CONTROVERSIES IN NEUROLOGY

SECTION EDITOR: E. S. ROACH, MD

Mild Cognitive Impairment as a Clinical Entityand Treatment TargetRonald C. Petersen, PhD, MD; John C. Morris, MD

M ILD COGNITIVE

i m p a i r m e n t(MCI) has beenactively investi-gated for the

past decade.1 , 2 The term wascoined in the late 1980s by theNew York University group toidentify individuals who were notcognitively normal for age and yetdid not have overt dementia, andtheir outcomes were described inan article examining predictors ofdementia by Flicker et al3 in theearly 1990s. Flicker et al character-ized MCI as equivalent to a GlobalDeterioration Scale rating of 3 andfound that a diagnostic approachcombining a careful, structuredinterview and the appropriate neu-ropsychological tests can discrimi-nate those individuals with MCIlikely to experience cognitive dete-rioration from those with a benignprognosis.3

Recently MCI has been givenmore specific criteria involvingthe features outlined in Figure 1of the comments by Gauthier andTouchon4 abstracted from Petersenet al.5 In conjunction with in-creased attention to the early stagesof development of Alzheimer dis-ease (AD), these criteria served tocatalyze intense interest in MCI as apossible prodromal stage of AD.6 Sev-

eral studies on longitudinal popu-lations were conducted, some ret-rofitting criteria to longitudinalstudies of aging, and several othersare either underway or completedand have demonstrated a progres-sion from MCI to AD at an elevatedrate over the base rate in the popu-lation.7-9 Most of these studies ex-amined the amnestic subtype ofMCI, in which memory impair-ment is a key feature. However, otherstudies, often encompassing abroader definition of MCI, havedemonstrated a reversion to nor-mal in some subjects, implying a lackof stability for the construct overtime.10-12 This has led to a closer in-spection of the underlying pre-mises of MCI and a refinement of cri-teria to recognize that MCI is aheterogeneous condition.13

Several factors have emerged ascontributing to the variability in theoutcomes of several of these stud-ies. Among these are the followingfactors: (1) criteria for MCI, (2)implementation of the criteria, (3)source of subjects, and (4) refer-ence standards for normal perfor-mance. Each of these points de-serves discussion as it pertains to thecomments raised by Gauthier andTouchon.4

CRITERIA FOR MCI

The criteria for amnestic MCI havebeen implemented in a variety ofstudies. In general, they character-ize a subset of individuals with MCIwho are likely to progress to clini-cally probable AD. In fact, many ofthe recently completed random-ized clinical trials on MCI have used

these criteria and have been able toapply them on a multicenter basis,and in some cases, international ba-sis, with good reliability.14,15 The cri-teria have been reasonably specificas well. In the recently completedAlzheimer’s Disease CooperativeStudy Clinical Trial on MCI, 212 ofthe 214 subjects who progressedfrom MCI to dementia were char-acterized as having possible or prob-able AD, indicating that when rig-orously applied, the criteria can behighly specific.14,16

However, subsequent work hasalso indicated that the amnestic MCIsubtype (Figure 1 of the Gauthierand Touchon article) may not en-compass all of the prodromal statesof dementia. In fact, other syn-dromic subtypes of MCI can be pos-tulated that likely lead to other de-mentias.13 At a recent internationalconference on MCI in Stockholm,Sweden, a more general approach tothe diagnosis of MCI was pro-posed, which included the consid-eration of multiple types of cogni-tive impairment in addition to thememory impairment that character-izes amnestic MCI.13,17 As shown inFigure 1, an algorithm can beused to subdivide MCI into themajor subdivisions of amnestic andnonamnestic MCI followed by fur-ther subclassification into single-or multiple-domain categories.This scheme is being implementedin several longitudinal studies ofaging, including the newly initi-ated Alzheimer’s Disease Neuro-imaging Initiative sponsored bythe National Institute on Aging(Bethesda, Md).

Author Affiliations: Alzheimer’s DiseaseResearch Center, Mayo Clinic College ofMedicine, Rochester, Minn(Dr Petersen); Alzheimer’s DiseaseResearch Center, Washington UniversitySchool of Medicine, St Louis, Mo(Dr Morris).

(REPRINTED) ARCH NEUROL / VOL 62, JULY 2005 WWW.ARCHNEUROL.COM1160

©2005 American Medical Association. All rights reserved.

Once the physician has made thedelineation of the clinical subtype ofMCI, the next step involves the de-termination of the proposed etiol-ogy of the syndrome as outlined inFigure 2. This approach is similarto that used in delineation of mostneurological diagnoses, includingdementia. If by history, examina-tion, and laboratory testing, one con-cludes that the amnestic subtype islikely caused by a neurodegenera-tive process, then the probable out-come will be AD, as has been sup-ported by progression to more severestages of dementia and by neuro-pathological findings.18,19 The in-sidious onset of symptoms that typi-fies AD suggests that virtually allaffected individuals will experi-ence MCI as the earliest clinical ex-pression of the underlying AD pro-cess. Therefore, if a physician wereinterested in designing a random-ized clinical trial for AD, the appro-priate clinical subtype of MCI, am-nestic MCI, could be combined withthe suspected degenerative etiol-ogy and inclusion and exclusion cri-teria for the trial could be designedappropriately. This basic approachhas been successful in characteriz-ing a uniform group of individualsto participate in randomized clini-cal trials for MCI, using drugs thatare purported to have a mechanismrelevant to the underlying AD patho-logic features.

It is likely that the lack of con-sideration of the etiology of the clini-

cal subtypes of MCI has led to someof the variability in the literature.Epidemiology studies characterizeMCI subtypes clinically and evalu-ate the subjects longitudinally. How-ever, as Figure 2 demonstrates, someetiologies of amnestic MCI (eg, ma-jor depression) would be expectedto improve. Therefore, the generalstatement that MCI is “unstable” isinappropriate since it implies a prob-lem with the construct of MCI. Onthe contrary, some types of MCI ofcertain etiologies should improve.However, if one were to restrict thediscussion to the amnestic subtypeof presumed degenerative etiology,then the predictive value of the con-struct as prodromal AD is borne out.It is only when the subtypes of MCIare discussed broadly without ref-erence to the etiology that the con-fusion arises. The appreciation of thisinterpretation of the literature re-flects increasing sophistication of theconstruct over the years.

IMPLEMENTATIONOF CRITERIA

Some variability is introduced by themanner in which the criteria for MCIare used. Recently, the construct ofcognitive domains has been intro-duced to subclassify the varioustypes of MCI. Inherent in this no-tion is the definition of a domain andthe instruments used to measure thatdomain. While there is some agree-ment on which domains are rel-

evant, there is no consensus. How-ever, this is not different from theapplication of the criteria for the di-agnosis of dementia or AD.20 Thesedefinitions also require impair-ment of memory and other cogni-tive domains (eg, attention, lan-guage, visuospatial skills), as well asa functional impairment. There-fore, the challenge is similar; thephysician must determine what con-stitutes a given cognitive domain,such as memory, and use appropri-ate instruments for measuring thatdomain. There are no specific rec-ommendations in the criteria for de-mentia or AD with respect to whatconstitutes specific domains or howthe physician should measure them.

In addition to the definition of adomain, the physician must decidehow to assess the domains. Does thephysician use 1, 2, or more mea-sures of memory, for example, to de-termine if memory is impaired?Clearly the more measures the phy-sician uses, the more reliable themeasurements will be. Some of thevariability seen in the epidemio-logic studies concerning MCI mayrelate to the tendency to use a singlemeasure to assess a domain such asmemory. If that domain forms thecornerstone for the definition of MCIand the subjects with MCI as a groupare shown to be unstable over time,then the conclusion may be drawnthat the construct of MCI is not re-liable when, in reality, it may be thepsychometric measure “instabil-

Cognitive Complaint

Not Normal for AgeNo Dementia

Cognitive DeclineEssentially Normal Functional Activities

MCI

Amnestic MCI

Amnestic MCIMultiple Domain

Amnestic MCISingle Domain

Nonamnestic MCIMultiple Domain

Nonamnestic MCISingle Domain

Nonamnestic MCI

MemoryImpairment Only?

Single NonmemoryCognitive Domain

Impaired?

Memory Impaired?Yes No

Yes YesNo No

Figure 1. Flowchart for diagnosis of mild cognitive impairment (MCI) subtypes.

Etiology

Amnestic MCI

Nonamnestic MCI

Clin

ical

Cla

ssifi

catio

n SingleDomain

Degenerative Vascular PsychiatricMedical

Conditions

MultipleDomain

SingleDomain

MultipleDomain

AD

AD

FTD

DLB VaD

Depr

DeprVaD

Figure 2. Scheme for combining clinical subtypes with presumed etiology.AD indicates Alzheimer disease; Depr, depression; DLB, dementia with Lewybodies; FTD, frontotemporal dementia; MCI, mild cognitive impairment; andVaD, vascular dementia.

(REPRINTED) ARCH NEUROL / VOL 62, JULY 2005 WWW.ARCHNEUROL.COM1161

©2005 American Medical Association. All rights reserved.

ity” that is responsible for the groupvariability over time.10

Another important aspect of theimplementation of the criteria con-cerns the manner in which the cri-teria are applied. That is, do psycho-metric cutoff scores determine thediagnosis or are clinical methodsused, including careful interviewswith the subject and an informant?As mentioned earlier, to the extentthat one relies on psychometric cut-off scores rather than clinically in-formed judgment, the reliability ofthe subsequent diagnoses may becompromised. Therefore, the phy-sician needs to assess the literatureon MCI from the perspective of thecontext in which the study was per-formed. If a physician believes thatthe study was derived from a set-ting that more closely simulates hisor her practice, the data from re-search clinics might be more appro-priate than those derived from theepidemiologic setting.

SOURCE OF SUBJECTS

A major source of variability amongstudies arises from the origin of thesubjects for the study. There is lessheterogeneity and hence less vari-ability in studies arising frommemory or dementia clinics thanfrom community-based epidemio-logic studies. The former often in-volve a group of individuals who areseeking help for a problem. Demen-tia clinics usually have inclusion andexclusion criteria that help to re-fine the subjects entering into thestudy, and the subjects typically willbe more impaired at the time of ini-tial examination. When unselectedpopulations are approached througha random selection procedure, thereis much more variability in both thesyndromic symptoms and in the eti-ology of the clinical syndromes. Thecombination of these factors plus thepractical requirement of perform-ing less exhaustive evaluations andlikely relying more heavily on psy-chometric criteria all contribute tothe variability of the classificationsand ultimate outcomes in the epi-demiologic setting. This is not un-reasonable since the 2 settings arelikely addressing related but differ-ent questions. The dementia clinicis addressing clinical phenotypes and

predictors of outcome whereas theepidemiologic study focuses on ratesof clinical conditions and progres-sion.

REFERENCE GROUPS

Finally, the reference points for thecomparison to normal can varyamong studies. Variability can be in-troduced by reference standards fornormal cognitive function, and thesereference points can include age andeducation-based norms, compari-son with performance of young in-dividuals, and a change in perfor-mance over time, as well as otherbases for comparison. None of theseis perfect and each introduces asource of bias, yet these can be im-portant. Since the most difficult clini-cal decision often involves the dis-tinction between the cognitivechanges of normal aging and MCI,this decision involves a recognizedset of assumptions regarding nor-mal cognition. This decision oftenis based on predetermined assump-tions regarding performance in nor-mal aging.

POTENTIAL TREATMENT

A difficult issue arises concerningpotential treatments for patients withMCI. First, it must be recognizedthat no treatments have been ap-proved by the Food and Drug Ad-ministration for the indication ofMCI of any subtype. Nevertheless,if a person has the profile of amnes-tic MCI of a presumed neurodegen-erative etiology, this subtype willprogress to clinically probable ADwith a high likelihood. Moreover, theneurobiological phenotype of thissubset of MCI represents a biologi-cal substrate that is intermediatebetween physiological changes ofaging and fully developed AD: over-representation of the apolipopro-tein E4 allele,21,22 volumetric loss inentorhinal cortex and hippocam-pus as measured by magnetic reso-nance imaging23,24 neuronal countsin postmortem samples,25 increasedbrain markers of oxidative stress,26

and abnormalities of the choliner-gic system.27 Recognizing that broadlydefined, MCI is heterogeneous andthat not all individuals will progressto AD or other dementing illness,

nonetheless experienced physiciansusing a combination of clinical andneuropsychological methods in-creasingly and accurately detect thesubset of MCI that very likely willprogress to fully expressed symp-tomatic AD. The remaining discus-sion focuses on this subset alone.

All of the caveats regarding het-erogeneity must be considered andphysicians and patients alike mustrealize that MCI remains a researchconstruct and that the uncertain-ties of outcome include the possi-bility of clinical stability. However,if the patient’s clinical syndrome isconsidered likely to progress to clini-cal AD, then the patient can be coun-seled accordingly. For example,since individuals with MCI are quitecompetent and aware of their diffi-culties, they may want to engage infinancial planning, retirement con-siderations, and other discussions re-garding their futures. In addition, asnoted in Figure 2, some etiologiesof MCI may be reversible (eg, de-pression or medical comorbidi-ties), and these conditions wouldneed to be treated. Nonpharmaco-logical interventions, such as cog-nitive rehabilitation, might be use-ful at this stage as well.

Some physicians may choose touse pharmacological interventions atthis point. For example, if the treat-ing physician is convinced that theclinical condition harbors the un-derlying pathophysiologic featuresof incipient AD, the physician maywish to discuss this with the pa-tient from the perspective of initi-ating AD therapy early. Althoughthere is no Food and Drug Admin-istration approval for the use ofsymptomatic treatments for ADdrugs prior to diagnosis, the physi-cian may wish to consider the indi-vidualized use of cholinesterase in-hibitors for potential symptomaticbenefit at this point in the diseaseprocess. Memantine is also ap-proved for AD treatment but for laterstages of progression. There is noevidence that these medications haveany impact on the underlying dis-ease process or that they will post-pone any clinically meaningful endpoints. The relative expense of treat-ment needs to be weighed againstpotential benefits. For individualswith MCI for whom their symp-

(REPRINTED) ARCH NEUROL / VOL 62, JULY 2005 WWW.ARCHNEUROL.COM1162

©2005 American Medical Association. All rights reserved.

toms are judged clinically to becaused by AD, it may be reasonableto consider symptomatic treatmentat this point in the overall course.

SUMMARY

Mild cognitive impairment is anevolving construct. While appropri-ate questions have been raised con-cerning its definition, outcome, andpotential treatments, considerableprogress has been made. We agreewith Gauthier and Touchon4 thatMCI is heterogeneous. The hetero-geneity reflects a refinement of theentity rather than a weakness. Justas the construct of dementia hasbeen refined over the years to re-flect increasing sophistication in di-agnostic skill, so has the character-ization of MCI been refined. If theapproach to classifying the syn-drome of MCI into multiple sub-types is combined with the pre-sumed etiology, much of thevariability in the literature can be re-solved.

The construct is useful both forclinical as well as research pur-poses. If used in a systematic fash-ion, it can help characterize the pro-dromal states of several types ofdementia. If specific criteria are usedincluding etiology, it can predictclinically probable AD with highspecificity, and this provides physi-cians and investigators an opportu-nity to recognize, counsel, and ul-timately treat the condition beforeit advances to fully expressed de-mentia. In essence, we have to beable to move the diagnostic thresh-old for these disorders back to ear-lier stages of impairment with the an-ticipation of developing appropriatetherapies to intervene at these stages.Ideally, we would like to developmarkers that predict the future de-velopment of dementia and inter-vene with safe and effective thera-pies to truly prevent AD and theother dementing disorders.

Accepted for Publication: March 31,2005.Correspondence: Ronald C. Pe-tersen, PhD, MD, Department of

Neurology, Mayo Clinic College ofMedicine, 200 First St SW, Roches-ter, MN 55905 (peter8@mayo.edu).Author Contributions: Study con-cept and design: Petersen and Mor-ris. Acquisition of data: Petersen andMorris. Analysis and interpretation ofdata: Petersen and Morris. Draftingof the manuscript: Petersen and Mor-ris. Critical revision of the manu-script for important intellectual con-tent: Petersen and Morris. Statisticalanalysis: Petersen and Morris. Ob-tained funding: Petersen and Mor-ris. Administrative, technical, and ma-terial support: Petersen and Morris.Study supervision: Petersen andMorris.Funding/Support: This study wassupported by grants UO1 AG 06786,P50 AG 16574, and U01 AG 10483(Dr Petersen) and P01 AG 03991and P50 AG 05681 (Dr Morris) fromthe National Institute on Aging,Bethesda, Md.

REFERENCES

1. Petersen RC. Mild cognitive impairment.Continuum. 2004;10:9-28.

2. Petersen RC, ed. Mild Cognitive Impairment: Ag-ing to Alzheimer’s Disease. New York, NY: Ox-ford University Press; 2003.

3. Flicker C, Ferris SH, Reisberg B. Mild cognitive im-pairment in the elderly: predictors of dementia.Neurology. 1991;41:1006-1009.

4. Gauthier S, Touchon J. Mild cognitive impair-ment is not a clinical entity and should not betreated. Arch Neurol. 2005;62:1164-1166.

5. Petersen RC, Doody R, Kurz A, et al. Current con-cepts in mild cognitive impairment. Arch Neurol.2001;58:1985-1992.

6. Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tan-galos EG, Kokmen E. Mild cognitive impairment:clinical characterization and outcome. Arch Neurol.1999;56:303-308.

7. Bennett DA, Wilson RS, Schneider JA, et al.Natural history of mild cognitive impairment inolder persons. Neurology. 2002;59:198-205.

8. Fisk JD, Merry HR, Rockwood K. Variations in casedefinition affect prevalence but not outcomes ofmild cognitive impairment. Neurology. 2003;61:1179-1184.

9. Ganguli M, Dodge HH, Shen C, DeKosky ST.Mild cognitive impairment, amnestic type: an epi-demiologic study. Neurology. 2004;63:115-121.

10. Larrieu S, Letenneur L, Orgogozo JM, et al. Inci-dence and outcome of mild cognitive impair-ment in a population-based prospective cohort.Neurology. 2002;59:1594-1599.

11. Ritchie K, Artero S, Touchon J. Classification cri-teria for mild cognitive impairment: a population-

based validation study. Neurology. 2001;56:37-42.

12. Unverzagt FW, Gao S, Baiyewu O, et al. Preva-lence of cognitive impairment: data from the In-dianapolis Study of Health and Aging. Neurology.2001;57:1655-1662.

13. Petersen RC. Mild cognitive impairment as a di-agnostic entity. J Intern Med. 2004;256:183-194.

14. Grundman M, Petersen RC, Ferris SH, et al. Mildcognitive impairment can be distinguished fromAlzheimer’s disease and normal aging for clini-cal trials. Arch Neurol. 2004;61:59-66.

15. Petersen RC. Mild cognitive impairment clinicaltrials. Nat Rev Drug Discov. 2003;2:646-653.

16. Petersen RC, Thomas RG, Grundman M, et al.Donepezil and vitamin E in the treatment of mildcognitive impairment. N Engl J Med. In press.

17. Winblad B, Palmer K, Kivipelto M, et al. Mild cog-nitive impairment—beyond controversies, to-wards a consensus. J Intern Med. 2004;256:240-246.

18. Storandt M, Grant EA, Miller PJ, Morris JC. Ratesof progression in mild cognitive impairment andearly Alzheimer disease. Neurology. 2002;59:1034-1041.

19. Morris JC, Storandt M, Miller JP, et al. Mild cog-nitive impairment represents early-stage Alzhei-mer’s disease. Arch Neurol. 2001;58:397-405.

20. McKhann G, Drachman D, Folstein M, KatzmanR, Price D, Stadlan EM. Clinical diagnosis of Alz-heimer’s Disease: report of the NINCDS-ADRDAwork group under the auspices of Department ofHealth and Human Services Task Force on Alz-heimer’s Disease. Neurology. 1984;34:939-944.

21. Dik MG, Jonker C, Bouter LM, Geerlings MI, vanKamp GJ, Deeg DJ. APOE-epsilon4 is associatedwith memory decline in cognitively impairedelderly. Neurology. 2000;54:1492-1497.

22. Petersen RC, Smith GE, Ivnik RJ, et al. Apolipo-protein E status as a predictor of the develop-ment of Alzheimer’s disease in memory-impaired individuals. JAMA. 1995;273:1274-1278.

23. Jack CR Jr, Petersen RC, Xu YC, et al. Predictionof AD with MRI-based hippocampal volume in mildcognitive impairment. Neurology. 1999;52:1397-1403.

24. Korf ES, Wahlund LO, Visser PJ, Scheltens P.Medial temporal lobe atrophy on MRI predicts de-mentia in patients with mild cognitive impairment.Neurology. 2004;63:94-100.

25. Kordower JH, Chu Y, Stebbins GT, et al. Loss andatrophy of layer II entorhinal cortex neurons in el-derly people with mild cognitive impairment. AnnNeurol. 2001;49:202-213.

26. Pratico D, Clark CM, Liun F, Rokach J, Lee VY,Trojanowski JQ. Increase of brain oxidative stressin mild cognitive impairment: a possible predic-tor of Alzheimer disease. Arch Neurol. 2002;59:972-976.

27. DeKosky ST, Ikonomovic M, Styren S, et al.Upregulation of choline acetyltransferase activ-ity in hippocampus and frontal cortex of elderlysubjects with mild cognitive impairment. AnnNeurol. 2002;51:145-155.

(REPRINTED) ARCH NEUROL / VOL 62, JULY 2005 WWW.ARCHNEUROL.COM1163

©2005 American Medical Association. All rights reserved.