Digital Pathology Will Usher in a New Era of Personalized Medicine

by: Joseph Krueger

The Human Genome Project (HGP) was a 13-year project coordinated by the U.S. Department of Energy and the National Institutes of Health in cooperation with genetic bioscience giants Celera Genomics (NASDAQ:CRA), Incyte (NASDAQ:INCY), and Human Genome Sciences (HGSI). The Human Genome Project aimed to tell us what the sequence of the average person’s genetic code is “supposed” to be, and gave the necessary information for publicly funded research as well as for-profit research to begin understanding how the differences between individuals DNA affect health and disease. Indeed, this information gave rise to the concept of “personalized medicine”, which seeks to use precise genetic information about individual patients to custom tailor their mode of therapy for a disease. This concept of personalized medicine has fueled the growth of multi-billion dollar companies like Myriad Genetics (NASDAQ:MYGN), Sequenom (NASDAQ:SQNM), Illumina (NASDAQ:ILMN), and Affymetrix (NASDAQ:AFFY).

We have been hearing about the promise of “personalized medicine” for over a decade now, but what has come of it? It has long been predicted that diagnostics based on genetic information brought forward by the human genome project would bring more targeted and cost-effective therapies. However, to date, almost 20 years later, there are only a few examples that have come to fruition. In general, large pharma has been slow to implement useful companion tools to apply the concept of personalized medicine to their clinical drugs. The idea seems simple: Along with the drug, create a companion diagonostic that will determine the correct course of therapy for the patient. More specifically, the companion diagnostic will predict ahead of time whether or not patients will respond to the drug.

This is why companion diagnostics are becoming an essential part of a new drug launch. Patient stratification is key to successful trials, approval, and patient welfare. But developing a companion diagnostic along with a drug is a challenge. High-throughput diagnositic companies like Qiagen (NASDAQ:QGEN), Quest (NYSE:DGX), BioRad (NYSE:BIO), Genzyme (GENZ), or LabCorp (NYSE:LH) are hesitant to make custom companion diagnostics, because the development expense is large and because the drug is not approved its utility is not secured. In addition, incorporation of the diagnostic into a clinical trial often requires the additional expense of contract research organizations like Covance (NYSE:CVD) and Quintiles (private), at the drug developer’s risk. However, large diagnostic companies are slowly beginning to respond to this need. LabCorp’s recent acquisition of companion diagnostics company Monogram Biosciences was aimed at strengthening LabCorp’s presence in personalized medicine. Monogram’s Trofile assay is the accepted companion diagnostic for Pfizer’s HIV drug Selzentry. Monogram also has a proprietary technology, VeraTag, which has been used to develop a sensitive means to assess HER-2 status in tissue samples.

A companion diagnostic is extremely important in oncology, as there are known, specific drivers of certain types of cancers. With the implementation of targeted therapies, which only target a single specific gene product, the patient’s response to a targeted drug is determined by the presence of this driver. Thus, patients need to be carefully selected prior to therapy. This is extremely beneficial to the patient, as they follow the right course of therapy; but it is also extremely beneficial to the drug manufacturer as they can identify patients that will respond best to their drug, in some cases approaching 90% success.

There has already been a couple success stories in this arena. In the case of Roche/ Genentech’s (OTCQX:RHHBY) Herceptin, an antibody against HER2, Genentech uses HER2 gene amplification to gate breast cancer patient treatment. Genentech partnered Dako to have a HER2 companion diagnostic kit developed for this. This companion diagnostic helped ensure that the right patients were receiving the right treatment; and also helped ensure that Herceptin would be successful in the trials. Another similar example is with Amgen's (NASDAQ:AMGN) Vectibix: Studies have revealed that anti-EGFR monoclonal like Amgen's Vectibix, an monoclonal antibody against EGFR, do not work in colorectal (CRC) tumors that have a mutated KRAS gene. In response to this, Amgen partnered with Qiagen to develop a companion diagnostic to identify patients with a KRAS mutation who will not respond to the drug. Similarly, Lilly (NYSE:LLY) partnered with Quest Diagnostics for a KRAS mutation companion diagnostic for Erbitux.

These examples demonstrate how the co-development of a companion diagnostic with the drug is useful to both patients and the company, as it helps ensure success in clinical trials and success in therapy. Indeed, the FDA is very supportive of drug companies co-developing companion diagnostics “very early in the drug development process”. The paradigm the FDA is pushing for is where a company develops a clinically relevant patient selection marker during phase II trials to be used during phase III trials. Between phase II and phase III, the company discusses with the FDA the methodologies and study design to proceed with the phase III trial which include the companion diagnostic. Upon successful completion of phase III, both the drug and companion diagnostic are ready for approval.

These types of partnerships are encouraging large diagnostic companies as well as small biotech companies to develop companion diagnostics for developing drugs. Partnerships are being formed earlier and earlier in the process. Smaller biotech companies seem to be more agile at identifying markers and methodologies which are useful in selecting which patients will respond to a particular drug. Beyond this, smaller companies are more motivated to develop a companion diagnostic for a likely-to-be marketed drug than the drug developer. The small biotech sees this R&D investment as means to generate a product with revenue and the pharmaceutical company may see this R&D investment as extra cost and a hinderence to the development of the drug in a competitive market. Thus, a partnership between the two is natural and productive.

However, one thing that all the above examples have in common is that they are genomic-based diagnostics. These tests (in fact most tests) are based on determining gene amplification, gene mutation, or a gene expression profile. This is because of the principle of modern, targeted therapies: The drug targets only one gene product, so the state of that gene is the only one which will determine response to the drug. Genetic based tests for a targeted therapy can help improve therapy by identifying patients who are susceptible to a particular disease [Myriad’s (MYGN) BRACAnalysis]; patients who will respond (Herceptin) or won’t respond (Vectibix) to a particular therapy; patients who are best suited for particular therapy regimen [Genomic Health’s (NASDAQ:GHDX) Oncotype Dx platform], and how patients individual gene sequence (MassARRAY SNP analysis from companies like Sequenom) or metabolism will affect the drug potency (pharmaocgenomics).

These tests are very useful in their appropriate application with targeted therapies, but leave a lot of patients left out . In the case of cancer, single gene mutations or amplification driving the disease are the minority. For example, in the case of Herceptin, only about 1/3 of breast cancer patients have the HER2 gene amplified. In the case of Erbitux or Vectibix or any of the many EGFR antibodies, KRAS mutation occurs about 40% of the time in colorectal cancer. So whereas these methodologies are good at including/excluding good sizes of the total patient pool, “personalized medicine” cannot be offered to the excluded patients. So what about the rest of these patients, who represent the majority?

This majority of patients will receive standard of care (SOC) without any companion diagnostic. As well, since these companion diagnostic tests cost an average of $3000, many health care providers may not cover this expense, and these patients will not receive the most appropriate care available even if they were eligible. The federal government’s drive to reduce health care costs will only exacerbate this. However, this change in policy may actually drive the development of more useful companion diagnostics: A useful diagnostic should improve care while reducing waste, and be compatible with the cost-sensitive environment of healthcare.

So how can the majority of cancer patients, who do not have a mutation, amplification, or polymorphism in single gene benefit from personalized medicine? What type of companion diagnostic can be integrated into the routine procedures utilized in cancer diagnosis, without significant extra cost or procedure for the patient and their health care provider? Fortunately, a new kind of diagnostics platform is emerging: Quantitative Immunohistochemistry. Immunohistochemisty (IHC) is the standard procedure used at all hospitals to diagnosis and stage cancer. In this case, a biopsy of the tumor is taken, the specimen is prepared into microscope slides, stained for a variety of markers, and examined by a pathologist.

In oncology, this assessment of the local tissue environment by the pathologist is essential, and part of the routine diagnosis. Once prepared, the tumor sample can be analyzed for the desired biomarker to be used as the companion diagnostic. The presence of absence of a certain marker on a certain kind of cell may predispose the patient to response from a targeted therapy. In this case, the marker profile could be independent of the targeted gene itself and thus cannot be determined by genetic tests. It may not have anything to do with gene amplification or mutation of a gene; it may have to do with the cellular location of a particular gene product or the makeup of the tumor. These are determinants that only pathology can assess.

In the development of companion diagnostics, much effort has been spent on soluble markers from the blood. Historically, in most cases, there are no surrogate markers present that can be used- there is simply not enough information in blood or other bodily fluids to provide adequate prognostic information. The information of how well a patient is responding or will respond to a given treatment must come from the local environment of the tumor, and this means tissue testing and pathology. Pathologists will survey a tissue under a microscope and access changes in the local tumor response or nearby tissue. Immunohistochemistry is often run on a number of markers to indicate localized protein changes. Until recently, pathology has been a historically qualitative discipline, with pathologists providing a visual assessment of grade levels of various types.

However, this process is laborious and requires the expertise of a pathologist. Fortunately, the routine collection of specimen and histological examination is already in place; all that is required is the addition of another histological assay that directly assesses a patients response to a targeted therapy. In this case, the companion diagnostic is the addition of a significant histology based biomarker. However, to perform well as a companion diagnostic, it most no require significant extra protocol or expense. Digitial Pathology techniques provide for this, as the analysis can be performed by a separate service provider with expedient efficiency with low cost due to the incorporation of electronic scanning and computer based algorithms which determine the assay endpoint. As well, in contrast to current companion diagnostics which give a digital yes/no answer, Quantitative Immunohistochemistry provides layers of detail to the pathologist which can guide the course of therapy in many ways.

This new technology, whole slide imaging, or digital pathology, is moving pathology closer towards an electronic science. Radiologists have worked exclusively from digital media for years, but pathology is just now moving in this direction. The entire tissue section on the glass slide is scanned in a high-throughput scanner, made by established digital pathology companies like Aperio, Bioimagene, Zeiss, and Hamamatsu, or newer entrants with even faster scanning technologies like Leica, Philips, and GE’s Omnyx subsidiary. The pathologist can now read and diagnose from a computer screen, eliminating the need to ship glass slides around during clinical trials. More importantly, the pathologist can run computer image analysis across the tissue section, generating quantitative information highly useful in the development of a companion diagnostics program. This methodology not only speeds up the pathological assessment process, reducing cost, it provides the means for a rapid and inexpensive companion diagnostic based on the “gold standard” assessment of pathology.

While I normally only write about public companies, I could not find any public companies that are leading the development of companion diagnostics using digital pathology! Large CRO’s like Quintiles will provide digital pathology services, but they do not appear to be developing companion diagnostics based on this methodology. This suggests that it is truly a ground breaking field. However, there are some private companies that appear to making strides:

AstraZeneca (NYSE:AZN) and privately held Danish immuno-diagnostic company Dako entered into a collaboration agreement to develop companion diagnostic tests for multiple AstraZeneca oncology projects, including biologics and small molecules, in various stages of discovery and development. The companies will work together to develop diagnostic tests to help physicians determine the most appropriate cancer treatment for patients. The companies did not disclose the financial terms of their deal.

Aureon, a private company focused on “enabling personalized patient care through predictive pathology” appears to be making strides in the digital pathology arena. The company developed PathoMetrix, an automated machine vision tissue image analysis system that uses advanced image processing algorithms in order to segment and measure properties of histopathological objects. The company launched Prostate Px+, a prostate scoring system earlier this year. It is said to be the first commercial test to predict prostate cancer progression and disease recurrence at the time of diagnosis.

One company I can provide much more detail on is a privately held company called Flagship Biosciences. My knowledge of this company is based on some first-hand diagnostics work with pathology image analysis which has been done by pathologist Dr. David Young and his team from OSI Pharmaceuticals, now a subsidiary of Astellas Pharmaceuticals (OTCPK:ALPMY). David Young started Flagship Biosciences, and co-founded it with longtime image analysis pathologist expert Dr. Frank Voelker, who most recently founded the tissue biomarker laboratory at Novartis in Cambridge. They added an experienced digital pathology executive, Steve Potts, who installed a nationwide digital pathology network while at Quest Diagnostics and then built the biopharma business for Aperio, the current hardware leader in digital pathology. On their board of directors is Dr. John Bloom, who was responsible for setting up the central lab model ten years ago at Eli Lilly that most of the industry followed, and that led to the dominance of CROs. Bloom has stated (see here) that he considers the development of tissue companion diagnostics of equal par in size and significance to the central lab problem he solved ten years previously that led to the industry giants of Covance and others.

One example of how a digital pathology companion diagnostic is already being developed at Astellas stems from OSI’s determination that the resistance to their EGFR kinase inhibitor, Tarceva, is linked to a certain state of the tumor cells called the epithelial-to-mesenchymal transition (EMT). This work involved a difficult assessment of one marker for epithelial cells and one marker for mesenchymal cells. As shown here, the IHC values were compared between the computer and a veteran pathologist scoring manually. This work demonstrates how successful digital pathology companion diagnostics strategies can be easily integrated into oncology drug patient selection.

To date, there are only few examples of successful genetic based companion diagnostics. As such, the companion diagnostics space is not one that has been kind to venture capitalists, and public offerings of such companies has dwindles. However, the type of work needed to develop useful companion diagnostics requires a new paradigm like Digital Parhology, and is probably better served by smaller companies who are able to be more flexible in working between large pharmaceutical customers and established large CROs. The fact that most large CROs and reference labs have found it difficult to build a large digital pathology business may be proof of that. It takes large IT investment and risk management, constant dialog between image analysis experts and pathologists, niche marketing expertise, and a continually churning hardware and software vendor environment.

Utilizing digital pathology to develop a non-genome based companion diagnostic which can serve the majority patient population is key to making “personalized medicine” a reality. Implementing the companion diagnostic by integrating it into the customary pathological workflow assures widespread clinical use and payer acceptance. Small, private companies like Flagship Biosciences are paving the way- big pharma should be paying attention. I am continuing to look for small public companies that are working in this field. These types of companies are obvious acquisition targets by the larger pharmaceutical companies who will rely on Digital Pathology to provide a companion diagnostic for their blockbuster drug. I will be watching carefully for new targeted therapies coming through the big pharma pipelines which are ideal candidates for a Quantitative Immunohistochemistry companion diagnostic and small biotech companies that can fulfill these needs.

Disclosure: I do not have positions in any of the stocks mentioned