Astronaut in the desert looking at the moon
Credit: Orla / iStock / Getty Images Plus

Currently there are 51 companion diagnostic (CDx) tests approved by the U.S. Food and Drug Administration (FDA). The vast majority of these are used to recommend targeted therapies for hematological malignancies and solid tumors. In fact, of the 170 approved indications listed on the FDA website for companion diagnostics, only three are for non-oncology drugs.

More than a quarter century after Dako and Abbot gained CDx approval for its tests to recommend the breast cancer drug Herceptin, the field is spreading its wings beyond those first immunohistochemistry assays. Next-generation sequencing technology has led companies like Foundation Medicine, Guardant, and others to develop CDx tests that include hundreds of biomarkers which have been serially approved for a range of malignancies.

Further, CDx approvals are bubbling up for indications outside of oncology. For instance, at the end of April 2024, testing giant Labcorp announced that its nAbCyteTM Anti-AAVRh74var HB-FE Assay received FDA approval as a CDx for the Pfizer drug Beqvez which treats hemophilia B. The approval was also notable as it marked the first CDx approved for a gene therapy.

But where is the field of CDx development headed? Today, machine learning and artificial intelligence (AI) are beginning to make an impact on pharma-diagnostic co-development. And the recent Labcorp approval is an early indication that CDxes are moving beyond cancer indications to include hereditary diseases, autoimmune disorders, and other indications that require big data and robust computing power to drive complex biomarker signatures.

Digital pathology moves from workflows to CDx

Michael Rivers
Michael Rivers
Roche Tissue Diagnostics

Among the emerging trends in CDx development is using digitized pathology slides to identify biomarkers and biomarker signatures that leverage the powerful image recognition capabilities of AI. Until recently, digital pathology was focused on improving workflows for overworked pathologists while providing some automation that aided the reading of slides. But these capabilities are now being developed in ways that are poised to make a significant impact on CDx development.

“Tissue diagnostics, and anatomic pathology in general, is going through a massive transformation from what has been a very manual, analog practice to a fully digitized, much more automated solution,” says Michael Rivers, head of Roche Tissue Diagnostics. “We see that our partners and pharma are looking at tissue diagnostics as an incredibly important element of the entire diagnostic workflow. Understanding what is happening in the tissue, in the tumor microenvironment, can only happen with digital pathology and AI.”

Paul Beresford, PhD, vice president and general manager of CDx with Agilent Technologies agrees. “The real innovation is coming on digitizing the slide and making that a dataset, essentially,” he says. “That makes it both more robust and reproducible and that data can be used in a machine learning strategy. Then you can apply a range of different strategies incorporating, for example, the H and E as well as the immunohistochemistry staining into a final algorithm.”

Paul Beresford
Paul Beresford,PhD
Agilent Technologies

From this base, Agilent will seek to establish the feasibility of these new methods of pinpointing existing known biomarkers, or even identifying novel biomarkers, into the early-stage strategy of working with pharmaceutical partners for CDx development. Diagnostics developers have long desired to work with pharma partners much earlier in the drug discovery and development process—as early as Phase I and Phase II studies—in order to get a jump on identifying the biomarkers that will eventually be used in the CDx. In many instances, however, that earlier work has landed in the lap of contract research organizations (CROs), which often kept diagnostics developers locked out until the launch of pivotal studies.

Now, via its digital pathology partnerships with Hamamatsu (slide imaging), Proscia (slide file management), and Visopharm (image analysis and algorithms), Agilent is positioning itself as a one-stop co-development partner in the CDx arena. “We are now able to provide that continuous engagement with pharma,” Beresford notes.

Roche, PathAI partnership

Having the capability to work with pharma sponsors throughout the biomarker and CDx development lifecycle was also the driving force behind an exclusive collaboration between Roche Tissue Diagnostics and digital pathology company PathAI announced in February.

According to Eric Walk, MD, chief medical officer of PathAI, the partnership with Roche—already one of the leading CDx developers worldwide—is a natural extension for the company.

“[PathAI founder] Andy Beck had the vision for creating a company that was focused on leveraging AI technology and focusing on challenges in histopathology, both in the pharma space but also in the clinical space,” Walk says. “From the very beginning, we felt that pharma development, including translational diagnostics, and eventually companion diagnostics, would benefit from AI technology and I think this partnership is a manifestation of that hypothesis and that philosophy.”

Eric Walk
Eric Walk, MD

The deal will see the two companies work to develop AI-enabled digital pathology algorithms with the aim of making the algorithms a central component of future companion diagnostics. Walk sees that there is a future pathway to developing AI-powered companion diagnostics, based on existing CDxes that rely on manually analyzing pathology slides.

“With the current generation of companion diagnostics, I use PDL-1 as an example, you have an IHC assay that is interpreted by a human pathologist. That pathologist generates a score based on a cutoff, that determines whether the patient is positive or negative and that is the basis for treatment,” Walk says. “So in the short term, AI and digital pathology can benefit the paradigm that I just described, simply through reproducibility—to help pathologists be more reproducible, be more accurate on the current generation of tests.”

Looking further into the future, Walk sees that AI pathology slide interpretation can provide a tiered approach for CDxes. The first, just described, can be for AI to assist the pathologist in their reading of the slide; another could be an AI-only approach whereby the diagnosis is made via an algorithm only; or a third approach where AI does all the upfront analysis, and the diagnosis is merely confirmed by a trained pathologist.

Medical Science Laboratory: Portrait of Beautiful Black Scientist Looking Under Microscope Does Analysis of Test Sample. Ambitious Young Biotechnology Specialist, working with Advanced Equipment
Credit: gorodenkoff / iStock / Getty Images Plus

“I think AI enablement will take different forms, and it will probably be an evolution over time,” Walks adds.

From Rivers’ view, the exclusive deal with PathAI made sense from two standpoints. First, PathAI already had strong, long-standing relationships with major pharmaceutical companies, which were already familiar with the company’s technology and capabilities. Second, Roche views PathAI as the leader among the many companies operating in the digital pathology space.

“We see an amazing growth in interest here,” Rivers points out. “We chose to pick one partner we felt we could work well with that would augment our capacity and help us to build these integrated assay and algorithm solutions that could then be deployed on the Roche navify digital pathology solution.”

Yet, while the potential is huge, Rivers notes there are still some hurdles to clear in broadly bringing these capabilities to market.

“The first step is we need to digitize the labs,” sayd Rivers.
“As exciting as these AI solutions are, we can’t use them until we have a digitization, scanning infrastructure, and the digital viewing environment and workflow for the pathologist to transition from the microscope to digital.”

Companion diagnostics for hereditary diseases

Currently, only three CDxes are approved for use outside the field of oncology: one for hemophilia A; one for hemophilia B; and a third for obesity.

But a collaboration between QIAGEN and population health company Helix announced in early 2023 seeks to change that. The companies are collaborating to develop companion diagnostics in a range of hereditary diseases such as Parkinson’s disease and cardiovascular or inflammatory disease like nonalcoholic steatohepatitis (NASH).

Each company brings a diverse set of expertise to the relationship, with QIAGEN already marketing eleven PCR-based companion diagnostics. Helix, whose work has focused mainly on population genomics services for health systems with more than 100,000 patients, has a deep database of whole-exome sequencing (WES) data that can facilitate patient recruitment for clinical trials while also leveraging real-world data and real-world evidence.

Huw Ricketts
Huw Ricketts, PhD

“Although not novel, WES has great potential to support patient testing in complex (non-oncology) diseases like Parkinson’s, Alzheimer’s and dementia,” says Huw Ricketts, PhD, senior director of CLIA business development at QIAGEN.

The partnership with Helix, he notes, allows for the companies to provide single-site IVD development for pharmaceutical partners and QIAGEN can then develop distributable CDx kits of the WES assays to labs for decentralized testing.

The WES testing that will be offered by the companies represents what the companies see as the “future of testing…non-oncology diseases,” Ricketts says, noting that PCR tests focus on a single biomarker, while WES can provide a platform for testing multiple biomarkers for multiple diseases.

“Metaphorically spoken, a WES (assay) offers the library to laboratories, it is then up to the lab to decide which aisle of the library they need, and which books are most relevant,” Ricketts adds.

Regulatory challenges

With the implementation of IVDR by EMA in Europe, as well as the renewed push by the FDA in the U.S. surrounding stricter regulation of laboratory developed tests (LDTs), the pathway to gaining approval for a CDx isn’t always clear.

In the case of the QIAGEN/Helix partnership, the companies’ strategy of developing a WES kit for regulatory approval should help drive adoption once brought to market. As Ricketts explains: “Currently, labs in the EU and the U.S. are mostly using WES assays they developed themselves, so-called LDTs. For LDTs, the labs have to take care of their IVD certification themselves. This is challenging, as the process needs many resources and regulatory expertise. Therefore, providing an already certified WES assay to the labs will facilitate and accelerate their workflows.”

For Roche and PathAI, Rivers notes that the path to developing the CDxes the two companies envision presents an “element of the unknown.” The crux here is that to date there is no template for bringing an algorithm-driven CDx through regulatory approval since it has never been done before.

As Rivers describes it, their development efforts will need to provide a clear characterization and control of every step of the process from the scanning of pathology slides, to the workflow, and including the user interface provided to the pathologist. There can be no so-called “black box” where the data is fed into it and an answer pops out. Everything will need to be spelled out in detail.

“This is how we have been operating with the FDA to date with our existing algorithms and our existing digital pathology solution,” Rivers notes. “They really want to look at the pixel path from the generation all the way through to the analysis and have clear provenance and understanding of where did these samples come from—even with regard to the training data that is used. That’s why a very tightly integrated control process is critical, I think, for success.”

FDA building and sign
The U.S. Food and Drug Administration, Public domain, via Wikimedia Commons

While much attention is paid to EMA and FDA, Beresford notes that the evolution of regulatory standards is global and includes China, which has produced new guidance documents including some related to immunohistochemistry cutoffs and the need a minimum number of patients at the cutoff. Closer to home, he notes there may be some pilot program from the FDA surrounding NGS-based testing that could allow smaller labs to bring forward tests of substantial equivalence to approved diagnostics that could potentially lower the burden to getting a test approved.

He adds that some companies have even approached Agilent to see if that same standard may be applied to IHC tests. “I’m not sure if that was the FDA’s initial thought, but we’ll see,” Beresford says.

The challenge for labs in U.S. could again boil down to objections to proposed LDT regulation from years past, that more stringent requirements for these tests will lock out smaller labs from using their homegrown tests to serve their patients.

“For smaller labs, where you might be doing one test a year for a very small patient population, do they need to put in place a full manufacturing path that’s going to be FDA compliant?” Beresford asks. “That’s undoable for most labs. It’s a headwind for the lab providers, but it may be an opportunity for some of the large IVD players to think through how we can help labs like that with some additional tool sets.


Chris Anderson, a Maine native, has been a B2B editor for more than 25 years. He was the founding editor of Security Systems News and Drug Discovery News, and led the print launch and expanded coverage as editor in chief of Clinical OMICs, now named Inside Precision Medicine.

Also of Interest