ARTICLE OPEN ACCESS

MRI load of cerebral microvascular lesions andneurodegeneration cognitive decline anddementiaRui Wang PhD Anna Laveskog MD Erika J Laukka PhD Gregoria Kalpouzos PhD Lars Backman PhD

Laura Fratiglioni MD PhD and Chengxuan Qiu PhD

Neurologyreg 201891e1487-e1497 doi101212WNL0000000000006355

Correspondence

Dr Qiu

chengxuanqiukise

or Dr Wang

ruiwangkise

AbstractObjectiveTo explore the differential associations of neurodegeneration and microvascular lesion loadwith cognitive decline and dementia in older people and the modifying effect of the APOEgenotype on these associations

MethodsA sample of 436 participants (age ge 60 years) was derived from the population-based SwedishNational study on Aging and Care in Kungsholmen Stockholm and clinically examined atbaseline (2001ndash2003) and 3 occasions during the 9-year follow-up At baseline we assessedmicrovascular lesion load using a summary score for MRI markers of lacunes white matterhyperintensities (WMHs) and perivascular spaces and neurodegeneration load for markers ofenlarged ventricles smaller hippocampus and smaller gray matter We assessed cognitivefunction using the Mini-Mental State Examination (MMSE) test and diagnosed dementiafollowing the Diagnostic and Statistical Manual of Mental Disorders 4th edition criteria Weanalyzed data using linear mixed-effects mediation and random-effects Cox models

ResultsDuring the follow-up 46 participants were diagnosed with dementia Per 1-point increase inmicrovascular lesion and neurodegeneration score (range 0ndash3) was associated with multipleadjusted β-coefficients of minus035 (95 confidence interval minus051 to minus020) and minus044 (minus056 tominus032) respectively for the MMSE score and multiple adjusted hazard ratios of 168(112ndash251) and 235 (158ndash352) respectively for dementia carrying APOE e4 reinforced theassociations with MMSE decline WMH volume changes during the follow-up mediated 669and 127 of the total association of MMSE decline with the baseline microvascular score andneurodegeneration score respectively

ConclusionsBoth cerebral microvascular lesion and neurodegeneration loads are strongly associated withcognitive decline and dementia The cognitive decline due to microvascular lesions is exacer-bated by APOE e4 and is largely attributed to progression and development of microvascularlesions

These authors contributed equally to this work

From the Department of Neurobiology (RW EJL GK LB LF CQ) Care Sciences and Society Aging Research Center Karolinska Institutet and Stockholm University Division ofRadiology (AL) Department of Clinical Science Intervention and Technology Karolinska University Hospital at Huddinge Department of Neuroradiology (AL) KarolinskaUniversity Hospital Stockholm and Stockholm Gerontology Research Center (LF) Stockholm Sweden

Go to NeurologyorgN for full disclosures Funding information and disclosures deemed relevant by the authors if any are provided at the end of the article

The Article Processing Charge was funded by Swedish Research Council

This is an open access article distributed under the terms of the Creative Commons Attribution License 40 (CC BY) which permits unrestricted use distribution and reproduction in anymedium provided the original work is properly cited

Copyright copy 2018 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology e1487

Cerebral small vessel disease (SVD) is visualized on MRI aswhite matter hyperintensities (WMHs) lacunes perivascularspaces (PVSs) microbleeds and brain atrophy1 Studies haveshown that SVD plays a pivotal role in cognitive decline anddementia2ndash4 but most studies have examined individualmarkers Because SVD markers often occur concurrently inolder people a cluster of SVDmarkers may better capture theextent and severity of microvascular and neurodegenerativedamage than single markers and better reflect their cognitivephenotypes5 Cross-sectional studies have suggested a strongcorrelation of heavy SVD load with cognitive deficits67

whereas longitudinal data are still lacking

Furthermore neuropathologic studies have correlated MRImarkers of regional and global brain atrophy (eg hippo-campal atrophy and enlarged ventricles) with neurodegener-ative pathologies89 Notably neurodegenerative andmicrovascular pathologies share common mechanisms (eginflammation and oxidative stress) and have a reciprocal re-lationship For instance cerebral arteriosclerosis and hypoxiaimpair perivascular drainage of amyloid and increase neuro-degeneration whereas Alzheimer-related pathologies causeauxiliary vascular damage10ndash12 Thus SVD or atrophicmarkers are likely to reflect a mixture of microvascular andneurodegenerative pathology in the brain

Therefore several issues on SVD load and cognitive pheno-types remain to be clarified (1) whether various SVDmarkers which represent different dominant pathologies suchas intrinsic arteriolar disease (lacunes and WMH) and neu-rodegeneration (hippocampal atrophy) have differentialeffects on cognition is unclear (2) the extent to which theassociation of SVD load with cognitive phenotypes is attrib-uted to subsequent structural brain changes is unknown and(3) whether APOE e4 interacts with SVD load to affectcognitive decline and dementia has yet to be determined Inthis population-based cohort study we seek to assess thedifferential associations of microvascular lesion and neuro-degeneration load with cognitive decline and dementia and toexamine the effect of APOE e4 on these relationships

MethodsStudy participantsParticipants in this population-based cohort study were de-rived from the Swedish National study on Aging and Care inKungsholmen (SNAC-K) a multidisciplinary study of agingand health13 Briefly the SNAC-K participants consisted of an

age-stratified random sample of people who were aged 60years or older and living either at home or in institutions in theKungsholmen district of central Stockholm Sweden Thesample included 3 younger age cohorts each with a 6-yearinterval (60 66 and 72 years) that were followed up withevery 6 years and 8 older age cohorts each with a 3-yearinterval (78 81 84 87 90 93 96 and 99 + years) that werelongitudinally assessed every 3 years At baseline (March 2001to August 2004) 3363 of the 4590 eligible participants(733) were examined Participants of the SNAC-K MRIstudy represented a subsample (n = 555) of the SNAC-Kparticipants who were noninstitutionalized and free of dis-ability and dementia and were recruited between September2001 and October 20031415 In this study we used 9-yearfollow-up data for the SNAC-K MRI sample (2001ndash2003through 2010ndash2013) Of the 555 participants 57 were ex-cluded because of incomplete MRI data or suboptimal imagequality (n = 56) or missing data on the Mini-Mental StateExamination (MMSE) test at baseline (n = 1) Of theremaining 498 participants the 431 participants who had atleast 1 cognitive assessment at follow-up were included in theanalysis concerning cognitive decline and the 436 partic-ipants who had a diagnostic assessment for dementia atfollow-up were included in the analysis involving dementia Inaddition 321 of the 498 participants (645) had at least 1follow-upMRI scan Figure 1 shows the flowchart of the studyparticipants Compared with participants who did not un-dergo brain MRI the SNAC-K MRI participants wereyounger and more educated and the 2 groups did not differ insex and APOE e4 allele distribution as previously reported15

Standard protocol approvals registrationsand patient consentsThe Ethics Committee at the Karolinska Institutet or theRegional Ethics Review Board in Stockholm Sweden ap-proved all parts of the SNAC-K study including linkage withpatient register and death certificates All participants pro-vided written informed consent

MRI acquisition and reading protocolAt baseline participants were scanned on a Philips Intera 15TMRI system (Eindhoven The Netherlands)1415 The coreMRI protocol included an axial 3D T1-weighted fast-field-echo sequence (time of repetition [TR] 15 ms time to echo[TE] 7 ms flip angle [FA] 15deg field of view [FOV] 20 matrix256 times 256) a fluid-attenuated inversion recovery [FLAIR]sequence (TR 6000 ms TE 100 ms inversion time 1900 msFA 90deg echo train length 21 FOV 184 times 230 matrix 204 times256) and a proton densityT2-weighted fast spin-echo

GlossaryATC = Anatomical Therapeutic Chemical BMI = body mass index FA = flip angle FLAIR = fluid-attenuated inversionrecovery FOV = field of view MMSE = Mini-Mental State Examination PVS = perivascular spaces SNAC-K = SwedishNational study on Aging and Care in Kungsholmen SVD = small vessel disease TE = time to echo TR = time of repetitionWMH = white matter hyperintensity

e1488 Neurology | Volume 91 Number 16 | October 16 2018 NeurologyorgN

sequence (TR 4000 ms TE 1890 ms FA 90deg echo trainlength 6 FOV 1875 times 250 matrix 192 times 256 5 mm sliceswithout the use of gap and angulation)1415

A single rater (GK) manually drew global WMH volumeon FLAIR images and further interpolated on the corre-sponding T1 images to compensate for the gap betweenslices in FLAIR (Dice coefficients ranged 074ndash078 andthe mean Dice coefficient was 076) as previously repor-ted16 We segmented T1 images into gray matter whitematter and CSF using SPM12 in MATLAB R2012b Aspecialist in neuroimaging analysis (GK) visuallyinspected all segments Hippocampal volume in bothhemispheres was manually delineated using the region ofinterest tool and the lateral ventricular volumes were es-timated using a region-growing-based semiautomatic toolin HERMES MultiModality14 We also used automatedFreeSurfer segmentation to measure hippocampal volumeat baseline17 The manually and automatically measured

hippocampal volumes were highly correlated (r = 0862 plt 0001) In this study we used the manual measurementof hippocampal volume as we used in previous studies1415

We adjusted all volumetric measurements by total in-tracranial volume

We defined lacunes as small lesions with CSF signal on allsequences and surrounding high signals on FLAIR sequenceWe used a visual rating scale to evaluate PVS as fully reportedelsewhere18 Briefly we used T1 and T2 images to assess PVSin different brain areas (eg frontal lobe parieto-occipitallobe basal ganglia thalami and cerebellum) in each brainhemisphere In each region the number of visible PVS wasmeasured on a 0ndash3 point scale 0 no visible PVS 1 1ndash5 PVS2 6ndash10 PVS or 3 gt10 PVS Then we summed up the pointfrom all brain regions to obtain a global PVS score A clinicalneuroradiologist (AL) made all the assessments of PVS Thevisual rating scale for PVS had the weighted κ statistic of 077for both intrarater and inter-rater reliability18

Figure 1 Flowchart of the study participants in SNAC-K MRI 2001ndash2003 to 2010ndash2013

MMSE = Mini-Mental State Examination SNAC-K = Swedish National study on Aging and Care in Kungsholmen

NeurologyorgN Neurology | Volume 91 Number 16 | October 16 2018 e1489

Load of MRI markers for microvascular lesionsand neurodegenerationWe assessed the cerebral microvascular lesion load and neu-rodegeneration load separately with 2 MRI scores (1) themicrovascular lesion score summarized the MRI markers oflacunes WMH and PVSs because these MRI markers areassumed to reflect predominantly microvascular damage inthe brain1920 and (2) the neurodegeneration score countedtheMRImarkers of enlarged ventricles smaller hippocampusand smaller total gray matter because these MRI markers areusually indicative of Alzheimer-related pathology orneurodegeneration8920 We calculated each of the 2 MRIscores by summing up the total points of the correspondingMRI markers that were concurrently present in a participantWe assigned 1 point to each of the 6 MRI markers as follows(1) presence of lacunes being in the fourth quartile of (2)WMH volume (3) global PVS score and (4) lateral ven-tricular volume and being in the first quartile of (5) hippo-campal volume and (6) total gray matter volume Both themicrovascular lesion score and neurodegeneration scoreranged 0ndash3

Global cognitive function and dementiaWe used the MMSE test to assess global cognitive function atbaseline (2001ndash2003) for all participants at 3-year(2004ndash2007) 6-year (2007ndash2010) and 9-year(2010ndash2013) follow-ups for participants aged ge78 years at6-year and 9-year follow-ups for participants aged 72 yearsand at 6-year follow-up for participants aged 60 and 66 years

We used the Diagnostic and Statistical Manual of Mental Dis-orders 4th edition criteria to define dementia according toa validated 3-step diagnostic procedure1321 Briefly the ex-amining physician made a first preliminary diagnosis of de-mentia based on interviews clinical examination andcognitive testing then a reviewing physician independentlymade a second preliminary diagnosis in case of disagreementbetween the 2 preliminary diagnoses a third opinion froma senior physician was sought to make the final diagnosis Wedefined Alzheimer disease according to the National Instituteof Neurological and Communicative Disorders and Strokeand the Alzheimerrsquos Disease and Related Disorders Associa-tion criteria For participants who had died before the sub-sequent follow-up examination physicians thoroughlyreviewed medical records and death certificates to determinewhether the participant died with dementia or Alzheimerdisease

CovariatesWe collected data on potential confounding factors (ascovariates in the analysis) through face-to-face interviewsclinical examinations and laboratory tests The confoundingfactors included demographics (eg age sex and education)lifestyles (eg smoking alcohol consumption and physicalactivity) medical history (eg hypertension diabetes heartdisease and stroke) use of medications (eg antihypertensiveagents oral hypoglycemic agents and cholesterol-lowering

drugs) weight height blood pressure and biomarkers (egtotal cholesterol HbA1c and APOE genotype)22 Educationwas measured by the maximum years of formal schooling Weclassified all medications according to the Anatomical Ther-apeutic Chemical (ATC) classification system Body massindex (BMI) was calculated as measured weight (kg) dividedby height (m) squared and obesity was defined as a BMIge30 kgm2 We dichotomized smoking status as neverformer vs current smoking alcohol consumption as heavy vsno or moderate alcohol intake and physical activity as in-activity vs fitness-enhancing activity We defined hyper-tension as arterial blood pressure ge14090 mm Hg orcurrent use of antihypertensive drugs (ATC codes C02C03 and C07-C09) diabetes as having a self-reportedhistory of diabetes use of oral hypoglycemic agents or in-sulin (ATC code A10) records of diabetes in patient reg-ister or HbA1c ge 65 and high total cholesterol as totalcholesterol gt622 mmolL or current use of cholesterol-lowering drugs (ATC code C10)

Statistical analysisWe compared the baseline characteristics of participants byAPOE e4 status using logistic regression models adjustingfor age We used multiple linear mixed-effects models toestimate the β-coefficient and 95 confidence interval (CI)of average annual changes in the MMSE score over thefollow-up period related to baseline MRI scores for cerebralmicrovascular lesions and neurodegeneration15 The mod-els included the baseline MRI score follow-up time andtheir interaction term Because we were interested in theeffects of brain MRI load on MMSE changes during follow-up we reported only β-coefficient (95 CI) for the MRIscore times follow-up time interaction term which reflects ad-ditional effect of the MRI score on annual MMSE changesWe used random-effects Cox proportional hazards modelsto assess the association of MRI scores at baseline withincident dementia with follow-up time as the time scale Weverified the proportional hazards assumption by plottinglog-log survival curves and the Schoenfeld residual test Wetested interactions between an MRI score and APOE ɛ4allele by adding the 3-way product term of the MRI score timesfollow-up time times APOE ɛ4 allele to the model Whena statistical interaction was detected we further performedstratifying analysis to verify the direction and magnitude ofthe interaction We reported the results from 4 modelsmodel 1 was controlled for age sex education cardiovas-cular risk factors (eg smoking diabetes and hyperten-sion) and APOE e4 allele in model 2 we simultaneouslyentered the baseline MRI scores for both cerebral micro-vascular lesions and neurodegeneration into model 1 toassess their independent effects on cognitive outcomes toassess the effect of structural brain changes that occurredduring the follow-ups on the association of baseline MRIscores with cognitive decline and dementia in model 3 weadded follow-up WMH volume (a marker of microvascularlesions) to model 1 and in model 4 we added follow-uptotal gray matter volume (a marker of neurodegeneration)

e1490 Neurology | Volume 91 Number 16 | October 16 2018 NeurologyorgN

to model 3 Finally for the analysis involving cognitivedecline we also performed mediation analysis to test andquantify the mediation effect of structural brain changesoccurred during the follow-up on the association betweenbaseline MRI score and MMSE decline in which we usedbootstrapping methods to estimate the 95 CI We usedStata 140 for Windows (StataCorp LP College StationTX) for all analyses

Data availabilityData related to the current study are derived from the SNAC-K and SNAC-KMRI projects (snac-kse) Access to theseanonymized SNAC-K and SNAC-KMRI data will be avail-able from the corresponding author (CQ) upon reasonablerequest and approval by the SNAC-K data management andmaintenance committee at the Aging Research Center Kar-olinska Institutet Stockholm Sweden

ResultsAt baseline the mean age of the 436 participants was 703years (SD = 91) and 610 were women After controllingfor age carriers and noncarriers of the APOE e4 allele did notsignificantly differ for all baseline characteristics examinedexcept the total cholesterol level where the APOE e4 allelecarriers had higher total cholesterol than the noncarriers(table 1) In addition the age- and sex-adjusted partial cor-relation coefficient between the baseline microvascular lesionscore and neurodegeneration score was 005 (p = 030)

During the 9-year follow-up period the average annual rate ofMMSE changes was minus063 (95 CI minus075 to minus051 p lt0001) after controlling for age sex and education A higherMRI score for both microvascular lesions and neuro-degeneration at baseline was associated with faster MMSE

Table 1 Baseline characteristics of study participants (n = 436) by APOE e4 allele status

Characteristics

Total sample APOE laquo4 allelea

(n = 436) No (n = 315) Yes (n = 115) p Valued

Age (y) mean (SD) 703 (91) 705 (92) 696 (86) 0365

Female n () 266 (610) 192 (610) 73 (635) 0590

Education n ()

Elementary school 51 (117) 39 (124) 12 (104)

Middle school 197 (452) 145 (460) 52 (452)

University 188 (431) 131 (416) 51 (444) 0694

Hypertension n () 313 (718) 231 (733) 78 (678) 0374

Diabetes mellitus n () 28 (64) 23 (73) 4 (35) 0167

BMI (kgm2) mean (SD)b 260 (41) 261 (43) 258 (34) 0448

Total cholesterol (mmolL) mean (SD)b 62 (11) 61 (11) 64 (09) 0048

Current smoking n () 50 (115) 38 (121) 11 (96) 0391

Heavy alcohol drinking n () 79 (181) 57 (181) 21 (183) 0970

Physical inactivity n () 83 (190) 66 (210) 16 (139) 0120

MMSE score mean (SD) 291 (11) 291 (11) 290 (11) 0495

Lacunar infarcts n () 51 (117) 41 (130) 9 (78) 0174

PVS score median (IQR) 30 (24ndash34) 30 (25ndash34) 30 (23ndash33) 0554

WMH volume mean (SD)c 042 (074) 040 (074) 048 (076) 0161

Total gray matter volume mean (SD)c 3697 (362) 3686 (364) 3728 (358) 0529

Hippocampal volume mean (SD)c 045 (006) 045 (005) 045 (006) 0178

Ventricular volume mean (SD)c 182 (108) 182 (108) 180 (108) 0796

Abbreviations BMI = body mass index IQR = interquartile range MMSE = Mini-Mental State Examination PVS = perivascular space WMH = white matterhyperintensitiesa Information on the APOE genotype was missing in 6 participantsb The numbers of participants with missing value were 1 for BMI and 6 for total cholesterol The missing values were replaced with their mean in furtheranalysesc Volume was corrected for total intracranial volume and data are mean (SD) of percentage of the volume over the total intracranial volumed p Value was adjusted for age

NeurologyorgN Neurology | Volume 91 Number 16 | October 16 2018 e1491

decline independent of demographics cardiovascular riskfactors and APOE e4 status (p for trend lt 001) (table 2model 1) Furthermore when simultaneously entering base-line MRI scores into the model the linear association (β-co-efficient) withMMSE decline remained statistically significantfor both MRI scores although the association with the neu-rodegeneration score appeared to be stronger than that for themicrovascular lesion score (table 2 model 2)

Of the 498 baseline participants 321 (645) undertook atleast 1 follow-up MRI scan and data on follow-up WMH andtotal graymatter volumes were available During the follow-upperiodWMHvolume increased at the average rate of 145 mLper year (95 CI 114ndash176 p lt 0001) (standardizedβ-coefficient 017) whereas the total gray matter volumedecreased at the average rate of 1103 mL per year (95 CIminus1203 to minus1003 p lt 0001) (standardized β-coefficientminus015) When entering follow-up WMH volume variable intothe model the associations of both MRI scores at baselinewith MMSE decline during the follow-up were substantiallyattenuated (table 2 model 3) notably the linear relationshipwith MMSE decline remained statistically significant for the

neurodegeneration score (p for linear trend lt001) but not forthe microvascular lesion score Further entering follow-uptotal gray matter volume variable into model 3 had almost noeffect on the results (table 2 model 4)

In the mediation analysis after controlling for a range of po-tential confounders 669 of the total association of MMSEdecline with the baseline microvascular lesion score but only127 of the total association with the baseline neuro-degeneration score was attributed to WMH changes that oc-curred over the follow-up period In contrast changes in totalgray matter volume that occurred during the follow-up periodcould explain only 128 and 22 of the total associations ofMMSEdecline with the baselinemicrovascular lesion score andneurodegeneration score respectively (figure 2)

There was a reliable interaction of APOE e4 status with scoresfor both microvascular lesions and neurodegeneration onMMSE decline (p for interaction lt001) A stratifying analysisrevealed a linear association between microvascular lesionsand MMSE decline only among APOE e4 allele carriers(figure 3) In addition carrying the APOE e4 allele magnified

Table 2 Associations of baselineMRI scores formicrovascular lesions andneurodegenerationwith annual changes in theMMSE score

Baseline MRI scores formicrovascular lesions andneurodegeneration

No ofparticipants

β-coefficient (95 confidence interval) average annual changes in the MMSE score

Model 1a Model 2b Model 3c Model 4d

Microvascular lesion score

Continuous (range 0ndash3) 431 minus035 (minus051 to minus020)g minus020 (minus036 to minus005)f minus003 (minus016 to 009) minus002 (minus015 to 011)

Categorical

0 245 0 (reference) 0 (reference) 0 (reference) 0 (reference)

1 131 minus042 (minus068 to minus016)g minus021 (minus047 to 005) minus019 (minus039 to 002) minus017 (minus037 to 003)

ge2 55 minus067 (minus104 to minus031)g minus038 (minus074 to minus001)f 002 (minus028 to 032) 005 (minus025 to 035)

p Value for trende lt0001 0023 0477 0620

Neurodegeneration score

Continuous (range 0ndash3) 431 minus044 (minus056 to minus032)g minus039 (minus051 to minus027)g minus022 (minus032 to minus011)g minus021 (minus032 to minus011)g

Categorical

0 235 0 (reference) 0 (reference) 0 (reference) 0 (reference)

1 104 minus049 (minus076 to minus023)g minus046 (minus073 to minus019)g minus044 (minus064 to minus023)g minus044 (minus064 to minus023)g

ge2 92 minus105 (minus134 to minus076)g minus093 (minus124 to minus063)g minus040 (minus065 to minus015)g minus039 (minus064 to minus014)g

p Value for trende lt0001 lt0001 lt0001 lt0001

Abbreviation MMSE = Mini-Mental State Examinationa Model 1 adjusted for demographic factors cardiovascular risk factors (ie hypertension obesity physical inactivity high cholesterol current smokingheavy drinking and diabetes) and APOE e4 alleleb Model 2 the 2 baseline MRI scores for cerebral microvascular lesions and neurodegeneration were added simultaneously to model 1c Model 3 follow-up white matter hyperintensity volume was added to model 1d Model 4 follow-up total gray matter volume was added to model 3e p Value is for the test of a linear trend for association between the baselineMRI score and theβ-coefficient of average annual changes in theMMSE score thatwas derived from the model when the score was entered as a continuous variable (coded as 0 1 and 2 if the score ge2)f p lt 005g p lt 001

e1492 Neurology | Volume 91 Number 16 | October 16 2018 NeurologyorgN

the association between the neurodegeneration load andMMSE decline

During the 9-year follow-up period 46 participants were di-agnosed with dementia of these 27 were classified to haveAlzheimer disease Increased MRI scores for microvascularlesions and neurodegeneration were linked to an increasedrisk of dementia (table 3 model 1) when simultaneouslyentering the 2 MRI scores into the model the linear trend ofan association with the risk of dementia was statistically sig-nificant only for the neurodegeneration score (table 3 model2) Further adding the variables of follow-up WMH and totalgray matter volumes to the model did not substantially alterthe association between either score at baseline and risk ofincident dementia (table 3 models 3 and 4) There was nosignificant interaction between the baseline MRI scores andAPOE e4 allele on the risk of dementia The results for theassociations between the baseline MRI scores and risk ofAlzheimer disease were similar to those for dementia ingeneral (table e-1 linkslwwcomWNLA702)

Finally we repeated the analyses only among participants whoundertook at least 1 follow-up brain MRI scan and who haddata available on global WMH and total gray matter volumes(n = 321) which yielded results that were consistent withthose reported in tables 2 and 3 (data not shown)

DiscussionThis population-based cohort study of older adults revealedthat (1) an increasing load for both cerebral microvascularlesions and neurodegeneration is strongly associated witha faster decline of global cognitive function and a greater riskof dementia and Alzheimer disease and neurodegenerationload appears to have a stronger and more prominent associ-ation with cognitive decline and dementia risk than that ofmicrovascular lesion load (2) the association of microvas-cular lesion load with subsequent cognitive decline is largely(67) attributed to the progression and development ofWMH and (3) carrying APOE e4 allele strengthens the

Figure 2Mediation effects of changes in WMH and total gray matter volumes during the follow-up on the associations ofMMSE decline with the baseline MRI score for cerebral microvascular lesions (A) and for neurodegeneration (B)

Mediation models were adjusted for demographic factorscardiovascular risk factors and APOE e4 allele MMSE =Mini-Mental State Examination WMH = white matter hyper-intensities p lt 001 p = 0066

NeurologyorgN Neurology | Volume 91 Number 16 | October 16 2018 e1493

association of brain MRI load with cognitive decline espe-cially with regard to the cerebral microvascular lesion load

A correlation of cerebral SVD load with global cognitive im-pairment and dementia has been reported in cross-sectionalstudies of community-dwelling older people67 In the currentcohort study we categorized all 6 MRI markers of brainlesions into microvascular lesions and neurodegeneration andrevealed a more prominent association of neurodegenerationload with cognitive decline and risk of dementia than that ofmicrovascular lesion load The classification of MRI markersis based on the facts that lacunes and WMH are considered tobe of vascular origin because of cerebral arteriosclerosis andhypoperfusion1920 and enlarged PVS can signal microvas-cular lesions23 whereas postmortem MRI studies correlatedhippocampal atrophy with local increased burden of Alz-heimer pathologies (eg β amyloid and tau protein) and

sclerosis924 and ventricular enlargement was a summarymarker of global gray and white matter atrophy825 Howevercerebral microvascular and neurodegenerative pathologiesshare both risk factors (eg smoking hypertension andAPOE gene) and biological mechanisms (eg inflammationand oxidative stress)2627 and neurodegenerative pathologiesin aging may cause cerebral perfusion deficits that precedevolume loss28 Also cerebral microvascular and neurodegen-erative pathologies often coexist in older people829 andneuropathologic studies have linked volumetric brain meas-urements (eg enlarged ventricles and gray matter atrophy)with mixed Alzheimer and cerebrovascular pathologies (egneuritic plaques neurofibrillary tangles arteriosclerosis andinfarcts)820 Thus we should be aware that MRI markers forcerebral microvascular lesions and neurodegeneration do notnecessarily represent distinct brain pathologies Toward thisend our findings are in line with the view that the total burden

Figure 3 Association of the baseline MRI scores for cerebral microvascular lesions (A) and neurodegeneration (B) withaverage annual changes in MMSE score by APOE e4 status

MMSE = Mini-Mental State Examina-tion WMH = white-matter hyper-intensities Model 1 adjusted fordemographic factors and cardiovas-cular risk factors model 2 the 2baseline MRI scores for microvascu-lar lesions and neurodegenerationwere added simultaneously tomodel1 p lt 005 p lt 001

e1494 Neurology | Volume 91 Number 16 | October 16 2018 NeurologyorgN

of cerebral mixed vascular and neurodegenerative alterationsis the strongest determinant of cognitive phenotypes inaging1530ndash32 The interactive effect of SVD load with APOEe4 allele on cognitive decline is consistent with findings ofinteractions between APOE e4 allele and individual markersfor microvascular (eg WMH) and neurodegenerative (egneuritic plaques and neurofibrillary tangles) lesions on cog-nitive phenotypes in aging8103334

Both cerebral microvascular lesion and neurodegenerationloads had a strong association with subsequent cognitive de-cline and dementia risk in older adults Because MRI markersfor neurodegeneration are more likely to reflect mixed mi-crovascular and neurodegenerative pathologies the apparentstronger effect of neurodegeneration load than that of mi-crovascular lesions on global cognitive decline and the risk ofdementia suggests that clusters of various brain MRI markerswith different underlying neuropathologies may act in-teractively to contribute to substantial cognitive decline anddementia risk35 This is consistent with the clinical datashowing that markers of neurodegeneration are more prom-inent determinants of global cognitive deficits than those ofmicrovascular lesions (eg WMH microbleeds and

lacunes)36ndash38 A cohort study of middle-aged people witha family history of Alzheimer disease also demonstrated thatclusters of MRI and CSF markers for various types of brainpathology were differentially related to patterns of cognitivedecline39

Because nearly two-thirds of participants had follow-up MRIdata in our study we were able to explore the effect ofstructural brain changes on the association of baseline cerebralSVD load with cognitive decline and dementia Our media-tion analysis suggested that the association of microvascularlesion load with subsequent cognitive decline was largelymediated by follow-up changes in WMH volume whereas theassociation of neurodegeneration load to cognitive declinewas independent of changes in volumes of both WMH andtotal gray matter during the follow-up This implies that theassociation between cerebral microvascular lesions and sub-sequent cognitive decline is attributed to the progression ornew development of cerebral microvascular lesions suggest-ing a vascular nature of the cognitive decline This is a criticalfinding as it may imply that interventions aiming to slow theprogression and development of microvascular lesions (egWMH) might attenuate cognitive decline in aging However

Table 3 Associations of baseline MRI scores for microvascular lesions and neurodegeneration with incident dementia

Baseline MRI scores formicrovascular lesions andneurodegeneration

No of casesno ofparticipants

Hazard ratio (95 confidence interval) dementia

Model 1a Model 2b Model 3c Model 4d

Microvascular lesion score

Continuous (range 0ndash3) 46436 168 (112ndash251)f 169 (110ndash262)f 177 (116ndash270)g 181 (117ndash279)g

Categorical

0 15247 100 (reference) 100 (reference) 100 (reference) 100 (reference)

1 18132 126 (056ndash283) 089 (038ndash209) 132 (058ndash299) 108 (048ndash245)

ge2 1357 267 (104ndash682)f 199 (076ndash520) 299 (111ndash802)f 325 (121ndash877)f

p Value for trende 0049 0170 0038 0036

Neurodegeneration score

Continuous (range 0ndash3) 46436 235 (158ndash352)g 227 (153ndash336)g 238 (158ndash358)g 218 (145ndash328)g

Categorical

0 4236 100 (reference) 100 (reference) 100 (reference) 100 (reference)

1 11106 458 (127ndash1644)f 444 (123ndash1600)f 471 (129ndash1716)f 391 (110ndash1385)f

ge2 3194 1393(387ndash5017)g

1356(368ndash5002)g

1441(393ndash5288)g

1275(353ndash4608)g

p Value for trende lt0001 lt0001 lt0001 lt0001

a Model 1 adjusted for demographic factors cardiovascular risk factors (ie hypertension obesity physical inactivity high cholesterol current smokingheavy drinking and diabetes) and APOE e4 alleleb Model 2 the 2 baseline MRI scores for cerebral microvascular lesions and neurodegeneration were added simultaneously to model 1c Model 3 follow-up white matter hyperintensity volume was added to model 1d Model 4 follow-up total gray matter volume was added to model 3e p Value is for the test of a linear trend for association between the baseline MRI score and the hazard ratio of incident dementia that was derived from themodel when the score was entered as a continuous variable (coded as 0 1 and 2 if the score ge2)f p lt 005g p lt 001

NeurologyorgN Neurology | Volume 91 Number 16 | October 16 2018 e1495

we found little evidence that structural brain changes (WMHand total gray matter) influenced the link of brainMRI load tosubsequent risk of dementia

Major strengths of our study include the population-baseddesign longitudinal assessment of structural brain integrityand comprehensive control of potential confounders Ourstudy also has limitations First the study sample had highersocioeconomic positions than the average Swedish pop-ulation and was younger and healthier than the target pop-ulation Thus one should be cautious in generalizing ourfindings to other populations Furthermore the MMSE testmay not be sensitive to subtle cognitive changes and may besubject to practice effects when administered in a series ofassessments Third MRI sequences for additional MRImarkers for brain lesions such as cerebral microbleeds andmicroinfarcts were not available which might have under-estimated the effect of cerebral microvascular lesion load oncognitive outcomes Finally some MRI markers that we usedfor neurodegeneration (eg enlarged ventricle) may correlatealso with brain vascular pathology which needs to be kept inmind when interpreting our findings

Overall our population-based cohort study demonstratesa strong association of MRI load for both microvascularlesions and neurodegeneration with global cognitive declineand risk of dementia in older people The cognitive con-sequences of cerebral microvascular lesion load largely stemfrom its progression or development of new cerebral micro-vascular lesions This suggests that a structural MRI-basedscore approach for brain lesions may help identify individualsat risk for accelerated cognitive decline and dementia forclinical trials as well as for preventive interventions targetingvascular pathways that aim to delay the onset of late-lifecognitive impairment and dementia

Author contributionsStudy concept and design R Wang A Laveskog EJ LaukkaL Backman L Fratiglioni and C Qiu Data acquisition ALaveskog G Kalpouzos EJ Laukka and L Fratiglioni Dataanalysis R Wang Interpretation of data all authors Draftingof themanuscript RWang and C Qiu Critical revision of themanuscript all authors R Wang had full access to all data inthis study and takes responsibility for the integrity of data andthe accuracy of data analysis

AcknowledgmentThe authors thank all the SNAC-K participants and theircolleagues in the SNAC-K Study Group for their collabora-tion in data collection and management

Study fundingSNAC-K is supported by the Swedish Ministry of Health andSocial Affairs the participating County Councils and Mu-nicipalities and the Swedish Research Council This work wasfurther supported by grants from the Swedish ResearchCouncil (grants nos 2015-02531 2016-06658 and 2017-

05819) the Swedish Research Council for Health WorkingLife and Welfare (grant no 2014-01382) the KarolinskaInstitutet Stockholm Sweden and the European UnionHorizon 2020 Framework Programme for Research and In-novation (grant no 667375) L Backman was supported byan Alexander von Humboldt Research Award and a donationfrom the af Jochnick Foundation The funders had no role instudy design data collection and analysis preparation of themanuscript or decision to publish

DisclosureR Wang was supported by a grant from the Swedish ResearchCouncil (grant no 2016-06658) A Laveskog E Laukka andG Kalpouzos report no disclosures relevant to the manu-script L Backman was supported by an Alexander vonHumboldt Research Award and a donation from the af Joc-hnick Foundation L Fratiglioni was supported by the Euro-pean Union Horizon 2020 Framework Programme forResearch and Innovation (grant no 667375) C Qiu wassupported by grants from the Swedish Research Council(grants nos 2015-02531 and 2017-05819) the Swedish Re-search Council for Health Working Life and Welfare (grantno 2014-01382) and the Karolinska Institutet StockholmSweden Go to NeurologyorgN for full disclosures

Publication historyReceived by Neurology January 5 2018 Accepted in final formJuly 12 2018

References1 Wardlaw JM Smith EE Biessels GJ et al Neuroimaging standards for research into

small vessel disease and its contribution to ageing and neurodegeneration LancetNeurol 201312822ndash838

2 Qiu C Fratiglioni L A major role for cardiovascular burden in age-related cognitivedecline Nat Rev Cardiol 201512267ndash277

3 Ding J Sigurethsson S Jonsson PV et al Space and location of cerebral microbleedscognitive decline and dementia in the community Neurology 2017882089ndash2097

4 Ding J Sigurethsson S Jonsson PV et al Large perivascular spaces visible on magneticresonance imaging cerebral small vessel disease progression and risk of dementia the agegeneenvironment susceptibility-reykjavik study JAMA Neurol 2017741105ndash1112

5 Lopez OL Klunk WE Mathis C et al Amyloid neurodegeneration and small vesseldisease as predictors of dementia in the oldest-old Neurology 2014831804ndash1811

6 Staals J Booth T Morris Z et al Total MRI load of cerebral small vessel disease andcognitive ability in older people Neurobiol Aging 2015362806ndash2811

7 Xu X Hilal S Collinson SL et al Validation of the total cerebrovascular diseaseburden scale in a community sample J Alzheimers Dis 2016521021ndash1028

8 Erten-Lyons D Dodge HH Woltjer R et al Neuropathologic basis of age-associatedbrain atrophy JAMA Neurol 201370616ndash622

9 Kotrotsou A Schneider JA Bennett DA et al Neuropathologic correlates of regionalbrain volumes in a community cohort of older adults Neurobiol Aging 2015362798ndash2805

10 Kester MI Goos JD Teunissen CE et al Associations between cerebral small-vesseldisease and Alzheimer disease pathology as measured by cerebrospinal fluid bio-markers JAMA Neurol 201471855ndash862

11 Gregg NM Kim AE Gurol ME et al Incidental cerebral microbleeds and cerebralblood flow in elderly individuals JAMA Neurol 2015721021ndash1028

12 Raz L Knoefel J Bhaskar K The neuropathology and cerebrovascular mechanisms ofdementia J Cereb Blood Flow Metab 201636172ndash186

13 Welmer AK Rizzuto D Qiu C Caracciolo B Laukka EJ Walking speed processingspeed and dementia a population-based longitudinal study J Gerontol A Biol SciMed Sci 2014691503ndash1510

14 Zhang Y Qiu C Lindberg O et al Acceleration of hippocampal atrophy in a non-demented elderly population the SNAC-K study Int Psychogeriatr 20102214ndash25

15 Wang R Fratiglioni L Kalpouzos G et al Mixed brain lesions mediate the associationbetween cardiovascular risk burden and cognitive decline in old age a population-based study Alzheimers Dement 201713247ndash256

16 Kohncke Y Laukka EJ Brehmer Y et al Three-year changes in leisure activities areassociated with concurrent changes in white matter microstructure and perceptualspeed in individuals aged 80 years and older Neurobiol Aging 201641173ndash186

17 Gerritsen L Kalpouzos G Westman E et al The influence of negative life events onhippocampal and amygdala volumes in old age a life-course perspective Psychol Med2015451219ndash1228

e1496 Neurology | Volume 91 Number 16 | October 16 2018 NeurologyorgN

18 Laveskog A Wang R Bronge L Wahlund LO Qiu C Perivascular spaces in old ageassessment distribution and correlation with white matter hyperintensities AJNRAm J Neuroradiol 20183970ndash76

19 Young VG Halliday GM Kril JJ Neuropathologic correlates of white matterhyperintensities Neurology 200871804ndash811

20 Jagust WJ Zheng L Harvey DJ et al Neuropathological basis of magnetic resonanceimages in aging and dementia Ann Neurol 20086372ndash80

21 Fratiglioni L Grut M Forsell Y Viitanen M Winblad B Clinical diagnosis of Alz-heimerrsquos disease and other dementias in a population survey Agreement and causes ofdisagreement in applying Diagnostic and Statistical Manual of Mental DisordersRevised Third Edition Criteria Arch Neurol 199249927ndash932

22 Wang R Fratiglioni L Laukka EJ et al Effects of vascular risk factors and APOE e4 onwhite matter integrity and cognitive decline Neurology 2015841128ndash1135

23 Potter GM Doubal FN Jackson CA et al Enlarged perivascular spaces and cerebralsmall vessel disease Int J Stroke 201510376ndash381

24 Apostolova LG Zarow C Biado K et al Relationship between hippocampal at-rophy and neuropathology markers a 7TMRI validation study of the EADC-ADNIHarmonized Hippocampal Segmentation Protocol Alzheimers Dement 201511139ndash150

25 Dong C Nabizadeh N Caunca M et al Cognitive correlates of white matter lesionload and brain atrophy the Northern Manhattan Study Neurology 201585441ndash449

26 Akinyemi RO Mukaetova-Ladinska EB Attems J Ihara M Kalaria RN Vascular riskfactors and neurodegeneration in ageing related dementias Alzheimerrsquos disease andvascular dementia Curr Alzheimer Res 201310642ndash653

27 Santos CY Snyder PJ Wu WC Zhang M Echeverria A Alber J Pathophysi-ologic relationship between Alzheimerrsquos disease cerebrovascular disease andcardiovascular risk a review and synthesis Alzheimers Dement (Amst) 2017769ndash87

28 Mattsson N Tosun D Insel PS et al Association of brain amyloid-β with cerebralperfusion and structure in Alzheimerrsquos disease and mild cognitive impairment Brain2014137(pt 5)1550ndash1561

29 Attems J Jellinger KA The overlap between vascular disease and Alzheimerrsquos diseaselessons from pathology BMC Med 201412206

30 Qiu C Sigurdsson S Zhang Q et al Diabetes markers of brain pathology andcognitive function the age geneenvironment susceptibility-Reykjavik Study AnnNeurol 201475138ndash146

31 Kawas CH Kim RC Sonnen JA et al Multiple pathologies are common and relatedto dementia in the oldest-old the 90+ Study Neurology 201585535ndash542

32 White LR Edland SD Hemmy LS et al Neuropathologic comorbidity and cognitiveimpairment in the nun and honolulu-Asia aging studies Neurology 2016861000ndash1008

33 Nicoll JA Savva GM Stewart J et al Association between APOE genotype neuro-pathology and dementia in the older population of England and Wales NeuropatholAppl Neurobiol 201137285ndash294

34 JackCR JrWiste HJWeigand SD et al Age sex and APOE e4 effects onmemory brainstructure and β-amyloid across the adult life span JAMA Neurol 201572511ndash519

35 Boyle PA Yang J Yu L et al Varied effects of age-related neuropathologies on thetrajectory of late life cognitive decline Brain 2017140804ndash812

36 Chui HC Zarow C Mack WJ et al Cognitive impact of subcortical vascular andAlzheimerrsquos disease pathology Ann Neurol 200660677ndash687

37 Ye BS Seo SW Kim JH et al Effects of amyloid and vascular markers on cognitivedecline in subcortical vascular dementia Neurology 2015851687ndash1693

38 Arba F Quinn T Hankey GJ Ali M Lees KR Inzitari D Cerebral small vessel diseasemedial temporal lobe atrophy and cognitive status in patients with ischaemic strokeand transient ischaemic attack Eur J Neurol 201724276ndash282

39 Racine AM Koscik RL Berman SE et al Biomarker clusters are differentially asso-ciated with longitudinal cognitive decline in late midlife Brain 20161392261ndash2274

NeurologyorgN Neurology | Volume 91 Number 16 | October 16 2018 e1497

DOI 101212WNL0000000000006355201891e1487-e1497 Published Online before print September 19 2018Neurology

Rui Wang Anna Laveskog Erika J Laukka et al and dementia

MRI load of cerebral microvascular lesions and neurodegeneration cognitive decline

This information is current as of September 19 2018

ServicesUpdated Information amp

httpnneurologyorgcontent9116e1487fullincluding high resolution figures can be found at

References httpnneurologyorgcontent9116e1487fullref-list-1

This article cites 39 articles 9 of which you can access for free at

Citations httpnneurologyorgcontent9116e1487fullotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionmriMRI

httpnneurologyorgcgicollectioncohort_studiesCohort studies

httpnneurologyorgcgicollectionalzheimers_diseaseAlzheimers disease

httpnneurologyorgcgicollectionall_cognitive_disorders_dementiaAll Cognitive DisordersDementiafollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpwwwneurologyorgaboutabout_the_journalpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnneurologyorgsubscribersadvertiseInformation about ordering reprints can be found online

ISSN 0028-3878 Online ISSN 1526-632XWolters Kluwer Health Inc on behalf of the American Academy of Neurology All rights reserved Print1951 it is now a weekly with 48 issues per year Copyright Copyright copy 2018 The Author(s) Published by

reg is the official journal of the American Academy of Neurology Published continuously sinceNeurology