5—
Implementation
The preceding chapters identify three major research initiatives that can yield breakthroughs in understanding the aging mind and practical applications for improving the functioning of older adults. In some instances, we have identified issues of implementation specific to particular research initiatives (for example, the need to develop innovative funding mechanisms for basic cognitive research needed to develop new technology to improve the adaptivity of older adults). This chapter addresses a set of implementation issues that cut across the research initiatives and offers some specific recommendations. They concern the organization of research support under the initiatives and the needs to promote interdisciplinary research, support infrastructure, and collaborate across agencies.
ORGANIZATION OF RESEARCH SUPPORT
In our judgment, the best way to support research under the three initiatives will vary. In some research areas (for example, the application of new mathematical techniques to cognitive aging), it may be sufficient at first to request proposals from individual investigators. In other areas, it may be advisable to organize competitions that require applicants to work together across disciplines. For example, the NIA might ask cognitive psychologists and engineers to work together on proposals for research on technological supports for cognitive performance in older adults, or neuroscientists and behavioral scientists to work together to propose research linking neural and behavioral phenomena at high resolution. This strategy may also prove useful
for encouraging the early stages of research on other topics that are likely to require unusual interdisciplinary collaborations (e.g., the neurobiology of training and practice). Appendix D offers a detailed discussion of the need for interdisciplinary collaboration on research on technological support, the benefits that may be gained, and some of the challenges.
In certain areas it may be worthwhile to hold competitions for small research teams or centers that would bring together cognitive scientists, behavioral scientists, and neuroscientists around a common problem. It is especially worth considering this support option for the initiative on cognition in context. An emerging interdisciplinary field such as the neurobiology of life experience might develop most rapidly as the result of the work of such interdisciplinary research groups.
To pursue the research initiatives successfully, it may be advisable for the NIA to establish some new research programs or program offices, or at least to establish special emphasis panels to review the proposals. The initiative on cognition in context may be an example, as it encompasses an unusually wide range of disciplines. We encourage the NIA to consider these organizational issues carefully and to consult with the research community as it develops the research initiatives and to seek forms of organization that will encourage the needed interdisciplinary integration (see next section).
PROMOTING INTERDISCIPLINARY RESEARCH
As this report makes clear, the study of cognitive aging is highly interdisciplinary and becoming more so. Each of the recommended research initiatives requires collaborations among neuroscientists, cognitive scientists, and behavioral scientists. Progress on each initiative depends on the development of research from multiple points of view (multidisciplinary research) and on efforts to integrate these points of view into comprehensive understandings of cognitive aging (interdisciplinary research). These needs extend even beyond the already interdisciplinary field of cognitive science, which has developed by linking such diverse specialties as linguistics, artificial intelligence, and neuroscience with behavioral science.
A great diversity of approaches is appropriate given the research problems, but it presents challenges of training and coordination because of the required breadth of knowledge. The field of cognitive neuroscience in particular requires the rethinking of traditional disciplinary training programs. The investigation of the relationship between cognition and brain processes demands knowledge of behavioral techniques for identifying cognitive processes and architecture, as well as techniques for measuring structural and functional aspects of brain. This interdisciplinary approach is especially germane to cognitive aging research for which age-related changes highlight the dynamic relationship between brain and cognitive processes.
Recent advances in imaging technology have generated enormous interest in brain-behavior relationships, increasing the importance of training for neuroscientists in experimental techniques for investigating cognitive processes. Graduate training in neuroscience, however, often emphasizes techniques for investigating brain structure and process with little emphasis on behavioral techniques that are fundamental to the investigation of memory, language, attention, and other cognitive functions and less on linking neural phenomena and fundamental cognitive processes to adaptive behavior in real-life situations.
The NIA should support postdoctoral fellowships, conferences, workshops, and summer institutes to encourage individual scientists to broaden their knowledge and technical capability in order to address interdisciplinary issues that are central to the new research initiatives.
For example, these mechanisms should be used to encourage neuroscientists and neuropsychologists to increase their training in cognitive experimental techniques and cognitive models. They should be used to encourage applicants with graduate degrees in cognitive psychology, linguistics, artificial intelligence, and the study of socioeconomic, educational, and cultural differences to strengthen their knowledge of brain structure and their ability to use techniques for investigating brain function.
The NIA, perhaps in conjunction with other institutes, should organize a special competition that would provide multiyear support of a few small multidisciplinary research centers or teams focused on analytical problems posed by the research initiatives that require the simultaneous application of multiple perspectives (e.g., neurobiology of training or cultural difference, development of adaptive technology).
The mechanism of interactive research program grants may also be useful for encouraging collaboration on common problems by researchers approaching them from different disciplinary perspectives.
RESEARCH INFRASTRUCTURE
Progress on the three major research initiatives depends on supportive research infrastructure, as already noted. For example, we have noted the need for access to aged animals for research. To ensure adequate access to such animals requires support beyond that provided for individual research projects. Such support is a public good for research.
The NIA should help support the maintenance of colonies of patho-
gen-free aged animals (including primates, rats, and mice) in regional centers.
This support should probably be in collaboration with other research funding agencies. It will involve a long-term commitment, especially for primates, and the support of highly skilled researchers and animal care staff. The research community should be consulted for advice on which animal strains are most important to keep available for research purposes.
The rest of this section discusses two major types of infrastructure that require support for the benefit of research on all the recommended initiatives: general-use databases, and brain imaging research capability.
Building General-Use Databases
Strategic investments in general-use data can support research on all three recommended initiatives. In our judgment, the greatest opportunity for such support lies in developing and expanding long-term, longitudinal studies.
There is little dispute that longitudinal research is essential to investigate processes of age-related change directly, rather than indirectly in the form of cross-sectional differences that may reflect a mixture of the products of change and of preexisting differences. Nevertheless, largely because of the greater time and expense, there are currently far more reports of cross-sectional comparisons of cognitive variables than longitudinal comparisons.
For many years the dominant interpretation has been that cross-sectional and longitudinal comparisons in cognition yield substantially different results, with much smaller, and later-occurring, age-related declines in longitudinal contrasts than in cross-sectional contrasts. However, some recent data call this view into question (e.g., Hultsch et al., 1998; Zelinski and Burnight, 1997; Sliwinski and Buschke, 1999). It is clear that the relation between cross-sectional and longitudinal age trends is still not fully understood, and thus it is important to have more linkages between the two major methods of investigating age differences. Longitudinal studies are certainly not the only type of research that can yield valuable information about aging, but they provide unique information that is not available from other research designs.
Several high-quality longitudinal studies focusing on cognition are currently under way in the United States (e.g., Baltimore, Seattle, Los Angeles) and elsewhere (e.g., Victoria, British Columbia; Manchester, England; Berlin, Germany). However, there are a number of features not represented in most existing studies that would be very desirable to include in future longitudinal studies.
First, at least in the initial assessment, the sample of participants should reflect major demographic variations in the population, and not just healthy middle-class whites as is the case in most current U.S. studies. Obtaining a
truly representative sample may require that the same protocol be implemented in several different sites, but the benefits of a broader base of generalization are likely to justify the additional costs and effort. In addition, it will be valuable for some purposes to oversample groups of particular interest, such as members of ethnic and racial minority groups, people with particular life experiences (e.g., certain occupations) or health conditions believed to predispose to cognitive changes, people with known risk factors for dementia, and the oldest old. Knowledge about cognitive change in the oldest old is limited because most past longitudinal research started in middle age, with the result that by ages 85 and beyond, samples were too small to allow good generalizations. It might therefore be valuable to expand some longitudinal studies by adding respondents at older ages.
Second, to examine precursors to changes and to allow more complete investigation of trajectories over time, the research participants should span a broad age range, ideally beginning in the 20s instead of the 50s or later, and extending to the end of life so as to investigate reported associations of cognitive status with mortality. Beginning the longitudinal assessments at early ages seems warranted because cross-sectional analyses have revealed substantial age-related differences in some cognitive abilities before age 50 (e.g., see the meta-analysis in Verhaeghen and Salthouse, 1997). Furthermore, at least one recent report found that cognitive variables in early adulthood predicted cognitive status more than 50 years later (Snowden et al., 1996). In addition, as noted in Chapter 2, there is reason to believe that neural changes occurring in early or mid-adulthood may be the first signs of cognitive changes that are not clinically observable until much later.
Third, a wider variety of variables should be obtained from the research participants than has been typical in past research. Within the cognitive domain, several different types of cognitive abilities should be assessed, not just memory or intellectual abilities, as in some current projects. It is also desirable to include variables reflecting specific theoretical processes in attention, memory, and language instead of only psychometric variables that probably reflect an unknown mixture of processes. It is essential, of course, that these theoretically more precise variables be established to have high levels of reliability in order to be included in longitudinal research protocols.
In order to relate changes in cognitive tasks to changes in daily functioning, it would be valuable to include some measures of everyday, or ecologically valid, activities. These should include a mixture of activities for which a neural substrate has been identified and others known to require cognitive activity, but for which the neural substrate remains unknown. Data should also be collected on life experience factors that may affect cognitive aging. These include basic sociodemographic information (ethnicity, education, socioeconomic status, etc.), but also more specific experiential factors that may underlie group differences in cognitive aging (see Chapter 3).
Other types of variables should also be included in future longitudinal studies. Several existing longitudinal studies include assessments of physical health, personality, lifestyle, hormone levels, and genetic markers, and new studies should continue and expand on those efforts to understand the mind through the life course, including influences from the prenatal environment and correlated physical conditions (e.g., blood pressure, respiratory flow, use of medications). This effort would support the research initiative on cognition in context. Including measures of brain structure and function along with the cognitive and physical assessments would support the research initiatives on the structure of the aging mind and on neural health. Among the possibilities that might be considered are volume estimates from structural MRIs (e.g., Raz et al., 1998) and variables obtained from event-related potentials (ERPs) or functional MRIs.
As noted in Chapter 4, moderate to strong correlations have been reported between cognitive variables and sensory-motor variables in cross-sectional studies (e.g., Anstey et al., 1993, 1997; Baltes and Lindenberger, 1997; Lindenberger and Baltes, 1994; Salthouse et al., 1996, 1998). To pursue the research opportunity afforded by examining these correlations, it would be helpful for new longitudinal studies to include assessments of sensory and motor capabilities.
A body of longitudinal studies that includes this variety of variables would provide the fundamental database for a successful research initiative on the structure of the aging mind, as well as significant support to the other two research initiatives. Analyses of the data could focus not only on changes in mean levels of performance in individual variables, but also on changes in the interrelations among variables. Researchers could examine change in various cognitive measures to determine which antecedent conditions predict cognitive change. They could also examine changes in associations among variables, particularly after temporal lags. For example, they could consider whether a change in the relation between two variables from one measurement occasion to the next is predictive of later change in a third variable. This sort of analysis would be particularly interesting when the early and late variables are from different domains, such as neural and cognitive.
Research on rich longitudinal datasets could clarify where in the structure of variables age-related changes are most pronounced. Some research has suggested that longitudinal change operates on a general factor (e.g., Hertzog and Schaie, 1988; Hultsch et al., 1998). However, relatively little research has been conducted attempting to investigate the structure of change across different types of variables. Information of this type is potentially very important because it can clarify how specific or broad an explanation will be needed to account for age-related cognitive changes. Finally, powerful new analytical models, such as latent growth modeling and latent change analysis, could be applied that might yield insights not possible with cross-sectional data.
The NIA should undertake a major effort to expand or develop large-scale longitudinal studies of cognitive aging. The studies should cover the range of variation in the population and should support research aimed at understanding the relationships among neural, cognitive, behavioral, sensory-motor, health, and life experience variables as they affect cognitive aging. Other institutes of the National Institutes of Health should be invited to cooperate in this effort, as they may benefit from the type of comprehensive longitudinal research being developed.
This recommendation is for an expansion of research activity and is not intended to detract from existing laboratory-based research on cognitive aging. In fact, longitudinal research may require an expansion of laboratory research in order to develop, validate, and continually improve measures of cognitive processes that are essential for use in longitudinal and other research on cognitive aging.
Considerable attention needs to be devoted from the outset to how best to implement this recommendation. We are recommending the study of a very large number of variables in a very broad population, and practical limitations will severely restrict the number of variables that can be measured for each individual participant in a study. There are trade-offs to be made between the needs for in-depth analysis of particular cognitive functions and for understanding these functions in context (e.g., the roles of sensory-motor, health, and cultural variables). In addition, it is necessary for any measure to pass screens of reliability and validity to be included in a major longitudinal study.
The NIA and cooperating institutes should engage in structured discussions with the research community, perhaps through a series of workshops, to address the problems involved in using resources effectively to create a broadly useful base of longitudinal data on cognitive function and its neural, behavioral, physical, and experiential correlates.
These discussions should address the following issues, among others:
-
effective ways to combine large-scale and more focused research so as to continually strengthen the base and the value of general-use data;
-
selection of which neural, behavioral, physical, experiential, and other variables to include in large-scale longitudinal studies and which to use in smaller, focused studies;
-
assessment of the adequacy of particular measures or indicators for the variables selected for inclusion in longitudinal studies;
-
priorities for developing and validating measures and indicators that might be included in future longitudinal studies;
-
the value of adopting standard procedures for screening or categorizing research participants with regard to status on key health and cognitive variables so as to increase comparability across studies;
-
appropriate evaluation methods, including the possibility of special review, for research proposals within the effort to expand longitudinal research, particularly in light of possible tensions between the needs of long-term longitudinal research and the evaluation criteria traditionally employed by study sections.
Building Research Capability for Using Data from Brain Imaging
The potential of brain imaging technology for research on cognitive aging will not be realized without focused investments in infrastructure. Three key elements of infrastructure are missing.
A Common Database for Human Brain Imaging Data Linked to Behavioral Characteristics
The research community could greatly benefit from an effort to standardize measures made from human brain imaging data, particularly MRI data, that can be used for brain-behavior research so that data collected for different purposes can be combined into a growing general database. To achieve this benefit, standard procedures would have to be developed for collecting and reporting such data, which would probably begin with records that contain structural brain imaging data, particularly MRI data, and behavioral characteristics of the same individuals.
The NIA should support a consensus conference that would discuss limitations and concerns specific to functional imaging and aging and develop standard procedures for collecting and reporting human brain imaging data, specifically including MRI data, usable for studying brain-behavior relationships during the aging process.
The conference would aim to identify and reach consensus about specific problems, including age effects on the coupling of neural and hemodynamic responses, time of day for MRI observations (e.g., the inappropriateness of testing older adults at night), the need for a stereotaxic atlas for the brains of older adults, and the implications of structural images for interpreting functional images. The conference would publish a report recommending certain procedures as standard, thus making a common database possible. The approach could be similar to one used to develop standards for electrophysi-
ological measurement of event-related potentials (Donchin et al., 1977; Picton et al., in press).
Centers for Monkey Brain Imaging Research
At the present time, the most pressing need is for centers that include capability for fMRI research on monkeys. Because such research is best carried out in a custom-designed, vertical bore magnet, progress in all research fields that can benefit from real-time imaging of the aging brain depends on the availability of this expensive equipment. The cost of the equipment provides the rationale for establishing research centers, possibly in collaboration with other institutes at the National Institutes of Health (NIH). Centers would presumably be established in places where there is existing expertise in cognitive neuroscience and in the techniques needed for research with awake, behaving monkeys. Optimally, the monkey facilities would be sited in close approximation to human fMRI facilities, so as to permit the sharing of core personnel and computer software.
The NIA, working with other institutes at NIH, should establish a monkey brain imaging facility with FMRI capability at NIH and support a few similar centers elsewhere.
Adequate Access to MRI Physicists
The rapidly increasing amount of MRI research underlines the shortage of qualified MRI physicists to work on research teams studying cognition in aging. Although this problem is broader in scope than the research concerns of the NIA, it is likely to slow research progress unless appropriate action is taken. The NIA should work with other federal agencies on ways to address the problem.
COLLABORATION ACROSS INSTITUTES
We have specifically recommended that the NIA collaborate with other institutes to establish research centers, including one at NIH, where monkey brain imaging procedures, specifically including fMRI, can be carried out in a custom-designed, vertical bore magnet. We believe that this is a single example of a more general point.
The NIA should seek additional opportunities to participate with other institutes such as the National Institute of Mental Health, the National Institute of Neurological Diseases and Stroke, and the National Eye
Institute to develop initiatives in neuroscience and cognitive aging as a way to increase the power of its research investments.
Several examples illustrate the kinds of collaborations that are likely to be beneficial. The NIA might share in the support of animal colonies and benefit by improved access to aging animals for research purposes. Research to clarify the interpretation of fMRI signals and to develop protocols for recording data and statistical techniques for analyzing them would be beneficial not only to the NIA but also to other institutes that support research that relies on these signals. Collaborations with other NIH institutes that deal with disabled populations (e.g., the National Institute of Neurological Diseases and Stroke, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, the National Institute of Deafness and Other Communication Disorders) would probably help advance research on adaptive technology. And interinstitute collaborations can also help in carrying out the large-scale longitudinal studies recommended in this chapter. These studies, sometimes with the inclusion of a few additional variables in the research protocol, can provide valuable information on a range of issues relating to health and well-being. Collaboration with other institutes and agencies may make possible studies that could not be fielded with NIA support alone.