Advertisement

Tissue-based research is a critical diagnostic component of drug discovery containing information on efficacy and toxicity signals, but getting this information is tedious and time-consuming. The task is compounded by the fact that good tissue-based research requires the expertise of pathologists specially trained in the field. There are never enough pathology experts to go around, and they are rarely located where the tissue is. Most biotech companies do not have pathologists on staff, relying instead on larger contract research organizations (CROs) or independent pathologists.

Companies who understand and value their pathologists and tissue-trained biologists have substantial competitive advantages in the pharmaceutical industry. Despite recent trends to automate pathology and tissue biomarker measurements, anyone familiar with solid tumor research knows that it isn’t just the expression of a biomarker that matters, but where and how it was expressed. The pathologist must remain a central figure on the road to new therapeutics.

Digital pathology uses whole-slide imaging to allow an entire glass microscope slide to be viewed on a computer screen from 20x to 100x oil resolution. A pathologist can view and annotate the image from any location rather than using a microscope and glass slides. Slides are scanned at the histology lab, and then viewed by one or more pathologists from any location. Essential pathology activities such as peer review can be conducted remotely with a full electronic record. Because the entire tissue section is scanned as a single image, computer analysis can be run to quantitate toxicology lesion size or measure protein expression.

All of the 13 largest pharmaceutical companies currently use digital pathology in biomarker discovery and preclinical tissue-based research. Nearly all of the major preclinical CROs offering pathology services have adopted digital pathology.
Let’s take a closer look at the level of decision making that can be facilitated with pathology support across discovery, preclinical, and oncology clinical trials, and see how digital pathology can help biotechnology companies make precise decisions faster at each stage.

Decision making in discovery
The primary goal in discovery with tissue-based research is to get quantitative efficacy data on compounds as quickly as possible. Tissue screening is unsurpassed in the biological relevance of the information, even if other modalities like cell-based screening lead in throughput. Any system that automates tissue screening must be built upon the findings of pathologists or expert biologists, with the tissue and data flowing around their key observations. Two examples from anti-angiogenesis and neuroscience illustrate this approach.

Quantifying anti-angiogenesis inhibitor effects

inst-1.jpg

click to enlarge

Figure 1. Using whole slide imaging and analysis, an entire skin matrigel section can be scanned and analyzed for vascular measurements. Software identifies stained endothelial cells and then computes vessels and statistics. New morphology biomarkers like lumen area and vessel wall thickness can be used for making go/no go decisions in anti-angiogenesis programs. (Source: Novartis NBRI) 
To decide whether to kill an anti-angiogenesis compound or move it forward, microvessel density measurements are frequently employed in xenografts or matrigels (Figure 1). This is a very tedious manual-counting process that can take weeks to complete. Not many pharmaceutical companies can afford to pay pathologists to perform manual counting so the question becomes, “Can a computer be trained by a pathologist to more accurately and precisely count vessels than a non-pathology scientist?” In a recent presentation  by Novartis, Pathology Experts, and Aperio Technologies, two PhD-level scientists manually counted vessels in matrigels and compared this with an automated whole-slide scanning approach. The scientists identified 60% of the vessels in common, but in 40% of the cases they disagreed.

This approach was then compared to whole-slide scanning and automated whole-section analysis. Each slide was scanned at 20x, then an automated image analysis algorithm was applied to find and count the microvessels. Unlike manual counting, the computer had perfect precision and repeatability, and could be tuned by the pathologist to within 90% of his manual count. In addition to the standard microvessel density measurement, other useful morphology biomarkers like cell-wall thickness and lumen area could also be calculated, something the scientists could not do manually.

This level of precise automation allows an executive decision-maker looking at more than one anti-angiogenesis compound to determine not only efficacy, but whether the compounds have different effects and potentially should be considered in a therapeutic cocktail. The time required to scan and analyze vessels by a computer was 10 times faster than manual counting, and the scanning and computer analysis could be fully automated and parallelized if required.

Measuring efficacy of neuroscience compounds 

inst-2.jpg

click to enlarge

Figure 2. In Alzheimer’s research, the dark brown regions of amyloid plaque can be measured on a whole slide image, or on only certain regions of a brain section. (Source: Pfizer, Inc.)  
Neuroscience discovery programs regularly screen compounds against large numbers of rodent models to provide an early indicator of efficacy. In Alzheimer’s models, this usually involves manual sampling with a microscope of small brain sections for measurement or counting of lesions like amyloid plaque, tau protein, or microfibular tangles (Figure 2). A high-volume core imaging laboratory at Pfizer investigated the time savings of implementing whole slide imaging and analysis, comparing standard microscope based approaches with newer whole slide imaging and whole tissue analysis. Their findings indicated that a 2.5 month rodent study analysis measuring amyloid plaque could be completed in only one month with a whole slide imaging solution.

Quantitative preclinical decision-making
More precise decision making in tissues extends beyond discovery into the preclinical stage. The opportunity to provide fast and effective decisions becomes more pronounced as an entire organization adopts digital pathology to integrate multiple sites digitally.  Pathologists on different continents but at the same pharmaceutical company can come to agreement on scoring, discuss differences, and allocate pathologists to projects regardless of geography.

The impact of digital pathology was investigated at Novartis Institutes for Biomedical Research (NIBR) in Cambridge, Mass. Pathologists routinely analyze biomarkers to guide decisions related to efficacy, toxicity, and mechanisms of action and timely turnaround of results was always a challenge. Novartis digitized over 20,000 slides in the first two years of use of the technology, with subsequent image analysis. The department experienced an over 30-fold improvement in turnaround time. 
Digital pathology is now being utilized at Novartis operations in the U.S., Europe, and China. Across the global organization, Novartis has adopted standardized terminology and metrics for efficacy or toxicity, and is able to make decisions faster.

International Oncology Clinical Trials
Similar to the discovery and preclinical areas, precise measurements in tissue are required for effective decision making in clinical trials. The impact of digital pathology in clinical trials is only in its infancy. In oncology in particular, the trials are expanding internationally while at the same time laboratory tests are becoming increasingly complex.

Pathology is used to confirm patient diagnosis during enrollment, as well as validate outcome during and at the end of the trial. With international trials, fewer patients per site and more sites makes pathology standardization of high importance. Pathologists using digital slides can compare results across sites, have a consolidated digital database for viewing histopathology results, and run image analysis tools to make precise measurements. An anatomic pathologist at a CRO can to do confirmatory reading during patient enrollment at a remote international location without the cost of time and travel.

Both communication and logistics improve between the pathologists at the CRO and the study directors and oncologists at the pharmaceutical sponsor. Biomarkers can be measured for supplemental information in the clinical trial. For example, Chicago-based Targeted Molecular Diagnostics (Westmont, Ill.), a CRO acquired by Quintiles Transnational Corp. (Research Triangle Park, NC), uses digital pathology to multiplex downstream phosphomarker measurements as efficacy readouts in kinase inhibitor clinical trials. 

The local tissue response to treatment is a key diagnostic component at every stage of drug development. Adoption of digital pathology leads to more precise and faster decisions across discovery, preclinical, and clinical trials.

About the Author
Dr. Steven Potts holds a PhD in biological engineering and an MBA from UC Davis, and a BS in physics from Wheaton College, Illinois.

References
1. Potts SJ, et al. Performance of a novel automated microvessel analysis algorithm across whole slide digital images. Society of Toxicologic Pathology Annual Meeting 2008.
2. Milici AJ, et al. Comparison of Plaque Burden in Tg2576 Mice Using the Aperio and TurboScan Imaging Platforms. Pfizer Global Research & Development. Pathology Visions Conference 2008.
3. Potts SJ. 2009. Digital pathology in drug development: multiste integration. Drug Discovery Today, In Press.
4. Novartis NIBR Application Note. Available at www.aperio.com.
5. Deeds J, Gardner H. Multi-Site Integration with Digital Pathology: Three Countries, Four Time Zones, One Oncology. Pathology Visions Conference 2008.
6. Hill J. The Use of Image Analysis in Assessing Biomarkers: Implications to Clinical Trials, Drug Development and Patient Treatment. Targeted Molecular Diagnostics. Pathology Visions Conference 2008.

This article was published in Drug Discovery & Development magazine: Vol. 12, No. 10, November/December 2009, pp. 24-27.

Advertisement
Advertisement