Nearly 100 years have passed since Alois Alzheimer first described his index patient Auguste D and the pathologic changes found in her brain in the postmortem examination. Since then, intensive research has allowed us to unravel some of the key mechanisms of AD and to identify a number of promising potential biomarkers for its diagnosis and prognosis. Although the search for new and better biomarkers needs to be continued, the full potential of the existing biomarkers needs to be explored in large prediction trials. However, for such trials to produce meaningful and generally applicable results, a further standardization, not only of the criteria for diagnosis but also of the measurement techniques of surrogate biomarkers, will be necessary. Most studies already use standardized criteria for the diagnosis of AD, eg, the NINCDS-ADRDA criteria. In contrast, the definitions of early stages or presymptomatic stages of AD or even healthy control subjects are less rigorous. The terms mild cognitive impairment
or cognitively impaired nondemented
, for example, have been used with varying definitions, and efforts to define standardized criteria have only recently been made [16
]. The same is true for the definition of healthy elderly controls
, ie, the term needs to be defined—if normal means healthy in every aspect or only absence of cognitive impairment or neurologic disease but otherwise allowing for the entire range of diseases frequently found in the elderly population. In addition to standardization of diagnostic criteria, it is also necessary to identify and then strictly apply optimized measurement parameters of the already-existing biochemical and neuroim-aging biomarkers. The ADNI will be a first important step toward this goal.
An important issue for trials searching for the most powerful combination of diagnostic and prognostic biomarkers is also the identification and, if possible, correction of factors that increase measurement variability. Generally, there are 2 sources of measurement variability: variability owing to limitations of the measurement technique and variability owing to subject-related factors. The former can be best addressed by the development of optimized and standardized measurement protocols. Subject-related measurement variability is more difficult to account for because it is usually multifactorial. Factors contributing to subject-related measurement variability can, for example, be genetic, eg, Apo-ε4 carrier status or environmental, eg, exposure to putative neurotoxins or dietary deficiencies, concomitant diseases and their treatment, eg, cerebrovascular disease or diabetes mellitus. In addition to these common sources of subject-related measurement variability, there are 2 sources that are also of potentially diagnostic and prognostic importance. The first is effects of AD treatments. Although there is currently no definite proof that any of these drugs could prevent AD or at least significantly slow down its progression, more and more patients with possible AD or even amnestic MCI are treated with cholinesterase inhibitors, or high doses of α-tocopherol, and even healthy subjects take antioxidants or vitamin preparations as protection against AD. It might even be possible that some of the drugs originally developed for the treatment of AD enhance cognitive performance in healthy subjects and are thus taken for this purpose only by otherwise completely healthy people. Therefore, it will become increasingly difficult in the future to observe the natural course of AD, and it is possible that such treatments influence the expression of some biomarkers and render them useless for diagnostic or prognostic purposes. Finally, the probably most important source of measurement variability is the variability caused by the disease itself. The fact that despite the intensive research efforts the etiology of AD is still elusive suggests that the disease processes in AD are either very complex or AD is just the common expression of several different etiologic processes. If the latter is true, it is very well possible that the different forms also vary in the expression of AD biomarkers.
Ultimately, the goal is diagnosis and treatment in the earliest possible stage of AD. For this purpose, it is also necessary to differentiate early AD from early manifestations of other dementias. Therefore, it will be necessary to test those biomarkers found to be useful for diagnosis and prognosis in AD and MCI patients also in other risk groups, eg, cognitively nonimpaired patients or young Apo-ε4 carriers, asymptomatic subjects with strong family history for AD, subjects with cognitive impairment in non-memory domains, or subjects suffering from other neurodegenerative diseases, eg, vascular dementia or Lewy body disease. However, although expanding the study population is essential to explore the full potential of a biomarker, ethical questions are also raised, particularly in subjects that are included because they already belong to a group with a higher risk for AD, eg, Apo-ε4 carriers. Some of these subjects will be found to be positive for the biomarker in question, and thus their risk for AD or the risk that they may already be in a very early stage of AD might be even higher than previously assumed. This raises the question if these subjects should be informed about their increased risk and if they should even be offered treatment with one of the currently available AD drugs or treatment in a study protocol. A similar problem occurs in subjects recruited as healthy controls who are positive for a single biomarker or a combination of different biomarkers that have been found to be highly predictive for the development of AD.
Although the currently available surrogate biomarkers need to be further explored, the search for new biomarkers and for less-invasive, safer, and, for the patient, more acceptable alternative measurement techniques needs to be continued. PET, for example, seems to be one of the most sensitive imaging markers for AD in the very early or even presymptomatic stage, particularly if the new tracers for amyloid imaging are as reliable as promised. Efforts to develop MR techniques allowing for the measurement of brain amyloid concentrations are under way, but none of them have progressed far enough to be tested in patients [114
]. Functional MRI techniques such as perfusion MRI o r BOLD fMRI may also be useful. The diagnostic and prognostic potential of other already-existing imaging modalities, eg, DTI, MRS o r PET with acetylcholine analogues needs to be explored further particularly regarding their usefulness to detect very early disease manifestations or contraindications for some forms of treatment, eg, risk of bleeding after amyloid vaccination. Finally, the knowledge about the earliest molecular pathomechanisms gained from studies in animal models of AD needs to be translated in new biochemical and neuroimaging biomarkers to allow the detection of the disease in its presymptomatic stage.