REVIEW

Vascular Risk Factors and Depression in Later Life:A Systematic Review and Meta-Analysis

Vyara Valkanova and Klaus P. Ebmeier

Reports of the association between cardiovascular risk factors and depression in later life are inconsistent; to establish the nature of theirassociation seems important for prevention and treatment of late-life depression. We searched MEDLINE, EMBASE, and PsycINFO forrelevant cohort or case control studies over the last 22 years; 1097 were retrieved; 26 met inclusion criteria. Separate meta-analyses wereperformed for Risk Factor Composite Scores (RFCS) combining different subsets of risk factors, Framingham Stroke Risk Score, and singlefactors. We found a positive association (odds ratio [OR]: 1.49; 95% confidence interval [CI]: 1.27–1.75) between RFCS and late-lifedepression. There was no association between Framingham Stroke Risk Score (OR: 1.25; 95% CI: .99–1.57), hypertension (OR: 1.14; 95%CI: .94–1.40), or dyslipidemia (OR: 1.08; 95% CI: .91–1.28) and late-life depression. The association with smoking was weak (OR: 1.35; 95%CI: 1.00–1.81), whereas positive associations were found with diabetes (OR: 1.51; 95% CI: 1.30–1.76), cardiovascular disease (OR: 1.76;95% CI: 1.52–2.04), and stroke (OR: 2.11; 95% CI: 1.61–2.77). Moderate to high heterogeneity was found in the results for RFCS, smoking,hypertension, dyslipidemia, and stroke, whereas publication bias was detected for RFCS and diabetes. We therefore found convincingevidence of a strong relationship between key diseases and depression (cardiovascular disease, diabetes, and stroke) and betweencomposite vascular risk and depression but not between some vascular risk factors (hypertension, smoking, dyslipidemia) anddepression. More evidence is needed to be accumulated from large longitudinal epidemiological studies, particularly if complementedby neuroimaging.

Key Words: Cardiovascular, major depressive disorder, meta-analysis, old age, risk, stroke

Late-life depression (LLD) and particularly late-onset depres-sion (LOD) have been conceptualized as distinct fromdepression with early onset (EOD) (1–4). Compared with

EOD, LOD is more often associated with no family history ofdepression and depressive ideation but more psychomotorretardation (5,6), cognitive impairment (especially executivedysfunction [7–9]), lack of insight, poor response to treatment(1), and a greater chance of progression to dementia (10,11). Inaddition, magnetic resonance imaging studies have demon-strated higher rates and greater severity of white matterhyperintensities in LOD compared with healthy volunteers (12–15)and with EOD patients (15,16).

The differences between early and late-life depression mightbe due to different underlying pathophysiological mechanisms(17). The term vascular (or subcortical ischemic) depressionpostulates a link between cerebrovascular disease and later lifedepression (18–20). It implies that micro-damage to small vesselscompromises the integrity of the frontal-subcortical circuitsinvolved in mood regulation (6,16,21–25). The vascular depres-sion hypothesis can explain increased risk of depression afterstroke and myocardial infarction (26–28) and the association ofLLD with brain scans suggestive of subclinical cerebrovasculardisease (12,21,23). However, studies of common cardiovascularrisk factors—such as smoking, hypertension, dyslipidemia anddiabetes, and depression—have yielded mixed results. Somestudies provide support for an association (29–35), whereasothers fail to do so (36–46). There is also strong evidence for a

From the Department of Psychiatry, University of Oxford, Oxford, United

Kingdom.

Address correspondence to Klaus P. Ebmeier, M.D., Department of

Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3

7JX, United Kingdom; E-mail: klaus.ebmeier@psych.ox.ac.uk.

Received Jul 6, 2012; revised Oct 22, 2012; accepted Oct 31, 2012.

0006-3223/$36.00http://dx.doi.org/10.1016/j.biopsych.2012.10.028

reciprocal relationship (47–52). Recent meta-analyses report thatdepression predicts incident myocardial infarction and earlierdeath, coronary artery disease, stroke, other cardiovasculardiseases (CVDs), and diabetes; apart from common causes ofboth CVD and depression, potential mechanisms for depressioncausing CVD span the depressive stress response, lifestyle factorssuch as exercise and food intake, as well as aspects of thetreatments used (47,53).

The importance of vascular risks and diseases precedingdepression might not be greater than that of other chronicdiseases. Vascular diseases might be associated with depression,not because of associated pathology (i.e., small or large brainvessel disease) but because of their effect on function and theresulting poor quality of life. Consistent with the chronic illnesshypothesis, the relationship between vascular risks or diseasesand depression was significantly attenuated after controlling forpresence of chronic illness (37,54,55), although attribution ofvariability to one (chronic illness) or the other (vascular risk anddisease) will be arbitrary or at least uncertain in most cases.Depression seems associated with poor general health (56),chronic obstructive pulmonary disease (27), chronic renal disease(57), arthritis (27), and loss of hearing or vision (27). Furthersupport for a causal relationship between general chronicillness and depression is provided by a recent prospectivecohort study that found an equally strong association betweenlong-term nonvascular conditions and risk of depressivesymptoms (46).

Although several systematic reviews have focused on thevascular depression hypothesis (13,20,58–62), the relationshipbetween vascular risk factors (VRFs) or vascular diseases anddepression has not been quantified. This systematic review andmeta-analysis aims to provide an overview of the literature todate, to quantify the extent to which VRFs or vascular diseasesmight be associated with or might be risk factors for depressionin late life, and to consider the contribution of the associateddisability. If vascular risks and pathological changes are etiologi-cal factors for depression, we expect to find significant associa-tions even after controlling for the complex effects of chronicillness and disability. Establishing the nature of the relationship

BIOL PSYCHIATRY 2013;73:406–413& 2013 Society of Biological Psychiatry

V. Valkanova and K.P. Ebmeier BIOL PSYCHIATRY 2013;73:406–413 407

between VRFs or diseases and LLD is important both in terms ofprevention and treatment of depression.

Methods and Materials

Search StrategyWe systematically searched for studies that investigated the

association between VRFs and depression in late life. Studiesconsidering vascular diseases such as coronary heart diseasetogether with VRFs were included, because it is a commonfeature of risk scales to include previous disease. The MEDLINE,EMBASE, and PsycINFO databases were searched for publicationsin all languages between 1990 and May 2012. The search termswere: [“depress*”] AND [“late life” OR “late onset” OR “olderadults” OR “geriatric”]; and second: [“depress*”] AND [“vasculardiseases” OR “vascular risk factors” OR “cerebrovascular riskfactors” OR “vascular”]. Additional studies were identified fromreference lists of relevant reviews and studies. Unpublishedliterature was identified from the DART Europe E-thesis Portal(dissertations and thesis), ZeTOC (conference proceedings), andOpen Grey (Grey Literature) databases. A total of 1097 resultswere retrieved. After screening of titles and abstracts 140 studieswere considered potentially relevant. The inclusion criteria were:cohort or case control studies, age ‡50 years, and frequencyor new cases of depression reported with and without VRFs,respectively. After review of the full text, 26 studies met theinclusion criteria. Common reasons for exclusion were reviewarticles, dual publications, or insufficient data to calculate out-come measures. Further reasons for exclusion were “exposure tovascular risk factors not reported” and “depression not reportedas an outcome” (Figure 1). Where there was an overlap insamples between studies, the study of higher quality or theone providing stronger evidence was included (e.g., moreparticipants, longitudinal design) (Figure 1).

The quality of the studies was assessed by scoring on a self-devised checklist (Table S1 in Supplement 1) that included thefollowing parameters: sample representativeness, study design,quality of reporting, VRFs measurement, outcome measurement,

Figure 1. Identification and attrition of studies. VRF, vascular risk factor.

and confounding factors (Supplement 1). Following the recom-mendations of the Meta-Analysis of Observational Studies inEpidemiology guidelines, we performed a sensitivity analysisexcluding studies with a score below 8 (Table 1).

Depression in studies was defined as: 1) diagnosis of majordepression, minor depression, or dysthymia according to the DSM-III R, DSM-IV, or other standard psychiatric diagnostic criteria; 2)depressive disorder or depressive symptoms, as defined by scoresabove a cutoff point on a standard mood rating scale (Centers forEpidemiologic Studies Depression Scale, Hamilton Rating Scale forDepression, or Geriatric Depression Scale). Of the studies included,three did not use these criteria. In two studies depressivesymptoms were identified through a single question from aquestionnaire (31,33), and one study used data recorded bygeneral practitioners in problem lists of patients (44).

Data ExtractionData were extracted in a systematic fashion as follows:

1) study characteristics (name, authors, publication year); 2) studydesign; 3) sample source; 4) sample characteristics (e.g., age, gender);5) inclusion and exclusion criteria; 6) definition and measures ofexposure; 7) definition and measures of outcome; and 8) analysisstrategy (statistical models, measures of effect size, confoundersthat were controlled). Data were extracted independently by bothauthors, and inconsistencies were resolved by consent.

Data AnalysisA meta-analysis was performed for studies that use a compo-

site measure of vascular risk (Risk Factor Composite Score [RFCS]).The RFCSs included different subsets of risk factors, and differentstudies used different RFCS groupings (e.g., two, three, or fourgroups; the low-risk group in some studies comprised participantswithout risk factors, whereas in other studies it included partici-pants with one risk factor). To make studies comparable, the datawere organized into two categories representing low vascular risk(0 or 1 risk factor) and high vascular risk (2 or more risk factors).A separate analysis was performed for studies using the Framing-ham Stroke Risk Score (FSRS), because it has been specificallydeveloped for assessing the risk of cerebrovascular disease(especially stroke). The FSRS is also well-validated and widelyused (63,64), although whether it predicts incident depression inlater life is not known. Most importantly, the studies using theFSRS used the same subset and definition of risk factors, thusincreasing the reliability of the results. Separate meta-analyseswere also conducted for the single factors smoking, hypertension,diabetes, dyslipidemia, CVD, and stroke.

Data were analyzed with Comprehensive Meta-Analysis, version2.2 (65). First, odds ratios (ORs) with confidence intervals (CIs) wereextracted or calculated from the available data (e.g., percentages orfractions, w2 with 1 df). When it was not possible to compute ORdirectly, standardized mean differences (Cohen’s d) were computedfrom means and SDs or from regression coefficient and transformedto ORs with conventional formulae (66). A random-effects model wasused to calculate the pooled mean effect size. The random-effectsmodel was preferred over a fixed effect model, because the includedstudies are heterogeneous in terms of population characteristics,definition and measurement of vascular risk, and outcomes (implicat-ing that the true effect size varies from one study to another) andalso to allow generalization of the results (67). Heterogeneity acrossstudies was assessed with the Cochrane Q statistic (p � .10 wasconsidered to indicate statistically significant heterogeneity) and theI2 statistic (25%, 50%, and 75% were considered to represent low,medium, and high heterogeneity, respectively). Publication bias was

www.sobp.org/journal

Table 1. Weighted Mean Effect Sizes and Measures of Heterogeneity and Bias

Vascular

Risk

Factor

Number

of Studies Odds Ratio

95% CI

(lower and

upper limits) p

Heterogeneity

Begg & Mazumdar

Rank Correlation

Egger’s Regression

Intercept

Q p I2 Kendall‘s t p (2-tail) t df p (2-tail)

D 10 1.08 .91–1.28 .40 15.3 .08 41.0 .07 .79 .37 8 .72

HTN 14 1.14 .94–1.40 .19 32.7 .002 60.3 �.21 .30 .80 12 .44

Smoking 10 1.35 1.00–1.81 .05 27.1 .001 66.8 .16 .53 .20 8 .85

DM 15 1.51 1.30–1.76 �.0005 18.7 .18 25.0 �.42 .03 2.07 13 .06

DMa 5 1.46 1.14–1.86 .003 6.3 .18 36.5 .20 .62 .82 3 .47

CVD 10 1.76 1.52–2.04 �.0005 12.1 .21 25.7 �.16 .53 .27 8 .80

CVDa 6 1.40 1.08–1.80 .01 10.6 .06 53.0 .07 .85 .09 4 .93

Stroke 10 2.11 1.61–2.77 �.0005 21.9 .01 58.9 �.11 .65 .40 8 .70

Strokea 5 1.80 1.24–2.62 .002 7.3 .12 45.4 .4 .33 2.10 3 .13

RFCS 18 1.49 1.27–1.75 �.0005 71.3 �.0005 76.2 �.10 .57 3.94 16 .001

RFCSa 8 1.15 1.02–1.28 .02 10.2 .18 31.2 .07 .80 1.10 6 .32

FSRS 5 1.25 .99–1.57 .06 7.2 .12 44.6 .2 .62 1.98 3 .14

CI, confidence interval; CVD, cardiovascular disease; D, dyslipidemia; DM, diabetes mellitus; HTN, hypertension; FSRS, Framingham Stroke Risk Score;RFCS, Risk Factor Composite Score.

aAdjusted for chronic illness/disability.

408 BIOL PSYCHIATRY 2013;73:406–413 V. Valkanova and K.P. Ebmeier

assessed with funnel plots with the Duval and Tweedie trim-and-fillmethod, Begg and Mazumdar’s rank correlations, and Egger’sregression intercept test (65). Finally, sensitivity and subgroupanalyses were performed to estimate the effect of study andparticipant characteristics on the results, such as mean age, studydesign (cross sectional vs. longitudinal), source of sample (communityvs. hospital), measure of vascular risk (self-report vs. clinical examina-tion), definition of vascular risk (patient with stroke included orexcluded), and outcome measure (rating scale vs. clinical interview). Inaddition, to assess the influence of chronic illness and disability onthe relationship between VRFs and LLD, subgroup analyses includingstudies that controlled for these factors were performed for diabetes,CVD, stroke, and RFCS.

Results

The literature search identified 1097 studies. Twenty-six studiesmet the inclusion criteria (29–33,35,37–39,41,44,46,54,55,68–79). Theirbaseline characteristics are summarized in Tables S2 (20 cross-sectional studies) and S3 (6 longitudinal studies) in Supplement 1.

Composite Measure of Vascular RiskThe RFCS were available from 18 studies, yielding a total

sample of 17,899 participants (29–33,35,37,39,41,44,46,54,68,69,71,75,77,79). The random-model pooled OR showed that theodds of LLD are 1.49� greater in participants with high vascularrisk compared with participants with low vascular risk (95% CI:1.27–1.75; p � .0005) (Figure S1 in Supplement 1). The studieswere highly heterogeneous (Q ¼ 71.3; p � .0005; I2 ¼ 76.2),suggesting that the variability of effect sizes is caused not bysampling error but by systematic differences between studies.Begg and Mazumdar rank correlation showed no significantpublication bias (Kendall’s t ¼ �.10, two-tailed p ¼ .57), butthe more sensitive Egger’s regression intercept detected bias(t(16) ¼ 3.94; two-tailed p � .0005), and analysis with Duval &Tweedie’s trim-and-fill method (65) trimmed eight studies andresulted in a lower but still significant pooled OR (OR: 1.10; 95%CI: 1.05 to 1.16) (Table 1; Figure S2 in Supplement 1).

The pooled effect sizes in subgroups defined by differentstudy characteristics are summarized in Table 2. Generally, higher

www.sobp.org/journal

vascular risk was associated with greater odds of depression in allsubgroups, particularly in studies with cross-sectional design(OR: 1.54; 95% CI: 1.27–1.87; p � .0005), noncommunity samples(OR: 1.60; 95% CI: 1.20–2.12; p ¼ .001), self-report of vascular risk(OR: 1.75; 95% CI: 1.50–2.04; p � .0005), clinical diagnosis ofdepression (OR: 1.63; 95% CI: 1.19–2.25; p ¼ .003), and studiesthat excluded patients with stroke (OR: 1.53; 95% CI: 1.21–1.93;p � .0005). When the analysis was confined to the five studiesthat used FSRS, the mean effect size decreased greatly, resultingin a nonsignificant relationship between FSRS and LLD (OR: 1.25;95% CI: .99–1.57; p ¼ .06). The strength of the associationbetween vascular risk and LLD weakened when only studies thatadjusted for sociodemographic variables were considered (OR:1.37; 95% CI: 1.15–1.63; p � .0005). It was further attenuatedwhen the analysis was confined to the eight studies thatcontrolled for chronic illness but remained statistically significant(OR: 1.15; 95% CI: 1.02–1.28; p ¼ .02). Sensitivity analysis limitedto the studies with quality score ‡8 (30,35,37,39,41,46,54,71,75,77,79) also showed an attenuated but significant association(OR: 1.34; 95% CI: 1.12–1.16; p � .0005).

Individual VRFs/Vascular DiseasesSmoking. Ten studies reported data on the association

between smoking and LLD (30,35,38,39,55,74–77,79). In totalthere were 20,120 participants. The pooled random-model ORwas 1.35 (95% CI: 1.00–1.81; p = .05). Studies were significantlyheterogeneous (Q = 27.1; p = .001; I2 = 66.8) (Table 1).

Hypertension. Fourteen studies compared the prevalence orincidence of LLD between individuals with and without hyper-tension (30,35,38,39,44,70,72–79). This yielded a total sampleof 20,197. After pooling these 14 studies, the random-modelOR was � 1 but statistically nonsignificant (OR: 1.14; 95% CI:.94–1.40; p = .19). The effects were moderately heterogeneous(Q = 32.7; p = .002; I2 = 60.3) (Table 1).

Dyslipidemia. Data on the association between dyslipidemiaand LLD were available from 10 studies, with 17,957 participantsin total (30,39,55,72–75,77–79). After pooling these 10 studies,the random-model OR was 1.08 (95% CI: .91–1.28; p = .4).A moderate heterogeneity was found with an I2 = 41.0 (Q = 15.26;p = .08) (Table 1).

Table 2. Vascular Risk and Likelihood of Late-Life Depression in Different Subgroups

Studies Included n OR (95% CI) p

All Studies 18 1.49 (1.27–1.75) �.0005

Study Design

Longitudinal studies (30,35,37,46,75) 5 1.38 (1.06–1.79) .02

Cross-sectional studies (29,31–33,39,41,44,54,68,69,71,77,79) 13 1.54 (1.27–1.87) �.0005

Age: Studies that Include Persons ‡65 yrs (30–32,35,39,46,54,69,71) 9 1.47 (1.17–1.85) .001

Source of Sample

Studies that used community-based samples (30–33,35,39,41,46,68,71,75,79) 12 1.45 (1.19–1.75) �.0005

Studies that used non community-based samples (29,37,44,54,69,77) 6 1.60 (1.20–2.12) .001

Measures of Exposure

Self-report (31–33,35,68) 5 1.75 (1.50–2.04) �.0005

Examination and blood tests (29,30,37,39,41,44,46,54,69,71,75,77,79) 13 1.38 (1.15–1.65) �.0005

Measures of Outcome

Studies that measured outcome with rating scale:(30,31,33,35,39,41,44,46,69,71) 10 1.42 (1.18–1.71) �.0005

Studies that measured outcome with clinical interview (29,32,37,54,68,75,77,79) 8 1.63 (1.19–2.25) .003

Studies that Controlled for SDV (30,31,35,37,39,41,46,54,71,75,77,79) 12 1.37 (1.15–1.63) �.0005

Studies that Controlled for Chronic Illness/Disability (30,35,37,39,44,46,71,75) 8 1.15 (1.02–1.28) .02

Studies that Used FSRS (37,46,54,75,77) 5 1.25 (.99–1.57) .06

Studies that Exclude Patients with Stroke (29–31,33,39,44,46,69,77) 9 1.53 (1.21–1.93) �.0005

Odds ratios (ORs) with 95% confidence intervals (CIs).FSRS, Framingham Stroke Risk Score; SDV, sociodemographic variables.

V. Valkanova and K.P. Ebmeier BIOL PSYCHIATRY 2013;73:406–413 409

Diabetes mellitus. Fifteen studies, with a total sample of24,466 participants, reported data on prevalence or incidence ofLLD in individuals with and without diabetes (30,35,38,39,44,55,70,72–79). The pooled effect sizes revealed that the likelihoodof LLD was 1.51� greater in individuals diagnosed with diabetes(OR: 1.51; 95% CI: 1.30–1.76; p � .0005; random-effects model).The studies were homogeneous (Q = 18.7; p = .18; I2 = 25.0).When the analysis was confined to the five studies thatcontrolled for chronic illness (30,44,55,75,76), the overall ORwas 1.46 (95% CI: 1.14–1.86; p = .003) (Table 1).

CVD. Ten studies compared the prevalence or incidence ofLLD between individuals with and without CVD (30,35,44,55,70,73–76,78). In total there were 21,841 participants. The random-model pooled OR showed 1.76� greater odds of LLD in indivi-duals diagnosed with CVD (95% CI: 1.52–2.04; p � .0005). Theeffects were homogeneous (Q = 12.1, p = .21, I2 = 25.7). When theanalysis was confined to the six studies that control for chronicillness (30,44,55,73,75,76), the overall effect size decreased butremained statistically significant (OR: 1.40; 95% CI: 1.08–1.80;p = .01) (Table 1)

Stroke. Ten included studies compared the prevalence orincidence of LLD between individuals with and without stroke(35,38,39,41,55,69,71,73,74,78). The total sample consisted of16,221 participants. A diagnosis of stroke was associatedwith 2.11� greater likelihood of LLD (OR: 2.11; 95% CI:1.61–2.77; p � .0005, random-effects model). Measures of hetero-geneity were significant (Q = 21.9; p = .01; I2 = 58.9). Fivestudies controlled for chronic illness (38,39,55,71,73) with apooled random-model OR of 1.80 (95% CI: 1.24–2.62; p = .002)(Table 1).

Publication BiasThere was no evidence of publication bias except for the studies

examining the association between diabetes and LLD, in whichthere was significant bias as measured by Begg and Mazumdar rankcorrelation (Kendall’s t ¼ �.42; two-tailed p ¼ .03) and Egger’sregression intercept (t ¼ 2.07; two-tailed p ¼ .06) (Table 1).

Discussion

The central question of this review was whether VRFs aredirectly related to LLD or whether the observed relationshipis nonspecific. A significant association was found betweenthe composite measure of vascular risk and depression inlater life. The positive association also persisted and remainedstatistically significant across several subgroups stratified by studycharacteristics, such as study design, source of sample, measuresof exposure, and measures of outcome. These results support thevascular—or subcortical ischemic—hypothesis (18–20).

However, although the association remained significant, itsstrength was considerably attenuated when only studies control-ling for chronic illnesses were included. A recent review reportedan association between LLD and nonvascular as well as vascularchronic conditions (27). If nonspecific chronic illness can, to agreat extent, account for the association between vasculardisorders and LLD, this challenges the vascular hypothesis. Therelationship between chronic illness and vascular health iscomplex. Vascular disease might be only one of several patho-physiological steps leading to depression. Causes or conse-quences of many chronic nonbrain illnesses might affectbehavior and might promote neurohumoral (immune or endo-crine) responses, influencing vascular integrity and ultimately thecortico-limbic circuits implicated in depression. Patients with highmedical burden are more likely to have vascular disease, andtherefore controlling for overall medical burden might removea significant part of the impact of vascular disease.

We found a lack of association between the FSRS and LLD.In a risk score like the FSRS acute vascular diseases are less likelyto be considered. This implies that vascular risk without acutevascular disease might not precede depression and is in keepingwith the other results of this meta-analysis. Alternatively, thisnegative finding—in the context of a robust relationshipbetween CVD and depression—might simply suggest that theFSRS is not suitable for assessing the relationship betweenperipheral VRFs and depression, as opposed to stroke, and a

www.sobp.org/journal

410 BIOL PSYCHIATRY 2013;73:406–413 V. Valkanova and K.P. Ebmeier

better instrument is still needed. The use of the FSRS made itmore likely that risks were measured in a systematic way, andthus studies were more comparable, leading to a moderatevariability in effect size. This contrasts with the highly hetero-geneous results for the RFCS on the basis of different subsets ofrisk factors.

When individual risk factors were considered, no associationwas found between hypertension and LLD. Dyslipidemia was alsonot associated with LLD, whereas the effects for smoking justreached significance. Diabetes, heart disease, and stroke werestrongly associated with LLD, and the association was stillsignificant after adjustment for chronic illnesses. Therefore, themeta-analysis demonstrates a clear effect of current vasculardisease on the rates of depression, which is in agreement withearlier meta-analyses (26–28,80). The strong association withdiabetes, heart disease, and stroke is consistent with a vascularmechanism for LLD, but the weak association with smoking andthe lack of association with hypertension and dyslipidemia arenot supportive.

A number of explanations could account for the discrepantfindings. First, it is possible that certain risk factors are more closelylinked to LLD than others. However, all the aforementioned riskfactors are associated with small-vessel brain disease (81), whichhas been implicated in the pathogenesis of vascular depression(6,16,21–25). The negative finding for hypertension generatesparticular challenges for the vascular depression hypothesis.Hypertension is strongly related to small-vessel brain disease (82);has greater predictive power for atherothrombotic disease in thecerebrovascular territory compared with other risk factors (e.g.,smoking) (83); and increases the risk of white matter hyperinten-sities (60,84–87), which have been linked to LLD (12–16).

Second, LLD might result only from severe vascular pathology,which is typically associated with overt vascular disease. Thesignificant association found with the composite score whereparticipants with high risk (two or more risk factors) werecompared with participants with low risk (no or one risk factors)and the lack of association with some individual risk factorsimplies that the risk of depression increases when the vascularburden increases. The relationship between severity of vascularlesions and vascular symptoms, however, is not straightforward.Krishnan et al. (21) found that only 16% of those with severevascular lesions demonstrated vascular symptoms, whereas ina later study they found that participants with and withoutmagnetic resonance imaging-defined subcortical ischemic depressioncan have similar vascular scores (19). These findings suggest that therisk of depression might not be linearly related to the severity ofvascular pathology but that vascular changes are only associated withdepression, if they become severe enough to compromise organfunction. This might explain the absence of associations withhypertension, dyslipidemia, and the weak association with smokingas single factors. In addition to severity of lesions, the important factormight be lesion location. In this case, vascular pathology in thefrontal-subcortical circuits that are related to mood regulation wouldbe predictive of depression (23,88–90).

Third, risk-factor studies assume that brain changes correspondwith peripheral changes. Atherosclerosis, the main cause ofvascular disease, is a systemic disease that affects arteries simulta-neously in different vascular territories, but it might develop atdifferent rates of progression in different locations (83). Althoughthere is evidence that cerebral and extra-cerebral atherosclerosisare associated (91,92), the degree of cerebral atherosclerosis mightnot always correlate with the degree of extra-cerebral athero-sclerosis. Furthermore, much evidence supporting the vascular

www.sobp.org/journal

depression hypothesis comes from neuroimaging studies, in whichcerebral atherosclerosis is measured as the presence and severityof white matter lesions (12–15,21,90). Therefore, it is possible thatonly cerebral atherosclerosis increases the risk of depression in oldage. For instance, Lee et al. (93) found that white matterhyperintensities in patients with ischemic stroke were associatedwith intracranial rather than extra-cranial atherosclerosis, whereasin a large longitudinal study measures of extra-cerebral athero-sclerosis did not increase the risk of LLD (45).

Finally, it is possible that depression in later-life is a hetero-geneous condition, so patients with vascular depression repre-sent a small subgroup of elders with depressive disorders. Notonly can different neuropathological pathways lead to illness, butVRFs themselves are linked to depression through varyingpathophysiological mechanisms. There is evidence for the roleof structural and functional disconnection, chronic low-gradeinflammation, and hypoperfusion (94–97). These processesdevelop to a different degree in patients with one or more riskfactors and might explain the weakness of associations ofdepression with individual risk factors.

Methodological LimitationsDefinitional problems made comparison between studies diffi-

cult. There is disagreement about the age cutoff (50, 60, or 65years) to differentiate between early- and late-onset depression, soparticipants classified as LOD in one study would be consideredEOD in another. Furthermore, many studies did not control for ahistory of depression. Even though a substantial number ofparticipants with LLD might have a late age of onset, some ofthem might have suffered from a recurrent depressive episode. Theinterpretation of results was further complicated, because aspectsof each individual risk factor—such as duration, appropriatetreatment, and adherence to treatment—were not considered,which might result in measurement error. Not all studies controlledfor chronic illness, and some of the studies that did includedvascular disease within the overall medical burden, thus minimizingthe relationship between vascular disease and depression. Therewere inconsistencies between studies in the definitions of VRFs.Vascular risk in some studies was conceptualized as the presence ofone or a number of VRFs, such as smoking or hypertension,whereas others included in the RFCS definite vascular diseases,such as stroke. Moreover, VRFs were grouped together andregarded as having an equal impact on depression. However, otherstudies (42,73,75,76) and this meta-analysis found that the effects ofVRFs on depression were heterogeneous, challenging the use ofnonweighted composite scores to quantify risk. Because manystudies considered only a limited range of risk factors, undetectedvascular conditions in the low-risk group that would attenuate atrue relationship between vascular risk and LLD cannot beexcluded. Finally, there was moderate-to-high heterogeneity inthe effects for RFCS, smoking, hypertension, dyslipidemia, andstroke, which might be due to study design, sample size,participant characteristics, and definitions or measurement of riskfactors and LLD. In addition, publication bias was detected for RFCSand diabetes. The results of the meta- analysis therefore have to beinterpreted with caution.

Conclusions

We found convincing evidence of a strong relationshipbetween key diseases—such as CVD, diabetes, and stroke—andbetween composite vascular risk and depression but notbetween some VRFs (hypertension, smoking, dyslipidemia) and

V. Valkanova and K.P. Ebmeier BIOL PSYCHIATRY 2013;73:406–413 411

depression. More evidence needs to be accumulated from largelongitudinal epidemiological studies that consider biological(vascular and nonvascular), psychological, and social risk factorsas well as mediators on their relationship in the context ofmechanism-specific investigations. The main problem in theassociative studies is that they use VRFs as proxies for brainchanges, assuming that evidence of peripheral vascular changesis associated with brain changes, which might not necessarilybe true. Risk-factor studies should therefore be complemented byneuroimaging to examine brain processes more directly. Combin-ing these two approaches might help to identify pathogeneticallydistinct pathways to LLD, which is crucial in terms of earlyintervention and specific treatment.

KPE received grant and other support from the Medical ResearchCouncil (United Kingdom), Gordon Edward Small Charitable Trust,Norman Collisson Foundation, HDH Wills 1965 Charitable Trust, andNational Institute for Health Research (England).

The authors report no biomedical financial interests or potentialconflicts of interest.

Supplementary material cited in this article is available online.

1. Alexopoulos GS, Schultz SK, Lebowitz BD (2005): Late-life depression:A model for medical classification. Biol Psychiatry 58:283–289.

2. Hickie IB, Naismith SL, Norrie LM, Scott EM (2009): Managingdepression across the life cycle: New strategies for clinicians andtheir patients. Intern Med J 39:720–727.

3. Thomas AJ, Gallagher P, Robinson LJ, Porter RJ, Young AH, Ferrier IN,et al. (2009): A comparison of neurocognitive impairment in youngerand older adults with major depression. Psychol Med 39:725–733.

4. Alexopoulos GS, Kiosses DN, Heo M, Murphy CF, Shanmugham B,Gunning-Dixon F (2005): Executive dysfunction and the course ofgeriatric depression. Biol Psychiatry 58:204–210.

5. Alexopoulos GS, Meyers BS, Young RC, Kakuma T, Silbersweig D,Charlson M (1997): Clinically defined vascular depression. Am JPsychiatry 154:562–565.

6. Alexopoulos GS, Kiosses DN, Klimstra S, Kalayam B, Bruce ML (2002):Clinical presentation of the “depression-executive dysfunction syndrome”of late life. Am J Geriatr Psychiatry 10:98–106.

7. Lockwood KA, Alexopoulos GS, van Gorp WG (2002): Executivedysfunction in geriatric depression. Am J Psychiatry 159:1119–1126.

8. Hickie IB, Naismith SL, Ward PB, Little CL, Pearson M, Scott EM, et al.(2007): Psychomotor slowing in older patients with major depression:Relationships with blood flow in the caudate nucleus and whitematter lesions. Psychiatry Res 155:211–220.

9. Herrmann LL, Goodwin GM, Ebmeier KP (2007): The cognitive neurop-sychology of depression in the elderly. Psychol Med 37:1693–1702.

10. Kohler S, Thomas AJ, Barnett NA, O‘Brien JT (2010): The pattern andcourse of cognitive impairment in late-life depression. Psychol Med40:591–602.

11. Saczynski JS, Beiser A, Seshadri S, Auerbach S, Wolf PA, Au R (2010):Depressive symptoms and risk of dementia: The Framingham HeartStudy. Neurology 75:35–41.

12. de Groot JC, de Leeuw FE, Oudkerk M, Hofman A, Jolles J, Breteler MM(2000): Cerebral white matter lesions and depressive symptoms inelderly adults. Arch Gen Psychiatry 57:1071–1076.

13. Baldwin RC (2005): Is vascular depression a distinct sub-type ofdepressive disorder? A review of causal evidence. Int J GeriatrPsychiatry 20:1–11.

14. Taylor WD, MacFall JR, Payne ME, McQuoid DR, Steffens DC,Provenzale JM, et al. (2005): Greater MRI lesion volumes in elderlydepressed subjects than in control subjects. Psychiatry Res 139:1–7.

15. Herrmann LL, Le Masurier M, Ebmeier KP (2008): White matterhyperintensities in late life depression: A systematic review. J NeurolNeurosurg Psychiatry 79:619–624.

16. O‘Brien J, Desmond P, Ames D, Schweitzer I, Harrigan S, Tress B(1996): A magnetic resonance imaging study of white matter lesionsin depression and Alzheimer‘s disease. Br J Psychiatry 168:477–485.

17. Seldenrijk A, van Hout HP, van Marwijk HW, de Groot E, Gort J,Rustemeijer C, et al. (2011): Carotid atherosclerosis in depression andanxiety: Associations for age of depression onset. World J BiolPsychiatry 12:549–558.

18. Alexopoulos GS, Meyers BS, Young RC, Campbell S, Silbersweig D,Charlson M (1997): ‘Vascular depression‘ hypothesis. Arch GenPsychiatry 54:915–922.

19. Krishnan KR, Taylor WD, McQuoid DR, MacFall JR, Payne ME,Provenzale JM, et al. (2004): Clinical characteristics of magneticresonance imaging-defined subcortical ischemic depression. BiolPsychiatry 55:390–397.

20. Culang-Reinlieb ME, Johnert LC, Brickman AM, Steffens DC, Garcon E,Sneed JR (2011): MRI-defined vascular depression: A review of theconstruct. Int J Geriatr Psychiatry 26:1101–1108.

21. Krishnan KR, Hays JC, Blazer DG (1997): MRI-defined vascular depres-sion. Am J Psychiatry 154:497–501.

22. Lyness JM, Caine ED, Cox C, King DA, Conwell Y, Olivares T (1998):Cerebrovascular risk factors and later-life major depression. Testing asmall-vessel brain disease model. Am J Geriatr Psychiatry 6:5–13.

23. Greenwald BS, Kramer-Ginsberg E, Krishnan KR, Ashtari M, Auerbach C,Patel M (1998): Neuroanatomic localization of magnetic resonanceimaging signal hyperintensities in geriatric depression. Stroke 29:613–617.

24. Wager-Smith K, Markou A (2011): Depression: A repair response tostress-induced neuronal microdamage that can grade into a chronicneuroinflammatory condition? Neurosci Biobehav Rev 35:742–764.

25. Steffens DC, Bosworth HB, Provenzale JM, MacFall JR (2002):Subcortical white matter lesions and functional impairment ingeriatric depression. Depress Anxiety 15:23–28.

26. Whyte EM, Mulsant BH (2002): Post stroke depression: Epidemiology,pathophysiology, and biological treatment. Biol Psychiatry 52:253–264.

27. Huang CQ, Dong BR, Lu ZC, Yue JR, Liu QX (2010): Chronic diseasesand risk for depression in old age: A meta-analysis of publishedliterature. Ageing Res Rev 9:131–141.

28. Robinson RG, Spalletta G (2010): Poststroke depression: A review. CanJ Psychiatry 55:341–349.

29. Hickie I, Scott E, Naismith S, Ward PB, Turner K, Parker G, et al. (2001):Late-onset depression: Genetic, vascular and clinical contributions.Psychol Med 31:1403–1412.

30. Mast BT, Miles T, Penninx BW, Yaffe K, Rosano C, Satterfield S, et al.(2008): Vascular disease and future risk of depressive symptomatol-ogy in older adults: Findings from the Health, Aging, and BodyComposition study. Biol Psychiatry 64:320–326.

31. Yochim B, Mast BT, Lichtenberg PA (2003): Cerebrovascular riskfactors and depressed mood in inner city older adults. ClinicalPsychologist 7:11–20.

32. Naarding P, Schoevers RA, Janzing JG, Jonker C, Koudstaal PJ,Beekman AT (2005): A study on symptom profiles of late-lifedepression: The influence of vascular, degenerative and inflammatoryrisk-indicators. J Affect Disord 88:155–162.

33. Yochim BP, Kerkar SP, Lichtenberg PA (2006): Cerebrovascular riskfactors, activity limitations, and depressed mood in African Americanolder adults. Psychol Aging 21:186–189.

34. Holley C, Murrell SA, Mast BT (2006): Psychosocial and vascular riskfactors for depression in the elderly. Am J Geriatr Psychiatry 14:84–90.

35. Zimmerman JA, Mast BT, Miles T, Markides KS (2009): Vascular risk anddepression in the Hispanic Established Population for the Epidemio-logic Study of the Elderly (EPESE). Int J Geriatr Psychiatry 24:409–416.

36. Lyness JM, Caine ED, King DA, Conwell Y, Cox C, Duberstein PR (1999):Cerebrovascular risk factors and depression in older primary carepatients: Testing a vascular brain disease model of depression. Am JGeriatr Psychiatry 7:252–258.

37. Lyness JM, King DA, Conwell Y, Cox C, Caine ED (2000): Cerebrovas-cular risk factors and 1-year depression outcome in older primary carepatients. Am J Psychiatry 157:1499–1501.

38. Stewart R, Prince M, Mann A, Richards M, Brayne C (2001): Stroke,vascular risk factors and depression: Cross-sectional study in a UKCaribbean-born population. Br J Psychiatry 178:23–28.

39. Kim JM, Stewart R, Shin IS, Yoon JS (2002): Previous stroke but notvascular risk factors are associated with depression in a cognitivelyimpaired older Korean population. Int J Geriatr Psychiatry 17:453–458.

www.sobp.org/journal

412 BIOL PSYCHIATRY 2013;73:406–413 V. Valkanova and K.P. Ebmeier

40. Kim JM, Stewart R, Shin IS, Yoon JS (2004): Vascular disease/risk andlate-life depression in a Korean community population. Br J Psychiatry185:102–107.

41. Licht-Strunk E, Bremmer MA, van Marwijk HW, Deeg DJ, Hoogendijk WJ,de Haan M, et al. (2004): Depression in older persons with versuswithout vascular disease in the open population: Similar depressivesymptom patterns, more disability. J Affect Disord 83:155–160.

42. Cervilla J, Prince M, Rabe-Hesketh S (2004): Vascular disease riskfactors as determinants of incident depressive symptoms: A prospec-tive community-based study. Psychol Med 34:635–641.

43. Naarding P, Tiemeier H, Breteler MM, Schoevers RA, Jonker C,Koudstaal PJ, et al. (2007): Clinically defined vascular depression inthe general population. Psychol Med 37:383–392.

44. Nuyen J, Spreeuwenberg PM, Beekman AT, Groenewegen PP, van denBos GA, Schellevis FG (2007): Cerebrovascular risk factors andsubsequent depression in older general practice patients. J AffectDisord 99:73–81.

45. Newson RS, Hek K, Luijendijk HJ, Hofman A, Witteman JC, Tiemeier H(2010): Atherosclerosis and incident depression in late life. Arch GenPsychiatry 67:1144–1151.

46. Kivimaki M, Shipley MJ, Allan CL, Sexton CE, Jokela M, Virtanen M,et al. (2012): Vascular risk status as a predictor of later-life depressivesymptoms: A cohort study. Biol Psychiatry 72:324–330.

47. Knol MJ, Twisk JW, Beekman AT, Heine RJ, Snoek FJ, Pouwer F (2006):Depression as a risk factor for the onset of type 2 diabetes mellitus. Ameta-analysis. Diabetologia 49:837–845.

48. Ariyo AA, Haan M, Tangen CM, Rutledge JC, Cushman M, Dobs A,Furberg CD (2000): Depressive symptoms and risks of coronaryheart disease and mortality in elderly Americans. CardiovascularHealth Study Collaborative Research Group. Circulation 102:1773–1779.

49. Larson SL, Owens PL, Ford D, Eaton W (2001): Depressive disorder,dysthymia, and risk of stroke: Thirteen-year follow-up from theBaltimore epidemiologic catchment area study. Stroke 32:1979–1983.

50. Carney RM, Freedland KE (2003): Depression, mortality, and medicalmorbidity in patients with coronary heart disease. Biol Psychiatry 54:241–247.

51. van den Akker M, Schuurman A, Metsemakers J, Buntinx F (2004): Isdepression related to subsequent diabetes mellitus? Acta PsychiatrScand 110:178–183.

52. Ford DE, Mead LA, Chang PP, Cooper-Patrick L, Wang NY, Klag MJ(1998): Depression is a risk factor for coronary artery disease in men:The precursors study. Arch Intern Med 158:1422–1426.

53. Van der Kooy K, van Hout H, Marwijk H, Marten H, Stehouwer C,Beekman A (2007): Depression and the risk for cardiovasculardiseases: Systematic review and meta analysis. Int J Geriatr Psychiatry22:613–626.

54. Sanders ML, Lyness JM, Eberly S, King DA, Caine ED (2006):Cerebrovascular risk factors, executive dysfunction, and depressionin older primary care patients. Am J Geriatr Psychiatry 14:145–152.

55. Stewart R, Hirani V (2010): General health status and vasculardisorders as correlates of late-life depressive symptoms in a nationalsurvey sample. Int J Geriatr Psychiatry 25:483–488.

56. Prince MJ, Harwood RH, Blizard RA, Thomas A, Mann AH (1997):Impairment, disability and handicap as risk factors for depression inold age. The Gospel Oak Project V. Psychol Med 27:311–321.

57. Hedayati SS, Yalamanchili V, Finkelstein FO (2012): A practicalapproach to the treatment of depression in patients with chronickidney disease and end-stage renal disease. Kidney Int 81:247–255.

58. Baldwin RC, O‘Brien J (2002): Vascular basis of late-onset depressivedisorder. Br J Psychiatry 180:157–160.

59. Camus V, Kraehenbuhl H, Preisig M, Bula CJ, Waeber G (2004): Geriatricdepression and vascular diseases: What are the links? J Affect Disord 81:1–16.

60. Kales HC, Maixner DF, Mellow AM (2005): Cerebrovascular disease andlate-life depression. Am J Geriatr Psychiatry 13:88–98.

61. Samaras N, Rossi G, Giannakopoulos P, Gold G (2010): Vasculardepression. An age related mood disorder. Eur Geriatr Medicine 1:220–225.

62. Sneed JR, Culang-Reinlieb ME (2011): The vascular depression hypoth-esis: An update. Am J Geriatr Psychiatry 19:99–103.

63. Wolf PA, D‘Agostino RB, Belanger AJ, Kannel WB (1991): Probability ofstroke: A risk profile from the Framingham Study. Stroke 22:312–318.

www.sobp.org/journal

64. D‘Agostino RB, Wolf PA, Belanger AJ, Kannel WB (1994): Stroke riskprofile: Adjustment for antihypertensive medication. The FraminghamStudy. Stroke 25:40–43.

65. Biostat WB (2006): Comprehensive Meta-Analysis 2.2.048. Englewood,New Jersey: Biostat.

66. Ellis P (2010): The Essential Guide to Effect Sizes: Statistical Power, Meta-Analysis, and the Interpretation of Research Results, 1st ed. Cambridge,UK: Cambridge University Press.

67. Borenstein M, Hedges L, Higgins J, Rothstein H (2009): Introduction toMeta-Analysis, 1st ed Cichester, UK: John Wiley and Sons.

68. Oldehinkel AJ, Ormel J, Brilman EI, van den Berg MD (2003): Psychosocialand vascular risk factors of depression in later life. J Affect Disord 74:237–246.

69. Mast BT, MacNeill SE, Lichtenberg PA (2004): Post-stroke andclinically-defined vascular depression in geriatric rehabilitationpatients. Am J Geriatr Psychiatry 12:84–92.

70. Taylor WD, McQuoid DR, Krishnan KR (2004): Medical comorbidity inlate-life depression. Int J Geriatr Psychiatry 19:935–943.

71. Vinkers DJ, Stek ML, van der Mast RC, de Craen AJ, Le Cessie S,Jolles J, et al. (2005): Generalized atherosclerosis, cognitive decline,and depressive symptoms in old age. Neurology 65:107–112.

72. Rainer MK, Mucke HA, Zehetmayer S, Krampla W, Kuselbauer T,Weissgram S, et al. (2006): Data from the VITA study do not supportthe concept of vascular depression. Am J Geriatr Psychiatry 14:531–537.

73. Kim JM, Stewart R, Kim SW, Yang SJ, Shin IS, Yoon JS (2006): Vascularrisk factors and incident late-life depression in a Korean population.Br J Psychiatry 189:26–30.

74. Almeida OP, Flicker L, Norman P, Hankey GJ, Vasikaran S, vanBockxmeer FM, et al. (2007): Association of cardiovascular risk factorsand disease with depression in later life. Am J Geriatr Psychiatry 15:506–513.

75. Luijendijk HJ, Stricker BH, Hofman A, Witteman JC, Tiemeier H (2008):Cerebrovascular risk factors and incident depression in community-dwelling elderly. Acta Psychiatr Scand 118:139–148.

76. Wu KY, Liu CY, Chau YL, Chang CM (2010): Transient ischemic attackand incidence of depression in old age: Evidence from a population-based analysis in Taiwan. Am J Geriatr Psychiatry 18:382–387.

77. Dalby RB, Chakravarty MM, Ahdidan J, Sorensen L, Frandsen J, JonsdottirKY, et al. (2010): Localization of white-matter lesions and effect ofvascular risk factors in late-onset major depression. Psychol Med 40:1389–1399.

78. Kim BS, Lee DH, Lee DW, Bae JN, Chang SM, Kim S, et al. (2011):The role of vascular risk factors in the development of DED syndromeamong an elderly community sample. Am J Geriatr Psychiatry 19:104–114.

79. Park JE, Lee JE (2011): Cardiovascular disease risk factors anddepression in Korean women: Results from the fourth KoreanNational Health and Nutrition Examination Survey. Psychiatry Res 190:232–239.

80. Nouwen A, Winkley K, Twisk J, Lloyd CE, Peyrot M, Ismail K, et al.(2010): Type 2 diabetes mellitus as a risk factor for the onset ofdepression: A systematic review and meta-analysis. Diabetologia 53:2480–2486.

81. Breteler MM, van Swieten JC, Bots ML, Grobbee DE, Claus JJ, van denHout JH, et al. (1994): Cerebral white matter lesions, vascular riskfactors, and cognitive function in a population-based study: TheRotterdam Study. Neurology 44:1246–1252.

82. Slavin H, McManus J, Stott DJ (2012): Small vessel cerebrovasculardisease in older adults. Rev Clin Gerontology 22:184–194.

83. Lahoz C, Mostaza JM (2007): [Atherosclerosis as a systemic disease].Rev Esp Cardiol 60:184–195.

84. Dufouil C, de Kersaint-Gilly A, Besancon V, Levy C, Auffray E,Brunnereau L, et al. (2001): Longitudinal study of blood pressureand white matter hyperintensities: The EVA MRI Cohort. Neurology 56:921–926.

85. van Dijk EJ, Breteler MM, Schmidt R, Berger K, Nilsson LG, Oudkerk M,et al. (2004): The association between blood pressure, hypertension,and cerebral white matter lesions: Cardiovascular determinants ofdementia study. Hypertension 44:625–630.

86. Firbank MJ, Wiseman RM, Burton EJ, Saxby BK, O‘Brien JT, Ford GA(2007): Brain atrophy and white matter hyperintensity change in older

V. Valkanova and K.P. Ebmeier BIOL PSYCHIATRY 2013;73:406–413 413

adults and relationship to blood pressure. Brain atrophy, WMHchange and blood pressure. J Neurol 254:713–721.

87. Hoptman MJ, Gunning-Dixon FM, Murphy CF, Ardekani BA, Hrabe J,Lim KO, et al. (2009): Blood pressure and white matter integrity ingeriatric depression. J Affect Disord 115:171–176.

88. MacFall JR, Payne ME, Provenzale JE, Krishnan KR (2001): Medial orbitalfrontal lesions in late-onset depression. Biol Psychiatry 49:803–806.

89. Taylor WD, MacFall JR, Steffens DC, Payne ME, Provenzale JM,Krishnan KR (2003): Localization of age-associated white matterhyperintensities in late-life depression. Prog NeuropsychopharmacolBiol Psychiatry 27:539–544.

90. Gunning-Dixon FM, Hoptman MJ, Lim KO, Murphy CF, Klimstra S,Latoussakis V, et al. (2008): Macromolecular white matter abnormal-ities in geriatric depression: A magnetization transfer imaging study.Am J Geriatr Psychiatry 16:255–262.

91. Ikeda K, Ohno H, Tamura M, Kashihara H, Anan K, Hosozawa K, et al.(2002): Cerebral atherosclerosis and coronary calcification. Stroke 33:2345–2346; author reply 2345–2346.

92. Rosano C, Naydeck B, Kuller LH, Longstreth WT Jr, Newman AB (2005):Coronary artery calcium: associations with brain magnetic resonance

imaging abnormalities and cognitive status. J Am Geriatr Soc 53:609–615.

93. Lee SJ, Kim JS, Chung SW, Kim BS, Ahn KJ, Lee KS (2011): White matterhyperintensities (WMH) are associated with intracranial atherosclero-sis rather than extracranial atherosclerosis. Arch Gerontol Geriatr 53:e129–e132.

94. Alexopoulos GS, Morimoto SS (2011): The inflammation hypothesis ingeriatric depression [published online ahead of print March 2]. Int JGeriatr Psychiatry.

95. Liao Y, Huang X, Wu Q, Yang C, Kuang W, Du M, et al. (2012): Isdepression a disconnection syndrome? Meta-analysis of diffusiontensor imaging studies in patients with MDD. J Psychiatry Neurosci 37:110180.

96. Colloby SJ, Vasudev A, O‘Brien JT, Firbank MJ, Parry SW, Thomas AJ(2011): Relationship of orthostatic blood pressure to white matterhyperintensities and subcortical volumes in late-life depression.Br J Psychiatry 199:404–410.

97. Naismith SL, Norrie LM, Mowszowski L, Hickie IB (2012): Theneurobiology of depression in later-life: Clinical, neuropsychological,neuroimaging and pathophysiological features. Prog Neurobiol 98:99–143.

www.sobp.org/journal