Many research groups are trying to develop prediction models for AD. There are several studies designed to refine clinical criteria for the earliest presentation of AD, and possibly move the diagnostic threshold back to earlier stages [6
]. However, as we do that, we may run the risk of reducing specificity. That is, we may be able to increase the sensitivity of predicting individuals who are likely to progress from MCI to AD, but in doing so, we may lose some of the specificity of that prediction. Studies on MCI in a community population setting have indicated that, the earlier we make the diagnosis of cognitive impairment, the more likely we are to encounter neuropathological heterogeneity [7
]. However, if we wait until the full dementia syndrome is developed, while we may be more specific with regard to the underlying neuropathologic substrate, but we likely have waited until the point at which a great deal of degeneration has occurred in the central nervous system.
The challenge is to improve sensitivity and specificity, and current thinking in the field suggests that biomarkers may aid in this process. While this is likely true, there are cautions that need to be considered with respect to this approach, as well. Among the most promising biomarkers at present include quantitative measures on MRI, cerebrospinal fluid (CSF) markers for tau and Aβ, genetic susceptibilities such as those found with apolipoprotein E genotypes, FDG-PET metabolism patterns, plasma markers for Aβ ratios and proteomic signatures, and molecular imaging such as amyloid imaging with Pittsburgh Compound B (PiB) [10
]. It makes intuitive sense that one or more of these markers coupled with a clinical phenotype may prove very useful.
One proposal to assess this approach would include using the criteria for amnestic MCI (aMCI) of a presumed etiology as the clinical phenotype and biomarkers could be added to increase the specificity that this clinical presentation will progress to clinical AD [16
]. Currently, some studies suggest this by documenting higher rates of progression from MCI to AD in those individuals who are ApoE4 carriers or have atrophic hippocampal formations [13
]. Similarly, CSF biomarkers have been shown to predict who is more likely to progress from MCI to AD more rapidly [12
]. Markers such as amyloid imaging have suggested an important role for this modality in predicting who will progress, but few longitudinal data are available at present to inform us on this issue.
The real challenge in predicting progression in MCI is to compare these markers to each other. That is, a head-to-head comparison is necessary to validate the predictive utility of each of these markers. Ideally, the true utility of these approaches should be validated in a population-based sample of MCI subjects, since clinic-based samples already have a distorted probability of developing AD in the near future. Consequently, it is difficult to predict from a referral population the true value of these multiple predictors in a more general setting, and, as such, community-based studies are necessary before these types of recommendations can be adopted for implementation in larger populations.
A similar strategy could be applied to developing prediction models for asymptomatic individuals, as well. Studying MCI is an important strategy, but as Leon had intended, this is not the ultimate goal of our clinical research. We should strive for assessing individuals even earlier in the clinical spectrum, prior to MCI. The same strategy would apply, but now, instead of using aMCI criteria, we might employ an even more subtle clinical presentation such as a modest memory impairment while individuals are still functioning in the normal range. To that, one might consider a positive family history for AD, certain ApoE genotypes, MRI measures, FDG-PET indices, plasma markers, CSF measures, and/or amyloid imaging. However, to make this feasible, the value of each of these clinical presentations and biomarkers would have to be validated in longitudinal studies. These types of investigations require the assessment of large longitudinally followed populations over extended periods of time. These studies are expensive and labor intensive, but necessary. The current funding climate makes these types of studies difficult to conduct and nearly impossible to support financially. At least one or two of these markers may be evaluated in a particular clinical setting, but what is needed is a longitudinal study of thousands of persons with many of these markers available simultaneously.
Alternatively, a more realistic approach might include a stratified screening technique which would entail the utilization of relatively simple screening techniques at first pass followed by more invasive and expensive technologically advanced procedures. For example, perhaps a brief cognitive measure could be combined with a positive family history, genetic susceptibility profile, and a screening blood test such as, a plasma Aβ ratio. Based on the results of the combination of these markers, subjects could then be stratified for more in-depth assessments using imaging techniques, CSF markers, and other more expensive technologies. Again, however, this approach would need to be validated, preferably in population-based samples to establish their utility as screening techniques for early therapeutic intervention.
To implement these approaches, we will need to shift our focus from small clinic-based assessment to large population-based approaches. In the interim, smaller population-based studies can inform us on the most useful measures to pursue on a larger scale. The current NIH-style research study based on five-year epoch in time makes it difficult to evaluate these various predictive markers. Newer mechanisms for supporting these larger efforts need to be considered.
Ultimately, of course, we are moving toward an approach to individualized medicine. We could envision a situation in which each person has his/her own fingerprint with respect to the likelihood of developing degenerative diseases. This propensity could then be modified with lifestyle interventions or possibly therapies that were disease modifying. We are moving in this direction for chronic conditions such as AD.
Leon Thal was a visionary in our field and championed clinical research designed to modify the diagnostic thresholds for the detection of AD. His work through the ADCS was oriented in this direction, and he had plans of moving the effort to larger audiences, and the field owes him a great debt of gratitude for his leadership toward that goal.