Seeing the Future with Imaging Science: Interdisciplinary Research Team Summaries (2011)

“Biomarker” is a term often used in the biomedical disciplines for a characteristic that can be used as an indicator of some biological condition or outcome that is ultimately of interest but difficult to ascertain directly, at least with respect to applications in human disease. A number of implicit and explicit definitions of “biomarker” are in common circulation. There are also quite different uses of the term in other disciplines (National Research Council report, Opportunities in Neuroscience for Future Army Applications, referenced in “Suggested Reading”). To avoid confusion, the IDR Team may wish to adopt the following definition, published by the Biomarkers Definitions Working Group of the National Institutes of Health (Atkinson et al., 2001, p. 91, referenced under Reading):

Therefore, biomarkers should be proven surrogate endpoints that can accurately predict clinical endpoints such as how a patient feels, functions, or whether or not the patient survives. In the field of psychiatry, diseases are categorized and identified based on clinical symptoms that may not specifically reflect the underlying pathophysiology present. Biomarkers have the potential to probe physiological processes to provide quantitative methods for differentiating illnesses with similar clinical symptoms resulting from unique neurobiological mechanisms, such as schizophrenia and other forms

of psychosis. Furthermore, many psychiatric diseases have early asymptomatic or transitional phases that can last years. In the case of schizophrenia, many patients experience several pre-psychotic phases prior to full onset of symptoms. Novel biomarkers could potentially identify those individuals most likely to progress to the illness in order to target preventive or early treatment. In addition to prediction and early detection, longitudinal biomarker measures may be used to assess the trajectory of the disease process. Those biomarkers that are sensitive to changes in the trajectory caused by certain drugs could be used to identify potential responders to specific pharmacological interventions (personalized medicine).

Currently there are many potential biomarkers in development for diseases such as Alzheimer’s disease, depression and schizophrenia. In depression, measures of quantitative electroencephalogram (EEG) concordance are being investigated. Potential biochemical markers include measurement of amyloid plaques and other proteins or hormones in cerebrospinal fluid or plasma. A large body of research is also focused on the search for genetic markers of these diseases. Currently, these markers have been promising in identifying individuals who are at high risk for illness; however, many lack specificity to predict those who will progress to full symptoms among the vulnerable individuals.

Neuroimaging provides several non-invasive tools for the determination of potential psychiatric biomarkers. Structural imaging such as anatomic magnetic resonance imaging (MRI) can detect lesions and atrophy, as well as cerebrovascular disease, associated with a variety of psychiatric illnesses. Functional imaging such as positron emission tomography (PET) has been used to find distinctive patterns of glucose hypometabolism in dementia. Also, changes in brain function in response to specific cognitive tasks measured with functional MRI (fMRI) may be indicators of disease. White matter tract integrity can be measured with MRI diffusion tensor imaging and has been found to correlate with drug response in patients with depression and in early stages of schizophrenia. Another MRI technique, MR spectroscopy is useful in measuring biochemical levels in a single voxel or across the brain to assess integrity of specific metabolic pathways that may be altered in disease. These and other imaging biomarkers may be combined with complementary physiological and biochemical measures for the most sensitive and specific indicators of illness.

The safe and non-invasive nature of many of these imaging techniques such as MRI make them attractive for early and repeated screenings of large populations. However, while these neuroimaging biomarkers of psychiatric

disease are promising, there are many obstacles that must be overcome in order to make them clinically feasible. These include, but are not limited to, the following: imaging and processing methods must be standardized and repeatable; results must be validated and interpretable in relation to standard norms of brain maturation and aging; drug effects on potential measures must be fully elucidated; and studies must be performed with large populations of well-characterized patients and at-risk individuals. The challenge here is to identify all of the qualifications of clinically useful imaging biomarkers for psychiatric disease and the tools and methods required to develop these efficiently.

• How can imaging biomarkers be used to demonstrate neurobiological mechanisms of disease?

• How can they be used to determine disruptions in connectivity across brain networks that may be the underlying cause of psychiatric brain disorders?

• What new technologies are required to allow validation and increase predictive power of imaging biomarkers of psychiatric diseases?

• What methods/qualifications should be used in determining which imaging modality is most useful for the given application? How do we jointly optimize the biomarker and the imaging system?

• How can imaging biomarkers be most efficiently utilized in conjunction with more invasive biomarkers?

• What can we learn from on-going large mutli-site studies such as the ADNI study (http://www.adni-info.org/Home.aspx)?

Atkinson AJ Jr., Colburn WA, DeGruttola VG, DeMets DL, Downing GJ, Hoth DF, Oates JA, Peck CC, Schooley RT, Spilker BA, Woodcock J, and Zeger SL. Biomarkers and surrogate endpoints: Preferred definitions and conceptual framework. Clin Pharm Ther 2001;69(3):89-95. Abstract accessed online June 15, 2010.

Cedazo-Minguez A and Winblad B. Biomarkers for Alzheimer’s disease and other forms of dementia: clinical needs, limitations and future aspects. Exp Gerontol 2010;45:5-14. doi: 10.1016/j.exger.2009.09.008. Accessed online June 15, 2010.

Correll CU, Howser M, Auther AM, Cornblattt BA. Research in people with psychosis risk syndrome: a review of the current evidence and future directions. J Child Psychol Psychiatry 2010;51(4):390-431. doi: 10.1111/j.1469-7610.2010.02235.x. Abstract accessed online June 15, 2010.

Kumar A, Ajilore O. Magnetic resonance imaging and late-life depression: potential biomarkers in the era of personalized medicine. Am J Psychiatry 2008;165(2):166-8. doi:10.1176/appi.ajp.2007.07111771. Accessed online June 15, 2010.

National Research Council. Opportunities in neuroscience for future army applications. National Academies Press; Washington, DC, 2009.

Pantelis C, Yucel M, Bora E, Fornito A, Testa R, Bewer WJ, Velakoulis D, Wood SJ. Neurobiological markers of illness onset in psychosis and schizophrenia: the search for a moving target. Neuropsychol Rev 2009;19:385-98. doi: 10.1007/s11065-009-9114-1. Accessed online June 15, 2010.

• Owen T. Carmichael, University of California, Davis

• Dennis W. Choi (IOM), Simons Foundation

• Marcel Adam Just, Carnegie Mellon University

• Linda J. Larson-Prior, Washington University in St. Loius

• Carolyn C. Meltzer, Emory University

• Andrew B. Raij, University of Memphis

• A. Ravishankar Rao, IBM Research

• James M. Rehg, Georgia Institute of Technology

• Bruce R. Rosen, Massachusetts General Hospital

• Kathleen M. Raven, University of Georgia

Kathleen Raven, NAKFI Science Writing Scholar, University of Georgia

Brain images of neurological and psychiatric disorders are needed to help research, make diagnoses, track disease progression, and monitor treatment. One problem that neurological researchers face is how to detect a disease as early as possible. The safe, non-invasive nature of imaging technology available today makes them attractive for repeated screenings of large populations in order to identify images that mark the onset or presence of neurological diseases. In order to create image biomarkers, researchers would collect data using technology such as magnetic resonance imaging (MRI) and positron emission tomography (PET). Currently scientists and physicians can detect clear neurobiological changes visible from a stroke, for example. In the future, similar changes in the brain as a result of depression may be visible.

Image biomarkers would complement already established biochemical and genetic biomarkers for diseases such as Alzheimer’s and depression. For example, a biochemical marker of Alzheimer’s disease is the presence of amyloid plaque in a patient’s brain visible through MRI scans. The trouble with biochemical and genetic markers is that they work best only in high-risk individuals; many lack the specificity needed to predict who will progress to have full disease symptoms later.

Although image biomarkers of psychiatric diseases are promising, many obstacles must be overcome in order to make them clinically feasible. These include, but are not limited to, the following: the imaging and processing methods must be standardized and repeatable; the results must be validated and able to be interpreted in relation to standard norms of brain maturation and aging; the drug effects on potential measures must be fully elucidated; and studies must be performed with large populations of well characterized patients and at-risk individuals. The challenge here is to identify all the qualifications of clinically useful imaging biomarkers for psychiatric disease and the tools and methods required to develop these efficiently. An interdisciplinary research team (IDR) at the National Academies Keck Futures Initiative Conference on Imaging Science debated these and other challenges surrounding the tools and validation methods required to develop clinically useful non-invasive imaging biomarkers of neurodegenerative and psychiatric disease.

For the purposes of IDR team 6’s discussion, the group agreed to the biomarker definition provided by the Biomarkers Definitions Working Group of the National Institutes of Health:

The group conceded that image biomarkers have several distinct applications. These include identification of the disease state, tracking disease progression, and evaluation of therapeutic response. Because of the brain’s structural, anatomical, and functional complexity, the group advised against searching for single image biomarkers. Instead, a biomarker panel should be used. The panel should incorporate data from structural imaging (e.g., MRI and MR spectroscopy) and functional imaging (e.g., PET scans). A successful biomarker panel would also take into account changes in the brain’s anatomy, physiology, metabolism, electrophysiology, and neurochemistry over time. Examples of indirect biomarkers—these could also be thought

of as “proto-biomarkers”—could be markers that predict the disease course or therapeutic response. The team cited blood pressure as a classic indirect biomarker—although unrelated to the field of psychiatric diseases—in the prevention of cardiovascular disease.

One of the main concerns shared by the group in determining image biomarkers is how to identify relevant information within brain structure as image resolution becomes higher and higher. The team predicted a massively expanding role for structural imaging as image resolution approaches the 10μ (micron) scale. Researchers will contend with even more image information and will need to make additional decisions about what is useful and useless. However, greater magnification of brain structure could potentially shrink the continuum between “neurobiological” and “psychiatric” diseases.

One team member who specializes in autism research suggested that data points gathered from behavioral patterns should be strongly correlated with image biomarkers. Autism research has long been tied to behavioral markers, such as the direction and length of a child’s gaze on the moving lips of a person talking. Novel sociological means of capturing such information could include such devices as goggles for toddlers designed to track eye gaze to conversation-monitoring devices on mobile phones.

Imaging biomarker tools and methods should take into account temporal changes in the brain across time spans as short as image measurement time to as long as years. Longitudinal data can help researchers distinguish significant from insignificant changes in brain images. Similarly, the team suggested the assessment of “microstates” beyond just arousal, or nonresting/sleeping, state of the brain. The tracking of these states could lead to predictive or disease markers and possibly guide therapeutic treatment.

Additional recommendations for imaging biomarkers made in the final presentation are as follows:

• Explore the role for public-private partnerships (e.g., the Foundation for the National Institutes of Health and the European Union’s Innovative Medicine Initiative) in driving the coordinated development of optical imaging technology and novel molecular probes

• Emphasize methods for imaging through skull and gaining portability

• Combine structural and functional modalities to develop increas­ingly precise, multidimensional baseline images of regions and pathways altered in psychiatric diseases (e.g., reward, mood, and theory of mind)

• Assess both highly selected and representative—or comorbid—populations in search for biomarkers

• Emphasize adaptation of large-scale informatics, computational, and feature extraction approaches to novel forms of imaging and behavioral datasets (e.g., graphs)

The group identified priorities within two types of biomarkers that could have significant implications for public health and would help, through early detection and treatment, reduce the cost of health care. Within predictive biomarker research, much work remains to be done on the placebo response to antidepressant therapy. Identifying biomarkers associated with the emergence of earliest symptoms in presymptomatic individuals would allow earlier intervention. Opportunities exist to determine the predictive markers of weight gain, which is the most common side effect associated with atypical antipsychotic agents. Therapeutic biomarker priorities include mood disorders, deficient social interactions, symptoms of schizophrenia, and alcoholism.

In conclusion, the group cautioned against confusing biomarkers of disease with byproducts of disease. A common example is the formation of amyloid plaque found in patients diagnosed with Alzheimer’s disease. It is unclear if the plaque is an indicator of the disease or the result of a disease mechanism. A recurring theme throughout the discussion was the important role that computer software developers currently have, and will continue to have, in processing new information gathered from the ways discussed above. As more information is collected, the sophistication of computational methods and algorithms to make sense of data will, by necessity, increase.

This page intentionally left blank.