A British AI company called C the Signs is pursuing FDA De Novo clearance for a tool that scans a patient’s existing medical records, physician notes, prescriptions, and test results to flag people who may be at risk of cancer, without ordering a single new scan or lab test. If cleared, the company says it would be the first device of its kind authorized in the United States, since the FDA has told the company it has not identified a comparable product already on the market.
Mining Records Instead of Ordering New Tests
Most AI cancer-detection tools work by analyzing an image, a scan, or a tissue sample that has to be newly generated. C the Signs takes a different approach, applying machine learning to data clinicians have already collected, essentially re-reading the medical record for patterns a busy doctor might not connect across multiple visits, specialists, or years of history. Dr. Bea Bakshi, the company’s co-founder and CEO, has framed the tool as a way to catch risk signals hiding in plain sight, in information that already exists but is scattered across a patient’s chart rather than assembled into a single risk picture.
A Track Record Built in the U.K.
C the Signs has been used by more than 11,000 healthcare professionals across the United Kingdom’s National Health Service, where the company says it has helped identify 75,000 cancer patients and reduced diagnostic timelines by 21%. That is a substantial base of real-world usage data for the company to bring to U.S. regulators, though the NHS’s centralized, single-payer record-keeping system differs meaningfully from the fragmented, multi-vendor electronic health record landscape in American hospitals, where the same patient’s data is often split across systems that do not talk to each other.
Why the De Novo Pathway, and Why That’s Notable
The FDA’s De Novo pathway is reserved for novel, lower-to-moderate risk devices that do not have an existing comparable product to serve as a predicate, meaning the agency has to build a new regulatory classification rather than measuring the tool against something it has already cleared. The FDA’s acknowledgment that it hasn’t identified a similar device already on the market suggests C the Signs would be establishing a new device category for records-based cancer risk detection, a designation that, if granted, could open the door for competitors to follow using the classification as a new predicate.
The Skeptical Case
Tools that mine existing medical records for risk signals face a validation challenge distinct from image-based diagnostics: their output is only as good as the data already in the chart, and incomplete, inconsistent, or delayed documentation across different providers and specialties can degrade performance in ways that are hard to test for in a controlled study. Critics of AI risk-scoring tools broadly have also raised concerns about false positives generating unnecessary follow-up testing and patient anxiety, and about whether performance validated on a relatively homogeneous NHS population and record-keeping system will hold up amid the far more fragmented data landscape of U.S. healthcare, where the same patient’s records might live across three or four unconnected systems.
Scaling the Evidence Base
To address exactly that generalizability question, the company has said it plans to include 250,000 Americans in a study this year, a considerably larger cohort than many AI diagnostic validation studies rely on, aimed at demonstrating the tool performs consistently across the more varied U.S. healthcare data environment before regulators sign off. The company previously drew a notable investment from Khosla Ventures to help fund this expansion into the American market.
What’s Next
C the Signs says it hopes to launch commercially in the U.S. next year if the De Novo clearance comes through, a timeline that would put a records-mining cancer risk tool into American clinics well within the next 18 months. Given how much of American healthcare’s cancer-detection problem stems from patients falling through the cracks between specialists rather than from a lack of underlying data, a tool that surfaces existing warning signs without requiring new tests could prove attractive to primary care practices looking for a low-cost way to catch missed diagnoses, assuming the pending U.S. validation study delivers results as strong as the company’s U.K. numbers.
