Let’s start with what we know. For years, the state of pharmaceutical drug development has been in decline in terms of new molecular entities or biologics that are clearing FDA approval.

Most agree that targeted therapies are the key to the future success of drug development, especially with respect to cancer treatments. However, currently we have just a handful of companion diagnostics for just a handful of targeted therapies, with many of these companion diagnostics far from reaching their mark of achieving desirable false positive or negative results (as recently reported in the New York Times “Cancer fight: Unclear Tests for New Drug”). The majority of patients who are administered targeted therapies still do not respond in the way that we would hope. Pharmaceutical R/D spending, while much of it is becoming outsourced, has actually increased inversely to the success of clinical development programs.

Being Disruptive Will Only Get You So Far

I have been watching life science technologies evolve for about 20 years. Every 6 years or so, a new disruptive technology takes hold. I am thinking about technologies like gel-based DNA sequencing, reverse transcription PCR, 96-well DNA capillary sequencing, MALDI TOF, DNA microarrays, and now next generation DNA sequencing. What I have seen in the previous examples (with exception of reverse transcription PCR), is that the closer that these technologies get to producing data that has real clinical utility, the harder it is for the technology to be used in a clinical setting, including drug development. This is because most of these technologies are too complex, too expensive, take too long to prepare samples, and often give too much information to use in a routine clinical environment.

Digital Pathology – the New Kid in Town

For drug developers that follow the latest and greatest technology trends, there is a “new kid in town.” Digital pathology was originally introduced as a way for pathologists to teach, collaborate, and review cases remotely. It has also produced a means to view, store, and analyze images. Recently, digital pathology has been implemented into the routine clinical diagnostic setting. But can digital pathology really make a significant future impact on drug development and clinical diagnostics?

Definiens sub-cellular classification of Her2 stained breast tissue, showing nucleus, cytoplasm, and membrane classification.

It’s All in the Translation

Let’s take a look at the “translate-ability” factor of digital pathology; the ability to take novel findings and “translate” them into the clinic to perform relatively simple, cheap, fast, and accurate diagnostics without having to drastically alter existing workflows. While most life science technology platforms start out in the basic or translational research settings and seek to become implemented into a diagnostic environment, digital pathology started out as a clinical tool, and is making its way “backwards” into the translational research setting. Currently, there are at least 200 different laboratories that are applying digital pathology to translational research. This information, when discovered, will become more easily translated into the clinical environment, because all software algorithms needed to identify clinically relevant biomarker and / or morphological features can be deployed into a CLIA laboratory setting without needing to radically alter workflows or technology platforms.

The Tissue is the Issue

With digital pathology, a patient’s biopsy can be digitally imaged and quantitatively data mined. This information can then be stored electronically, whether as part of a patient case record with other imaging modalities, or as part of a comprehensive research study. For example, after performing digital pathology image analysis on a cohort of patient digital slides, one can ask the question, “show me all patients who have an average tumor cell area of greater than 50 square microns, and who have a predominance of at least a medium Her2 expression level on the membrane of these cells.” Asking more insightful questions will undoubtedly lead to better drug development results. Additionally, the reproducibility of digital pathology will be key in the development of more accurate diagnostics, and by extension, increased patient response rates to targeted therapies.

Definiens Tissue Studio is a new software application that is designed for digital pathology image analysis in the translational research setting. Scientists can perform retrospective studies on cohorts of patients from past clinical trials and data mine tissue materials that were collected, fixed, and stained for markers of interest; to determine correlations with biomarker expression, cellular morphology (or both) with clinical outcome. As applications like Definiens Tissue Studio proliferate, 1000’s of scientists will soon begin discovering how various combinations of biomarkers, biomarker localization, and morphologic properties (i.e., circularity, length, width, area, elliptical fit, and so forth) correlate with clinical outcome. Once validated, these algorithms can then be implemented in a CLIA laboratory setting via the Definiens XD platform; a progression that has already been implemented in 2 large clinical service laboratories to date.

Definiens Tissue Studio analysis of Her2 stained breast tumor tissue showing individual cellular biomarker evaluation, left; as well as individual cellular morphological information, right. Heavily stained cells are dark brown; medium stained cells, orange; lightly stained cells, yellow.

Improving Tissue-based Diagnostic Accuracy

What about existing immunohistochemistry-based diagnostics? Definiens recently conducted a clinical study with a major pharmaceutical company where tumor biopsy slides from 80 patients from a Phase II breast cancer clinical trial were stained using the Dako Herceptest and evaluated by a panel of 6 pathologists. Definiens developed an algorithm to classify individual cells as 1+, 2+ or 3+, and then generated a final H-score for each patient resulting in a 30% improvement in accuracy compared to the original evaluation. That is a major improvement, and this type of objective approach is sure to have a positive impact on clinical trial inclusion and exclusion; as well as facilitating more accurate treatment decisions not only for Her2, but other immunohistochemistry-based biomarker assays.

Definiens cell-by-cell classification of the Dako Herceptest. Original images, top; Definiens classified images, bottom. 3+ cells are in purple; 2+ cells, orange; 1+ cells, yellow; no stain present, green. This method has achieved as much as a 30% accuracy improvement as compared to a panel of 6 pathologists.

Decision Support for Drug Development

Finally, digital pathology image analysis makes it possible to directly correlate cellular morphology with the effect of a drug on a drug target; or surrogate biomarkers for toxicity. Insights into these two vital areas in drug development are needed to enable drug prioritization or terminate a drug development program altogether. Digital pathology image analysis ensures higher confidence in making these decisions. Definiens digital pathology image analysis solutions are currently being used by several leading pharmaceutical companies to facilitate more intelligent, more informed go/no-go decisions in both pre-clinical and early clinical development.

The Train is Leaving the Station – All Aboard!

Can digital pathology save drug development? My answer to this question is that I am optimistic that digital pathology will certainly make a positive impact on reducing drug attrition rates. But one thing is for sure: The digital pathology train is leaving the station, and drug developers should be taking a very close look at the impact that this technology can make on their translational research and drug development programs.

Mr. Duncan joined Definiens in December of 2008 and serves a dual role as Director, Business Development Clinical Diagnostics, and Global Marketing Manager. Mr. Duncan is the creator and editor of Digital Pathology Insights, a leading online news and discussion portal for digital pathology and personalized medicine. Prior to Definiens, Mr. Duncan served as Vice President, Business Development at Aureon Laboratories and was responsible for several collaborations around applying morphometric imaging techniques to multiplexed immunofluorescence assays applied to formalin-fixed, paraffin embedded tumor tissue.