response
“Rather than navigating a litany of drop-downs, you can describe the study in natural language, and the system aligns it to how the underlying data are structured and labeled.” —Buchbinder
With real-time alerts enabled, the recruitment timeline compresses dramatically. As Buchbinder explained during the demo, “the window between diagnosis and treatment initiation can be as short as two to three weeks,” so timing is key. Aperture addresses this by notifying trial teams immediately when eligible patients are identified. In the live demonstration, when a new patient match appeared, Buchbinder noted: “This patient was just diagnosed. Their pathology workup was just completed at one of our network labs, and we’re being notified now before they’ve started any treatment that would make them ineligible.”
The infrastructure angle
Aperture’s timing claim rests on plumbing,. Modern precision trials often hinge on criteria that first coalesce in pathology: morphology, IHC readouts and linked molecular results. With clinical-grade digital pathology now cleared for primary diagnosis in the U.S. and adoption widening, those signals exist in structured form inside the lab before they ripple out to EHRs or payer datasets.
The trend is to meet the data where it’s born, standardize it and stream it. Concentriq, for instance, sits in the diagnostic workflow, where each case generates whole-slide images, pathologist interpretations and metadata. Aperture layers AI on top to extract and normalize features across staining protocols and scanners, tokenize patient records and make cohorts queryable in natural language.
Proscia describes tokenization of PHI, lab-controlled recontact workflows back to treating physicians, and audit trails suitable for regulated use. Sponsors will still expect validation artifacts, versioning, and clear provenance across slide, stain, algorithm and report. That is, as the Access2025 event noted, a cohort can continue identifying matches over time as new patients are diagnosed, which means engaging patients before treatment begins and eligibility windows close.
Filed Under: Oncology
