Company Spotlight: ArteraAI Breast - Predicting Metastasis from H&E Slides

Maya: Imagine being able to predict a patient's 5- and 10-year risk of breast cancer metastasis using just standard pathology slides and a few basic clinical variables.

Sam: Today we are looking at the exact FDA clearance letter for a tool that aims to do just that. It is the 510(k) clearance letter and Indications for Use form for ArteraAI Breast, published by the U.S. Food and Drug Administration on May 4, 2026.

Maya: And before we get started, please note that our voices are AI-generated, and this episode is a summary of a publicly available FDA document, not company marketing material.

Sam: Right, we are purely looking at what the FDA put on the public record in document K254115. So, Maya, as our resident clinician-turned-analyst, let us start from the top. What exactly is this device?

Maya: According to the letter, ArteraAI Breast is a Class II device. The FDA classifies it as a software algorithm device analyzing digital images for cancer prognosis. And importantly, the indications for use specify that it is a software only device.

Sam: Software only. So there is no proprietary scanner or new physical lab equipment involved here?

Maya: Exactly. In fact, the document states that ArteraAI Breast is intended to use whole slide images acquired from FDA-cleared interoperable scanners and file formats that have been validated for use with it.

Sam: Okay, so a hospital or lab scans a slide on an FDA-cleared scanner they already have, and then this software takes over. What exactly is it looking at on those slides?

Maya: The software analyzes a scanned histopathology whole slide image from treatment-naive breast resection specimens. And it is very specific about the tissue preparation. It has to be prepared from formalin fixed paraffin-embedded tissue, often called FFPE, and stained using Hematoxylin and Eosin stains.

Sam: Hematoxylin and Eosin. That is the standard of care in pathology labs everywhere.

Maya: It really is. That is what makes this interesting for the boardroom and the clinical ward. It integrates into an existing, ubiquitous tissue preparation workflow. But the image alone is not the only input.

Sam: Wait, really? What else does the software need to make its prediction?

Maya: The indications for use list three additional inputs that must be provided by the physician. Those are patient age, tumor size, and nodal status.

Sam: So it combines the digital image analysis of the standard slide with the patient age, the size of the tumor, and the nodal status. And what does it spit out at the end?

Maya: The output is twofold. It provides 5- and 10-year risks of distant metastasis and an ArteraAI risk score.

Sam: A 5- and 10-year risk of distant metastasis. That is a heavy piece of information to hand to an oncologist and a patient. Who exactly is this for? Is it for anyone who walks into a clinic with breast cancer?

Maya: No, the patient population is extremely specific, and this is where clinicians need to pay close attention. It is for adult patients with early-stage invasive breast cancer that is HR-positive, HER2-negative, and N0 or N1.

Sam: Let me break that down. HR-positive, HER2-negative. N0 or N1 means they have either no lymph node involvement or very limited node involvement, right?

Maya: Correct. And the document adds a few more guardrails. It specifies these patients must be without clinically or pathologically defined metastases after surgical tumor resection. Furthermore, they must be candidates for standard of care adjuvant therapy.

Sam: So we are talking about early-stage patients who just had surgery, have no known metastases yet, and the doctor is trying to figure out their risk over the next decade to guide what happens next.

Maya: Exactly. The stated intended use is to assist physicians with prognostic risk-based decisions along with other clinicopathological factors.

Sam: Okay, that explains what it does and who it is for. Now let us talk about the proof. If this is an FDA clearance, it means it was cleared based on being substantially equivalent to a predicate device. What was the predicate device they compared it to?

Maya: Because we are looking strictly at the administrative clearance letter today, and not the full 510(k) summary, it does not actually say. The letter simply notes the FDA determined the device is substantially equivalent to legally marketed predicate devices that were on the market prior to May 28, 1976, or to reclassified devices.

Sam: Right, it is just the official green light. It is an administrative document, not the full clinical dossier. So I am guessing we will not find the performance testing numbers or accuracy metrics in this specific document either.

Maya: Exactly. The clearance letter does not show the sensitivity, the specificity, or the positive predictive value. It does not state the size of the test dataset or the demographics of the patients used to train the software.

Sam: That is a very important distinction for our listeners. To be clear, we are not saying the product was not rigorously tested. Obviously, the FDA reviewed extensive data to grant this clearance, but a standard clearance letter naturally omits the real numbers and the dataset details.

Maya: Precisely. And that is why we always remind our audience that FDA clearance means substantial equivalence was demonstrated to the agency's satisfaction, but you have to look beyond the clearance letter, usually to the 510(k) summary once it is available, to find the clinical validation data.

Sam: Let us shift gears to a regulatory aspect of this letter that really stood out to me. There is a whole section in here about a Predetermined Change Control Plan or PCCP. What is that?

Maya: This is a critical concept for health-tech companies building AI. The FDA states that their substantial equivalence determination included the review and clearance of a Predetermined Change Control Plan.

Sam: Why does a software company want one of those plans?

Maya: Because machine learning models evolve. The letter explains that under the Act, a new premarket notification is not required for a change to a cleared device if the change aligns with an established Predetermined Change Control Plan.

Sam: So if they want to tweak the algorithm, or update the model, they do not have to go through the whole FDA submission process again, as long as it fits within the plan they already agreed on.

Maya: Exactly. But there are strict limits. The FDA reminds them in the letter that they do need a new submission if there is a major change or modification in the intended use of a device.

Sam: What else triggers a new submission?

Maya: A new submission is required if a modification could significantly affect the safety or effectiveness of the device. The letter gives specific examples, like a significant change or modification in design, material, chemical composition, energy source, or manufacturing process.

Sam: And what happens if a company strays outside that predetermined plan without telling the FDA?

Maya: The FDA does not mince words. The letter states that failure to submit a new premarket submission in those cases would constitute adulteration and misbranding under the Act.

Sam: Adulteration and misbranding. Those are the kinds of words that keep regulatory affairs executives awake at night.

Maya: They certainly are. And the letter actually provides titles of two FDA guidance documents to help companies figure this out. One is about deciding when to submit a 510(k) for a change to an existing device, and the other is specific to deciding when to submit for a software change.

Sam: It is interesting how much of this clearance letter is actually just laying out the ongoing rules of the road. It also goes into the Quality Management System Regulation, right?

Maya: Yes, it spends a whole paragraph on it. It explicitly states the device is subject to 21 CFR Part 820. It even highlights that the manufacturer must follow specific ISO 13485 clauses for design controls, nonconforming products, corrective actions, and preventative actions.

Sam: That is incredibly specific. For the founders and operators listening, that is your checklist. You cannot just build a cool AI model; you have to have the enterprise-grade quality management systems to back it up.

Maya: And it goes further. The FDA notes that regardless of whether a change requires premarket review, the quality management rules require device manufacturers to review and approve changes to device design and production and document those changes in the Medical Device File.

Sam: So even if your plan says you can make a change without telling the FDA, your internal paperwork still has to be flawless.

Maya: Exactly. The letter also reminds Artera that they must comply with the Unique Device Identification System rule, or UDI Rule, which means the device needs a unique identifier on its label and package unless an exception applies.

Sam: Alright, zooming out for a second. We have a software-only device that gives 5- and 10-year metastasis risk scores for early-stage breast cancer patients using H&E slides and three clinical data points. As a clinician, how do you see this fitting into the broader picture of oncology?

Maya: It shows a clear trend of trying to extract more prognostic value from the diagnostic materials we already collect. An H&E slide is standard. If software can pull a 10-year risk profile from that without needing entirely new, complex tissue processing, that could be very useful for risk-based decisions.

Sam: But with the massive caveat that we do not know the performance numbers from this document.

Maya: Exactly. A clinician cannot evaluate whether to trust this specific tool without seeing the clinical validation data, the dataset diversity, and the confidence intervals, none of which are in this clearance letter.

Sam: Before we wrap, I noticed a funny little detail at the very end of the document on the Indications for Use form. There is a Paperwork Reduction Act statement.

Maya: Yes, the FDA estimates the burden time for the collection of information on that specific form to average 79 hours per response.

Sam: 79 hours just for the paperwork burden of that form! A little reality check for anyone thinking the regulatory path is quick and easy.

Maya: Indeed. If companies have questions about these regulations, the letter actually advises contacting the Division of Industry and Consumer Education, or DICE.

Sam: Well, time for our crisp recap. ArteraAI Breast received FDA clearance as a software-only device to analyze standard H&E digital slides, along with age, tumor size, and nodal status, to predict 5- and 10-year distant metastasis risks for a specific cohort of early-stage breast cancer patients.

Maya: The clearance includes a Predetermined Change Control Plan, offering a regulatory pathway for future algorithm updates. However, because this is just the clearance letter, it does not disclose the predicate device name or any performance testing data.

Sam: As always, you can find a link to the full FDA document in our show notes so you can read it for yourself.

Maya: And remember, this podcast is for informational purposes only and is not medical advice. Thanks for listening.