BIOMARKERS IN ALZHEIMER'S DISEASE

Introduction - the need for biomarkers in AD

In 1999 the US National institute of Health convened a workshop to try and resolve some of the perceived ambiguity regarding the purpose and status of biomarkers. As a consequence a set of definitions was derived that have subsequently become widely adopted (1). A biomarker, according to these definitions is " a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacological responses to a therapeutic intervention". A biomarker might be used for diagnosis, or staging of disease or to follow disease progression or monitor response to therapy. A clinical endpoint is "a characteristic or variable that reflects how the patient feels, functions or survives." A surrogate endpoint on the other hand, is a sub-set of biomarkers that are intended to substitute for clinical end-points. "A surrogate end-point is expected to predict clinical benefit (or harm or lack of benefit or harm) based on epidemiologic, therapeutic, pathophysiologic or other scientific evidence". Biomarkers have been gradually adopted as surrogate endpoints in clinical trials in recent years, for example, HIV RNA and CD4 count being accepted as surrogate markers for trials of HIV/AIDS interventions.

In the Alzheimer's disease (AD) field there has been considerable progress in the search for a biomarker with promising markers being developed using a wide variety of neuroimaging techniques and biochemical changes in CSF, blood and other tissues and fluids. However progress in translating these findings to surrogate endpoints is in its infancy with no evidence yet that regulatory authorities are prepared to accept putative AD biomarkers as surrogate endpoints. Establishing that a biomarker is a true surrogate endpoint is challenging both statistically and practically. To be used as a surrogate endpoint the biomarker should ideally both predict clinical endpoints and capture effects of treatment on these endpoints. In other fields, such as HIV/AIDS, where surrogate endpoints have been accepted for licensing purposes then large numbers of trials were available for meta-analysis where the proportion of the treatment effect explained by the biomarker (such as CD4 counts) could be calculated (2). There are alternative statistical approaches to the evaluation of surrogacy (3,4) but whatever methods are used the practical challenges remain considerable as sample sizes are likely to be, if anything, larger than those required for phase III studies (5).

However the need for surrogates in AD is pressing. In AD, in contrast to many other diseases including those such as HIV/AIDS where surrogates are used, the clinical endpoints are themselves difficult to assess and only poorly translated to quantifiable measures. Thus the gold standard for clinical endpoints in AD trials, scales of cognition, function and behaviour, are unable to distinguish between interventions of symptomatic relief and interventions of disease modifying effect without elaborate and as yet untested trial strategies such as randomised start and withdrawal. Finally by the time AD is clinically apparent disease is already well-established, limiting the possible effectiveness of any therapeutic intervention designed to slow progression by reducing pathological processes. And yet in the early phases of AD, in Mild Cognitive Impairment (MCI) and in the pre-MCI stage the clinical endpoint slope is either very shallow or immeasurable (figure 1) making clinical trials in these early stages very difficult. In the absence of a surrogate endpoint the only option for such trials is to increase the time frame of the trial to a point when change in the clinical endpoints becomes possible. Consequently very large trials are needed because of the likelihood of including people without the disease and because of high drop out rates.

In summary then biomarkers for AD are a pressing and urgent concern and are required for many purposes. Both diagnostic biomarkers and surrogate biomarkers could be of potential utility in clinical trials. Biomarkers for diagnosis, both differential diagnosis and early diagnosis would be useful in refining the inclusion and exclusion criteria. A subset of diagnostic biomarkers include those that predict conversion from MCI (or indeed control status) to AD or predict more rapid clinical progression. A biomarker might be useful in postlicensing effectiveness evaluation and prescribing decisions. For example the recent UK NICE draft evaluation of the cholinesterase inhibitors suggests that although efficacious, the effectiveness (and cost-benefit effectiveness in particular) might be insufficient to warrant recommending prescribing these class of drugs in the National Health Service. A biomarker that identifies those likely to benefit might, in these situations, be of considerable utility. However the 'holy grail' of biomarkers in relation to AD is to identify a surrogate that might replace a clinical end-point and thus facilitate the search for disease modification therapies in AD.

Criteria for biomarkers in AD

A NIA consensus group defined the core qualities desirable in a biomarker as "detect[ing] a fundamental feature of neuropathology ... validated in neuropathologically-confirmed cases; ... a sensitivity >80% for detecting AD and a specificity of >80% for distinguishing other dementias.... reliable, reproducible, non-invasive, simple to perform, and inexpensive" (6). This council of perfection is unlikely to be attainable in all respects. Moreover the desirable qualities of a biomarker for diagnosis differ from those for a biomarker intended as a surrogate endpoint. A further NIA consensus conference noted that it would be useful if a biomarker captured the beneficial effects of a therapeutic intervention (7). From a regulatory perspective this is the primary and essential quality of a surrogate biomarker.

The advances being made in biomarker discovery using biochemical analyses of accessible fluids is summarised below. A series of authoritative reviews and consensus documents have evaluated the current data regarding biomarkers in AD and these will be referred to in preference to the primary data wherever possible.

Biochemical and genetic biomarkers

As the pathogenetic basis of AD and related dementias becomes increasingly understood then potential biomarkers suggest themselves. Various proteins and genes known to be associated with AD risk and pathology have been examined as putative biomarkers. Some of these markers are trait and not state markers and thus might be biomarkers of risk rather than occurrence of AD. Thus APOE status clearly increases risk of AD and increases the likelihood of MCI conversion to AD. Indeed APOE status may even predict altered brain function decades before the onset of AD. Trait markers such as APOE might therefore have some utility in clinical trials enriching for a population at high risk of AD. Trait markers can also increase the diagnostic specificity but at the expense of sensitivity. No other genetic marker has the reproducibility of APOE but it is estimated that there are likely to be 5-7 other genetic variants associated with late onset AD and a number of genomic linkage regions have been identified and partially replicated. The AlzGene website is probably the best source of current information on genetic variants associated with AD. It is possible that as these susceptibility variants become identified then singly or in combination they might provide a biomarker of risk.

Proteins can be trait markers also. Thus the reproducible finding of homocysteine levels associated with AD is almost certainly a trait marker of environmental influences or genetic variability not directly linked to AD, that increases risk of suffering from dementia. Measuring levels of homocysteine is therefore unlikely to be a biomarker of state and therefore unlikely to be a diagnostic or surrogate marker. Nonetheless as for genetic biomarkers, protein state markers may have a role in enriching for populations at risk of AD or for conversion from a prodromal state.

Genes and proteins have been sought in relation to AD both by the candidate approach and by systematic whole-genome or proteome based methods. There are advantages and drawbacks to both approaches. Linkage studies are somewhat more frequent than their equivalent proteomic studies largely as the methodology for proteomic data is less advanced. Proteomic approaches to biomarker discovery include surface-enhanced laser desorption/ionization-time of flight-mass spectrometry (SELDI), a method particularly useful for the analysis of small molecular weight proteins with the advantage of high throughput analysis but which has the disadvantage of identifying patterns of change but not the proteins responsible for the change. Alternatives include separation of proteins by electrophoresis followed by mass spectrometry analysis a method that is not high throughput and is best for relatively large proteins but which has the advantage of actually identifying the proteins that are responsible for pattern differences between sample groups allowing further development of high throughput sensitive assays such as ELISA. Other methods are available and being developed including mass-tagging of samples and direct quantitative mass spectrometric analysis (see for example reference 8).

One of the possible limitations to state biomarker discovery by proteomic methods is intra-individual variability in proteins. It is reassuring then that for CSF inter-individual variability is greater than intra-individual variability (9).

CSF markers of neuropathology

A very considerable body of evidence now attests to an increase in tau protein and a reduction in amyloid in AD (10). Tau is, in contrast to AB, an intracellular protein and unlikely to be secreted by neurons. Thus any tau in CSF is most likely to be a measure of neuronal lysis (which may be a non-specific marker of neurodegeneration) or possibly release from extracellular neurofibrillary tangles which may be related more to factors, currently unknown, governing turnover of such lesions.

The large number of studies on total-tau have been comprehensively reviewed by Blenow and Hampel who identified more than 35 studies that had, in 2003, reported CSF levels in AD patients compared to controls (11). Almost all of these used an ELISA produced by Innogenetics and more than half showed a sensitivity of more than 80% versus controls. Some evidence suggests discriminant properties for AD vs other dementias (12). However some studies show tau increases with non-AD dementias (13) and in non-dementia neuronal damage such as that incurred during stroke (14). One study however indicate that CSF tau in FTD may be lower than in AD (15).

A formal meta-analysis confirms AD vs control differences (16) in total tau levels but these levels appear to be stable or at least not to show a predictable relationship with deterioration in AD (17). Various phospho-epitopes of tau have been examined in relation to AD (reviewed in (11)) and of these Threonine 231 appears to be most consistently associated with AD and may show the most promise a marker correlating with progression or severity (18). Phospho-tau may also be a better discriminant between different dementias than total tau (13,19-21). Both total and phospho-tau are elevated in very mild AD and in MCI but do not reliably predict progression from MCI to AD alone (11,22).

In contrast to tau the levels of Ab decline in AD although this decline may be non-linear. A review of studies to 2003 found more than half had a sensitivity of 90% or more for AD vs controls 11,23, specificity ranging from 42 to 88% (7). Aß levels might be a better marker for monitoring progression (24).

Reference values for normal tau and Ab levels with ageing have been established (25) and the commercially available ELISAs for both protein are highly reproducible although, importantly for trials, sample processing does affect assay results (26). Other methods including multi-modal assays are being developed (27).

In summary there is overwhelming evidence that total-tau, some phospho-tau assays and Aß discriminate subjects with AD from controls with acceptable levels of sensitivity and specificity (>80%). The assays are reliable, commercially available and the samples are stable and relatively resistant to experimental error (but note some caution re freeze-thaw cycles and Aß). Even so development work is seeking to improve these assays. The evidence however that these CSF markers of pathology might be surrogate end-points is not so robust with the most promise being for Thr231-tau. In the context of clinical trials CSF measures of tau or AΒ might be most useful in identifying people with very early AD and in enriching for likely converters in the MCI group. It remains to be seen whether CSF tau or Aß might be surrogate markers that respond to therapeutic intervention.

Other CSF biomarkers

A large number of other proteins have been examined in CSF including in the last year alone markers of glycation, oxidative stress, inflammation and phosholipid metabolism (24,28-33). These and other markers previously reported all show promise as do non-candidate protein marker studies using various proteomics based approaches. However a considerable amount of work is required to progress any of these putative markers to the point where they may be useful as surrogates in disease modification trials.

Plasma markers of candidate proteins in AD

Lumbar puncture is relatively easy to perform and has a very low incidence of mild adverse effects even in the frail elderly. However it remains a relatively invasive procedure. An ideal biomarker, especially one designed for clinical trials would be based on a sampling technique that had no adverse effects and was more in line with routine assessments. Accessible fluids for such a biomarker include blood, urine, tears, saliva and even hair samples. There is no biomarker yet available for dementia but considerable efforts are being expended in the search for a biomarker based on these fluids that could be used in clinical rials as a surrogate end point or as a diagnostic marker. It remains to be seen whether the blood or a similar peripheral fluid will reflect changes occurring in the closed system of brain and CSF. However some evidence does suggest either that the blood-brain barrier is not as impermeable as was once thought and/or that some of the changes of AD may be systematic metabolic changes that could be examined peripherally. In line with this two proteomic based approaches in small samples of AD cases were successfully identified relative to controls (34,35).

Recent studies have suggested isoprostanes, markers of oxidative potential, cholesterol and its associated metabolites, inflammation and others as potential blood based biomarkers in AD (36-53). However, although promising, none of these yet reach the levels of replicability required for a diagnostic or a surrogate marker. There is some evidence of altered Aß or APP metabolites in plasma or serum in AD. However as plasma Aß is at least partially inherited 54 this may represent a trait rather than a state marker and technical considerations make assays of Aß difficult.

Summary and suggestions for the development of surrogate markers in AD for the use in disease modifying trials

In conclusion, of the various biochemical measures examined, consensus reviews and more recent literature shows that CSF measures of tau and Aß either separately or together are markers of neurodegeneration and probably early markers of disease. Whether they are specific markers of AD or not remains to be fully determined. As the various dementias show considerable pathological overlap this is to some extent a tautological position - if the markers are true biomarkers they will only be as specific as the disorders are pathologically distinct. Neither CSF measures of tau or Aß appear to be useful as surrogate endpoints as they do not appear to show convincing, if any, change with progression although there may be some role for phospho-tau as a surrogate endpoint. Peripheral fluids have been shown to show biochemical changes in AD relative to controls both using proteomic and candidate protein and other metabolite changes. This raises the promise of a low-intervention, low-cost surrogate marker, which would be utilisable in clinical trials.

The need for such a biomarker was recently recognised by the European Federation of Pharmaceutical Agencies (EFPIA) representing large pharma and SMEs who identified the lack of biomarkers and surrogate markers of disease as a specific barrier to new drug development (1). Although this is apparent in many areas of medicine it is especially true for AD and other disorders of the brain which is largely inaccessible to direct assessment. In their submission to the EU 6th Framework Programme for Research thematic priority area "Life Sciences, Genomics and Biotechnologies for Health", EFPIA decided to use Alzheimer's Disease as a model system to search for such disease markers, both neuroimaging based and molecular, in a submission (InnoMed) led by the UK. Similarly in the US a major initiative is underway to identify biomarkers, surrogate end points in particular, using a variety of approaches including peripheral fluid based biochemistry (http://www. alzheimers.org/nianews/nianews70.html).

In conclusion

* A trait marker or a biomarker of prediction may have utility in clinical trials especially to enrich for populations at risk of conversion (from MCI or from unaffected)

* A diagnostic biomarker may have utility to increase homogeneity of subject populations

* A surrogate endpoint is the most useful biomarker in the clinical trial context but achieving the status of a surrogate acceptable to regulatory authorities is exceedingly difficult and there are only few examples in medicine.

* Diagnostic markers are not necessarily surrogate markers and identifying markers of progression is at least as important (if not more so) than identifying markers that distinguish between AD, other dementias and controls.

* A marker of progression or a surrogate marker may show very poor sensitivity and specificity and vice versa a highly sensitive marker of AD may have no value as a surrogate.

* It is likely that for a marker to achieve the status of a surrogate, especially one recognised by regulatory authorities, then very large studies will be needed. These may only be achieved by meta-analysis. Guidelines foir the use of metaanalyses for diagnostic markers have been published (55).

* Incorporating biomarker discovery and validation research into longitudinal studies and clinical trials will increase the availability of data and samples necessary for establishing a surrogate marker

* Attention should be paid to standardising sample collection and curation. This is essential within studies but there may be some scope for standardisation across studies.

* Attention should be paid to multifactorial analysis of putative biomarkers - useful surrogates may emerge from more than one biomarker used in combination

* Both candidate protein and metabolite and non a priori (protemomic, lipidomics, metabolomic and transcriptomic) based approaches should be employed to maximise the chances of discovery of a useful surrogate biomarker.

* The AlzGene database might be usefully replicated for biomarkers as the publically available data increases

* Validation and replication of data and publishing of data according to widely accepted guidelines will become increasingly important as the field matures (56,57) as has occurred in other disorders (58).

References

References

1. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin.Pharmacol.Ther. 69,89-95 (2001).

2. Petrella, M., Montaner, J., Batist, G., & Wainberg, M. A. The role of surrogate markers in the clinical development of antiretroviral therapy: a model for early evaluation of targeted cancer drugs. Cancer Invest 22, 149-160 (2004).

3. Buyse, M., Molenberghs, G., Burzykowski, T., Renard, D., & Geys, H. The validation of surrogate endpoints in meta-analyses of randomized experiments. Bioslatistics. 1, 49-67 (2000).

4. Molenberghs, G. et al. Statistical challenges in the evaluation of surrogate endpoints in randomized trials. Control Clin.Trials 23,607-625 (2002).

5. Fleming, T. R. & DeMets, D. L. Surrogate end points in clinical trials: are we being misled? Ann.Intem.Med. 125,605-613 (1996).

6. Alzheimers Assoc & Natl Inst Aging. Consensus report of the Working Group on: "Molecular and Biochemical Markers of Alzheimer's Disease". Neurobiol.Aging 19, 109-116 (1998).

7. Frank, R. A. et al. Biological markers for therapeutic trials in Alzheimer's disease. Proceedings of the biological markers working group; NIA initiative on neuroimaging in Alzheimer's disease. Neurobiol.Aging 24, 521-536 (2003).

8. Zhang, J. et al. Quantitative proteomics of cerebrospinal fluid from patients with Alzheimer disease. J Alzheimers.Dis. 7, 125-133 (2005).

9. Hu, Y., Malone, J. P., Pagan, A. M., Townsend, R. R., & Holtzman, D. M. Comparative proteomics analysis of intra- and inter-individual variation in human cerebrospinal fluid. Mol.Cell Proteomics. (2005).

10. Wiltfang, J. et al. Consensus paper of the WFSBP Task Force on Biological Markers of Dementia: the role of CSF and blood analysis in the early and differential diagnosis of dementia. World J.Biol.Psychiatry 6, 69-84 (2005).

11. Blennow, K. & Hampel, H. CSF markers for incipient Alzheimer's disease. Lancet Neurol. 2, 605-613 (2003).

12. Parnetti, L. et al. CSF phosphorylated tau is a possible marker for discriminating Alzheimer's disease from dementia with Lewy bodies. Phospho-Tau International Study Group. Neurol.Sci. 22, 77-78 (2001).

13. Nagga, K., Gottfries, J., Blennow, K., & Marcusson, J. Cerebrospinal fluid phosphotau, total tau and beta-amyloid(l-42) in the differentiation between Alzheimer's disease and vascular dementia. Dement.Geriatr.Cogn Disord. 14, 183-190 (2002).

14. Hesse, C. et al. Transient increase in total tau but not phospho-tau in human cerebrospinal fluid after acute stroke. Neurosci.Lett. 297, 187-190 (2001).

15. Grossman, M. et al. Cerebrospinal fluid profile in frontotemporal dementia and Alzheimer's disease. Ann.Neurol 57, 721-729 (2005).

16. Sunderland, T. et al. Decreased beta-amyloid 1-42 and increased tau levels in cerebrospinal fluid of patients with Alzheimer disease. JAMA 289, 2094-2103 (2003).

17. Sunderland, T. et al. Longitudinal stability of CSF tau levels in Alzheimer patients. BioLPsychiatry 46, 750-755 (1999).

18. Hampel, H. et al. Tracking of Alzheimer's disease progression with cerebrospinal fluid tau protein phosphorylated at threonine 231. Ann.Neurol. 49,545-546 (2001).

19. Riemenschneider, M. et al. Phospho-tau/total tau ratio in cerebrospinal fluid discriminates Creutzfeldt-Jakob disease from other dementias. MoI.Psychiatry 8, 343-347 (2003).

20. Hampel, H. et al. Measurement of phosphorylated tau epitopes in the differential diagnosis of Alzheimer disease: a comparative cerebrospinal fluid study. Arch.Gen.Psychiatry 61, 95-102 (2004).

21. Buerger, K. et al. Differential diagnosis of Alzheimer disease with cerebrospinal fluid levels of tau protein phosphorylated at threonine 231. Arch.Neurol. 59, 1267-1272 (2002).

22. Blennow, K. CSF biomarkers for mild cognitive impairment. J.Intern.Med. 256, 224-234 (2004).

23. Hampel, H. et al. Core biological marker candidates of Alzheimer's disease perspectives for diagnosis, prediction of outcome and reflection of biological activity. J.Neural Transm. 111, 247-272 (2004).

24. Ganzer, S. et al. CSF-tau, CSF-Abeta 1 -42, ApoE-genotype and clinical parameters in the diagnosis of Alzheimer's disease: combination of CSF-tau and MMSE yields highest sensitivity and specificity. J.Neural Transm. 110, 1149-1160 (2003).

25. Sjogren, M. et al. Tau and Abeta42 in cerebrospinal fluid from healthy adults 21-93 years of age: establishment of reference values. Clin.Chem. 47, 1776-1781 (2001).

26. Schoonenboom, N. S. et al. Effects of processing and storage conditions on amyloid beta (1-42) and tau concentrations in cerebrospinal fluid: implications for use in clinical practice. Clin.Chem. 51,189-195 (2005).

27. Olsson, A. et al. Simultaneous measurement of beta-amyloid(l-42), total tau, and phosphorylated tau (Thr181) in cerebrospinal fluid by the xMAP technology. Clin.Chem. 51, 336-345 (2005).

28. Ahmed, N. et al. Protein glycation, oxidation and nitration adduct residues and free adducts of cerebrospinal fluid in Alzheimer's disease and link to cognitive impairment. J.Neurochem. 92,255-263 (2005).

29. Cepek, L. et al. Follow-up investigations of tau protein and S-100B levels in cerebrospinal fluid of patients with Creutzfeldt-Jakob disease. Dement.Geriatr.Cogn Disord. 19, 376-382 (2005).

30. Finehout, E. J., Franck, Z., & Lee, K. H. Complement protein isoforms in CSF as possible biomarkers for neurodegenerative disease. Dis.Markers 21,93-101 (2005).

31. Lindberg, C. et al. Soluble interleukin-1 receptor type II, IL-18 and caspase-1 in mild cognitive impairment and severe Alzheimer's disease. Neurochem.ini. 46, 551-557 (2005).

32. Vuletic, S. et al. Reduced CSF PLTP activity in Alzheimer's disease and other neurologic diseases; PLTP induces ApoE secretion in primary human astrocytes in vitro. J.Neurosci.Res. 80, 406-413 (2005).

33. Yamagishi, S., Nakamura, K., Inoue, H., Kikuchi, S., & Takeuchi, M. Serum or cerebrospinal fluid levels of glyceraldehyde-derived advanced glycation end products (AGEs) may be a promising biomarker for early detection of Alzheimer's disease. Med.Hypotheses 64, 1205-1207 (2005).

34. Lopez, M. F. et al. High-resolution serum proteomic profiling of Alzheimer disease samples reveals disease-specific, carrier-protein-bound mass signatures. Clin.Chem. 51, 1946-1954 (2005).

35. Zhang, R. et al. Mining biomarkers in human sera using proteomic tools. Proteomics. 4, 244-256 (2004).

36. Pulido, R., Jimenez-Escrig, A., Orensanz, L., Saura-Calixto, F., & Jimenez-Escrig, A. Study of plasma antioxidant status in Alzheimer's disease. Eur J Neurol 12, 531-535 (2005).

37. Straface, E. et al. Oxidative imbalance and cathepsin D changes as peripheral blood biomarkers of Alzheimer disease: a pilot study. FEES Lett. 579, 2759-2766 (2005).

38. Yamagishi, S., Nakamura, K., Inoue, H., Kikuchi, S., & Takeuchi, M. Serum or cerebrospinal fluid levels of glyceraldehyde-derived advanced glycation end products (AGEs) may be a promising biomarker for early detection of Alzheimer's disease. Med.Hypotheses 64, 1205-1207 (2005).

39. James, V. J. et al. Fibre diffraction of hair can provide a screening test for Alzheimer's disease: a human and animal model study. Med.Sci.Monit. 11, CR53-CR57 (2005).

40. Montine, T. J. et al. F2-isoprostanes in Alzheimer and other neurodegenerative diseases. Antioxid.Redox.Signal. 7, 269-275 (2005).

41. Mocali, A. et al. Increased plasma levels of soluble CD40, together with the decrease of TGF beta 1, as possible differential markers of Alzheimer disease. Exp.Gerontol. 39, 1555-1561 (2004).

42. Steinacker, P. et al. Heart fatty acid binding protein as a potential diagnostic marker for neurodegenerative diseases. Neurosci.Lett. 370, 36-39 (2004).

43. Kadioglu, E., Sardas, S., Asian, S., Isik, E., & Esat, K. A. Detection of Oxidative DNA damage in lymphocytes of patients with Alzheimer's disease. Biomarkers 9, 203-209 (2004).

44. Polidori, M. C. et al. Plasma antioxidant status, immunoglobulin g oxidation and lipid peroxidation in demented patients: relevance to Alzheimer disease and vascular dementia. Dement.Geriatr Cogn Disord. 18, 265-270 (2004).

45. Yamagishi, S., Inagaki, Y., Takeuchi, M., & Sasaki, N. Is pigment epithelium-derived factor level in cerebrospinal fluid a promising biomarker for early diagnosis of Alzheimer's disease? Med.Hypotheses 63, 115-117 (2004).

46. Emanuele, E. et al. Plasma osteoprotegerin as a biochemical marker for vascular dementia and Alzheimer's disease. Int J Mol.Med. 13, 849-853 (2004).

47. Montine, K. S. et al. Isoprostanes and related products of lipid peroxidation in neurodegenerative diseases. Chem.Phys.Lipids 128, 117-124 (2004).

48. Sun, Y. X. et al. Inflammatory markers in matched plasma and cerebrospinal fluid from patients with Alzheimer's disease. Dement.Geriatr Cogn Disord. 16, 136-144 (2003).

49. Tully, A. M. et al. Low serum cholesteryl ester-docosahexaenoic acid levels in Alzheimer's disease: a case-control study. Br.J Nutr. 89, 483-489 (2003).

50. DeKosky, S. T. et al. Plasma and cerebrospinal fluid alpha 1-antichymotrypsin levels in Alzheimer's disease: correlation with cognitive impairment. Ann.Neurol 53, 81-90 (2003).

51. Choi, J., Malakowsky, C. A., Talent, J. M., Conrad, C. C., & Gracy, R. W. Identification of oxidized plasma proteins in Alzheimer's disease. Biochem.Biophys.Res.Commun. 293, 1566-1570 (2002).

52. Pratico, D. et al. Increased 8,12-iso-iPF2alpha-VI in Alzheimer's disease: correlation of a noninvasive index of lipid peroxidation with disease severity. Ann.Neurol 48, 809-812 (2000).

53. Schipper, H. M. et al. Evaluation of heme oxygenase-1 as a systemic biological marker of sporadic AD. Neurology 54, 1297-1304 (2000).

54. Ertekin-Taner, N. et al. Heritability of plasma amyloid beta in typical late-onset Alzheimer's disease pedigrees. Genet.Epidemiol. 21, 19-30 (2001).

55. Irwig, L. et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann.Intern.Med. 120, 667-676 (1994).

56. Bossuyt, P. M. et al. Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD initiative. Standards for Reporting of Diagnostic Accuracy. Clin.Chem. 49, 1-6 (2003).

57. Greenhalgh, T. How to read a paper. Papers that report diagnostic or screening tests. BMJ 315, 540-543 (1997).

58. McShane, L. M. et al. REporting recommendations for tumor MARKer prognostic studies (REMARK). Nat.Clin.Pract.Oncol. 2, 416-422 (2005).