measurements regional cerebral cognitiveperformance in ...montaldi,brooks,mccoll,...

Journal of Neurology, Neurosurgery, and Psychiatry 1990;53:33-38

Measurements of regional cerebral blood flow andcognitive performance in Alzheimer's disease

D Montaldi, D N Brooks, J H McColl, D Wyper, J Patterson, E Barron, J McCulloch

AbstractSingle photon emission computedtomography (SPECT) with 99mTc-HMPAO was used to image 26 patientswith dementia of the Alzheimer type(DAT) and 10 healthy controls. Regionalcerebral blood flow (rCBF) data indi-cated a relative sparing of the occipitalregions in DAT. Normalisation to occi-pital flow illustrated highly significantCBF deficits in a number of corticalregions, particularly in the left and rightposterior-temporal cortex in DAT com-pared to controls. The cognitive perfor-mance of DAT patients was measuredusing a clinical cognitive assessmentprocedure (CAMCOG) and numerous

correlations between these scores andrCBF were obtained. The implicationsand value of this investigative techniqueare discussed.

Departments ofPsychologicalMedicine, andStatistics, GlasgowUniversityD MontaldiD N BrooksJ H McCollE Barron

Department ofClinical Physics,Southern GeneralHospital, Glasgow andthe Wellcome SurgicalInstitute, GlasgowUniversityD WyperJ PattersonJ McCullochCorrespondence to:D Montaldi,Wellcome NeuroscienceGroup,Gartnavel Royal Hospital,1055 Great Westem Road,Glasgow G12 OXH,Scotland.

Received 18 April and inrevised form 17 July 1989.Accepted 1 September 1989.

Alzheimer's disease is the most prominentform of dementia today affecting 5%,0 of thoseover the age of 65 and 2500 of those over 80years old.' The cognitive deficits resultingfrom this disease can range from more focaldeficits to an overall intellectual decline.23 Thepathophysiological changes accompanyingAlzheimer's disease involve both reductions inneurotransmitters and cell loss particularly inthe medial temporal and frontal regions.4 Invivo imaging studies of the disease usingpositron emission tomography (PET) andsingle photon emission computed tomography(SPECT) techniques have analysed patterns ofcerebral metabolism and regional cerebralblood flow (rCBF) respectively. The results ofsuch studies reveal reductions in metabolismand flow in many cortical regions,57 however,parietal reductions are those most frequentlyreported.9The aims of this study were three-fold.

Firstly to attempt to replicate some previousfindings using a dedicated tomographic brainscanner and radiopharmaceutical and a novelquantitation technique. Secondly, to identifythose cerebral regions where bloodflow is par-ticularly reduced in our Alzheimer population.Finally to relate these blood flow data tocognitive performance.Table I Patient and control details

Methods and procedureAll patients and controls form part of a longi-tudinal investigation into Alzheimer's diseaseinvolving six-monthly assessments on a seriesof cognitive tests and measurements of rCBFusing SPECT.

Patient population. Twenty six patients (17females) diagnosed as having dementia of theAlzheimer type (DAT) were studied (table 1).The mean age of the group was 76 years (54-90). Diagnosis ofAlzheimer's disease remains aserious problem; one which was tackled byapplying a very strict classification systemusing the Cambridge Diagnostic Examinationfor the Elderly (CAMDEX) developed by Rothet al (1986).'o Briefly, the CAMDEX criteriainclude a gradual intellectual deterioration forat least six months with the exclusion ofreversible dementias, other psychiatric dis-orders (for example alcohol abuse, majordepression), other neurological diseases (forexample, multi-infarct dementia, communi-cating hydrocephalus), diabetes and malig-nancy and cases of severely impaired hearing orsight. Apart from standard laboratory tests, allpatients also received ECG, EEG, and CTscans. Patients were selected for the DATgroup only when no other possible cause ofdementia could be considered. Although thepatients varied in both severity and age atonset, they were grouped together for thepurpose of this study.

Control population. Ten normal elderlycontrols (eight females) who were recruitedfrom a local social centre were screenedmedically before entry into this group. Themean age of the group was 71 years (50-80).Informed consent was obtained for all patientsand controls.Neuroimaging technique. The radio-

pharmaceutical (Technetium-99m labelledHMPAO) was prepared by adding 1200 MBqof 99m-Tc pertechnetate in 5 ml saline to thefreeze-dried mixture of the ligand (Ceretec,Amersham International). The patient dose of500 MBq was injected intravenously while thepatient was sitting in a quiet room with eyesclosed and ears unplugged.HMPAO is readily removed from the blood

during its first pass through the brain."Thereafter, a fast decomposition of themolecule takes place causing entrapment in the

Age MMSE CAMCOG

Total Praxis Mem LangN Mean Range Mean (SD) Mean (SD) Mean Mean Mean

DAT 26 76 54-90 9-6 (5-5) 35 0 (19-7) 5-1 5 0 15-3Controls 10 71 50-80 28-0 (1-8) 94-8 (6 6) 11-0 23-8 27-7

33

on August 23, 2021 by guest. P

rotected by copyright.http://jnnp.bm

j.com/

J Neurol N

eurosurg Psychiatry: first published as 10.1136/jnnp.53.1.33 on 1 January 1990. D

ownloaded from

http://jnnp.bmj.com/

Montaldi, Brooks, McColl, Wyper, Patterson, Barron, McCulloch

brain and producing a distribution whichreflects rCBF.'2Imaging was carried out five minutes to one

hour after injection and was performed on aNovo 810 dedicated neuro SPECT imager.The patient lay unrestrained on the imagercouch during the scanning session which lastedon average 20 to 25 minutes. During this timefive slices, 12 mm thick, were obtained in aplane parallel to the orbital-meatal line and atpositions 30, 40, 50, 60 and 70 mm superior tothis line. Acquisition time for each slice wasthree minutes.The Novo 810 imager uses the Harvard

multi-detector scanning design utilising 12 setsof scintillation detectors with focussingcollimators, arranged at 300 intervals aroundthe field of view. Each detector scans bothtangentially and radially to the field of viewresulting in relatively uniform spatial resolu-tion, sensitivity and slice thickness throughout.the slice. The image quality of the braintomographs produced are comparable withthose produced by most PET scanners curren-tly in use.The five acquired images were compared

with the anatomical patterns presented in atypical brain atlas for the purpose of definingtwo particular anatomical levels and for theidentification of different brain regions. Thelower of the two slices (the Standard slice),most usually at OM + 40, was defined by thepresence of the putamen, thalamus andoccipital cortex and by the absence ofcerebellum. The high slice (usually OM + 70)lay immediately superior to the corpuscallosum, and typically appeared as an ellipticalring ofhigh activity in the cortex with a furtherline of high activity running continuouslyanterior to posterior representing cortex oneach side of the inter-hemispheric fissure.

Regions of interest. Using the scanner'scomputer system, 14 regions of interest (ROI)outlining different brain regions were thendrawn on each of these slices. (figs 1 and 2) Tenregions of interest (five on each side) weredrawn on the standard slice and four on thehigh slice, with a separate outline encompass-ing the whole cerebral activity in both cases.

Corresponding right and left regions weresymmetrically identical. The ROI boundarieswere drawn along the outside surfaces of thebrain and internally followed the division be-tween the "grey" and "white" matter. On thestandard slice the five regions were described asfrontal, temporal, posterior temporal, occipital(fig 1), and basal ganglia; with high frontal andparietal on the high slice (fig 2). For each regionthe area and mean number of counts per pixelwere measured and expressed as a proportionof the activity in the slice as a whole (Nmean).To normalise these data further we divided theNmean for each ROI by the average Nmean ofthe occipital regions therefore producing anROI/occipital activity ratio (Nmean po). Thecerebrum/cerebellum ratio used in otherstudies'4 was not used here for two reasons.Firstly, using a dedicated neuro section scan-ner, a single image can be obtained within threeminutes. This facility for rapid data collection

Figure I SPET image of Standard slice, illustratingregions of interest.

provides a distinct advantage over the gammacamera in studies of dementia since refusalrates can be high with this patient population.The images used to extract cortical data cannotinclude the cerebellum and therefore additionalslice images would be required thus significan-tly increasing the imaging time and therebyprejudicing our very low refusal rate of 2%.The second reason for not adopting thecerebellum for normalisation is that cerebellarblood flow may be affected by crossedcerebellar diaschisis resulting from damage tothe cortico-ponto-cerebellar system.'5 Selec-tion of occipital flow as the best indicator ofunimpaired flow was based on reports of arelative sparing ofthe occipital lobes in DAT8 afinding strongly supported by some prelimin-ary results of our own.

Figure 2 SPET image of High slice, illustrating regionsof interest.

34



j.com/

J Neurol N


ownloaded from


Measurements of rCBF and cognitive performance in Alzheimer's disease

Figure 3 SPET scan illustrating bloodflow inStandard slice ofDA T (bottom) and control (top).

Examples of images. The images shown infigs 3 and 4 are SPECT scans of a patient withDAT, whose diagnosis was confirmed atnecropsy, (bottom) and a normal control (top).Figure 3 illustrates blood flow in the Standardslice indicating reductions in frontal, temporaland posterior temporal flow together with arelative occipital sparing in patients with DAT.Figure 4 illustrates blood flow in a High slicewhich apart from showing a deficit in higherfrontal flow, indicates a clear reduction inparietal blood flow in the DAT patient.

Cognitive assessment. Before carrying out afull neuropsychological battery, cognitivestatus was measured using the CAMCOG; acomponent of the CAMDEX. The CAMCOGis a short mental status examination similar tothe Mini-Mental State Examination(MMSE)" which is included in its design. It is,however, more extensive than the MMSE inthe areas of cognition that it covers. It com-prises the following eight subcomponents:1 Orientation 5 Memory2 Language 6 Calculation3 Attention 7 Abstract thinking4 Praxis 8 PerceptionThe CAMCOG can either provide a singletotal score of overall cognitive performance, or

Figure 4 SPET scan illustrating bloodflow in Highslice ofDA T (bottom) and control (top).

it can be broken down into subscores. For thepurpose of this study we considered the totalscore and three subscores; Language, Memoryand Praxis. We chose these subcomponents fortwo reasons: firstly, they are three cognitivefunctions known to be affected by DAT; andsecondly, they are the three largest subcom-ponents of the CAMCOG. Administration ofthe CAMCOG -takes approximately half anhour and was carried out within two weeks ofscanning.

ResultsPatterns of rCBFBefore starting the formal analysis we com-pared rCBF in DAT and controls with flownormalised to the slice alone. This showed ahighly significant sparing of the occipitalregions in DAT, lending support to ourdecision to study flow normalised to theoccipital cortex.The first part of our data analysis was

designed to compare rCBF in the DAT groupto that of controls. To do this we used twosample t-tests to compare the ROI/occipitalratios for each ROI across the two groups, andthe results are described in table 2. As the p-values clearly show there is strong evidence of

35



j.com/

J Neurol N


ownloaded from



Table 2 Regional cerebral bloodflow levels in patientswith DAT and normal controls

Nmean po

ROI DA T NC p-value

L posterior-temporal 0 9 1 07 <0 0001*R posterior-temporal 0 89 1 01 0.0006*L temporal 0 84 0 98 0.0007*Lparietal 0 93 1 04 0.0012*R frontal 0 83 0-96 0.0016*Lfrontal 0-82 0 94 0.0023*Rparietal 0 92 1 00 0 0055R temporal 0 87 0-98 0 0090R high frontal 0 84 0 96 0 0090L high frontal 0-85 0 96 0 0115L basal ganglia 1 02 1 12 0 060 (NS)R basal ganglia 1-02 1 08 0 125 (NS)

*Significance obtained at an overall 5% level when using aBonferroni correction.

reduced blood flow in all cortical regions ofinterest, but not the basal ganglia. The top 6 p-values, however, produced particularly strongevidence since significance was still obtained atan overall 5% level when applying a Bonferronicorrection (a very conservative statistic greatlyreducing chance significance).

rCBF and cognitive performance. The finalstage of our analysis investigated the associa-tion between rCBF data and cognitive per-formance in the DAT group. We measured thestrength ofthese associations using Spearman'srank correlation coefficient. Numerousassociations were found between rCBF andcognitive performance, and table 3 includes notonly those correlations where significance wasstill obtained at an overall 5% level whenapplying a Bonferroni correction but also thenumerous less significant associations. Despitethe large number of strongly significantcorrelations, there were, surprisingly, no suchcorrelations between rCBF and memory per-formance. This lack of strong correlations withmemory was investigated further by analysingCAMCOG total, and subscore distributions.As fig 5, 6 and 7 illustrate, the total score, andlanguage and praxis scores all show a broaddistribution. The memory scores, however,show no such distribution, having a markedfloor-effect (fig 8). Since correlational analysisrequires a reasonable distribution of scores toproduce meaningful results, the absence ofcorrelations with memory reflects this floor-effect.

DiscussionRegional cerebral bloodflow measurement

Table 3 Positive associations between cognitiveperformance and rCBF

ROI CAMCOG Memory Language Praxis

Rf * * *Lf * * *Rt * * *Lt ** ** *Rpt ** * ** **Lpt * **Rhf *Lhf *Rp ** * * **Lp ** * ** **

*Correlations significant at an individual 50o level.**Correlations significant at an overall 50°h level whenapplying a Bonferroni correction.

c8-

z

0-20 21-40 41-60 61-80 > 80

Score distribution

Figure 5 CAMCOG: Total scores.

using the NOVO 810 tomograph and 99-m TcHMPAO produces significantly differentresults in patients with Alzheimer's diseasewhen compared to elderly controls. Using anormalised mean value of bloodflow wereplicated the finding of a relative sparing ofthe occipital lobes8 and proceeded to use this tofurther normalise our data. Using this tech-nique we have provided strong evidence of theinvolvement of the posterior temporal regionsin Alzheimer's disease. While there appear tobe no specific reports of CBF deficits in thisregion, these results provide extremely strongevidence that it is clearly implicated in thedisease process. The well documented reduc-tion in parietal blood flow in Alzheimer'sdisease has received further support from thisstudy. The parietal regions are so frequentlyreported as implicated in the disease thatHolman proposed that their damage hasbecome the hallmark of Alzheimer's disease.'6However, while parietal bloodflow is sig-nificantly reduced in our patient population,(especially in the left hemisphere), the resultsof this study suggest that there is an evengreater involvement of both the left and rightposterior temporal regions in this disease.The involvement of the frontal lobes in

Alzheimer's disease provokes considerabledebate. There is no doubt that in many cases ofDAT, a degree of frontal pathology exists.7 1718However, many researchers like to separatepatients with frontal damage and label them as"frontal dementias" or possible Pick's dis-ease.'9 Our results identify a frontal componentin DAT since the rCBF deficits were highlysignificant in both the left and right frontalregions of a population showing no dispropor-tionate frontal signs clinically. Foster, Chaseand Mansi'8 suggest that anterior changes

16

12

4

0-0-5 6-10 11-15 16-20 >20

Score distribution

Figure 6 CAMCOG: Language scores.

36



j.com/

J Neurol N


ownloaded from


Measurements ofrCBF and cognitive performance in Alzheimer's disease

16-.

12

0

0-2 3-4 5-6 7.8 >8

Score distribution

occur late on in the disease process, and thissuggestion is supported by our findings, sincethe majority of our patients are moderately toseverely demented, (See MMSE scores table1). This does not, of course, exclude thepossibility that minimal/mild cases of DATcould show some frontal rCBF deficits. Wewould therefore suggest that the developingtrend towards using SPECT evidence ofexten-sive frontal pathology as an exclusion criterionfor Alzheimer's disease,20 is unwise and maymask interesting features of this disease. Theremay exist for example, a "frontal" subgroup inthe same way as Martin et al identified a"parietal" and a "temporal" subgroup on thebasis of both cognitive and metabolic profiles.Turning now to discuss the relationship

between SPECT data and cognitive perform-ance in DAT. While the highly significantcorrelations (ranging from 0:57 to 074), be-tween rCBF and cognitive performance asmeasured by the CAMCOG, support currenttheories of cortical organisation and cognitivefunction, there are two problems which makeinterpretation of the results particularly dif-ficult. The first problem is that of multiplecorrelations. Apart from the thirteen veryhighly significant correlations, numerous lesssignificant associations were also found. Thismultiplicity of correlations prevents the draw-ing of firm conclusions about the specificrelationship between rCBF deficits and cogni-tion in this population. The second problem isthat in spite of the strength of some of thecorrelations and the extensively reportedmemory deficit accompanying DAT, weobtained no such correlations between rCBFand memory performance on the CAMCOG.To avoid such confounding results in future

research we offer two alternative explanationsfor these problems. The first of these is that theCAMCOG and other cognitive screening testsare inappropriate tools for this form of inves-tigation. Due to the requirements of "brief'

Figure 8 CAMCOG:Memory scores.

12

c

8

0

z

Score distribution

cognitive screening tests, such procedures are,by nature non-specific and the subtests mayfrequently overlap in the cognitive domainsthey are assessing. Therefore, while this type ofassessment tool is ideal for providing a generalcognitive picture, much greater task specificityis required if meaningful correlations are to besought. The inappropriate use of the CAM-COG could therefore be responsible for theexcessive number of correlations found in thisand other studies.'42' Interestingly, this lack oftask specificity may also account for the failureto find correlations between rCBF andmemory. The non-specific nature of screeningtests greatly reduces their sensitivity andcapacity to identify subtle differences in cog-nitive performance between patients. Thissituation is clearly illustrated here by the floor-effect we found with the memory component ofthe CAMCOG (fig 8). Despite evidence fordifferential memory deficits in dementia (forexample, visual/verbal, implicit/explicit);memory on this test can only be expressed asone single score since subcomponent scoreswould be too small. Again, while these subcom-ponents adequately assess memory for screen-ing procedures they are far too generalised (forexample, they measure almost exclusively ver-bal memory), for use in correlational studies ofthis form. Without the use of specific tasks,floor-effects or similar confounding featureswill prevent both an accurate analysis and auseful interpretation of results. Thus, use ofthis type of tool to provide evidence of sub-groups,'4 particularly for studies of theheterogeneity of Alzheimer's disease, is veryquestionable and highly inadvisable.The second explanation for our results could

be the nature of our patients who are typicallyin the later stages of the disease. As with thereductions in blood flow, the cognitive impair-ments resulting from the later stages of thedisease are diffuse and appear to affect nearlyevery cognitive domain and numerous sig-nificant correlations may therefore be expec-ted. Unfortunately reports of similar studiesalso producing multiple correlations'4 21 do notdescribe the severity of the population underinvestigation. So far as the floor effect onmemory is concerned, it could be argued thatthis may have resulted solely from the severenature of our patient population. However,some preliminary results from our neuropsy-chological battery suggest that the lack of taskspecificity provides a much more plausibleexplanation. Although the memory componentof the CAMCOG produced a floor-effect, thesame moderate to severe patient populationreported here produced a substantial distribu-tion of scores on more specific tasks such asvisual memory span22 and paired associatelearning.23 This enables us to investigate morespecific relationships between rCBF andmemory, while avoiding the problem of mul-tiple correlations which swamp any specificeffects.We have illustrated rCBF deficits in a

moderate to severely demented population suf-fering from DAT, using HMPAO andSPECT. In particular, we have identified the

Figure 7 CAMCOG:Praxis scores.

37



j.com/

J Neurol N


ownloaded from



posterior temporal regions as those showing thelargest blood flow deficits, while a relativesparing is indicated in the occipital regions.Certainly the clinical and cognitive severity ofthe patient population must always be con-sidered when designing a study, and sub-sequently reported in publications. Mostimportantly, however, non-specific tools suchas the CAMCOG can produce results whichare difficult to interpret and which may lead theunwary to oversimplistic and inaccurate con-clusions. In the case of this study and otherslike it4 21 we have used very sophisticatedmeans ofmeasuring rCBF, yet extremely crudemeans of measuring cognitive performance. Infuture this type of research should adopt cog-nitive assessment tools compatible in sophis-tication to that of the neuroimaging technique.

This study was funded by The Wellcome TrustGrant 18738/1.19.We thank Mary-Teresa Hansen and ShonaWylie for their technical and nursing assis-tance.

1 Terry RD, Katzman R. Senile Dementia of the AlzheimerType. Annals of Neurology 1983;14:497-506.

2 Huppert FA, Tym E. Clinical and neuropsychologicalassessment of Dementia. Br Med Bull 1986;42 No 1:11-18.

3 Martin A, Browers P, Lalonde F, et al. Towards aBehavioural Typology of Alzheimer patients. J Clin ExpNeuropsychol 1986;8 No 5:594-610.

4 Van Hoesen GW, Damasio A. Neural correlates of cognitiveimpairment in Alzheimer's Disease. In: Mountcastle VB,Plum F, Geiger SR, eds. Handbook of Physiology 1987;Section 1, Vol 5:871-98.

5 Bonte FJ, Elliott DR, Hazem CH, Devous Sr. MD. SpectStudy of Regional Cerebral Bloodflow in Alzheimer'sDisease. Journal of Computer Assisted Tomography1986;10(4):579-83.

6 Johnson KH, Mueller ST, Walshe TM, et al. Cerebralperfusion imaging in Alzheimer's Disease with SPECT

and I-123 IMP. Neurology 1985;35 (Suppl 1):235.7 Duara R, Grady C, Haxby JV, et al. Positron emissiontomography in Alzheimer's Disease. Neurology1986;36:879-87.

8 Risberg J. Cerebral blood flow in dementias. Danish MedBull 1985;32(Suppl 1):48-50.

9 Frackowiak RSJ, Pozzili C, Legg NJ, et al. Regional cerebraloxygen supply and utilization in dementa: A clinical andphysiological study with oxygen-15 and positrontomography. Brain 1981:104:753-78.

10 Roth M, Tym E, Mountjoy CQ, et al. CAMDEX: Astandardised instrument for the diagnosis of mental dis-order in the elderly with special reference to the earlydetection of dementia. Br JPsychiatry 1986;149:698-709.11 Andersen AR, Friberg H, Knudsen K, et al. Extraction of""Tc d-L-HMPAO across the blood brain barrier. JCerebral Blood Flow Metab 1988:8:544-51.12 Neirinckx RD, Canning LR, Piper IM, et al. Technetium99m d-l HMPAO: A new radiopharmaceutical forSPECT Imaging regional cerebral blood perfusion. JNuclMed 1987;28:191-202.

13 Folstein MF, Folstein SE, McHugh PR. "Mini-mentalstate" a practical method for grading the cognitive state ofpatients for the clinician. JPsychiatr Res 1975;12:189-98.14 Burns A, Philpot MP, Costa DC, et al. The Investigation ofAlzheimer's disease with single photon emission tomo-graphy. J Neurol, Neurosurg Psychiatry 1989;52:248-53.15 Baron JC, Bousser, MG, Comar D, Castaigne P. Crossedcerebellar diaschisis in human supratentorial infarction.Ann Neurol 1980a:8: 128.

16 Holman BL. Perfusion and receptor SPECT in the demen-tias. George Taplin Memorial Lecture. J Nucl Med1986;27:855-60.

17 Grady CL, Haxby B, Horwitz M, et al. Longitudinal studyof the early neuropsychological and cerebral metabolicchanges in dementia of the Alzheimer type. J Clin ExpNeuropsychol 1988;10,(5):576-96.

18 FosterNL, ChaseTN, Mansi L, et al. Cortical abnormalitiesin Alzheimer's Disease. Ann Neurol 1984;16:649-54.19 Neary D, Snowden JS, Shields RA, et al. Single photonemission tomography using 'Tc-HM-PAO in the inves-tigation of dementia. J Neurol, Neurosurg Psychiatry1987;50: 1101-9.

20 Neary D, Snowden JS, Goulding P. Dementia offrontal lobetype. In: J Neurol Neurosurg Psychiatry 1988;51:353-61.21 Hunter R, McLuskie R, Wyper D, et al. The pattern offunction-related regional cerebral blood flow investigatedby single photon emission tomography with 9'Tc-HMPAO in patients with presenile Alzheimer's diseaseand Korsakoffs psychosis. Psychol Med (In press).22 Wilson L, Brodie E, Reinink E, et al. Memory for Patternand Path in Senile Dementia. J Clin Exp Neuropsychol1988;10(1):77.

23 Isaacs B, Walkey FA. A simplified Paired-Association Testfor Elderly Hospital Patients. Br J Psychiatry1964:110:80-3.

38



j.com/

J Neurol N


ownloaded from


measurements regional cerebral cognitiveperformance in ...montaldi,brooks,mccoll,...

Documents