INTERNATIONAL SYMPOSIUM ON OLFACTION AND TASTE

Olfactory Impairment in PresymptomaticAlzheimer’s Disease

Robert S. Wilson,a,b,c Steven E. Arnold,e Julie A. Schneider,a,b,d

Patricia A. Boyle,a,c Aron S. Buchman,a,b

and David A. Bennetta,b

aRush Alzheimer’s Disease Center, bDepartment of Neurological Sciences, cDepartment ofBehavioral Sciences, and dDepartment of Pathology, Rush University Medical Center,

Chicago, Illinois, USAeCellular and Molecular Neuropathology Program, Center for Neurobiology and

Behavior, and Department of Psychiatry, University of Pennsylvania,

Philadelphia, Pennsylvania, USA

Alzheimer’s disease (AD) impairs olfaction, but it is uncertain how early this occurs inthe disease process and whether the effect can be accounted for by other behavioral or ge-netic markers of the disease. We administered the Brief Smell Identification Test (BSIT)to 471 older people without dementia or cognitive impairment who then completedannual clinical evaluations and brain autopsy at death. BSIT score was associated withmore rapid decline in episodic memory and with increased risk of developing incidentmild cognitive impairment (MCI), even after controlling for baseline level of episodicmemory and possession of an apolipoprotein E ε4 allele. In 34 people who died withoutevidence of cognitive impairment, lower BSIT score was associated with higher level ofAD pathology, even after controlling for ε4 and for level of episodic memory functionwhen olfaction was assessed. These analyses suggest that among older people withoutclinical manifestations of AD or MCI, olfactory dysfunction is related to both the levelof AD pathology in the brain and the risk of subsequently developing prodromal symp-toms of the disease, and that these associations persist after accounting for the effectsof other recognized behavioral and genetic markers of the disease.

Key words: olfaction; Alzheimer’s disease; mild cognitive impairment

Introduction

Alzheimer’s disease (AD) is a leading causeof disability in old age, and the public healthchallenges posed by the disease are likely to in-crease in the coming decades with the aging ofthe U.S. population. Definitive classification ofAD currently requires a brain autopsy, under-scoring the need for biological or behavioralmarkers of the disease in the living. In particu-lar, markers are needed to support early diagno-

Address for correspondence: Robert S. Wilson, Ph.D., RushAlzheimer’s Disease Center, Rush University Medical Center, 600 SouthPaulina, Suite 1038, Chicago, IL 60612. Voice: 312-942-2354; fax: 312-942-2297. rwilson@rush.edu

sis because disease-modifying therapeutic com-pounds for AD are under development, and it isgenerally assumed that such compounds will bemost effective early in the disease course, beforepathology is widespread.1–3 In addition to cer-tain practical characteristics (such as being re-liable, noninvasive, and simple to perform), theideal biomarker should be related to AD neu-ropathology1 in those with little or no clinicalevidence of the disease. In view of the substan-tial clinical and neuropathologic heterogeneityof AD, accurate early diagnosis will most likelyrequire the use of several markers in conjunc-tion. In this context, therefore, markers thathave little or no correlation with other markerswill be most useful.

International Symposium on Olfaction and Taste: Ann. N.Y. Acad. Sci. 1170: 730–735 (2009).doi: 10.1111/j.1749-6632.2009.04013.x c© 2009 New York Academy of Sciences.

730

Wilson et al.: Olfaction in Presymptomatic AD 731

In the present study, we focus on olfactoryimpairment as an early sign of AD. With clinicaland neuropathologic data from the Rush Mem-ory and Aging Project, we previously showedthat difficulty identifying familiar odors pre-dicted subsequent development of mild cogni-tive impairment (MCI), a precursor to demen-tia in AD,4 and was robustly correlated withlevel of AD pathology on postmortem exami-nation.5 Here, we conduct further analyses ofindividuals without dementia or MCI at studyenrollment to test whether the association ofolfactory dysfunction with AD is evident inthis well functioning subgroup and whetherit persists after accounting for other recog-nized behavioral and genetic markers of thedisease.

Methods

Participants

All subjects were enrolled in the Rush Mem-ory and Aging Project, a longitudinal clinical-pathologic study of common chronic conditionsof old age.6 Eligibility for the present analy-ses required intact cognitive functioning anda valid score on the Brief Smell IdentificationTest (BSIT)7 at the time of study enrollment;471 individuals met these criteria, and analy-ses are based on this group; 383 were excludedbecause MCI or dementia was identified onthe baseline clinical evaluation (below). Studysubjects had a mean age of 79.3 (SD = 7.0) atbase line and a mean of 14.5 years of education(SD = 2.9); 76.2% were women and 91.3%were white and non-Hispanic (Table 1).

Clinical Evaluation

At base line and annually thereafter, partic-ipants had a uniform clinical evaluation thatincluded a medical history, a complete neu-rological examination, and administration ofa battery of 21 cognitive tests. Seven of theseassessed episodic memory: immediate and de-

TABLE 1. Descriptive Information on the 471Study Participants at Base Line∗

Characteristic Value

Age 79.3 (7.0; 55–100)Education 14.6 (2.9; 5–28)Women, % 76White non-Hispanic, % 92Brief Smell Identification Test 9.2 (1.9; 1–12)Episodic memory 0.39 (0.47; −0.87–1.85)Apolipoprotein E ε4, % 20

∗Data are presented as mean (standard deviation;range) unless otherwise indicated.

layed recall of the East Boston Story and StoryA from Logical Memory and Word List Mem-ory, Word List Recall, and Word List Recog-nition. As previously described, we formed acomposite measure of episodic memory by con-verting raw scores on each test to z scores,using the mean and standard deviation fromthe full cohort, and then averaging the z scoresto yield the composite.8,9 In addition, ratingsof impairment in five cognitive domains weremade by a neuropsychologist, guided by educa-tionally adjusted cutoff scores on a subset of thetests.10

Based on this evaluation, an experiencedclinician diagnosed MCI, dementia, and AD,as previously described.4,11 Dementia requireda history of cognitive decline and impairmentin at least two cognitive domains, one of whichhad to be memory to meet criteria for AD.12

Persons who did not meet criteria for demen-tia but who showed evidence of impairmentin at least one cognitive domain were clas-sified as MCI. These criteria for dementia,AD, and MCI have been widely used andpathologically validated in this and other13,14

cohorts.

Assessment of Odor Identification

Odor identification was assessed with theBSIT.7 For each item, a microcapsule contain-ing a familiar odor was scratched with a penciland placed under the nose of the participantwho matched the smell with one of four choices.

732 Annals of the New York Academy of Sciences

There are 12 items and the score is the numberof correct recognitions. In previous research,this score has been shown to correlate with the40-item University of Pennsylvania Smell Iden-tification Test, from which it was derived.15

Apolipoprotein E Genotyping

Apolipoprotein E genotyping was doneblinded to all other study data using methodsadapted from Hixson and Vernier,16 as previ-ously described.17,18 In all analyses, individualswere dichotomized into those with at least onecopy of the ε4 allele (i.e., ε2/4, ε3/4, or ε4/4)versus those without a copy (i.e., ε2/2, ε2/3, orε3/3).

Neuropathological Evaluation

A standard protocol was followed forbrain removal, sectioning and preserving thetissue, and quantifying pathologic findings,as reported in more detail elsewhere.13,14

Tissue from five brain regions (midfrontalgyrus, inferior parietal gyrus, middle tempo-ral gyrus, entorhinal cortex, and hippocampus[CA1/subiculum]) was cut into 0.5-cm-thickblocks, embedded in paraffin wax, sectioned at6 mm, and stained with modified Bielschowskysilver. For each of the five brain regions of in-terest, a neuropathologist or trained technician,blinded to all clinical data, separately countedneuritic plaques, diffuse plaques, and neurofib-rillar tangles in a 1-mm2 area using a 610 ob-jective (with 610 eyepiece) in the site judgedto have the most of a given type of pathol-ogy. For each type of pathology in each region,the raw count was divided by the SD of allcounts of that pathology in that region to yielda standard score. These standard scores wereaveraged to produce a composite measure ofcortical plaques and tangles and summary mea-sures of each type of pathology, as reportedpreviously.

Data Analysis

We used a proportional hazards model,19 totest the relation of odor identification score torisk of incident MCI. The analysis controlledfor age, sex, education, presence of the ε4 allele,and level of episodic memory function. We useda mixed-effects model20 to characterize changein a composite measure of episodic memoryand to test the relation of odor identificationscore to rate of memory decline, while account-ing for initial level of memory function and con-trolling for age, sex, education, and ε4. In a finalanalysis, we regressed a composite measure ofAD pathology on odor-identification score ina linear regression model adjusted for age atdeath, sex, education, time from olfactory test-ing to death, ε4, and episodic memory.

Results

At the time of study enrollment, scores onthe BSIT ranged from 1 to 12 (mean = 9.2,SD = 1.9), with higher values indicating betterability to identify familiar odors.

Odor Identification and Incidence of MCI

During up to 5.5 years of follow-up(mean = 2.7, SD = 1.4), 155 individuals(32.9%) developed MCI. To determine the re-lation of odor identification to risk of develop-ing MCI, we constructed a proportional haz-ards model. In addition to controlling for age,sex, and education, the model included termsfor two established disease markers: possessionof at least one copy of the apolipoprotein E ε4allele (present in 19.8%) and level of episodicmemory function (mean = 0.39, SD = 0.47), asindicated by a previously established compos-ite measure.8,9 In this analysis, risk of MCI wasassociated with odor identification test score(hazard ratio = 0.874; 95% confidence inter-val: 0.812, 0.941). Thus, a person who madefour errors (score = 8) was about 50% morelikely to develop MCI than a person who madeone error (score = 11).

Wilson et al.: Olfaction in Presymptomatic AD 733

Odor Identification and EpisodicMemory Decline

Decline in episodic memory is one of theearliest clinical manifestations of AD, often be-ginning several years before a clinical diagnosiscan be made.21 Table 1 shows the distributionof the composite measure of episodic mem-ory at base line. To test the relation of odoridentification to change in episodic memory,we constructed a mixed-effects model with thecomposite measure of episodic memory as theoutcome. The model also included terms tocontrol for the effects of age, sex, education, ε4,and level of episodic memory at base line. Inthis analysis, lower odor identification score wasrobustly associated with more rapid decline inepisodic memory (parameter estimate = 0.014,SE = 0.004, P < 0.001).

Odor Identification and Alzheimer’sDisease Pathology

An ideal early marker of AD would be re-lated to the underlying pathology in peoplewith little or no clinical evidence of the dis-ease. At the time of these analyses, we identi-fied 34 individuals who met two criteria: (i) noevidence of cognitive impairment on baselineor follow-up evaluations and (ii) died and un-derwent brain autopsy. These individuals diedat a mean age of 85.2 (SD = 6.4), they had amean of 14.0 years of education (SD = 2.2),67.7% were women, and 94.1% were whiteand non-Hispanic (Table 2). In a uniform neu-ropathologic examination done blinded to allclinical data, crude counts of neuritic plaques,diffuse plaques, and neurofibrillary tangles infive brain regions were converted to a stan-dard scale and averaged. Because the resultingdistribution was skewed, a square root trans-formation was used (see Table 2). In a linearregression model adjusted for age, sex, educa-tion, time from olfactory testing to death, ε4,and episodic memory function when olfactionwas tested, lower odor identification score wasassociated with higher level of AD pathology

TABLE 2. Descriptive Information on the 34 StudyParticipants who Died without Cognitive Impair-ment and Underwent Brain Autopsy∗

Characteristic Value

Age at death 85.2 (6.4; 67–98)Education 13.9 (2.2; 11–19)Women,% 68White non-Hispanic,% 94Brief Smell Identification Test 8.6 (1.8; 6–12)Episodic memory 0.35 (0.46; −0.63–1.23)Apolipoprotein E ε4, % 21Months from olfactory 26.0 (12.9; 3.5–47.1)

test to deathComposite AD pathology 0.58 (0.33; 0.05–1.18)

∗Data are presented as mean (standard deviation;range) unless otherwise indicated.

(parameter estimate = −0.063, SE −0.027,P = 0.028).

Discussion

We studied the relation of the ability to iden-tify familiar odors to the development of AD innearly 500 older persons without evidence ofcognitive impairment at enrollment. The anal-yses suggest that among older persons withoutthe clinical manifestations of AD or its pre-cursor, MCI, olfactory dysfunction is related toboth the level of AD pathology in the brain andthe risk of subsequently developing prodromalsymptoms of AD in the form of MCI and de-clining episodic memory.

It has long been recognized that olfaction isimpaired in persons with clinically diagnosedAD.22–27 Olfactory function is also impairedin MCI,26–31 a precursor of AD, and in thosewith at least one copy of the apolipoproteinE ε4 allele,28,32–34 a well established risk fac-tor for AD. In previous research in this cohort,we showed that odor recognition performancepredicted incidence of MCI and AD, rate ofcognitive decline, and level of AD pathologyon postmortem examination.4,5 The presentanalyses extend these findings by showing thatthe association of olfactory function with theclinical and pathologic manifestations of AD

734 Annals of the New York Academy of Sciences

can be seen in otherwise asymptomatic indi-viduals and that it persists even after account-ing for the effects of other recognized behav-ioral and genetic markers of the disease. Thesedata suggest that olfactory testing, when com-bined with other behavioral and biologic mark-ers, may contribute to early detection of AD.

Prior research in this cohort suggests that theassociation between olfactory dysfunction andclinical AD is largely due to the accumulationof AD pathology, particularly neurofibrillarytangles, in central olfactory regions, especiallyentorhinal cortex and hippocampus.5 The in-volvement of these sites is important becausethey are thought to be among the first areasaffected by the pathologic changes of AD.35–38

Thus, olfactory symptoms might conceivablyprecede cognitive impairment in AD by a sub-stantial period. In addition, decline occurs inother sensory systems with advancing age in as-sociation with cognitive decline.39 Because theentorhinal cortex processes input from multi-ple sensory modalities, it is possible that subtlechanges in other sensory functions might beearly signs of AD.

Confidence in these findings is strengthenedby several factors. The diagnoses of MCI andAD were based on uniform structured clinicalevaluations and widely accepted criteria. Ratesof participation in follow-up clinical evalua-tions and brain autopsy were high. Odor iden-tification was assessed with a standard scale.Episodic memory and AD pathology were as-sessed with previously established compositemeasures. An important limitation is that thecohort is selected so that the generalizability ofthe findings remains to be determined. In ad-dition, securely establishing the value of olfac-tory testing in early diagnosis will likely requiremore extensive assessment of olfactory process-ing conducted proximate to death in a largergroup of people.

Acknowledgments

This research was supported by National In-stitute on Aging Grants R01 AG17917, R01

AG022018, and K23 AG23040 and by the Illi-nois Department of Public Health. We thankthe many Illinois residents for participating inthe Rush Memory and Aging Project; TracyColvin, MPH, and Karen Skish for studycoordination; John Gibbons, MS, and GregKlein for data management; and Todd Beck,MS, and Woojeong Bang, MS, for statisticalprogramming.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. Frank, R.A., D. Galasko, H. Hampel, et al. 2003. Bio-logical markers for therapeutic trials in Alzheimer’sdisease. Proceedings of the biological markers work-ing group; NIA initiative on neuroimaging inAlzheimer’s disease. Neurobiol. Aging 24: 521–536.

2. Andreasen, N. & K. Blennow. 2005. CSF biomarkersfor mild cognitive impairment and early Alzheimer’sdisease. Clin. Neurol. Neurosurg. 107: 165–173.

3. Blennow, K. & E. Vanmechelen. 2003. CSF markersfor pathogenic processes in Alzheimer’s disease: di-agnostic implications and use in clinical neurochem-istry. Brain Res. Bull. 61: 235–242.

4. Wilson, R.S., J.A. Schneider, S.E. Arnold, et al. 2007.Olfactory identification and incidence of mild cogni-tive impairment in old age. Arch. Gen. Psychiatry 64:802–808.

5. Wilson, R.S., S.E. Arnold, J.A. Schneider, et al. 2007.The relationship between cerebral Alzheimer’s dis-ease pathology and odour identification in old age. J.

Neurol. Neurosurg. Psychiatry 78: 30–35.6. Bennett, D.A., J.A. Schneider, A.S. Buchman, et al.

2005. The Rush Memory and Aging Project: studydesign and baseline characteristics of the study co-hort. Neuroepidemiol. 25: 163–175.

7. Graham, J.E., K. Rockwood, B.L. Beattie, et al.1997. Prevalence and severity of cognitive impair-ment with and without dementia in an elderly popu-lation. Lancet 349: 1793–1796.

8. Wilson, R.S., L.L. Barnes & D.A. Bennett. 2003. As-sessment of lifetime participation in cognitively stim-ulating activities. J. Clin. Exp. Neuropsychol. 25: 634–642.

9. Wilson, R.S., L.L. Barnes, K.R. Krueger, et al. 2005.Early and late life cognitive activity and cognitivesystems in old age. J. Int. Neuropsychol. Soc. 11: 400–407.

Wilson et al.: Olfaction in Presymptomatic AD 735

10. Bennett, D.A., R.S. Wilson, J.A. Schneider, et al.2002. Natural history of mild cognitive impairmentin older persons. Neurology 59: 198–205.

11. Wilson, R.S., S.E. Arnold, J.A. Schneider, et al. 2006.Chronic psychological distress and risk of Alzheimer’sdisease in old age. Neuroepidemiology 27: 143–153.

12. McKhann, G., D. Drachmann, M. Folstein, et al.1984. Clinical diagnosis of Alzheimer’s disease. Re-port of the NINCDS-ADRDA Work Group underthe auspices of Department of Health and HumanServices Task Force on Alzheimer’s Disease. Neurology

34: 939–944.13. Bennett, D.A., J.A. Schneider, R.S. Wilson, et al.

2005. Mild cognitive impairment is related toAlzheimer’s disease pathology and cerebral infarc-tions. Neurology 64: 834–841.

14. Bennett, D.A., J.A. Schneider, Z. Arvanitakis, et al.2006. Neuropathology of older persons without cog-nitive impairment. Neurology 66: 1837–1844.

15. Doty, R.L., P. Shaman & M. Dann. 1984. Develop-ment of the University of Pennsylvania Smell Identi-fication Test: a standardized microencapsulated testof olfactory function. Physiol. Behav. 32: 489–502.

16. Hixson, J. & D.T. Vernier. 1990. Restriction isotypingof human apolipoprotein E by gene amplification andcleavage with Hhal. J. Lipid Res. 31: 545–548.

17. Wilson, R.S., J.A. Schneider, L.L. Barnes, et al. 2002.The apolipoprotein E ε4 allele and decline in differ-ent cognitive systems during a 6-year period. Arch.

Neurol. 59: 1154–1160.18. Wilson, R.S., J.L. Bienias, E. Berry-Kravis, et al. 2002.

The apolipoprotein E ε2 allele and decline in episodicmemory. J. Neurol. Neurosurg. Psychiatry 73: 672–677.

19. Cox, D.R. 1972. Regression models and life tables(with discussion). J. R. Soc. Stat. B. 74: 187–220.

20. Laird, N.M. & J.H. Ware. 1982. Random-effectsmodels for longitudinal data. Biometrics 38: 963–974.

21. Hall, C.B., R.B. Lipton, M. Sliwinski, et al. 2000.A change point model for estimating the onset ofcognitive decline in preclinical Alzheimer’s disease.Stat. Med. 19: 1555–1566.

22. Doty, R.L., P.F. Reyes & T. Gregor. 1987. Presenceof both odor identification and detection deficits inAlzheimer’s disease. Brain Res. Bull. 18: 597–600.

23. Murphy, C., M.M. Gilmore, C.S. Seery, et al. 1990.Olfactory thresholds are associated with degree ofdementia in Alzheimer’s disease. Neurobiol. Aging 11:465–469.

24. Mesholam, R.I., P.J. Moberg, R.N. Mahr & R.L.Doty. 1998. Olfaction in neurodegenerative dis-ease: a meta-analysis of olfactory functioning inAlzheimer’s and Parkinson’s diseases. Arch. Neurol. 55:84–90.

25. Larsson, M., H. Semb, B. Winblad, et al. 1999. Odoridentification in normal aging and early Alzheimer’s

disease: effects of retrieval support. Neuropsychology 13:47–53.

26. Serby, M., P. Larson & D. Kalkstein. 1991. The na-ture and course of olfactory deficits in Alzheimer’sdisease. Am. J. Psychiatry 148: 357–360.

27. Royet, J.P., B. Croisile, R. Williamson-Vasta, et al.2001. Rating of different olfactory judgments inAlzheimer’s disease. Chem. Senses 26: 409–417.

28. Wang, Q.S., L. Tian, Y.L. Huang, et al. 2002. Olfac-tory identification and apolipoprotein E episolon 4allele in mild cognitive impairment. Brain Res. 951:77–81.

29. Peters, J.M., T. Hummel, T. Kratzsch, et al. 2003.Olfactory function in mild cognitive impairmentand Alzheimer’s disease: an investigation using psy-chophysical and electrophysiological techniques. Am.

J. Psychiatry 160: 1995–2002.30. Eibenstein, A., A.B. Fioretti, M.N. Simaskau, et al.

2005. Olfactory screening test in mild cognitive im-pairment. Neurol. Sci. 26: 156–160.

31. Djordjevic, J., M. Jones-Gotman, K. de Sousa & H.Chertkow. 2008. Olfaction in patients with mild cog-nitive impairment and Alzheimer’s disease. Neurobiol.

Aging 29: 693–706.32. Gilbert, P.E. & C. Murphy. 2004. The effect of the

ApoE epsilon4 allele on recognition memory for ol-factory and visual stimuli in patients with patho-logically confirmed Alzheimer’s disease, probableAlzheimer’s disease, and healthy elderly controls. J.

Clin. Exp. Neuropsychol. 26: 779–794.33. Murphy, C., A.W. Bacon, M.W. Bondi & D.P.

Salmon. 1998. Apolipoprotein E status is associ-ated with odor identification deficits in nondementedolder persons. Ann. N. Y. Acad. Sci. 855: 744–750.

34. Saleron-Kennedy, R., S. Cusack & K.D. Cashman.2005. Olfactory function in people with genetic riskof dementia. Ir. J. Med. Sci. 174: 46–50.

35. van Hoesen, G.W., B.T. Hyman & A.R. Damasio.1991. Entorhinal cortex pathology in Alzheimer dis-ease. Hippocampus 1: 1–8.

36. Hyman, B.T., G.W. van Hoeson, C. Kromer & A.Damasio. 1984. Alzheimer’s disease: cell specificpathology isolates the hippocampal formation. Sci-

ence 225: 1168–1170.37. Price, J.L., P.B. Davies, J.C. Morris & D.L. White.

1991. The distribution of tangles, plaques and re-lated histochemical markers in healthy aging andAlzheimer’s disease. Neurobiol. Aging 12: 295–312.

38. Li, K.Z.H. & U. Lindenberger. 2002. Relationsbetween aging sensory/sensorimotor and cognitivefunctions. Neurosci. Biobehav. Rev. 26: 777–783.

39. Cushman, L.A., K. Stein & C.J. Duffy. 2008. De-tecting navigational deficits in cognitive aging andAlzheimer’s disease using virtual reality. Neurology 71:888–895.