September 25, 2018 – Pathology AI Under a Central Lab Model: a cost-effective alternative and smart intermediate step for risk mitigation of a distributed medical device.
A pathology AI system with an intended use for clinical diagnostics (Dx), prognostics (Px) or companion diagnostics (CDx) can be commercialized under a central lab model either as a lab developed test (LDT) or a single-site medical device.
Compared to the development of a commercially-distributed medical device, a central lab model can provide agile, lower-risk and cost-effective alternatives and a smart intermediate step (or two) to a commercially-distributed medical device. See our previous article, Pathology AI as a Medical Device, for a discussion on what it takes to commercialize a pathology AI system as a distributed medical device.
In a central lab model, clinical laboratories across a country send their tissue samples for testing to a national central laboratory and get the test reports back. A single-site laboratory can provide an effective distribution channel to an entire country! In the US, a similar reference lab model is already well established for complex IHC testing. See our previous article, Global Service Model for Tissue Image Analysis, for a discussion on what it takes to scale from a single-site central lab model to a global multi-site model.
Central laboratories (single-site and multi-sites) need to establish a quality management system (QMS) to perform digital pathology and pathology AI in a regulated environment. In this setting, only the central laboratories need to adopt digital pathology and buy a slide scanner. There is no dependency on digital pathology adoption for any other laboratory! Local laboratories only need a computer and internet connection to view images of the histology slides and access data and reports.
Clinical laboratories in the US can provide a pathology AI system as a lab developed test (LDT) under CLIA regulations and CAP standards.
An LDT can be assembled using different lab-specific components that do not need to be medical devices. Any slide scanner and 3rd party software, like Photoshop, can be used to build a validated LDT. In this scenario there is no need to manufacture the pathology AI system as a medical device!
Pathology AI systems can move seamlessly from research to a clinical laboratory. Pathology AI applications can be developed as research prototypes until they show clinical utility in an exploratory setting. The prototype can then be transferred into the clinical laboratory as an LDT and distributed under the central lab model. Implementing an LDT is relatively fast and cost-effective! Most of the effort and cost is in the CLIA or CAP validation of the LDT.
An LDT provides the opportunity to gather clinical utility and safety data in the real-world, but highly controlled setting. This helps in building confidence in this new technology for clinicians, regulators, and payors, alike. The clinical data will reveal any issues before engaging in the costly commercialization of a medical device. Therefore an LDT can be used as risk mitigation for the development of the medical device. Since the LDT and the medical device are based on the same algorithm producing the same data, the clinical data obtained by the LDT could be used to support the regulatory approval of the medical device.
Another material incremental step from an LDT to a commercially-distributed medical device is an FDA approved single-site medical device.
An FDA-approved, single-site medical device requires that a clinical laboratory adds design controls to the development (think assembly) of the LDT (not to the development of its components), as well as obtaining regulatory approval based on site-specific clinical studies that may be more comprehensive than the CLIA or CAP validation. However, there is still no need to manufacture the pathology AI system as a medical device. The increased rigor of design controls and validation studies build further confidence in the technology.
By expanding the use of an FDA-approved, single-site medical device with proven performance to an FDA-approved, commercially-distributed medical device, the regulatory approval is broken down into two smaller but meaningful steps. This also creates a good risk mitigation for the regulatory approval of the commercially-distributed medical device. With the assumption that the medical device shares the same intended use, medical device definition, and especially the same algorithms; the final burden to regulatory approval of the commercially-distributed medical device is the manufacturing of the pathology AI system as a medical device and performing multi-site studies. This is time-consuming and expensive work but very low risk, especially when compared to developing a medical device from scratch.
When there is a need for a medical device, but the business case for a commercially-distributed device is a challenge, a single-site medical device under a central lab model can provide a viable alternative. This is increasingly pertinent when patient populations are limited, as is the case of second line indications such as immuno-oncology and rare diseases.
Are you offering/using a tissue image analysis test as an LDT? What is your experience with a central lab model? Do you struggle with a business case for a Pathology AI Dx, Px or CDx product?
Flagship Biosciences has a CLIA and CAP-certified clinical laboratory where we run digital pathology and our in-house developed pathology AI system in a regulated environment. With that we have everything in place to commercialize our pathology AI system under a central lab model. We are also already working with the FDA on a single-site medical device and have implemented design controls in our quality management system (QMS). Currently we are in the process of submitting our first 510k to the FDA going down the path of establishing confidence in our pathology AI system as a medical device.
If you want to learn more about the technology behind tissue image analysis, check out our previous LinkedIn articles, where we presented a short lecture series about the different aspects about Pathology AI intended for a broad non-technical audience.
Holger Lange, PhD
Chief Technology Officer