[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.205.0.26. Please contact the publisher to request reinstatement.
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Table 1. 
Baseline Scores and Average Annual Decline in Subgroups With and Without Baseline Depression*
Baseline Scores and Average Annual Decline in Subgroups With and Without Baseline Depression*
Table 2. 
Association of Baseline Depression With Baseline Cognitive Function and Subsequent Decline: Random-Effects Models
Association of Baseline Depression With Baseline Cognitive Function and Subsequent Decline: Random-Effects Models
Table 3. 
Association of Depression With Cognitive Scores in Subgroups With and Without Baseline Depression
Association of Depression With Cognitive Scores in Subgroups With and Without Baseline Depression
1.
Nebes  RDButters  MAMulsant  BHPollock  BGZmuda  MDHouck  PRReynolds  CF Decreased working memory and processing speed mediate cognitive impairment in geriatric depression. Psychol Med 2000;30679- 691
PubMedArticle
2.
Butters  MAWhyte  EMNebes  RDBegley  AEDew  MAMulsant  BHZmuda  MDBhalla  RMeltzer  CCPollock  BGReynolds  DFBecker  JT The nature and determinants of neuropsychological functioning in late-life depression. Arch Gen Psychiatry 2004;61587- 595
PubMedArticle
3.
Kramer-Ginsberg  EGreenwald  BSKrishan  RRChristiansen  BHu  JAshtari  MPatel  MPollack  S Neuropsychological functioning and MRI signal hyperintensities in geriatric depression. Am J Psychiatry 1999;156438- 444
PubMed
4.
Lichtenberg  PARoss  TMillis  SRManning  CA The relationship between depression and cognition in older adults: a cross-validation study. J Gerontol B Psychol Sci Soc Sci 1995;50P25- P32
PubMedArticle
5.
Rapp  MADahlman  KSano  MGrossman  HHaroutunian  VGorman  J Neuropsychological differences between late-onset and recurrent geriatric major depression. Am J Psychiatry 2005;162691- 698
PubMedArticle
6.
Chen  PGanguli  MMulsant  BHDeKosky  ST The temporal relationship between depressive symptoms and dementia: a community-based prospective study. Arch Gen Psychiatry 1999;56261- 266
PubMedArticle
7.
Devanand  DPSano  MTang  MXTaylor  SGurland  BJWilder  DStern  YMayeux  R Depressed mood and the incidence of Alzheimer's disease in the elderly living in the community. Arch Gen Psychiatry 1996;53175- 182
PubMedArticle
8.
Green  RCCupples  LAKurz  AAuerbach  SGo  RSadovnick  DDuara  RKukull  WAChui  HEdeki  TGriffith  PAFriedland  RPBachman  DFarrer  L Depression as a risk factor for Alzheimer disease: the MIRAGE Study. Arch Neurol 2003;60753- 759
PubMedArticle
9.
Jorm  AF History of depression as a risk factor for dementia: an updated review. Aust N Z J Psychiatry 2001;35776- 781
PubMedArticle
10.
Small  BJFratiglioni  LViitanen  MWinblad  BBackman  L The course of cognitive impairment in preclinical Alzheimer's disease: three- and 6-year follow-up of a population-based sample. Arch Neurol 2000;57839- 844
PubMedArticle
11.
Elias  MFBeiser  AWolf  PAAu  RWhite  RFD’Agostino  RB The preclinical phase of Alzheimer's disease: a 22-year prospective study of the Framingham cohort. Arch Neurol 2000;57808- 813
PubMedArticle
12.
Chen  PRatcliff  RBelle  SHCauley  JADeKosky  STGanguli  M Patterns of cognitive decline in pre-symptomatic Alzheimer's disease: a prospective community study. Arch Gen Psychiatry 2001;58853- 858
PubMedArticle
13.
Butters  MABecker  JTNebes  RDZmuda  MDMulsant  BHPollock  BGReynolds  CF  III Changes in cognitive functioning followed treatment of late-life depression. Am J Psychiatry 2000;1571949- 1954
PubMedArticle
14.
Nebes  RDPollock  BGHouck  PRButters  MAMulsant  BHZmuda  MDReynolds  CF Persistence of cognitive impairment in geriatric depressed patients following antidepressant treatment: a randomized double-blind clinical trial with nortriptyline and paroxetine. J Psychiatr Res 2003;3799- 108
PubMedArticle
15.
Ganguli  MBelle  SRatcliff  GSeaberg  EHuff  FJvon der Porten  KKuller  LH Sensitivity and specificity for dementia of population-based criteria for cognitive impairment: the MoVIES project. J Gerontol 1993;48M152- M161
PubMedArticle
16.
Ganguli  MLytle  MReynolds  MDDodge  H Random vs volunteer selection for a community-based study. J Gerontol A Biol Sci Med Sci 1998;53M39- M46
PubMedArticle
17.
Radloff  LS The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1385- 401Article
18.
Ganguli  MGilby  JSeaberg  EBelle  S Depressive symptoms and associated factors in a rural elderly population: the MoVIES project. Am J Geriatr Psychiatry 1995;3144- 160Article
19.
Lassila  HStoehr  GPGanguli  MSeaberg  ECGilby  JEBelle  SHEchement  DA Factors associated with the use of prescription medications in an elderly rural population: the MoVIES Project. Ann Pharmacother 1996;30589- 595
PubMed
20.
Ganguli  MMulsant  BHRichards  SStoehr  GPMendelsohn  AB Antidepressant use over time in a rural older adult population: the MoVIES project. J Am Geriatr Soc 1997;451501- 1502
PubMed
21.
Morris  JCHeyman  AMohs  RCHughes  JPvan Belle  GFillenbaum  GMellits  EDClark  C The Consortium to Establish a Registry for Alzheimer's Disease (CERAD), I: clinical and neuropsychological assessment of Alzheimer's disease. Neurology 1989;391159- 1165
PubMedArticle
22.
Folstein  MFFolstein  SEMcHugh  PR “Mini-Mental State”: a practical method of grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12189- 198
PubMedArticle
23.
Rosen  WGMohs  RCDavis  KL A new rating scale for Alzheimer's disease. Am J Psychiatry 1984;1411356- 1364
PubMed
24.
Becker  JTBoller  FSaxton  JAMcGonigle-Gibson  KL Normal rates of forgetting verbal and non-verbal material in Alzheimer's disease. Cortex 1987;2359- 72
PubMedArticle
25.
Borkowksi  JGBenton  ALSpreen  O Word fluency and brain damage. Neuropsychologia 1967;5135- 140Article
26.
Benton  ALDe Hamsher  K Multilingual Aphasia Examination.  Iowa City University of Iowa1976;
27.
Kaplan  EFGoodglass  GHWeintraub  S The Boston Naming Test.  Media, Pa Lea & Febiger1983;
28.
Freedman  MLeach  LKaplan  E Clock Drawing: A Neuropsychological Analysis.  New York, NY Oxford University Press Inc1994;
29.
Reitan  RM The relation of the Trailmaking Test to organic brain damage. J Consult Psychol 1955;19393- 394
PubMedArticle
30.
Ganguli  MRatcliff  GBelle  SHuff  FJKancel  MJFischer  LKuller  LH Effects of age, gender, and education on cognitive tests in an elderly rural community sample: norms from the Monongahela Valley Independent Elders Survey (MoVIES). Neuroepidemiology 1991;1042- 52
PubMedArticle
31.
Ganguli  MRatcliff  GDeKosky  S Cognitive test scores in community-based rural elderly with and without dementia. Aging Ment Health 1997;1176- 180Article
32.
Ratcliff  GDodge  HHBirzescu  MAGanguli  M Tracking cognitive functioning over time: ten-year longitudinal data from a community-based study. Appl Neuropsychol 2003;1076- 88
PubMedArticle
33.
Ganguli  MDodge  HHChen  PBelle  SDeKosky  ST Ten-year incidence of dementia in a rural elderly US community population: the MoVIES Project. Neurology 2000;541109- 1116
PubMedArticle
34.
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R).  Washington, DC American Psychiatric Association1987;
35.
Hughes  CPBerg  LDanziger  WLCoben  LAMartin  RL A new clinical scale for the staging of dementia. Br J Psychiatry 1982;140566- 572
PubMedArticle
36.
Laird  NMWare  JH Random-effects models for longitudinal data. Biometrics 1982;38963- 974
PubMedArticle
37.
SAS Institute Inc, SAS/STAT User's Guide, Version 8: Software Manual.  Cary, NC SAS Institute Inc2000;
38.
Modrego  PJFerrandez  J Depression in patients with mild cognitive impairment increases the risk of developing dementia of the Alzheimer type: a prospective cohort study. Arch Neurol 2004;611290- 1293
PubMedArticle
39.
Airaksinen  ELarsson  MLundbgerg  IForsell  Y Cognitive functions in depressive disorders: evidence from a population-based study. Psychol Med 2004;3483- 91
PubMedArticle
40.
Alexopoulos  GS Frontostriatal and limbic dysfunction in late-life depression. Am J Geriatr Psychiatry 2002;10687- 695
PubMedArticle
41.
Krishnan  KRTaylor  WDMcQuoid  DRMacFall  JRPayne  MEProvenzale  JMSteffens  DC Clinical characteristics of magnetic resonance imaging–defined subcortical ischemic depression. Biol Psychiatry 2004;55390- 397
PubMedArticle
42.
Sheline  YISanghavi  MMintun  MAGado  MH Depression duration but not age predicts hippocampal volume loss in medically healthy women with recurrent major depression. J Neurosci 1999;195034- 5043
PubMed
43.
Dufouil  CFuhrer  RDartigues  JFAlperovitch  A Longitudinal analysis of the association between depressive symptomatology and cognitive deterioration. Am J Epidemiol 1996;144634- 641
PubMedArticle
44.
Vinkers  DJGussekloo  JStek  MLWesterdorp  RGvan der Mast  RC Temporal relationship between depression and cognitive impairment in old age: prospective population-based study. BMJ 2004;329881
PubMedArticle
45.
Ganguli  MDodge  HMulsant  BH Rates and predictors of mortality in an aging, rural, community-based cohort: the role of depression. Arch Gen Psychiatry 2002;591046- 1052
PubMedArticle
46.
Wilson  RSMendes de Leon  CBennett  DABienias  JLEvans  DA Depressive symptoms and cognitive decline in a community population of older persons. J Neurol Neurosurg Psychiatry 2004;75126- 129
PubMedArticle
47.
Blazer  DGHybels  CFPieper  CF The association of depression and mortality in elderly persons; a case for multiple, independent pathways. J Gerontol A Biol Sci Med Sci 2001;56M505- M509
PubMedArticle
Original Article
February 2006

Depressive Symptoms and Cognitive Decline in Late LifeA Prospective Epidemiological Study

Author Affiliations

Author Affiliations: Division of Geriatrics and Neuropsychiatry, Department of Psychiatry (Dr Ganguli and Mr Du), and Department of Medicine (Dr Chang), University of Pittsburgh School of Medicine; Department of Epidemiology, University of Pittsburgh Graduate School of Public Health (Drs Ganguli and Dodge); and Department of Neuropsychology, Healthsouth Harmarville Rehabilitation Hospital (Dr Ratcliff), Pittsburgh, Pa.

Arch Gen Psychiatry. 2006;63(2):153-160. doi:10.1001/archpsyc.63.2.153
Abstract

Context  Depression is associated with cognitive impairment and dementia. It is less clear whether depression contributes to further cognitive decline over time, independently of incipient dementia.

Objective  To examine the relationship between depressive symptoms and subsequent cognitive decline in a cohort of nondemented older adults, some of whom remained dementia free during follow-up and others in whom incident dementia eventually developed.

Design  Twelve-year prospective epidemiological study, including biennial measurement of cognition and depressive symptoms, biennial assessment of dementia, and comparison of cognitive function at baseline and over time in persons with and without baseline depressive symptoms in the dementia-free and eventual-dementia groups, using random-effects models.

Setting  A largely blue-collar rural community.

Participants  Population-based sample of 1265 adults 67 years and older without dementia at baseline.

Main Outcome Measures  Scores over time on each of several cognitive test composites.

Results  Among 1094 participants who remained dementia free, those with baseline depressive symptoms had significantly lower baseline scores on all cognitive composites than the nondepressed participants. Among the 171 individuals in whom dementia later developed, depression was associated with worse performance in some but not all baseline cognitive composites. Cognitive decline over time was minimal in the dementia-free group, whereas marked decline was seen in the eventual-dementia group. Depressive symptoms were not associated with rate of cognitive decline over time in either group.

Conclusions  Depressive symptoms are cross-sectionally associated with cognitive impairment but not subsequent cognitive decline. Substantial cognitive decline over time cannot be explained by depression and most likely reflects incipient dementia.

A frequent concern for clinicians is whether a depressed older adult's cognitive impairment and decline reflect an underlying dementia or can be explained solely by the depression. A substantial literature validates clinical observations that cognitive impairment often coexists with depression in elderly patients.15 In nondemented individuals, both depression and cognitive decline are associated with subsequent onset of dementia.612 Major depression has been associated with cognitive deficits that persist after the depression remits,13,14 but it is difficult to rule out confounding by incipient dementia, which may not become clinically manifest during the relatively short duration of most intervention studies. To determine whether depression itself independently predicts cognitive decline over time, it is essential to know whether dementia will develop in a given individual in the future.

In a representative, aging, community-based cohort, previous epidemiological work from our group suggested that a high level of depressive symptoms was a prodromal stage rather than an independent risk factor for Alzheimer disease.6 In the same cohort, we now focus on the association of depressive symptoms with subsequent decline in several cognitive domains, independent of the effect of incipient dementia. The availability of 12 years of follow-up data allowed us to detect and adjust our analytic models for subsequent incident dementia. In effect, we examined whether depressive symptoms at baseline predicted future cognitive decline among nondemented cohort members who remained dementia free during several years of follow-up (dementia-free group). For comparison, we addressed the same issue in the remaining cohort members in whom dementia subsequently developed (eventual-dementia group).

METHODS
STUDY SITE AND POPULATION

The Monongahela Valley Independent Elders Survey (MoVIES) was conducted from 1987 to 2002, in a population of low income and education levels in rural southwestern Pennsylvania. Sampling and recruitment procedures have been detailed previously.15 Study procedures were approved annually by the institutional review board at the University of Pittsburgh, Pittsburgh, Pa. Entry criteria were being 65 years or older, living in the community at the time of recruitment, being fluent in English, and having at least a sixth-grade education.15 A total of 1681 participants underwent assessment at study entry (wave 1, 1987-1989). These participants consisted of 1422 individuals randomly selected from the voter registration list and an additional 259 volunteers from the same area, all meeting the same entry criteria.16 Our analytic models always include a term for recruitment status, indicating whether the individual was from the random or the volunteer subgroup at study entry. At approximately 2-year intervals thereafter, subjects underwent reevaluation in a series of data collection waves. Between waves 1 and 2, 340 individuals died, relocated, or dropped out.

Wave 2 (1989-1991), when depression was assessed for the first time, served as the baseline for the present analyses. Cohort size at wave 2 was 1341; of these, we excluded 64 prevalent cases of dementia, ie, those with onset of dementia (see “Assessment of Dementia”) before wave 2.

DEPRESSION ASSESSMENT

After providing written informed consent, subjects underwent in-home screening, including a modified Center for Epidemiological Studies–Depression Scale (mCES-D),17,18 with higher scores reflecting a greater number of depressive symptoms. The modification includes all 20 original CES-D items but asks about their presence during most of the preceding week in a yes/no format coded as 1/0, so that the maximum possible score is 20. Following standard epidemiological practice, we used an actuarial approach to defining abnormality, ie, a percentile-based cutoff derived on cohort norms, to identify the most depressed 10th of the sample. Because the distribution of mCES-D scores in this community-based cohort was skewed to the right, depression was treated as a categorical variable with the cutoff score at the cohort's 90th percentile, which was a score of 5. That is, participants reporting the presence of 5 or more depressive symptoms during most of the previous week were categorized as depressed.18 Twelve participants with incomplete mCES-D data at baseline were excluded from these analyses. The mCES-D was administered at all waves, beginning with wave 2. In post hoc analyses, to capture stability of depressive symptoms over time, we further distinguished between transient depression (wave 2 only) and persistent depression (waves 2 and 3). Information about regular use of antidepressant drugs was obtained by self-report and inspection of medication bottle labels in the home.19,20

COGNITIVE ASSESSMENT

The MoVIES cognitive screening battery, administered at each biennial visit, incorporated the neuropsychological panel of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD).21 This battery included the Mini-Mental State Examination (MMSE)22; the immediate learning and delayed recall items of the 10-item CERAD Word List22,23; immediate retell and delayed recall of an 18-item story24; initial letter fluency for the letters P and S25; category fluency for the names of fruits and animals26; the 15-item CERAD version of the Boston Naming Test21,27; the 4-item CERAD Constructional Praxis task21,23; the Clock Drawing Test28; and Trail-Making Tests A and B.29 Detailed descriptions of these tests and cross-sectional and longitudinal norms from the MoVIES cohort have been reported previously.3032 Tests were grouped according to cognitive domain and were based on conceptual grounds and previous factor analysis.32 Composite cognitive scores were constructed by z-transforming the individual test scores using baseline means and standard deviations and averaging the z scores for each domain. The composites were designated as learning (immediate retell of the story and immediate learning of the word list), memory (delayed recall of the story and delayed recall of the word list), visuospatial ability (CERAD constructional praxis and clock drawing), language (CERAD version of the Boston Naming Test, initial letter fluency for the letters P and S, and category fluency for the names of fruits and animals), and executive function (Trail-Making Tests A and B), although we recognize that the Trail-Making Tests capture only some of the functions typically characterized as executive. The MMSE was not included in the composites but examined separately as a test of general mental status. The focus of the present analyses was scores over time (rate of change) on each composite and the MMSE during the length of time that each participant was involved in the study.

ASSESSMENT OF DEMENTIA

As previously described,15,33 participants operationally classified as cognitively impaired or declined on the basis of cognitive screening results were invited to undergo a clinical evaluation for dementia, following the assessment protocols established by CERAD21 and the University of Pittsburgh Alzheimer Disease Research Center and modified for field use. Clinical evaluations were performed with the examiners blinded to screening cognitive scores. A consensus diagnosis process established or ruled out the diagnosis of dementia according to the DSM-III-R34 and the stage of dementia according to the Clinical Dementia Rating Scale. Clinical Dementia Rating Scale scores (stages) of 0, 0.5, 1, 2, and 3 indicate no, questionable, mild, moderate, and severe dementia, respectively.35 Based on all available evidence, a date of symptom onset was estimated in each case.33 Participants receiving a Clinical Dementia Rating Scale score of 1 or more at wave 1 or wave 2 (prevalent cases at wave 2) were excluded. Incident dementia (Clinical Dementia Rating Scale score, ≥1) with onset subsequent to wave 2 was included as a covariate in the analyses, allowing us to compare these individuals with those who remained dementia free for the duration of their involvement in the study.

The cohort consisted of 1094 subjects who remained dementia free during follow-up (dementia-free group) and 171 subjects with eventual dementia (eventual-dementia group). The mean duration of follow-up from baseline to final assessment for each subject in these analyses varied across tests, but ranged from 1.9 to 11.8 years (mean follow-up, 7.4 years) in the dementia-free group. For the eventual-dementia group, the duration of follow-up ranged from 1.7 to 11.2 years (mean, 5.9 years; mean [SD] duration from baseline until estimated onset of dementia, 5.4 [3.1] years). For those classified as depressed (mCES-D, >4) at baseline, median (range) follow-up was 6.3 years (1.7-11.0 years); for those classified as nondepressed, 7.2 years (1.7-11.8 years).

STATISTICAL METHODS

Baseline characteristics of the cohort and subgroups were compared using t tests for continuous variables and χ2 tests for categorical variables (2-tailed tests). In each of the dementia-free and eventual-dementia groups, means and standard deviations of baseline scores on each cognitive test were calculated for depressed and nondepressed subgroups. Annualized decline (crude change) on each test was calculated as the difference in scores between baseline and the final assessment for each subject, divided by the number of years that the subject was followed up until his or her final assessment (before follow-up was terminated by death, dropout, or end of the study).

The random-effects modeling approach assumes that each participant's pattern of scores over time on a given measure is explained by his or her covariate patterns (fixed effects), except for person-specific random effects. The baseline level of cognition is explicitly modeled as the intercept. For scores on each cognitive composite and MMSE, we used a single random-effects model36 to examine the association of depressive symptoms with scores at the baseline and follow-up waves on that composite or the MMSE. Cognitive scores over time were modeled as a function of the following terms (coefficients) adjusted for age, sex, education, and recruitment status: depression (present or absent at baseline), time (years since baseline), incident dementia (present or absent), interaction terms depression × time, dementia × time, depression × dementia, and depression × dementia × time. The model includes multiple terms; for simplicity, we focus herein on the terms directly relevant to examining the effect of depressive symptoms on baseline and longitudinal cognitive function in the dementia-free and eventual-dementia groups. In effect, these are the terms that allow us to compare the depressed and nondepressed subgroups of the dementia-free and eventual-dementia groups.

For the dementia-free group, the term depression indicates the association between baseline depression and baseline cognitive score; the interaction term time × depression indicates the effect of depression on the annual rate of cognitive change. The statistical significance of these terms is indicated by their P values in the random-effects models.

For the eventual-dementia group, the sum of the terms depression and depression × dementia indicates the mean difference in baseline cognitive score associated with depression; the sum of the terms of depression × time and dementia × depression × time indicates the effect of depression on the annual rate of cognitive change. The significance of these summed effects is determined by examining the difference in −2 log likelihood ratio statistics (χ2 test with 2df) before and after including the additional terms depression and depression × dementia to examine the effect of depression on baseline scores and depression × time and dementia × depression × time to examine the effect of depression on slope of decline.

Three sets of post hoc analyses were performed. First, for antidepressant use, the analyses were repeated, including as a covariate the regular use of antidepressant drugs at baseline. Second, for persistent vs transient depression, analyses were repeated for those with mCESD scores of 5 or more at wave 2 only (transient) and at waves 2 and 3 (persistent). Finally, for age as random effect, to account for the possibility that the effect of time (duration since study entry) might vary according to age, we fit the models again as mixed models treating chronological age as a random rather than a fixed covariate.

To assess model fit, model assumptions were examined analytically and graphically. All analyses were performed using SAS statistical software, version 8.37

RESULTS
SAMPLE SIZE AND CHARACTERISTICS

After excluding 12 participants with incomplete mCES-D data, cohort size at baseline (wave 2) was 1265 participants 67 years and older. Their mean (SD) age was 74.6 (5.3) years; 60.8% were women; and 61.0% had greater than a high school education. At baseline, the 1094 dementia-free participants were significantly younger (mean [SD] age, 74.1 [5.2] years [P<.001, by t test]) than the 171 eventual-dementia subjects (mean [SD] age, 77.7 [5.3] years at baseline). The dementia-free group was more likely than the eventual-dementia group to be male (40.3% vs 32.2%; P = .04 by χ2 test) and to have education beyond high school (62.6% vs 50.9%; P = .002 by χ2 test). At baseline, 106 (9.7%) of the dementia-free group and 22 (12.9%) of the eventual-dementia group had depression, ie, substantial depressive symptoms as indicated by mCES-D scores of 5 or more. Antidepressant drug use was reported by 28 subjects at wave 2, consisting of 23 taking tricyclics, 3 taking fluoxetine hydrochloride, and 2 taking trazodone hydrochloride.

At waves 2, 3, 4, 5, and 6, the cohort size was 1265, 1071, 925, 758, and 595, respectively, with mean (SD) ages of 74.6 (5.3), 76.5 (5.1), 78.3 (4.8), 80.0 (4.3), and 81.8 (3.9) years, respectively. Women constituted 60.8%, 63.4%, 63.6%, 64.9%, and 66.4%, respectively, of the cohort at these successive waves. The proportions with depression were 10.1%, 7.7%, 7.0%, 5.9%, and 6.0%, respectively. The 128 individuals classified as depressed at wave 2 were significantly older than those classified as nondepressed (mean [SD] age, 75.8 [5.5] years, compared with 74.5 [5.1] years; t1 = 2.45; P = .02). Those with transient depression (depressed at wave 2 only) and persistent depression (depressed at waves 2 and 3) were not significantly different in baseline (wave 2) age (mean [SD] age, 76.8 [5.4] vs 78.6 [5.5] years; t1 = 1.81; P = .07). Among the 128 persons depressed at baseline, 13 (10.2%) died before wave 3, compared with 6.1% of the nondepressed persons. In a logistic regression model adjusting for age and sex, depression was significantly associated with death between waves 2 and 3 (odds ratio [OR], 1.94; 95% confidence interval [CI], 1.02-3.70). During follow-up, the cohort experienced 50.4% overall mortality. In generalized estimating equations combining data from waves 2 through 6, depression was not significantly associated with mortality between successive waves (OR, 1.11; 95% CI, 0.85-1.37).

BASELINE TEST SCORES (CROSS-SECTIONAL) IN DEPRESSED AND NONDEPRESSED SUBGROUPS

Mean scores at baseline on each cognitive test were calculated in depressed and nondepressed subgroups of the dementia-free and eventual-dementia groups that constituted our cohort (Table 1). Mean baseline scores on all tests were significantly lower in the eventual-dementia group than in the dementia-free group (P<.001; data not shown), although none of the participants met diagnostic criteria for dementia at the time.

We summarized the associations of baseline depressive symptoms with cognitive scores at baseline and over time. Table 2 provides background information for the summary in Table 3. Results shown for the random-effects models (Table 2) are restricted to the terms relevant to the specific comparisons reported herein. These coefficients (time, depression, depression × dementia, depression × time, and depression × dementia × time) and their corresponding P values for each cognitive measure correspond to the summary terms and P values shown in Table 3, as described in the “Methods” section.

Interpretation of Tables 2 and 3 will be illustrated using the learning composite as an example. For associations of depressive symptoms with baseline scores on the learning composite, the coefficient for depression (Table 2) is −0.211, which corresponds to the coefficient under the baseline score for the dementia-free group in Table 3. The sum of the coefficients for depression and depression × dementia (Table 2) is −0.134 and corresponds to the coefficient for the baseline score in the eventual-dementia group in Table 3. Similarly, for the rate of change in scores over time (ie, decline) on the learning composite, the coefficient for depression × time in Table 2 is −0.012, which corresponds to the coefficient for average decline in the dementia-free group in Table 3. The sum of the coefficients for depression × time and depression × dementia × time in Table 2 is 0.001 and corresponds to the coefficient for rate of change in the eventual-dementia group in Table 3. Significance levels were calculated as described in the “Methods” section and footnotes to Table 3.

BASELINE SCORES IN DEPRESSED AND NONDEPRESSED SUBGROUPS

In the dementia-free group, random-effects models showed depressive symptoms to be significantly associated with baseline scores on all composites and the MMSE after adjustment for covariates (age, sex, education, and recruitment status). In the comparison eventual-dementia group, depressive symptoms were also associated with significantly lower mean scores on the memory, visuospatial ability, and executive composites but not on the learning or the language composite or the MMSE (Table 3).

CHANGE AND RATE OF CHANGE IN SCORES OVER TIME IN DEPRESSED AND NONDEPRESSED GROUPS

Duration of follow-up varied among participants from about 2 to about 12 years. Annualized decline on each test, from baseline until the end of follow-up, unadjusted for covariates, was minimal and significantly less on all tests and composites in the dementia-free group than in the eventual-dementia group (P<.001, data not shown).

Annualized decline, unadjusted for covariates, was also calculated in the subgroups with and without baseline depressive symptoms in the dementia-free and eventual-dementia groups (Table 1). These raw decline scores are presented to provide clinical context; they are not the outcome variables in the random-effects models reported in the next paragraph.

Baseline depressive symptoms were not associated with cognitive decline on any composites in the dementia-free group or the eventual-dementia group (Table 3). These results are based on random-effects models, including depression, eventual dementia, age, sex, education, and recruitment status, in which the outcome variables were the z scores at multiple waves.

POST HOC ANALYSES
Transient vs Persistent Depression

Random-effects models fit separately for transient depression (depression at wave 2 only) and persistent depression (depression at waves 2 and 3) showed results identical to each other and to the models based on the entire dementia-free subgroup, ie, depression was associated with baseline scores on all composites and the MMSE, but not with decline on any, regardless of whether depression was transient or persistent (data not shown).

Antidepressant Use

The random-effects models for decline were fit once again for each test, including baseline antidepressant use (n = 28) as a covariate. In the dementia-free group, baseline MMSE scores were no longer associated with depression after adjusting for antidepressant use. In the eventual-dementia group, baseline memory and visuospatial ability composites were no longer associated with depression after adjusting for antidepressant use. Antidepressant use did not change the lack of association between baseline depression and subsequent cognitive decline in either group (data not shown).

Age as Random Covariate

In mixed models treating chronological age at baseline as a random rather than a fixed covariate, results remained unchanged with a single exception: the association between baseline depression and baseline MMSE scores was no longer significant (data not shown).

All models were examined graphically and analytically, and showed good fit.

COMMENT

In a community-based cohort followed up for as long as 12 years, we examined the relationship between depressive symptoms and subsequent cognitive test performance over time, adjusting for the presence of incipient dementia as indicated by eventual development of dementia during follow-up. Among individuals who remained dementia-free for the duration of follow-up, cognitive decline over time was minimal. Despite significant associations with lower baseline scores on all cognitive composites and general mental status, depressive symptoms were not associated with subsequent decline on any of these measures. In the comparison group in whom dementia later developed, marked decline was seen on all tests over time. However, baseline depressive symptoms had no impact on the rate of decline in this group either. Post hoc analyses examining different subgroups did not change these results. Our findings suggest that substantial cognitive decline does not occur in the absence of an incipient dementia, and therefore cannot be attributed to depression.

Several studies,68 including previous work from the MoVIES cohort,6 have shown that depression in a cohort as a whole is a precursor/prodrome or a susceptible state for the development of dementia. Depression in conjunction with mild cognitive impairment has been particularly implicated as a prognostic indicator for Alzheimer disease.38 Unlike previous studies, we were able to examine the effect of depressive symptoms on cognitive decline, independently of the effects of underlying dementing processes that were undetectable at baseline. Although our study participants were all nondemented at baseline, they had up to 12 years of follow-up, during which we could detect subsequent incident dementia and adjust for it in our models. By using random-effects models, we were in effect able to examine the associations between depressive symptoms and subsequent cognitive decline separately in dementia-free and eventual-dementia groups.

The absolute amounts of decline were small (Table 1), given that we examined a nonclinical cohort free of dementia at baseline, and that we annualized decline to allow comparisons among individuals followed up for varying lengths of time. This finding is consistent with the previous report that, in the MoVIES cohort as a whole, those who survived for the duration of the study and completed all tests at all waves showed very small changes in mean scores over time.6 Our previous work also showed that although mean scores remained stable, standard deviations grew progressively larger during the same period, indicating an increase in heterogeneity with aging, even in healthy survivors.32 The present work suggests that depressive symptoms may help explain group heterogeneity in test scores at any given time, but do not contribute to individual change over time.

Most previous reports15,13,14 have been derived from clinic-based samples of patients with major depression. In a cross-sectional community-based study of younger Swedish adults aged 20 to 64 years, individuals with depressive disorders had impairments in episodic memory and mental flexibility (Trail-Making Test B).39 Whether clinic or community based, cross-sectional studies have used a variety of different measures of cognitive functioning, making direct comparisons difficult. Furthermore, although the associations between depression and cognitive impairment have been consistently present in these studies, cause and consequence cannot be distinguished. Most prospective clinical studies of patients have had relatively short follow-up periods, eg, 12 weeks of antidepressant drug treatment for major depressive disorder. Remarkably, despite substantial variation in study populations, design, and assessments, most investigators have detected associations between depression (whether syndromes or symptom measures) and measures of memory and/or executive functioning. It has been suggested that the underlying impairment is slowed information processing2 and, because the deficits persist after successful treatment of depressive illness, that they might represent traits or reflect underlying cortical or subcortical degenerative or vascular disease.13,40,41 Other hypothesized causal mechanisms include disruptions to the hypothalamic-pituitary-adrenal axis in depression, with adverse consequences for hippocampal integrity and neurogenesis.9,42

Depressive symptoms detected in epidemiological studies of the population at large may represent a milder and less homogeneous affective state than that of patients with major depression seeking treatment in psychiatric facilities. One previous population-based prospective study of 1600 older adults in France found that high CES-D scores did not predict cognitive decline (defined as the loss of ≥5 points on the MMSE) over 3 years.43 In another prospective epidemiological study in the Netherlands, with a sample of 500 persons 85 years and older, those with depression at baseline did not show significantly accelerated cognitive decline during follow-up.44 Like cross-sectional studies, longitudinal studies of shorter duration could fail to detect and adequately control for incipient dementia. Here, our considerably longer follow-up confirms strong cross-sectional associations of depression symptoms with all cognitive composites in those who remained dementia free, but finds no associations with subsequent decline on any of them. However, depressive symptoms are a significant predictor of mortality,45 albeit not necessarily in the short term from wave to wave. It is possible that lack of decline over time represents survivor bias in many of these studies, ie, those who died during follow-up might have been the ones who would have declined had they survived. Only 1 prospective community-based study46 found a significant association between depressive symptoms and cognitive decline. That study, with an overall approach similar to ours, followed up 4392 older adults in the Chicago, Ill, area for an average of 5.3 years. That study did not focus on the effects of depression independent of the effects of dementia; it did not exclude cases of prevalent dementia or adjust for subsequent incident dementia.46

A potential limitation of the present study is that we did not collect data sufficient to diagnose specific depressive disorders but rather measured self-reported symptoms on a standard depression scale, as is typical of large community-based studies.43,44,46 Like depressive disorders, high levels of depression symptoms are themselves associated with adverse outcomes, including mortality.45,47 Also, we lacked neuroimaging data with which to determine the anatomy of the degenerative or cerebrovascular substrates, eg, hippocampal or frontal atrophy and white matter lesions, which others have reported in depression.39,40 Given the already low annual rate of cognitive decline observed in our dementia-free group, it remains possible that any additional changes attributable to depression might have been too small to detect, although our sample size provided more than sufficient power.

Our findings help explain some of the heterogeneity in cognitive functioning, but not of cognitive decline, among elderly nondemented individuals. At the level seen in a community-based cohort, the independent effect of depressive symptoms on cognition does not include clinically significant decline over time. Substantial cognitive decline in an older adult is unlikely to be due to depression alone, and most likely reflects an incipient dementia.

Back to top
Article Information

Correspondence: Mary Ganguli, MD, MPH, Western Psychiatric Institute and Clinic, 3811 O’Hara St, Pittsburgh, PA 15213-2593 (gangulim@upmc.edu).

Submitted for Publication: June 6, 2005; final revision received August 12, 2005; accepted August 18, 2005.

Funding/Support: This study was supported in part by grants R01 AG07562, K24 AG022035, and K01 AG023014 from the National Institute on Aging, National Institutes of Health, US Department of Health and Human Services, Bethesda, Md.

Acknowledgments: We thank all of the Monongahela Valley Independent Elders Survey (MoVIES) project staff from 1987 to 2002 for their efforts; the 1681 MoVIES participants in the Monongahela Valley, Pennsylvania, during the same period for their cooperation; and Meryl Butters, PhD, Judith Saxton, PhD, and Changyu Shen, PhD, for helpful comments on earlier versions of the manuscript.

References
1.
Nebes  RDButters  MAMulsant  BHPollock  BGZmuda  MDHouck  PRReynolds  CF Decreased working memory and processing speed mediate cognitive impairment in geriatric depression. Psychol Med 2000;30679- 691
PubMedArticle
2.
Butters  MAWhyte  EMNebes  RDBegley  AEDew  MAMulsant  BHZmuda  MDBhalla  RMeltzer  CCPollock  BGReynolds  DFBecker  JT The nature and determinants of neuropsychological functioning in late-life depression. Arch Gen Psychiatry 2004;61587- 595
PubMedArticle
3.
Kramer-Ginsberg  EGreenwald  BSKrishan  RRChristiansen  BHu  JAshtari  MPatel  MPollack  S Neuropsychological functioning and MRI signal hyperintensities in geriatric depression. Am J Psychiatry 1999;156438- 444
PubMed
4.
Lichtenberg  PARoss  TMillis  SRManning  CA The relationship between depression and cognition in older adults: a cross-validation study. J Gerontol B Psychol Sci Soc Sci 1995;50P25- P32
PubMedArticle
5.
Rapp  MADahlman  KSano  MGrossman  HHaroutunian  VGorman  J Neuropsychological differences between late-onset and recurrent geriatric major depression. Am J Psychiatry 2005;162691- 698
PubMedArticle
6.
Chen  PGanguli  MMulsant  BHDeKosky  ST The temporal relationship between depressive symptoms and dementia: a community-based prospective study. Arch Gen Psychiatry 1999;56261- 266
PubMedArticle
7.
Devanand  DPSano  MTang  MXTaylor  SGurland  BJWilder  DStern  YMayeux  R Depressed mood and the incidence of Alzheimer's disease in the elderly living in the community. Arch Gen Psychiatry 1996;53175- 182
PubMedArticle
8.
Green  RCCupples  LAKurz  AAuerbach  SGo  RSadovnick  DDuara  RKukull  WAChui  HEdeki  TGriffith  PAFriedland  RPBachman  DFarrer  L Depression as a risk factor for Alzheimer disease: the MIRAGE Study. Arch Neurol 2003;60753- 759
PubMedArticle
9.
Jorm  AF History of depression as a risk factor for dementia: an updated review. Aust N Z J Psychiatry 2001;35776- 781
PubMedArticle
10.
Small  BJFratiglioni  LViitanen  MWinblad  BBackman  L The course of cognitive impairment in preclinical Alzheimer's disease: three- and 6-year follow-up of a population-based sample. Arch Neurol 2000;57839- 844
PubMedArticle
11.
Elias  MFBeiser  AWolf  PAAu  RWhite  RFD’Agostino  RB The preclinical phase of Alzheimer's disease: a 22-year prospective study of the Framingham cohort. Arch Neurol 2000;57808- 813
PubMedArticle
12.
Chen  PRatcliff  RBelle  SHCauley  JADeKosky  STGanguli  M Patterns of cognitive decline in pre-symptomatic Alzheimer's disease: a prospective community study. Arch Gen Psychiatry 2001;58853- 858
PubMedArticle
13.
Butters  MABecker  JTNebes  RDZmuda  MDMulsant  BHPollock  BGReynolds  CF  III Changes in cognitive functioning followed treatment of late-life depression. Am J Psychiatry 2000;1571949- 1954
PubMedArticle
14.
Nebes  RDPollock  BGHouck  PRButters  MAMulsant  BHZmuda  MDReynolds  CF Persistence of cognitive impairment in geriatric depressed patients following antidepressant treatment: a randomized double-blind clinical trial with nortriptyline and paroxetine. J Psychiatr Res 2003;3799- 108
PubMedArticle
15.
Ganguli  MBelle  SRatcliff  GSeaberg  EHuff  FJvon der Porten  KKuller  LH Sensitivity and specificity for dementia of population-based criteria for cognitive impairment: the MoVIES project. J Gerontol 1993;48M152- M161
PubMedArticle
16.
Ganguli  MLytle  MReynolds  MDDodge  H Random vs volunteer selection for a community-based study. J Gerontol A Biol Sci Med Sci 1998;53M39- M46
PubMedArticle
17.
Radloff  LS The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1385- 401Article
18.
Ganguli  MGilby  JSeaberg  EBelle  S Depressive symptoms and associated factors in a rural elderly population: the MoVIES project. Am J Geriatr Psychiatry 1995;3144- 160Article
19.
Lassila  HStoehr  GPGanguli  MSeaberg  ECGilby  JEBelle  SHEchement  DA Factors associated with the use of prescription medications in an elderly rural population: the MoVIES Project. Ann Pharmacother 1996;30589- 595
PubMed
20.
Ganguli  MMulsant  BHRichards  SStoehr  GPMendelsohn  AB Antidepressant use over time in a rural older adult population: the MoVIES project. J Am Geriatr Soc 1997;451501- 1502
PubMed
21.
Morris  JCHeyman  AMohs  RCHughes  JPvan Belle  GFillenbaum  GMellits  EDClark  C The Consortium to Establish a Registry for Alzheimer's Disease (CERAD), I: clinical and neuropsychological assessment of Alzheimer's disease. Neurology 1989;391159- 1165
PubMedArticle
22.
Folstein  MFFolstein  SEMcHugh  PR “Mini-Mental State”: a practical method of grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12189- 198
PubMedArticle
23.
Rosen  WGMohs  RCDavis  KL A new rating scale for Alzheimer's disease. Am J Psychiatry 1984;1411356- 1364
PubMed
24.
Becker  JTBoller  FSaxton  JAMcGonigle-Gibson  KL Normal rates of forgetting verbal and non-verbal material in Alzheimer's disease. Cortex 1987;2359- 72
PubMedArticle
25.
Borkowksi  JGBenton  ALSpreen  O Word fluency and brain damage. Neuropsychologia 1967;5135- 140Article
26.
Benton  ALDe Hamsher  K Multilingual Aphasia Examination.  Iowa City University of Iowa1976;
27.
Kaplan  EFGoodglass  GHWeintraub  S The Boston Naming Test.  Media, Pa Lea & Febiger1983;
28.
Freedman  MLeach  LKaplan  E Clock Drawing: A Neuropsychological Analysis.  New York, NY Oxford University Press Inc1994;
29.
Reitan  RM The relation of the Trailmaking Test to organic brain damage. J Consult Psychol 1955;19393- 394
PubMedArticle
30.
Ganguli  MRatcliff  GBelle  SHuff  FJKancel  MJFischer  LKuller  LH Effects of age, gender, and education on cognitive tests in an elderly rural community sample: norms from the Monongahela Valley Independent Elders Survey (MoVIES). Neuroepidemiology 1991;1042- 52
PubMedArticle
31.
Ganguli  MRatcliff  GDeKosky  S Cognitive test scores in community-based rural elderly with and without dementia. Aging Ment Health 1997;1176- 180Article
32.
Ratcliff  GDodge  HHBirzescu  MAGanguli  M Tracking cognitive functioning over time: ten-year longitudinal data from a community-based study. Appl Neuropsychol 2003;1076- 88
PubMedArticle
33.
Ganguli  MDodge  HHChen  PBelle  SDeKosky  ST Ten-year incidence of dementia in a rural elderly US community population: the MoVIES Project. Neurology 2000;541109- 1116
PubMedArticle
34.
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R).  Washington, DC American Psychiatric Association1987;
35.
Hughes  CPBerg  LDanziger  WLCoben  LAMartin  RL A new clinical scale for the staging of dementia. Br J Psychiatry 1982;140566- 572
PubMedArticle
36.
Laird  NMWare  JH Random-effects models for longitudinal data. Biometrics 1982;38963- 974
PubMedArticle
37.
SAS Institute Inc, SAS/STAT User's Guide, Version 8: Software Manual.  Cary, NC SAS Institute Inc2000;
38.
Modrego  PJFerrandez  J Depression in patients with mild cognitive impairment increases the risk of developing dementia of the Alzheimer type: a prospective cohort study. Arch Neurol 2004;611290- 1293
PubMedArticle
39.
Airaksinen  ELarsson  MLundbgerg  IForsell  Y Cognitive functions in depressive disorders: evidence from a population-based study. Psychol Med 2004;3483- 91
PubMedArticle
40.
Alexopoulos  GS Frontostriatal and limbic dysfunction in late-life depression. Am J Geriatr Psychiatry 2002;10687- 695
PubMedArticle
41.
Krishnan  KRTaylor  WDMcQuoid  DRMacFall  JRPayne  MEProvenzale  JMSteffens  DC Clinical characteristics of magnetic resonance imaging–defined subcortical ischemic depression. Biol Psychiatry 2004;55390- 397
PubMedArticle
42.
Sheline  YISanghavi  MMintun  MAGado  MH Depression duration but not age predicts hippocampal volume loss in medically healthy women with recurrent major depression. J Neurosci 1999;195034- 5043
PubMed
43.
Dufouil  CFuhrer  RDartigues  JFAlperovitch  A Longitudinal analysis of the association between depressive symptomatology and cognitive deterioration. Am J Epidemiol 1996;144634- 641
PubMedArticle
44.
Vinkers  DJGussekloo  JStek  MLWesterdorp  RGvan der Mast  RC Temporal relationship between depression and cognitive impairment in old age: prospective population-based study. BMJ 2004;329881
PubMedArticle
45.
Ganguli  MDodge  HMulsant  BH Rates and predictors of mortality in an aging, rural, community-based cohort: the role of depression. Arch Gen Psychiatry 2002;591046- 1052
PubMedArticle
46.
Wilson  RSMendes de Leon  CBennett  DABienias  JLEvans  DA Depressive symptoms and cognitive decline in a community population of older persons. J Neurol Neurosurg Psychiatry 2004;75126- 129
PubMedArticle
47.
Blazer  DGHybels  CFPieper  CF The association of depression and mortality in elderly persons; a case for multiple, independent pathways. J Gerontol A Biol Sci Med Sci 2001;56M505- M509
PubMedArticle
×