Participants at the Barcelona meeting began a discussion of the broad areas to be addressed by the consortium. Elaboration of these focus areas will continue through the efforts of working groups and subsequent group meetings.
4.1. Cohort: Who?
The goals of this research enterprise demand that the study cohort reflect different cultures, genetic backgrounds, and access to services. This will require standardization across many centers around the world. Participants should also be available for serial measurements. Given the large number of existing cohorts assembled for the above-mentioned studies and others, participants questioned whether a new cohort is needed, or if existing cohorts can be combined, concatenated, or expanded. Another question concerned whether a subgroup of people who have aged successfully should be included, to study factors that contribute to healthy brain aging.
4.2. Domains: What?
In determining what to measure, a number of governing principles emerged. First was the need to remain flexible, because both our technology and our understanding of the biological basis of AD are rapidly evolving. Mechanisms for adapting to new knowledge and technology will have to be built into any study design. The second principle states that brain aging cannot be disentangled from aging in other physiologic systems, such as the kidney and heart. Worldwide, there will be a tremendous explosion of aging people in Asia, Latin America, and Africa, where people have different health problems than in the developed world. Therefore, we will need to understand brain function against the backdrop of other health changes. The third principle maintains that the amount of information from various measures and the amount of accompanying “noise” are not constant. Thus, a measure that provides much information at later stages of the disease may not give much information at early stages (e.g., certain imaging studies). Moreover, the relationship between biological and clinical phenotypes is not clear. Thus to extract information, measurements will be required across multiple domains, such that investigators can look for patterns of change. Individual change in one domain may not be very useful by itself, but in the context of other changes, it can become meaningful. It will be imperative to determine how changes relate to one another.
Keeping these basic principles in mind, the group identified a number of domains that will be important to measure. Some disagreement arose about the relative importance of measuring changes in the brain versus changes in function.
Cognition, of course, is the key domain, and within this domain, memory, attention, executive function, speed of processing, language, and visuospatial performance are all important. In a typical person with AD, memory problems become apparent first, followed by problems with executive function and language. Embedded within the area of memory, measures of short-term versus consolidated memory may be needed. Several issues were discussed involving the cognitive domain. Some participants expressed concern about a ceiling effect with cognitive measures. As the population becomes increasingly educated, it becomes more difficult to detect differences, especially in early stages of the disease. An ideal cognitive battery would contain an entry level (different at different ages), as well as a threshold where everyone fails in some aspect. Whatever measures are used, they will have to be designed so that they are neither education-dependent nor culture-dependent.
Other domains of the study include an assessment of mood and behavior, which can be affected up to 10 years before dementia is detected. Activities of daily living are recognized as a key metric that divides MCI from AD. However, no standardized measures are available for the predementia stages of AD. Motor function can be assessed using the timed-walk, up-and-go, and two-step treadmill tests. To supplement tests of motor function, we may want to include assessments of muscle mass and strength. In the cardiovascular area, at minimum, blood pressure should be measured. Other measures to consider in this area include arterial stiffness and pulse-wave velocity. In the metabolic domain, we should include body mass index and food-frequency questionnaires. It will also be important to collect data on family history. In the area of genetics, we will want to collect data regarding the question of what genetics can tell us about the early stages of the disease. Expression profiling may allow an assessment of patterns of genetic changes, and these may relate to other health problems.
The study will also involve the collection of biological materials, including blood and cerebrospinal fluid, and possibly neurological tissue for environmental studies. Serum samples would be used not only for biomarker studies, but also to evaluate the overall health of other organ systems through laboratory tests such as hemoglobin A1c (HbA1c), creatinine, and glucose. DNA from blood would be used to develop new genetic markers and evaluate existing genetic risk factors.
Another key component of the study will involve imaging. For this, the ADNI study can be used as a template, although it does not presently examine vascular dementia. The ADNI protocol was designed to focus on the conversion from MCI to AD, and from normal function to MCI. We would like to look more broadly at antecedents for cognitive dysfunction. Thus, although ADNI and European-ADNI have made significant progress in developing systematic protocols, we may need new technologies for imaging early changes in the brain, including new ligands for positron-emission tomography. Other poorly understood areas where imaging may provide useful tools include neurogenesis in the dentate gyrus, and synaptic and dendritic pruning.
4.3. Data gathering: How and when?
A primary goal of this enterprise is to make data-gathering a routine part of the general health examination, similar to the way blood-pressure and cholesterol measurements are performed presently. Cognitive assessment should also be included as a part of general medical practice, so that data are collected in a systematic way, and sampled in a standard way. General practitioners, or primary-care providers, will be especially important on the front line of the data-gathering effort. Thus protocols should be structured as simply as possible, to collect a representative sample. A user-friendly toolbox for general practitioners is essential. This may include computer-based or web-based assessments that can be set up at a kiosk, although an alternative instrument may be needed for rural or other areas that lack broadband internet access, or where computers are not available. Self-administered assessment tools could also be considered.
In designing data-gathering tools, it will be important to bear in mind the overall burden on participants and clinicians (particularly if primary-care providers are involved), because too heavy a burden could decrease compliance with the study. One possible strategy would involve a basic “toolkit” that collects a minimal amount of data from everyone, and then focuses on certain areas at specialized centers. Another issue relating to both participant and clinician burden pertains to the collection of bodily fluids. The collection and storage of samples will need to be standardized and monitored. One suggestion advocated an “optimal” standard as well as an “acceptable” standard, with samples coded so that investigators would know which samples are usable in certain studies. For example, for proteomics studies, sample collection needs to be tightly controlled.
The frequency of data collection from each participant will also need to be determined. The data-collection visits may increase as a participant ages.
4.4. Data management
Managing data presents another set of questions to be resolved. For example, should data be stored in a centralized location, or at multiple, decentralized locations? Should blood and tissue be stored in a central repository? Once again, looking to existing studies should provide some guidance in answering these questions. A minimal, international data set will be established, to ensure that comparisons can be made across sites.
Data sharing, data mining, and knowledge extraction emerged as possible barriers that will require creative solutions. The NIH require data from NIH-funded studies to be shared through qualified-access databases, but some European countries will not allow their scientists to share data. One possible solution would permit members of the consortium special exemptions to share data across international boundaries. Decisions will also have to be made regarding who will have access to data. A number of possible models exist. The ADNI, for example, provides access to almost everyone, and what people do with the data is unrestricted. However, this has resulted in some misinterpretation and misuse of data, as well as conflicting results from different investigators analyzing the same data. In the Genome-Wide Association Studies, everyone who contributes data can take their own data, but not aggregate data, back to their home site. Other studies restrict access to data, or provide only portions of the data to investigators outside the study.
The issues of data sharing and standardization need to be resolved at the outset. Convoluted mechanisms for sharing data interfere with scientific progress, and this has not been successfully addressed in existing or previous international collaborations. In Europe, databases with clinical data and biomaterials are difficult to share outside the researchers' home country. Moreover, when attempting to share data between countries, laws and ethical standards and issues of ownership, intellectual property, and privacy all require careful consideration and management.
4.5. Sites: Where?
Existing large cohort studies may provide the easiest and quickest path to start NASAU (i.e., the EU-NA enterprise). Potential United States study sites that may be invited to participate include the Mayo Clinic and the University of Pittsburgh Alzheimer Disease Research Center (ADRC), the Pittsburgh Collaboration with Barcelona, the Women's Health Initiative (multiple sites), the CHS (multiple sites), the Action to Control Cardiovascular Risk in Diabetics (ACCORD) study trial (multiple sites), and the Framingham Heart Study (Boston, MA). Potential sites in Canada include those in the C5R network.
In Europe, possible sites include the German network, the French Center for Excellence in London, Spain's clinical trial network (the Consorcios Asociados de Investigación Biomédica en Red), the Million Women Study in the United Kingdom, and the Monica project in France (Paris, Toulouse, and Strasbourg). Collaborating with the Sarkozy Initiative in France may also be possible.
As currently conceived, the EU-NA enterprise would involve joint programming and two parallel structures with overlapping leadership, to facilitate cross-talk between the EU and NA investigators. After we have demonstrated the practicality of this model through our working group of scientists, the NIH, the Canadian Institutes of Health Research, and the European Commission may be convinced to involve themselves and support the project. Presently, two cross-Atlantic models are being developed. One, a collaboration initiated between Bruno Vellas in France and the late Leon Thal's Alzheimer's Disease Cooperative Study (ADCS) in San Diego, continues to design a multidomain trial regarding prevention. The second model is the Pittsburgh-Barcelona collaboration. We expect to build on and expand these models.
The ADNI provides another model that may serve as a template for designing the administrative aspects of this study. Similar to the German Competence Network, the ADNI has a central administrative core with an external advisory board, an industry scientific advisory board, a resource-allocation committee, and centralized facilities for data banking, data mining, data storage, and several technology cores, such as the neuroimaging core at the University of California at Los Angeles. The ADNI has taken advantage of the ADCS for clinical coordination, and also has a data-coordinating center. Genetic data are stored at one site, which offers advantages in terms of distribution: more sophisticated users can obtain raw data, whereas other users can access data in other, more usable forms.
Informational technology administration can be particularly difficult to manage across countries and continents because of the different technologies that are used. This is another area to be addressed.