Is clinical systems pathology the future of pathology?
Ambulatory medical care
(Forecasts and trends)
Ambulatory medical care (Standards)
Clinical pathology (Forecasts and trends)
Clinical pathology (Standards)
|Publication:||Name: Archives of Pathology & Laboratory Medicine Publisher: College of American Pathologists Audience: Academic; Professional Format: Magazine/Journal Subject: Health Copyright: COPYRIGHT 2008 College of American Pathologists ISSN: 1543-2165|
|Issue:||Date: May, 2008 Source Volume: 132 Source Issue: 5|
|Topic:||Event Code: 010 Forecasts, trends, outlooks; 350 Product standards, safety, & recalls Computer Subject: Market trend/market analysis|
|Geographic:||Geographic Scope: United States Geographic Code: 1USA United States|
At present, the practice of diagnostic anatomic pathology is
fragmented by technologic and subspecialty divides and often generates
fragmented information. The pathologist as a diagnostician is well
placed and should be the integrator of all available relevant data into
a coherent whole (the disease entity). Disease entities are at the
moment the best way to predict the fate of the patient (prognosis) and
how best to treat the sick person. However, mechanistic classifications
of lesions, particularly of tumors, are increasingly driving therapy and
demand specialized reagents and expertise. However, more often than not,
because of specialization, technologic divides, and constraints of the
health care systems, the pathologist delivers information, often
descriptive, that needs to be integrated by the clinician-therapist. The
issue is not so much who is best placed intellectually to make a
synthesis, the issue is that the interpretation of what is seen on a
gross specimen, histologic section, or cytologic preparation depends to
a very significant extent on the clinical and biologic context of the
findings. That is why a slide is not read, rather it is interpreted, an
activity that is far from being objective.
Today, the gross anatomic features of a lesion are captured, in exquisite fashion, by diagnostic imaging. One can hardly look at slides without knowledge of the radiologic findings. Consequently, in many instances, the pathologist examining histologic slides must look at multi-modality images (magnetic resonance imaging, nuclear magnetic resonance spectroscopy, positron emission tomography, x-rays) to understand the gross morphology, and it has been argued that the activities of radiology and pathology should be under the roof of a single department.
The factors that have lead to this state of affairs and that, to a certain extent, maintain this situation are (1) the fact that we have evolved as a visual species and that our brains are remarkably efficient in recognizing patterns but not adept at extracting quantitative information from images (1); (2) 400 years of gross and microscopic morbid anatomy have lead to the accumulation of a large descriptive and correlative repository of information; (3) the generation and widespread adoption of new technologies to interrogate tissue has been slow; (4) the interpretation of a slide requires experiential knowledge (acknowledged by the fact that the bill for every anatomic pathology examination or test carries a professional component); and (5) the combined dynamics of factors shaping the discipline and its practitioners have been such that the practice of surgical pathology, even today, is eminence-based rather than evidence-based.
Acting against these factors are a series of forces for change, including (1) rapid progress in technologies that produce high-density data sets; (2) progress in information technology, computational science, and bioinformatics; (3) the ascent of complex-systems sciences coexisting with the reductionism that yielded the great successes of the past century; (4) access to the singularity of the individual (genome sequence); (5) globalization of medical issues and practice; and (6) health care economics.
I posit that one possible response to these forces is "clinical systems pathology." For the purpose of this discussion and future debate, I define it as follows: Clinical systems pathology seeks to understand perturbed physiologic systems and complex pathologies in their entirety by integrating all levels of functional and morphologic information into a coherent model. It is practiced as a combination of "bottom-up" data collection, often comprehensive (omics), and "top down" computational modeling and simulation. It enables the rational design and testing of effective personalized predictive medical intervention and preventive measures.
HOW IS CLINICAL SYSTEMS PATHOLOGY REDUCED TO PRACTICE?
First, systems pathology requires the integration of clinical data, in vivo imaging (including functional imaging), ex vivo morphology (including functional morphology), and comprehensive data sets (eg, genomics). This level of integration demands the application of knowledge engineering and the use of artificial intelligence. The major changes from the present standards of practice consist in shifting from a single modality of interrogating tissues or cells (conventional histology and immunohistochemistry) to a quantitative multichannel modality and using artificial intelligence for classification and model building. Eventually the integration of all data sets pertaining to one patient produces a more personalized and individually predictive diagnosis, and as a consequence, a therapy that is tailored to the individual (Figure 1). Techniques to improve the integration and analysis of data obtained with different platforms are constantly evolving and being improved. (2)
[FIGURE 1 OMITTED]
As proof of concept, we can examine an early example of clinical systems pathology applied to prostate cancer (3) conducted at Aureon Laboratories (Yonkers, NY).
The aim of this exercise was to test whether clinical systems pathology would better predict the prognosis of patients diagnosed with prostate cancer and help sort out the therapeutic options. For that, we integrated elements of the clinical data set, morphometric data, and quantitative molecular markers. In this case, rather than collecting a large number of data points (eg, gene expression profiles) we collected a limited set of data taking into account the spatial distribution of the values we capture. In other words, functional (molecular-level) information is put in the context of the tissue architecture (Figure 2). For example, in a single slide, the amount of androgen receptor protein in the nuclei of individual cells is measured in the stromal cell, in a tumor cell, or in the nontumoral epithelial cell compartment. It is important to underscore that the architectural features of the tissue are also captured and that a large number of features are subjected to analysis. In this fashion we practice "quantitative phenomics" in contrast to the "pattern phenomics" we have been traditionally accustomed to. The advantage is that high-density "quantitative phenomics" data sets can be effectively integrated with the rest of the comprehensive data concerning a specific patient. The ensemble of data is then processed by artificial intelligence to identify, in a training cohort of cases, the combination of data that most accurately predicts the outcome of a given therapy, and the predictive power is subsequently validated on an independent set of patients. Application of clinical systems pathology to prostatectomy specimens yields a significant improvement over the present standard of care in the prediction of prostate-specific antigen recurrence and disease recurrence for a particular individual. (4) Preliminary results indicate that the outcome of a surgical intervention (prostate-specific antigen recurrence and disease-free survival) can also be predicted from features analyzed in needle biopsy cores, a finding that may help physicians and patients choose among therapeutic options. Ongoing studies are focusing on predicting the response to androgen deprivation therapy and the prediction of the response to radiation.
The limited experience with clinical systems pathology suggests that it represents a viable adaptation to the forces influencing pathology and health care at the beginning of the 21st century. It also suggests that by interrogating tissues with sophistication we can shift from eminence-based practice to a personalized evidence-based practice. The ability to demonstrate complex causal interactions operating at the tissue level gives us the same type of evidence that DNA forensics brings to the criminal justice system. It links the culpable (eg, a disturbed regulatory network) to the crime (the disease), hopefully with the same certainty level (>99%) as DNA fingerprinting condemns or exonerates a suspect. The clinical setting and initial macroscopic and microscopic look at the lesion will guide the selection of technologies and algorithms to be used in working up the patient's diagnosis and personalized therapy. The answer will be objective and provide evidence of causality; "because I say so" will be replaced by "because it is so." The consequence of this scenario for the practicing pathologist is that rather than seeking the opinion of an expert she or he will be accessing technologies, perhaps remote, to solve a case and provide a rational therapeutic indication through clinical systems pathology.
[FIGURE 2 OMITTED]
The road ahead is neither straight nor simple. The infrastructure to access, test, and adapt technologies to the clinic is onerous. Converting discovery strategies to efficient diagnostic tests is expensive and risky. The conduct of diagnostic trials is difficult, but we can overcome obstacles by hitchhiking on therapeutic trials. More importantly, training programs, pregraduate and postgraduate, need to prepare the next generations for a different world, a medical world in which the weaving of theory, modeling, and empirical data sets create the cloth of medical progress. These are times of change, but the winds are favorable for those who study causes and mechanisms of disease to bring science to the art of diagnostics.
(1.) Rosai J. Why microscopy will remain a cornerstone of surgical pathology. Lab Invest. 2007;87:403-408.
(2.) Deng X, Geng H, Ali HH. Cross-platform analysis of cancer biomarkers: a bayesian network approach to incorporating mass spectometry and microarray data. Cancer Informatics. 2007;2:183-202.
(3.) Saidi O, Cordon-Cardo C, Costa J. Technology insight: will symptoms pathology replace the pathologist? Nat Clin Pract Urol. 2007;4:39-45.
(4.) Cordon-Cardo C, Kotsianti A, Verbel DA, et al. Improved prediction of prostate cancer recurrence through systems pathology. J Clin Invest. 2007;117:1876 1883.
(5.) Jenkins RB, Nakagawa T, Kollmeyer T, et al. Use of a tissue biomarker panel to predict which men with a rising PSA post-definitive prostate cancer therapy will have systemic progression. 2007 ASCO annual meeting proceedings, part 1. J Clin Oncol. 2007;25(18S):5017.
(6.) Sole R, Gonzalez-Garcia I, Costa J. Spatial dynamics in cancer. In: Deisboeck TS, Kresh JY, eds. Complex Systems Science in Biomedicine. New York, NY: Springer; 2006:557-572. Topics in Biomedical Engineering International Book Series.
Jose Costa, MD
Accepted for publication January 10, 2008.
From the Yale University School of Medicine, New Haven, Conn. Dr Costa is a member of the Board of Directors of Aureon Laboratories, Younkers, NY.
Presented at the College of American Pathologists Futurescape of Pathology Conference, Rosemont, Ill, June 9 and 10, 2007.
Reprints: Jose Costa, MD, Yale University School of Medicine, 20 York St, EP2-607A, New Haven, CT 06510 (e-mail: jose.costa@yale. edu).
|Gale Copyright:||Copyright 2008 Gale, Cengage Learning. All rights reserved.|