Tiemeier H, van Dijck W, Hofman A, Witteman JCM, Stijnen T, Breteler MMB. Relationship Between Atherosclerosis and Late-Life DepressionThe Rotterdam Study. Arch Gen Psychiatry. 2004;61(4):369-376. doi:10.1001/archpsyc.61.4.369
Depression in late life has been associated with vascular abnormalities.
Several studies have demonstrated that persons with brain infarcts are more
likely to have depressive disorders. Furthermore, depression is related to
the subsequent development of ischemic heart disease.
To investigate the relationship between atherosclerosis at different
locations and depression in the general population.
Cross-sectional population-based study.
The Rotterdam Study, a population-based cohort study.
In 4019 men and women 60 years and older, we assessed atherosclerosis
at different locations, including common carotid intima-media thickness, plaques
in the carotid arteries, the ankle-brachial blood pressure index, and aortic
atherosclerosis. An overall measure of extracoronary atherosclerosis was obtained
in 3747 persons by computing the principal component of these extracoronary
atherosclerosis measures. In a subgroup of 1986 persons, we additionally measured
Main Outcome Measure
All subjects were screened for depressive symptoms. Screen-positive
subjects had a psychiatric interview to diagnose depressive disorder.
More severe extracoronary atherosclerosis was associated with a higher
prevalence of depressive disorders. For every 1–standard deviation increase,
the prevalence increased by 30%. Furthermore, we found a strong relationship
of severe coronary and aortic calcifications with depressive disorders (odds
ratio, 3.89; 95% confidence interval, 1.55-9.77; and odds ratio, 2.00; 95%
confidence interval, 1.02-3.96, respectively).
Atherosclerosis and depression are associated in the elderly. This finding
is compatible with the vascular depression hypothesis. However, the cross-sectional
nature of the study does not allow causal inferences. In particular, earlier
depressive episodes may have contributed to the development of atherosclerosis.
Several lines of evidence suggest that there is a relationship betweenvascular factors and late-life depression. Both symptomatic and silent braininfarcts are associated with subsequent depression.1 Furthermore,subtle white-matter and deep gray-matter abnormalities were found more frequentlyin depressed than in nondepressed persons.2 Avascular depression hypothesis has been proposed. It postulates that structuralchanges in the brain due to atherosclerosis are of primary importance in late-lifedepression.3,4
On the other hand, depressive symptoms are also related to subsequentcardiovascular disease. Several prospective population-based studies of depressedpersons showed an increased risk of myocardial infarct.5- 8 Possibly,depression contributes to the development of atherosclerosis or increasesthe risk of cardiovascular events in individuals who developed atherosclerosisunrelated to depression. However, none of these studies included any physiologicalmeasurements such as assessment of atherosclerosis or neuroendocrine function.Thus, it remains unclear which pathophysiologic mechanisms could explain theobserved association between depression and incident myocardial infarct. Numerousclinical studies on the cardiovascular risk associated with depressive symptomsinvestigated the excess morbidity and mortality in patients with establishedcoronary heart disease.9 Most studies did notassess the degree of atherosclerosis but rather focused on other measureslike platelet function or heart rate variability.10- 12 Epidemiologicevidence of an association between atherosclerosis and depression in the generalpopulation is lacking. Moreover, the vascular depression hypothesis is mainlysupported by neuroimaging studies, which demonstrated hyperintensities. Althoughassumed to be vascular, the pathological basis of the hyperintensities inpeople with depression has not been determined.13
We examined the association between atherosclerosis at different locationsin the body and depression in the Rotterdam Study, a large, prospective, community-basedstudy. The aim of the Rotterdam Study is to investigate determinants of chronicand disabling diseases. The study started in 1990, when all inhabitants 55years or older of Ommoord, a district of Rotterdam, the Netherlands, wereinvited to participate. A total of 7983 men and women entered the study (78%of those eligible). A first follow-up survey was done in 1993 to 1994, anda second follow-up survey in 1997 to 1999. In addition to follow-up surveys,the total cohort is continuously being monitored for major morbidity and mortalitythrough linkage of general practitioner and municipality records.
The study was conducted as part of the second follow-up examination(1997 to 1999) in the Rotterdam Study. The Medical Ethics Committee of ErasmusUniversity approved the study, and written informed consent was obtained fromall participants. In this survey, we added assessment of depressive symptomsto the study protocol. Of the 5901 subjects invited, 4730 persons (responserate, 80.2%) participated in the home interview. Of these, 4019 came to theresearch center for noninvasive assessments of atherosclerosis. The 711 subjectswith no assessment of atherosclerosis were on average older (78 vs 72 years),were more likely to be female (69.4% vs 57.9%), and had more depressive symptoms(11.8% vs 6.9%; overall prevalence, 7.8%).
In addition, noninstitutionalized participants younger than 85 yearswho had completed the third survey were invited for a coronary atherosclerosisscan at a second research site. Of the 3371 eligible subjects, 2263 agreedto undergo electron-beam computed tomography (CT). Participation rate (67.1%)was only moderate, because the examination was performed in another part ofthe city and scheduled as an additional visit. Nonparticipants were more likelyto be female (62.9% vs 53.3%) and to be depressed (9.0% vs 5.8%). Becauseof archiving problems, the scores of 277 participants were lost and electron-beamCT scores of only 1986 persons were available for analysis.
Depressive disorders were assessed by a 2-step procedure. First, participantscompleted the Dutch version of the original Center for Epidemiological StudiesDepression Scale (CES-D) during the home interview. The CES-D is a 20-itemself-reported measure of symptoms scored on a scale of 0 to 3 points.14 We used a score of 16 as a cutoff to indicate depressivesymptoms. This cutoff had a very high sensitivity for major depression ina random sample of older subjects in the Netherlands.15 Previousstudies have verified that a score of 16 and above on the CES-D representsclinically significant depressive symptoms.16 Ina second step, screen-positive subjects had a psychiatric workup. They werestudied with the Dutch version of the Present State Examination, a semistructuredpsychiatric interview included in the Schedules for Clinical Assessment inNeuropsychiatry.17 All interviews were conductedby one of 2 experienced clinicians, a psychiatrist and a clinical psychologist.Psychiatric disorders were classified according to the DSM-IV criteria with an algorithm based on the Present State Examinationscores. The diagnostic criteria include minor depression as defined in theappendix of the DSM-IV.
Of the 4019 subjects included in the analyses, 285 (7.1%) screened positivefor depressive symptoms as measured by the CES-D. Demented persons with depressivesymptoms were excluded (n = 10). A psychiatric workup was performed in 259(94.2%) of the 275 remaining participants. Twelve subjects refused to participatein this evaluation; 4 screen-positive subjects could not be reached. A depressivedisorder as defined by the DSM-IV criteria was establishedin 119 cases (56 persons with major depression, 47 with minor depression,and 16 with dysthymia). The remaining subjects either were classified as havinganxiety disorders or other psychiatric diseases (n = 22) or did not meet thecriteria for an Axis I psychiatric disorder (n = 118, subthreshold depressivesymptoms).
To define late-onset depression, we used data from the baseline interviewwith a physician. All persons in the present analyses responded to the questionsabout psychiatric history. Participants who reported a history of depressionbefore age 55 years were considered to have had early-onset depression.
We measured atherosclerosis noninvasively with 4 established methods,ie, the ankle-brachial blood pressure index, intima-media thickness in thecommon carotid arteries, the presence of plaques in the carotid arteries,and aortic atherosclerosis. These 4 measures assess extracoronary atherosclerosisat different locations in the body. Furthermore, in a subgroup of participants,we measured coronary atherosclerosis. All measures are strongly associatedwith incident cerebrovascular and coronary artery disease.18- 21
Ankle-brachial blood pressure index is an indicator of peripheral arterialdisease.22 Systolic blood pressure at the rightbrachial artery was calculated as the mean of 2 consecutive measurements witha random-zero sphygmomanometer. A single systolic blood pressure reading wastaken at both the left and the right posterior tibial arteries with a Dopplerultrasound 8-MHz transducer (Huntleigh 500 D; Huntleigh Technology PLC, Bedfordshire,England) while the subject was in the supine position. In agreement with theapproach followed by Fowkes et al,22 peripheralarterial disease was considered present when the ankle-brachial blood pressureindex was lower than 0.90 in at least 1 leg. The sensitivity and the specificityof this cutoff are 90% and 98%, respectively, for an angiographically definedstenosis of 50% or more in a major leg artery.23
Intima-media thickness was measured by recording ultrasonographic imagesof both the left and right carotid arteries, using a 7.5-MHz linear-arraytransducer (ATL Ultra-Mark IV; Advanced Technology Laboratories, Bethel, Wash).24 The lumen-intima interface and the media-adventitiainterface of the distal common carotid artery were measured offline. The commoncarotid intima-media thickness was determined as the average of near and farwall measurements of both left and right sides.
The presence of plaques in the carotid artery was assessed by evaluatingthe ultrasonographic images of the common, internal, and bifurcation sitesof the carotid artery for the presence of atherosclerotic lesions.25 Plaques were defined as a focal widening relativeto adjacent segments and composed of calcified or noncalcified components.The total carotid plaque score of each subject was defined by summation ofthe presence of plaques at far and near walls of left and right sides at 3locations. For the analyses we used 4 categories of plaque scores (no plaquesand 1, 2-3, and 4 or more plaques). Results from a reproducibility study ofplaque and intima-media thickness of the carotid artery among 80 participantsin the Rotterdam Study who underwent a second ultrasound have been describedelsewhere. In short, between paired measurements, a κ of 0.67 was foundfor plaques and of 0.74 for intima-media thickness.26
Atherosclerosis of the abdominal aorta was determined with a lateralx-ray of the lumbar spine. Calcified plaques were considered present whenlinear densities were clearly visible in an area parallel and anterior tothe lumbar spine.18 Values for the extent ofcalcification were scored according to the length of the involved area (<1cm, 1-2.5 cm, >2.5-5 cm, and >5 cm). Radiographic assessment of aortic atherosclerosiswas shown to be highly specific and in most cases represents advanced intimalatherosclerosis as determined at autopsy. A comparison study with CT was performedin our department. In 56 unselected patients, abdominal radiography showedcalcifications in 32; all but 1 were confirmed by CT.27
We used coronary calcifications as a measure of coronary atherosclerosis.Imaging of the epicardial coronary arteries was performed with an electron-beamCT scan (C-150; Imatron, San Francisco, Calif). From the root of the aortathrough the heart, 38 images were obtained with 3-mm slice thickness. Withthe method of Agatston et al,28 a calcium scorewas obtained by multiplying each area of interest with a factor indicatingpeak density within the area. We summed the scores for the calcificationsto obtain a score for the entire epicardial system.
Age, sex, cognitive function, education, antidepressant medication,cigarette smoking, total cholesterol level, body mass index, blood pressure,diabetes mellitus, history of myocardial infarction, and stroke were consideredas possible confounders. Cognitive function was assessed by the Mini-MentalState Examination. Education was measured on an ordinal scale and later dichotomizedinto low (primary education only) and high education. Information on currentantidepressant medication was obtained during the home interview, which includeda cabinet check. Cigarette smoking was coded as pack-years ([number of cigarettesper day/20] × years) and in categories of current, former, and neversmoker. Quetelet body mass index (calculated as weight in kilograms dividedby the square of height in meters) was assessed at the research center. Diabetesmellitus was defined as use of insulin or oral blood glucose–loweringdrugs or serum glucose levels higher than 198 mg/dL (11.0 mmol/L). Fastingblood samples were obtained from all subjects in this analysis, and serumtotal cholesterol level was determined by an automatic enzymatic procedure.Histories of myocardial infarction and stroke were obtained by direct questioningand computerized linkage with general practitioner medical records. Historywas considered positive when verified by a physician.
The association of atherosclerosis with depression was analyzed in 2ways. First, we assessed the relationship of individual atherosclerosis measureswith subthreshold depressive symptoms and depressive disorders by means oflogistic regression. Intima-media thickness was entered into the model asa continuous variable divided by 1 SD. Ankle-brachial blood pressure index,carotid plaques, and aortic calcifications were analyzed in the categoriesdefined earlier. For the coronary calcifications, 3 calcium score categorieswere defined as described previously: 0 to 100, 101 to 500, and above 500.29
Second, we used the different extracoronary atherosclerosis measuresto obtain an overall atherosclerosis score. A principal component analysiswas performed. This is a standard procedure to combine related measures.30 Original variables were replaced with linear combinationsof these variables. For this aim, we used the continuous measure ankle-brachialblood pressure index, intima-media thickness, plaques in the carotid arteries,and aortic atherosclerosis. We were able to reduce the 4 measures in our originaldata set to 1 principal component. The association with extracoronary atherosclerosiswas studied by regressing depressive disorders and subthreshold depressivesymptoms on this continuous measure divided by its standard deviation. A principalcomponent analysis increases the power to reveal relationships. Unlike othermethods, such as the construction of a composite score, it does not rely onarbitrary cutoff points (categorizations). To investigate a possible dose-responserelationship, we repeated the analyses including cases with major depressiononly.
To control for confounding, we added the potential confounders intothe basic model, which already contained age (continuous) and sex. The analyseswere run with and without exclusion of persons with a history of stroke andmyocardial infarction. On the one hand, these conditions can be seen as anintermediate in the relationship between atherosclerosis and depressive disordersand should not be excluded. On the other hand, an analysis in a subset freeof overt cardiovascular disease gives additional information. It indicateswhether the relationship of depression with atherosclerosis is due to a psychologicalreaction after manifestation of disease. We also checked whether the associationsfound were different in men and women by performing stratified analyses. Furthermore,we reran the analysis excluding all persons who had had an early-onset depression.The background of this secondary analysis is the vascular depression hypothesis,which posits that atherosclerosis contributed to the onset of depression.Episodes of depression at younger ages may indicate a vulnerability to otherfactors than vascular ones.
In addition to the analyses described, which rely on a categorical classification,we used the CES-D scores as a continuous variable in an analysis includingall persons with an overall atherosclerosis score. Of the 3747 subjects whohad at least 3 measures of atherosclerosis performed, 742 had a missing valueon the fourth measure. To decrease possible bias, these missing atherosclerosisvalues can be filled in (imputed) by the multiple imputation method.31 Imputation is based on the correlation between themissing variable and other variables including the atherosclerosis measures.This correlation can be estimated by using the subjects in whom the measurementwas performed. We followed the method described by van Buuren et al,32 which accounts for the uncertainty around estimatedvalues. Imputations were performed with the Multivariate Imputations by ChainedEquations statistical package (TNO Prevention and Health, Leiden, the Netherlands).For the present analyses, imputed values were used for the principal componentanalyses only.
Table 1 presents characteristicsof the reference subjects who scored below the CES-D cutoff and the caseswith subthreshold depressive symptoms or depressive disorders. Persons withdepressive disorders were slightly older (74 years vs 72 years), more likelyto be female (72% vs 57%), and more likely to have had a stroke (8% vs 3%)than the persons in the reference group.
Table 2 shows the associationsof the extracoronary atherosclerosis measures with subthreshold depressivesymptoms and depressive disorders. There was no substantial relationship betweenindividual measures of atherosclerosis and subthreshold depressive symptoms.A consistent pattern was found for depressive disorders across the 4 locationsat which atherosclerosis was measured. The persons with a more severe diseaseprocess were more likely to have a depressive disorder. For intima-media thicknessand severe aortic calcifications, the relationship reached statistical significance.
Table 3 shows that subjectswith higher levels of coronary calcifications had substantially more depressivedisorders but no more subthreshold depressive symptoms than subjects withoutcoronary calcifications.
Table 4 presents the resultsbased on the overall measure of atherosclerosis. Extracoronary atherosclerosiswas related to depressive disorders. Furthermore, the results suggest a dose-responserelationship between extracoronary atherosclerosis and depression. The highestodds ratios were observed when only persons with major depressive episodeswere included as cases. However, the association was no longer significantbecause of the smaller number of cases.
The findings were very similar when imputed data from subjects witha missing atherosclerosis measurement were included. The odds ratio for thisall-case analysis was 1.27 (95% confidence interval, 1.01-1.56; P = .04, with 102 cases with depressive disorders and 3491 personswho scored below the CES-D cutoff). Exclusion of persons with stroke and myocardialinfarction substantially attenuated the association of atherosclerosis withsubthreshold depressive symptoms but not with depressive disorders or majordepression (Table 4). Furthermore,the findings were similar in men and women. However, sex was a strong confounder.The elderly women had less atherosclerosis but were more likely to be depressed.
Excluding participants who reported an early-onset depressive episode(17 persons with depressive disorders and 151 persons who scored below thecutoff) scarcely changed the association between atherosclerosis and depression.The odds ratio of depressive disorders was 1.33 (95% confidence interval,0.97-1.81; P = .06) in the fully adjusted model.
The analysis using the CES-D scores as a continuous variable furthersupports the notion of a dose-response relationship between atherosclerosisand depression. Persons with more severe atherosclerosis had higher scoreson the CES-D: 0.3 point per standard deviation of overall atherosclerosisscore (95% confidence interval, 0.02-0.55; P = .03).
In this population-based study, we found that subjects with atherosclerosiswere more likely to be depressed. A combined measure of extracoronary atherosclerosiswas related to depressive disorders, although at some of the different locationsthe association was only moderate and nonsignificant. A strong relationshipwas observed only between severe coronary and aortic calcifications and depressivedisorders.
The strengths of this study are the large number of elderly people participatingand its population-based design. Furthermore, the psychiatric workup in subjectswho screened positive on the CES-D enabled us to determine in which groupdepressive symptoms were due to depressive disorders. A previous study inan elderly Dutch population reported a high sensitivity with the same cutoffpoint, and misclassification of disease is therefore unlikely to have influencedour results.15
Some methodologic issues of this study must be discussed. First, thisis a cross-sectional study and cannot demonstrate the chronology of the observedrelationship. Second, the prevalence of subjects with depressive symptomsin this study (7.1%) was relatively low. However, it is comparable to thecommunity prevalence observed by Blazer et al16 inthe United States (9.0%), who also used the CES-D. Furthermore, the rate fallswithin the variable range recently reported in a review of depressive symptomsin the elderly (2.8%-35%).33
To our knowledge, the present study is the first to examine the relationshipbetween measures of atherosclerosis and depressive disorders in community-dwellingsubjects. Many clinical studies have been performed in patients with preexistingvascular disease.34- 36 Thesestudies generally show a high risk of comorbid depression on survival aftera cardiovascular event. So far, few studies have been conducted in community-dwellingsubjects.37,38 A recent studyin US Army personnel with the use of electron-beam CT observed no correlationbetween psychological factors and coronary atherosclerosis.38 Westudied an elderly population that encompassed a very broad spectrum of thedegree of atherosclerosis. This may help to detect an association not foundin a homogeneous population of middle-aged men. Furthermore, the previouspopulation-based research, which investigated potential pathophysiologic mechanisms,concentrated on personality traits rather than specific mood states.37- 39 It is not clear howfar a single trait (eg, negative affectivity) can account for the associationsof depressive disorders with increased cardiovascular morbidity. This makesit even more difficult to relate previous findings to the present study, whichmeasured clinical depressive disorders. Moreover, personality traits are muchmore stable than specific mood states such as depression. Studying distinctdiseases has been advocated to identify potential treatments.9,40
Several explanations for the results need to be discussed. First, ourfindings are compatible with the vascular depression hypothesis, which positsan etiologic role of vascular factors in the onset of depression. There isevidence from mostly cross-sectional studies that cerebrovascular diseasecontributes to the development or persistence of depression.13 Thesestudies, few of which were population based, have used different neuroimagingtechniques in depressed patients.2,4,41- 44 Itis assumed that the observed neuroradiologic findings such as white-matterlesions are due to vascular disease.3 Furthermore,depression is highly prevalent in patients with overt or silent stroke,45 which also suggests that people with cerebrovasculardisease are particularly vulnerable to depression.
Second, depression could be a risk factor for the development of atherosclerosis.Although in most participants in this study we have no evidence of earlierepisodes of depression, many of the depressed elderly will have had a chronicand intermittent course of depression. Previously, longitudinal population-basedstudies associated depression with cardiovascular disease later in life. Depressivestatus at baseline was an independent risk factor for ischemic heart diseaseand death.5- 7,46 Numerousbiological processes that have been associated with depression may underliethis increased cardiovascular risk: decreased myocardial perfusion, abnormalitiesin platelet reactivity, cardiac arrhythmia, inflammatory processes, and noradrenergichyperactivity.47- 50 Thesepathophysiologic changes are generally accompanied by increased atherosclerosis.Even if behavioral factors such as difficulties in smoking cessation or pooradherence to medical treatment account for the relationship of depressionwith subsequent vascular disease, one would expect to find increased levelsof atherosclerosis in a cross-sectional study of elderly depressed persons.
Of course, the explanations are not mutually exclusive. The causal pathwaymay be bidirectional.13 Furthermore, both depressionand atherosclerosis could be manifestations of a common pathological process.
The observed association was somewhat more marked for coronary calcifications.This should be interpreted carefully because electron-beam CT was not availablein all subjects. Subjects who underwent electron-beam CT were less likelyto be depressed. However, unless the depressed nonparticipants had less atherosclerosisthan the depressed participants, this will not have introduced bias. Moreover,the strong association between coronary calcifications and depression mayreflect the good assessment of atherosclerosis rather than indicating a specificcardiac process. A recent publication of the Cardiovascular Health Study showedthat some subclinical atherosclerosis detected by electron-beam CT was missedby measures of peripheral arterial disease and intima-media thickness.51
In summary, our findings in this large study of community-dwelling elderlyconfirm the postulated link between vascular factors and depression. Furthermore,a dose-response relationship was observed: more severe atherosclerosis wasassociated with more severe depressive symptoms. The present study cannotestablish a causal role of atherosclerosis, but provides evidence that a generalizedatherosclerotic process accompanies late-life depression. It is prematureto speculate about clinical implications. However, if substantiated throughfurther research, this knowledge may allow the prevention of the physicalconsequences of depression or new treatment strategies for late-life depression.52 Primarily, prospective studies are needed to furtherelucidate the pathophysiology underlying the association between depressivedisorders and vascular disease. These studies should include different biologicalmeasures and psychiatric interviews.
Corresponding author: Monique M. B. Breteler, MD, PhD, Departmentof Epidemiology and Biostatistics, Erasmus Medical Center, PO Box 1738, 3000DR Rotterdam, the Netherlands (e-mail: firstname.lastname@example.org).
Submitted for publication July 1, 2003; final revision received October31, 2003; accepted November 18, 2003.
The Rotterdam Study is supported by the Research Institute for Diseasesin the Elderly, funded by the Ministry of Education and Science, The Hague,and the Ministry of Health, Welfare, and Sports, The Hague, and through theNetherlands Organisation for Scientific Research (NWO), The Hague. Dr Bretelerwas supported by a fellowship of the Royal Netherlands Academy of Arts andSciences, Amsterdam. This work has been supported by an unrestricted grantfrom Numico Research BV, Wageningen, the Netherlands.
An abstract of this work was presented at the 34th Annual Meeting ofthe Society for Epidemiologic Research; June 19, 2002; Palm Desert, Calif.