Chemotherapy decision-making in early-stage HR+/HER2− patients remains a clinically contested gray zone, but the validation of an FDA-cleared digital pathology–based risk stratification tool for breast cancer provides a significant development in patient care.
Here, Calvin Chao, Vice President of Medical Science at Artera, discusses the benefits of this technology and the implications for clinical care.
What clinical or diagnostic gap in breast cancer pathology was this AI risk stratification tool designed to address?
HR+/HER2-negative early-stage breast cancer is a heterogeneous disease. Although adjuvant chemotherapy is recommended for many patients at high risk of recurrence, clinicians need accurate prognostic tools to guide treatment decisions. Genomic assays are widely used but must be interpreted alongside established pathologic features, including tumor grade. Even with this combined approach, many patients are classified as having intermediate-risk disease, leaving uncertainty about the benefit of chemotherapy.
In addition, genomic assays are typically performed as send-out tests to centralized laboratories, resulting in higher costs, longer turnaround times, and limited accessibility in some settings, particularly in resource-constrained healthcare systems.
AI-powered digital pathology offers a potential alternative. These models require only a routinely prepared pathology slide image, together with standard clinical variables, allowing prognostic assessment to be integrated into existing pathology workflows. Because testing can be performed locally on digital slides, results may be available more quickly at the time of diagnosis, potentially improving access to risk stratification while reducing reliance on centralized molecular testing.
How does the tool integrate into existing pathology workflows?
Receiving FDA clearance as Software as a Medical Device (SaMD), allows the tool to be deployed in pathology laboratories using compatible FDA-cleared digital slide scanners. FDA clearance was supported by clinical and analytical validation studies demonstrating the software's safety and performance.
The test is intended for use after a pathologist has established the diagnosis of breast cancer. To perform the analysis, the pathology laboratory uploads a digitized image of the breast resection specimen containing the highest-grade tumor, along with key clinical information, including patient age, tumor size, and nodal status. Results are typically available within approximately one hour, enabling the prognostic assessment to be reviewed by the pathologist and returned to the treating clinician alongside the pathology report.
What validation and clinical performance data were most important in achieving FDA clearance?
The FDA based its clearance on both clinical and analytical validation data. The agency evaluated whether the AI model could accurately stratify patients into risk categories, such as low- and high-risk disease, in contemporary patient populations. It also assessed the reproducibility of the model's results across compatible digital slide scanners and different operators to ensure consistent performance.
The clearance follows an earlier FDA authorization of an AI-powered digital pathology platform for prostate cancer, reflecting the agency's growing experience in evaluating AI-based prognostic tools for clinical use.
How significant is this FDA clearance for the broader adoption of AI-assisted digital pathology tools?
FDA clearance for AI-powered digital pathology tools in both breast and prostate cancer demonstrates the potential for this technology to support prognostic assessment across multiple tumor types. However, regulatory clearance is only one step toward clinical adoption. The FDA permits the software to be marketed and deployed by healthcare institutions. However, broader implementation depends on evidence that the technology addresses an unmet clinical need, integrates into existing workflows, and provides value comparable to or greater than current standards of care.
As additional validation and real-world evidence become available, adoption will likely be driven by clinician confidence, clinical utility, reimbursement, and incorporation into routine oncology practice.
What is the potential impact of this technology on cancer care?
The AI-powered digital pathology test is designed to support treatment planning after the diagnosis has been established by providing additional information about a patient's risk of recurrence. The results are intended to help treating clinicians determine which patients are most likely to benefit from adjuvant chemotherapy, with the goal of reducing both overtreatment and undertreatment.
How might the technology affect workflow efficiency?
Because the assessment can be performed using routinely prepared digital pathology slides and standard clinical information, results may be available at the time of diagnosis. Compared with tests that require samples to be sent to centralized laboratories, a shorter turnaround time could accelerate treatment planning, reduce delays in care, and provide patients and clinicians with prognostic information earlier in the decision-making process.
What barriers still exist for wider implementation of AI-enabled pathology tools in clinical laboratories?
A core foundational piece of our AI tool is digital pathology, but unfortunately, adoption is still not as widespread and entrenched as we would like. However, we’ve heard from many pathologists and lab directors that the appeal of tests like ours is what will help drive the adoption of digital pathology and hopefully remove that as a barrier in the years to come.
Looking ahead, how do you see AI-driven risk stratification influencing future cancer diagnostics and precision oncology?
AI and precision oncology go hand in hand. As we learn more about the complexities of cancer and as treatment options continue to proliferate, we will have to rely on AI tools to help optimize treatment approaches based on prognostic and predictive assessments. While AI technologies are transforming every aspect of our society, the key question for AI tools in health care is not necessarily how powerful it is, but rather how easy it will be to integrate into existing workflows.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
