INTERNATIONAL JOURNAL OF GERIATRIC PSYCHIATRY

Int J Geriatr Psychiatry 2003; 18: 407–411.

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/gps.845

CSF phosphorylated tau—does it constitute an accuratebiological test for Alzheimer’s disease?

Alex Mitchell1* and Nick Brindle2

1Lecturer in Psychiatry, University of Leeds, Leeds, UK2Consultant in Old Age Psychiatry, Leeds, UK

SUMMARY

Introduction There is considerable interest in developing a diagnostic test which could differentiate between early Alz-heimer’s disease (AD) and other causes of memory impairment with more than 80% sensitivity and 80% specificity.Objective To review the studies that have examined CSF phosphorylated tau as diagnostic test of AD vs clinically repre-sentative comparison groups.Method A critical review of the literature using Embase, Web of Science, Medline and Psychinfo databases supplementedby handsearching and contact with experts in the field.Results CSF phosphorylated tau is a marker of AD that improves upon the utility of CSF total tau and clinical examinationalone. Studies have found high levels of tau phosphorylated at Threonine 231 and/or Serine 199 in AD but not in other causesof dementia, in depression or in healthy elderly controls. Of particular interest, the test appears equally valid in cases of earlyAD as in moderate or late stages and may also be of use in predicting future decline in subjects with mild cognitive impair-ment.Conclusion CSF phosphorylated tau is a promising diagnostic test for AD but this requires replication using pathologi-cally confirmed cases. Copyright # 2003 John Wiley & Sons, Ltd.

key words— Alzheimer’s disease; phosphorylated tau; cerebrospinal fluid; diagnostic test; mild cognitive impairment

With improvements in longevity, the importance ofdegenerative dementia increases. In 1900, 1% of theworld’s population lived to the age of 65 years, a hun-dred years later this percentage has risen to over 5%and by 2050 it is estimated reach 25%. As the preva-lence of dementia roughly doubles for every five yearsabove the age of sixty, the future healthcare implica-tions of increases in the degenerative dementias willassume paramount importance (Hebert et al., 2001).More than ever before, an early accurate diagnosisof the diseases causing dementia facilitates manage-ment decisions such as selection of symptomatictreatments. It may also inform families with respectto heritable risks and allow patients and families timeto plan for the future.

DIAGNOSIS OF ALZHEIMER’S DISEASE

The commonest cause of dementia in the elderly isAlzheimer’s disease (AD). Traditionally, the diagno-sis of AD relies on a thorough clinical workup, withthe exclusion of reversible causes of cognitive impair-ment, although a definite diagnosis can only be con-firmed at autopsy (McKhann et al., 1984).Unfortunately, there is no simple and reliable biologi-cal marker that can unequivocally detect AD early inits course and distinguish it both from other causes ofdementia and normal ageing (see Consensus Reportof the Working Group on Molecular and BiochemicalMarkers of Alzheimer’s disease). Tests directedtoward clinical markers, such as changes in olfactoryperception (McShane et al., 2001), pupillary dilata-tion (Kardon, 1998) or tests based on neuroimaging(SPECT or MRI) (Jagust et al., 2002), genetic testing(apolipoprotein E genotype) (Mayeux et al., 1998) oralterations in cerebrospinal fluid (CSF) markers (CSF

Received 29 April 2002Copyright # 2003 John Wiley & Sons, Ltd. Accepted 29 January 2003

* Correspondence to: Dr A. Mitchell, Academic Unit of Psychiatry,University of Leeds LS2 9LT, UK. Tel: (0113) 2332703. Fax:(0113) 2433719. E-mail: research@doctors.org.uk

�-amyloid 42 or CSF somatostatin) (Blennow andWallin, 2002), generally lack the necessary specificityin that the biological tests are too often abnormalwhen an alternative disease or no disease is present.In some cases the test appears to be within normallimits when the disease is present, that is the test lackssensitivity. Some authors have suggested that an idealbiological test for Alzheimer’s should have at least an80% sensitivity and 80% specificity (ConsensusReport of the Working Group on Molecular and Bio-chemical Markers of Alzheimer’s disease, 1998;Khachaturian, 2002). Perhaps a more useful way oflooking at the value of a possible diagnostic test iswhether it brings added value to the clinical diagnosisof ‘possible Alzheimer’s disease’. The accuracy ofvarious diagnostic procedures in AD in comparisonto the clinical examination are summarized in Table 1.

CSF PHOSPHORYLATED TAU AS ABIOMARKER

Tau proteins are microtubule-associated proteins thathave well-defined functions in the CNS (Buee et al.,2000). Tau proteins are present in the human brain insix isoforms, encoded by a single gene consisting of16 exons on chromosome 17q21. Tau proteins areessential for life, but became abnormally phosphory-lated in AD at several phosphorylation sites and theseabnormal forms are found in high concentration inpaired helical filaments and neurofibrillary tangles.Tau related pathology in AD demonstrates one ofthe strongest clinicoanatomical correlations with sta-ging of dementia (Thal et al., 2000). On this basis atest was developed that could detect concentrationsof tau in the CSF (Vandermeeren et al., 1993). Inthe last ten years, at least 30 studies have examinedwhether elevated tau concentrations (total tau) in theCSF can be used as a diagnostic test for AD. In thelargest multicentre study, Shoji et al. (2002) studied

366 patients with various stages of AD in comparisonwith 168 patients with other dementias, 316 patientswith neurological diseases and 181 healthy controls.The test showed modest specificity and positive pre-dictive value in cases of both established and earlyAD (Table 1). Other large studies with clinicallyrepresentative comparison groups have also found adisappointing specificity which is not dissimilar tothe specificity that is achieved by clinical diagnosisalone (Hulstaert et al., 1999).

Thus CSF total tau does not appear to have much‘added value’ in the diagnosis of AD. However, inAD, unlike most other forms of dementia and neuro-logical illness, all six isoforms of tau protein areabnormally phosphorylated (Buee and Delacourte,1999). For this reason there is now considerable inter-est in using CSF phosphorylated-tau (phospho-tau) asa diagnostic test. This was first reported in byBlennow and colleagues in 1995. They examined 44patients with AD in comparison with 84 subjects withdepression, vascular dementia, frontal lobe dementia,Parkinson’s disease and normal aging. They recordedan encouraging sensitivity figure of 88% and a speci-ficity of 74%. Since then four other groups have repli-cated these results with larger sample sizes, involvinga total of more than 1000 subjects (Table 2). In the lar-gest study, Itoh and colleagues examined 236 patientswith Alzheimer’s disease (diagnosed according toNINCDS-ADRDA criteria with an MRI examination)and 334 neurological or healthy controls (Itoh et al.,2001). The CSF total-tau and CSF phospho-tau assay(phosphorylated at serine 199) was performed, doubleblind and in duplicate. Using receiver operating curveanalysis, the optimum cut-off for total tau produced a77% sensitivity and 78% specificity whereas phos-pho-tau produced a 85% sensitivity and 85% specifi-city. Of interest, there was no correlation betweenCSF phospho-tau and either Mini-Mental StateExamination scores or Apolipoprotein E (APOE)

Table 1. The accuracy of diagnostic tests of Alzheimer’s disease in comparison to the clinical examination

Criteria Gold standard Total Sensitivity Specificity Positive Negative Referencediagnosis sample predictive predictive

value value

ApoE "4 allele Neuropathology 2188 65% 68% 90% 31% Mayeux et al. (1998)Clinical Diagnosis Neuropathology 2188 94% 71% 94% 72% Mayeux et al. (1998)& ApoE "4 alleleCSF total tau Clinical 1031 89.3% 55.3% 52.4% 90.4% Shoji et al. (2002)Combined CSF total Clinical 413 81% 86% 67% 88% Hulstaert et al. (1999)tau & amyloid �42Clinical diagnosis Neuropathology 2188 93% 55% 90% 64% Mayeux et al. (1998)alone

408 a. mitchell and n. brindle

Copyright # 2003 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2003; 18: 407–411.

status in Alzheimer patients. One criticism of thisstudy was that healthy controls were not agedmatched although the authors replicated resultspost-hoc, with data analysis restricted to elderly con-trols. The authors also examined levels of CSF phos-pho-tau in relation to other degenerative conditionswhere tau forms a pathological deposit (progressivesupranuclear palsy, frontotemporal dementia, cortico-basal degeneration) as well as in acute neurologicaldisease (Creutzfeldt-Jacob disease and meningitis).CSF phospho-tau was elevated in a small minorityof these patients. Combining the results of the five stu-dies presented in Table 2, suggests a mean sensitivityof 80% and specificity of 84%, along with a positivepredictive value of 76% and negative predictive valueof 87%. The high negative predictive value means thata negative test rules out AD with almost 90% prob-ability.

LIMITATIONS OF THE CSF PHOSPHORYLATEDTAU TEST

The results from these early investigations into theCSF phospho-tau test are encouraging, but severalcaveats should be discussed. A common limitationof all studies in this area to date is their reliance onthe clinical examination to establish the diagnosis ofAD. Although most studies use established criteria(with or without neuroimaging), the validity of a testcan only be fully established when the gold standard(that is neuropathology) is used. Other studies thathave examined clinical diagnosis, the ApoE "4 allele,and neuroimaging as viable diagnostic tests haveemployed neuropathological examination.

Another possible weakness of several studies in thisarea, has been the reliance on comparison groups that

are not clinically representative. For example, the testmay adequately distinguish between people with ADand aged matched healthy controls but can it differ-entiate between AD and neurological controls orother causes of dementia? The large multicentre studyfrom Itoh and colleagues (2001) suggests that the onlyother group of patients who may also have elevatedCSF phospho-tau are those with Creutzfeldt Jacobdisease (CJD). As this is a rare differential diagnosisit is unlikely to be a major problem in most clinicalsettings.

One question, yet to be resolved, is which site ofabnormal phosphorylation gives the best discrimina-tion between AD and other groups. It is of note thatthe study with the lowest sensitivity examined thethreonine 181 phosphorylation site (Nagga et al.,2002). Other smaller studies that have also lookedat phosphorylation at this site have also been disap-pointing (Sjogren et al., 2001). A comparative studyis underway examining the different tests (phospho-tau 231 vs phospho-tau 181 vs phospho-tau 199) inthe same subjects and controls (Hampel et al., 2002).

CLINICAL IMPLICATIONS OF THE CSFPHOSPHORYLATED TAU TEST

Any diagnostic test for AD will be of most use if it candifferentiate between early AD and other causes ofsubjective memory loss. Several provisional reportshave examined CSF phospho-tau in subjects withmild cognitive impairment, a proportion of whomdeteriorate into dementia. Buerger and co-workers(2002) examined total tau and phospho-tau in 77patients with mild cognitive impairment. CSF phos-pho-tau but not total tau was significantly higher inthose who developed dementia compared to those

Table 2. Diagnostic Studies using CSF Phosphorylated Tau in Alzheimer’s disease*

Phosphorylation Gold standard Total Sensitivity Specificity Positive Negative Referencesite diagnosis sample predictive predictive

value value

Thr 18/Thr 231 Clinical 128 88.6% 73.8% 63.9% 92.5% Blennow et al. (1995)(NINCDS-ADRDA)

Thr 181 Clinical 202 39.7% 81.4% 54.7% 70.5% Nagga K et al. (2002)(NINCDS-ADRDA)

Ser 199 Clinical 570 85.2% 85% 80.1% 89% Itoh N et al. (2001)(NINCDS-ADRDA)

Thre 231 Clinical 192 90.2% 80% 77.1% 91.7% Buerger K et al. (2002)(NINCDS-ADRDA)

Ser 396/Ser 404 Clinical 195 84% 95.7% 88.7% 93.7% Hu YY et al. (2002)(not stated)

*Studies larger than the original Blennow report are listed.NINCDS-ADRDA—National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s disease and RelatedDisorders Association; Ser¼ Serine; Thr¼Threonine.

csf phosphorylated tau 409

Copyright # 2003 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2003; 18: 407–411.

who did not ( p< 0.037). However, both this studyand a smaller study from Arai et al. (2000) found thatnot everyone who deteriorated had high levels of CSFphospho-tau. A related question is whether the testbecomes more abnormal as the disease progresses.Although this is intuitive, it would be clinically moreuseful if the test was abnormal in the earliest stages ofthe disease. Two studies have examined whether thereis an association between CSF concentration of phos-pho-tau and scores on the mini-metal state examina-tion of patients with Alzheimer’s disease (Itoh et al.,2001; Buerger et al., 2002). Both found no relation-ship, in that the test was equally valid in patientswhose Mini-Metal State Examination scores wereabove 19 as those who scores were below 19.

Where does this leave the value of CSF sampling asa diagnostic test in AD? Several groups have docu-mented that CSF phospho-tau appears to be an accu-rate biological diagnostic test for AD using clinicallyrelevant comparison groups and based on establishedclinical criteria. More work is required to establishwhich site of phosphorylation (or combination ofsites) is most useful and also to replicate these find-ings with neuropathologically confirmed cases. Withthe prospect of potentially disease modifying drugs,there has been a resurgence of interest in the earlydiagnosis of AD. One of the weakness of both neurop-sychological and neuroimaging techniques as diag-nostic tests in Alzheimer’s is that changes are oftentoo subtle to detect in the early stages (Almkvistand Winblad, 1999). Early evidence from studies ofCSF phospho-tau suggests a dissociation from thedegree of cognitive impairment that may be unique.This is of particular interest given that perirhinaland entorhinal phospho-tau immunoreactivity hasbeen linked with declines in episodic memory(Mitchell et al., 2002). It may be that CSF phospho-tau reflects the earliest stages of tau hyperphosphory-lation in Alzheimer’s disease before many neurofibril-lary tangles have developed (Augustinack et al.,2002). Clinically, CSF phospho-tau is a promisingway of distinguishing elderly subjects with subjectivememory complaints from those with suspected mildAlzheimer’s disease.

The question remains—will measurement of CSFphospho-tau become a standard diagnostic test forAD in the near future? This question is difficult toanswer because in many instances clinicians feel con-fident enough about diagnosis without recourse to aninvasive lumbar puncture. However, in difficult casesof possible early dementia where diagnosis is not cer-tain, we may at last see one of the recent successes inthe laboratory translated into clinical practice.

REFERENCES

Almkvist O, Winblad B. 1999. Early diagnosis of Alzheimerdementia based on clinical and biological factors. Eur Arch Psy-chiatry Clin Neurosci 249(Suppl. 3): III/3–III/9.

Andreasen N, Minthon L, Clarberg A, et al. 1999. Sensitivity, spe-cificity, and stability of CSF-tau in AD in a community-basedpatient sample. Neurology 53(7): 1488–1494.

Arai H, Ishiguro K, Ohno H, et al. 2000. CSF phosphorylated tauprotein and mild cognitive impairment: a prospective study.Experiment Neurology 166: 201–203.

Augustinack JC, Schneider A, Mandelkow E-M, Hyman BT. 2002.Specific tau phosphorylation sites correlate with severity of neu-ron cytopathology in Alzheimer’s disease. Acta Neuropathol103: 26–35.

Blennow K, Wallin A, Agren H, et al. 1995. Tau protein in cerebrosp-inal fluid—A biochemical marker for axonal degeneration inAlzheimer disease? Molecular Chem Neuropath 26(3): 231–245.

Blennow K, Wallin A. 2002. Cerebrospinal fluid markers. In Evi-dence-Based Dementia Practice, Qizilibash N, Schneider LS,Chui H, et al. (eds). Blackwell Publishing: Oxford; 175–193.

Buee L, Delacourte A. 1999. Comparative biochemistry of tau inprogressive supranuclear palsy, corticobasal degeneration,FTDP-17 and Pick’s disease. Brain Pathology 9: 681–693.

Buee L, Bussiere T, Buee-Scherrer V, et al. 2000. Tau protein iso-forms, phosphorylation and role in neurodegenerative disorders.Brain Res Rev 33: 95–130.

Buerger K, Zinkowski R, Teipel SJ, et al. 2002. Differential diag-nosis of Alzheimer Disease with cerebrospinal fluid levels of tauprotein phosphorylated at threonine 231. Arch Neurol 59: 1267–1272.

Consensus Report of the Working Group on Molecular and Bio-chemical Markers of Alzheimer’s disease. 1998. The Ronaldand Nancy Reagan Research Institute of the Alzheimer’s Asso-ciation and the National Institute on Aging Working Group. Neu-robiol Aging 19(2): 109–116.

Hampel H, Buerger K, Teipel SJ, et al. 2002. Comparative study ofCSF tau protein phosphorylated at threonine 231, threonine 181and serine 199 in the diagnosis of Alzheimer’s disease. Neuro-biol Aging 23(1)(Suppl. 1): 1396.

Hebert LE, Beckett LA, Scherr PA, Evans DA. 2001. Annual Inci-dence of Alzheimer Disease in the United States Projected to theYears 2000 Through 2050. Alz Dis Assoc Dis 15: 169–173.

Hu YY, He S-S, Wang X-C, et al. 2002. Elevated levels of phos-phorylated neurofilament proteins in cerebrospinal fluid ofAlzheimer disease patients. Neurosci Lett 320: 156–160.

Hulstaert F, Blennow K, Ivanoiu A, et al. 1999. Improved discrimi-nation of AD patients using b-amyloid(1– 42) and tau levels inCSF. Neurology 52: 1555–1562.

Itoh N, Arai H, Urakami K, et al. 2001. Large-scale, multicenterstudy of cerebrospinal fluid tau protein phosphorylated at serine199 for the antemortem diagnosis of Alzheimer’s disease. AnnNeurology 50(2): 150–156.

Jagust W, Chui H, Lee A-Y. 2002. Functional imaging in dementia.In Evidence-Based Dementia Practice, Qizilibash N, SchneiderLS, Chui H, et al. (eds). Blackwell Publishing: Oxford; 162–170.

Kardon RH. 1998. Drop the Alzheimer’s drop test. Neurology50(3): 588–591.

Khachaturian ZA. 2002. The challenges of developing and validat-ing molecular and biochemical markers of Alzheimer’s disease.Neurobiol Aging 23: 509–511.

Mayeux R, Saunders AM, Shea S, et al. 1998. Utility of the apoli-poprotein E genotype in the diagnosis of Alzheimer’s disease.New Engl J Med 338(8): 506–511.

410 a. mitchell and n. brindle

Copyright # 2003 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2003; 18: 407–411.

McKhann G, Drachman D, Folstein M, et al. 1984. Clinical diag-nosis of Alzheimer’s disease: report of the NINCDS-ADRDAWork Group under the auspices of Department of Health andHuman Services Task Force on Alzheimer’s disease. Neurology34: 939–944.

McShane RH, Nagy Z, Esiri MM, et al. 2001. Anosmia in dementiais associated with Lewy bodies rather than Alzheimer’s pathol-ogy. J Neurol Neurosurg Psychiatr 70(6): 739–743.

Mitchell TW, Mufson EJ, Schneider JA, et al. 2002. Parahippocam-pal tau pathology in healthy aging, mild cognitive impairment,and early Alzheimer’s disease. Ann Neurol 51: 182–189.

Nagga K, Gottfries J, Blennow K, Marcusson J. 2002. Cerebrosp-inal fluid phospho-tau, total tau and �-amyloid 1–42 in the dif-ferentiation between Alzheimer’s disease and vascular dementia.Dem Geriatr Cogn Dis 14: 183–190.

Shoji M, Matsubara E, Murakami T, et al. 2002. Cerebrospinal fluidtau in dementia disorders: a large scale multicenter study by aJapanese study group. Neurobiol Aging 23: 363–370.

Sjogren M, Davidsson P, Tullberg M, et al. 2001. Both total andphosphorylated tau are increased in Alzheimer’s disease. J Neu-rol Neurosurg Psychiatry 70: 624–630.

Thal DR, Holzer M, Rub U, et al. 2000. Alzheimer-relatedtau-pathology in the perforant path target zone and in thehippocampal stratum oriens and radiatum correlates with onsetand degree of dementia. Experiment Neurology 163(1):98–110.

Vandermeeren M, Mercken M, Van de Voorde A, et al. 1993. Detec-tion of tau proteins in normal and Alzheimer’s disease cerebrosp-inal fluid with sensitive sandwich ELISA’s. J Neurochem 61:1828–1834.

csf phosphorylated tau 411

Copyright # 2003 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2003; 18: 407–411.