Abstract
In the earliest clinical stages of Alzheimer's Disease (AD), when symptoms are mild, clinical diagnosis will still be difficult. AD related molecular mechanisms precede symptoms. Biological markers can serve as early diagnostic indicators, as markers of preclinical pathological change, e.g. underlying mechanisms of action (MoA). Hypothesis based candidates are derived from structural and functional neuroimaging as well as from cerebrospinal fluid (CSF) and plasma. Unbiased exploratory approaches e.g. proteome analysis or rater independent fully automated imaging post-processing methods yield novel candidates. Recent progress in the validation of core feasible imaging and neurochemical biomarkers for functions such as early detection, classification, progression and prediction of AD is summarized. Single core feasible biomarkers can already be used to enrich populations at risk for AD and may be further enhanced using distinct combinations. Some biomarkers are currently in the process of implementation as primary or secondary outcome variables into regulatory guideline documents, e.g. regarding phase II in drug development programs as outcome measures in proof of concept or dose finding studies. There are specific biomarkers available depending on the hypothesized mechanism of action of a medicinal product, e.g. impact on the amyloidogenic cascade or on tauhyperphosphorylation. Ongoing large-scale international controlled multi-center trials will provide further validation of selected core feasible imaging and CSF biomarker candidates as outcome measures in early AD for use in phase III clinical efficacy trials. There is a need of rigorous co-development of biological trait- and state-marker candidates facilitated through planned synergistic collaboration between academic, industrial and regulatory partners.
Key words: Alzheimers's disease, AD, MCI, preclinical, presymptomatic, enrichment, biological markers, biomarkers, CSF, blood, neuroimaging, early detection, treatment, therapy, clinical trial, validation, qualification, guidelines, regulatory authorities
Introduction
Alzheimer’s disease (AD) is one of the most common illnesses of later life. To date, the only treatments available are symptomatic, and they are not administered until advanced stages of the disease process. At present, AD cannot be diagnosed until the dementia syndrome has been clinically established and the progressing cognitive deficits affect the patient’s ability to cope with functional demands of his or her social and professional life. This clinical “threshold value” varies greatly between individuals; it depends among other things on the patient’s premorbid cognitive and intellectual level and is to some extent arbitrary. There is hence considerable interest in the reliable early detection of AD, possibly in presymptomatic/predementia stages. Otherwise, preventive and disease modifying therapies cannot be administered. The concept of mild cognitive impairment (MCI) has been questioned by regulatory bodies (1), however, due to high diagnostic sensitivity but low specificity, new operationalized criteria as proposed by Dubois and colleagues may be better suitable to define homogeneous patient populations with underlying progressive AD pathology, even at earlier pathological stages of the disease. These recently proposed research criteria are focussing on a symptomatic (not syndromatic) clinical core of significant episodic memory impairment, in addition to at least one or combined abnormal biological marker modalities derived from MRI (structural), PET (functional and metabolic), and cerebrospinal fluid (CSF) analysis of amyloid beta or tau related proteins. To be implemented in clinical treatment trials to enrich target patient populations at risk or as secondary or even primary endpoints, however, such biomarkers should reflect a neuropathological characteristic of AD and should be sufficiently qualified and validated in patients with a neuropathological confirmation. Once the aim of optimized characterization of homogenous target patient populations in early AD can be achieved this will clearly improve the efficiency of a clinical trial to detect any therapeutic effects of a new medicinal product.
Core biomarkers for enrichment of clinical trial populations
A number of in vivo neurochemistry and neuroimaging techniques, which can reliably assess aspects of physiology, pathology, chemistry, and neuroanatomy, hold promise as biomarkers in AD. The sensitivity of the “ideal” biomarker to detect AD should be at least 85%. Its specificity to differentiate AD patients from controls of the same age and from patients with other forms of dementia should be at least 75%. In clinically diagnosed populations, a higher level of specificity for biomarkers will not be able to be achieved for methodological reasons, as even the gold standard, the clinical diagnostic criteria, cannot be absolutely specific. The same applies to controls of the same age, as some of them might have undetected incipient preclinical AD. In large groups, this will inevitably affect the specificity of the results of even the best mechanistic biomarker. Despite the increasing evidence of promising results, biomarkers of AD, and particularly the early stages, are at various stages of development and have so far not generally been established in clinical routine.
Cerebrospinal fluid (CSF) neurochemical markers
Many studies have shown that CSF total tau protein (t-tau) is increased, A 42 beeing decreased in patients with AD relative to age-matched controls. Whereas sensitivity is higher than 80 %, however, specificity, which is more important for selection of homogeneous patient populations (enrichment), has not been found sufficiently high (2, 3). However, A 42 and t-tau have demonstrated predictive properties in studies on MCI populations (4).
The increase of hyperphosphorylated tau (p-tau) has shown to be more specific, as documented by a great number of studies investigating tau phosphorylated at threonine 231 (p-tau231) and at threonine 181 (p-tau181). P-tau231 as well as p-tau181 showed promising results in distinguishing AD from both control groups and even from comparisons such as other primary types of dementia. Studies indicate that P-tau231 is in fact distinguishing AD from e.g. frontotemporal dementia (FTD) with “ideal” sensitivity/specificity rates of 85–90 % (5).
In MCI subjects, high p-tau231 concentrations correlated with cognitive decline and conversion to AD (2, 3). High p-tau231 concentrations at baseline examination also correlated with structural disease progression (rate of hippocampal atrophy). A recent European multicenter trial using CSF p-tau231 in MCI subjects has shown that the performance of p-tau231 in predicting AD in this at-risk group is indeed robust, reliable, stable and consistent throughout multiple centers, therefore useful for enrichment in trials. P-tau231 proved to be a powerful candidate predictor of AD in predementia MCI subjects even in a very short mean observation interval of only one to two years (6).
A particularly promising new approach in the CSF focuses on the detection and quantification of -site-cleaving enzyme 1 (BACE-1, -secretase), one of the key enzymes responsible for the cascade initiating pathological amyloidogenic cleavage of the amyloid precursor protein (APP). A significant increase was found in both BACE-1 concentration and activity in the CSF of MCI subjects who ultimately cognitively declined and converted to AD compared to MCI non-converters or healthy subjects. Highest increases were found in subjects carrying the APOE4 risk allel. CSF BACE1 is promising for enrichment of patient populations as well as in future clinical trials focussing on anti-amyloid strategies (7).
Isoprostanes are also being studied as candidate markers of lipid peroxidation. An increase was found in the CSF of MCI subjects compared with controls, and levels also increased over time. With regards to their diagnostic precision, the CSF markers isoprostanes and p-tau performed better than memory tests. Isoprostanes even improved the results obtained using hippocampal volumetry to distinguish between the groups (2, 3).
Neuroimaging
Magnetic Resonance Imaging (MRI)
Manual hippocampal volumetry is currently the best-established biomarker for AD in the field of structural imaging (9, 10), however, there is no harmonized rating protocol and the method is laborious and time-consuming. Automated data-driven, hypothesis and rater-independent methods are currently being investigated to detect subtle and unbiased regional changes, namely voxel-based volumetry (VBM), deformation-based morphometry (DBM), and the measurement of cortical thickness (9, 10). In a recent study our group used a multivariate principal component analysis (PCA) approach to analyse DBM data in order to calculate individual risk for the presence of AD in MCI subjects. The application of this method allowed a group separation of about 80% between AD patients and healthy controls. Interestingly, the accuracy in distinguishing between MCI subjects who developed dementia over a period of 1½ years and MCI subjects whose cognitive performance remained stable over time was 70–80% (9).
In perspective, a combination of automated analysis modalities using multivariate statistical methods are likely to come forward which may prove as effective as manual hippocampal volumetry in the early detection of AD in subjects with MCI and most likely will therefore be successfully used in pharmacological studies (10). Currently there are also promising efforts to replace manual with semi- or fully automated hippocampus volumetry approaches.
Positron Emission Tomography (PET)
PET using 18fluorodeoxyglucose (18FDG) is used to study cortical metabolism. In AD patients, 18FDG-PET shows a typical pattern of reduced cortical uptake in the region of the temporal and parietal association cortex, particularly in the region of the posterior cingulum; MCI subjects already show – to a lesser extent – a similar distribution of metabolic deficits which can predict conversion from MCI to AD with an accuracy of over 80%. Therefore many researchers regard 18FDG-PET as gold standard in the in vivo diagnosis of early stages of AD (3, 9).
Novel markers have recently been developed to image amyloid plaques using PET in AD patients. The most extensively studied radiotracer is Pittsburgh Compound B (PIB), which shows a specifically enhanced uptake in AD patients compared with healthy controls. It is not clear at present, however, whether the diagnostic accuracy of this method may be better than that of the more matured FDG-PET. However, its application in treatment studies to investigate amyloid-modifying strategies as a marker of a biological mechanism would be conceivable (3, 9).
Discussion
Biological marker research in AD is most advanced for the function of diagnosis (2, 3, 9). Neurochemical (t-tau, p-tau, A . 42) and MRI-based structural core feasible markers (hippocampal, whole brain volumetry) are currently undergoing multicenter evaluation in controlled diagnostic phase IIb studies to determine sensitivity and specificity and to obtain initial assessment of their positive and negative predictive value. Specific combination seems to further enhance diagnostic accuracy for early clinical AD with diagnostic sensitivity/specificity reliably higher than 80% (2, 3, 9). A number of other candidate markers are currently undergoing earlier phase I or IIa diagnostic validation, particularly BACE-1 concentration and enzymatic activity (7), A monomers and oligomers, isoprostanes, MR-based analysis of cortical thickness, DBM, VBM, DTI, markers derived from MR-spectroscopy and amyloid-and tau binding ligands using PET, and potentially diagnostic paradigms using fMRI (e.g. functional connectivity) (8, 9, 10). Promising work on hypothesis driven or exploratory blood candidate markers has begun to seriously emerge only recently and has not yet reached phase IIa diagnostic development, so clearly the gold standard for early detection and trial enrichment remains in the CSF neurochemical core feasible candidates.
Currently, the short-term objective clearly is to have reliably validated biomarkers ready for use as early AD diagnostics, to support development of disease-modifying treatments (e.g. for enrichment, stratification, proof-of-principle, and markers of MoA) and to be broadly available for clinical screening and preselection of at risk subjects for further in depth diagnostic investigation so that those patients who would benefit most from the forthcoming disease modifying strategies can be accurately and reliably identified and treated in time. Depending on the proposed mechanism of action of a new medicinal product the choice of a suitable biomarker protocol (using markers of biological activity [MoA]) will sure vary.
Emerging new guideline documents of regulatory authorities, such as FDA and EMEA, will most likely strongly recommend thorough qualification/validation of biological marker candidates as outcome measures in upcoming phase-II and -III treatment trials of compounds claiming disease modifying properties (1). To this end, there is a need of thorough and rigorous co-development of biological trait- and state-marker candidates from suitable modalities (neurochemical & imaging) with various biomarker target functions and roles during all stages of drug development. This can only be achieved through planned synergistic collaboration between academic, industrial and regulatory partners. Fortunately, such collaborative consortia have already been initiated. To date, biomarker research in neurodegenerative disorders is an exceptionally dynamic, fascinating and fast developing area, however, much may still be learned from even more matured interdisciplinary fields, such as oncology, immunology, and cardiovascular research.
Enrichment and disease modification requires new biomarkers and diagnostics to identify pre-symptomatic patients for primary or secondary prevention, and then to assess the benefit of therapy. Quantification of drug benefit via biomarkers is useful during drug development to inform protocol design and business decision-making, and for regulatory authorization. Development of that biomarker into a diagnostic enables use in clinical practice for identification of eligible patients, to encourage compliance by the patient, or to direct changes in therapy. The potential for risk assessment via pharmacogenetics is limited to diseases with small numbers of polymorphisms or haplotypes, good penetrance, and large affected populations, and does not allow for the fuller applications of biomarkers to assessment of outcomes. Figure is adapted from a slide provided by Richard Frank
Figure 1.

Financial disclosure: None of the authors had any financial interest or support for this paper.
References
- 1.Broich K. Outcome measures in clinical trials on medicinal products for the treatment of dementia: a European regulatory perspective. Int Psychogeriatr. 2007;19(3):509–524. doi: 10.1017/S1041610207005273. 10.1017/S1041610207005273 PubMed PMID: 17433121. [DOI] [PubMed] [Google Scholar]
- 2.Blennow K., Hampel H. Cerebrospinal fluid markers for incipient Alzheimer's disease. Lancet Neurol. 2003;2(10):605–613. doi: 10.1016/s1474-4422(03)00530-1. 10.1016/S1474-4422(03)00530-1 PubMed PMID: 14505582. [DOI] [PubMed] [Google Scholar]
- 3.Hampel H., et al. Core candidate neurochemical and imaging biomarkers of Alzheimer's disease. Alzheimers Dement. 2008;4(1):38–48. doi: 10.1016/j.jalz.2007.08.006. 10.1016/j.jalz.2007.08.006 PubMed PMID: 18631949. [DOI] [PubMed] [Google Scholar]
- 4.Hampel H., et al. Value of CSF ß-Amyloid 1–42 and tau as predictors of Alzheimer's disease in patients with mild cognitive impairment. Mol Psychiatry. 2004;9:705–710. doi: 10.1038/sj.mp.4001473. PubMed PMID: 14699432. [DOI] [PubMed] [Google Scholar]
- 5.Hampel H., et al. Measurement of phosphorylated tau epitopes in the differential diagnosis of Alzheimer's disease — a comparative CSF study. Arch Gen Psychiatry. 2004;61(1):95–102. doi: 10.1001/archpsyc.61.1.95. 10.1001/archpsyc.61.1.95 PubMed PMID: 14706948. [DOI] [PubMed] [Google Scholar]
- 6.Ewers M., et al. Multicenter assessment of CSF-phosphorylated tau for the prediction of conversion of MCI. Neurology. 2007;69(24):2205–2212. doi: 10.1212/01.wnl.0000286944.22262.ff. 10.1212/01.wnl.0000286944.22262.ff PubMed PMID: 18071141. [DOI] [PubMed] [Google Scholar]
- 7.Hampel, H. and Shen, Y. Beta-site amyloid precursor protein cleaving enzyme 1 (BACE1) as a biological candidate marker of Alzheimer's disease. Scand J Clin Lab Invest, 2008: p. 1–5. [DOI] [PubMed]
- 8.Teipel S.J., et al. Multivariate deformation-based analysis of brain atrophy to predict Alzheimer's disease in mild cognitive impairment. Neuroimage. 2007;38(1):13–24. doi: 10.1016/j.neuroimage.2007.07.008. 10.1016/j.neuroimage.2007.07.008 PubMed PMID: 17827035. [DOI] [PubMed] [Google Scholar]
- 9.Teipel S.J., et al. Novel MRI techniques in the assessment of dementia. Eur J Nucl Med Mol Imaging. 2008;35(Suppl1):S58–S69. doi: 10.1007/s00259-007-0703-z. 10.1007/s00259-007-0703-z PubMed PMID: 18205002. [DOI] [PubMed] [Google Scholar]
- 10.Jessen, F. and Hampel, H. MRI as a surrogat marker in clinical trials in Alzheimer's disease. In: Imaging the Aging Brain (Eds.: Jagust, W., D'Esposito, M,), Oxford University Press, UK, in press.
