Brainstem Aminergic Nuclei and Late-Life Depressive Symptoms

Importance  The neurobiologic basis of late-life depressive symptoms is not well understood.

Objective  To test the hypothesis that neurodegeneration and neuronal density in brainstem aminergic nuclei are related to late-life depressive symptoms.

Design, Setting, Participants, and Exposure  Longitudinal clinicopathological cohort study at residences of participants in the Chicago, Illinois, metropolitan area. Participants included 124 older persons without dementia in the Rush Memory and Aging Project who had annual evaluations for a mean (SD) of 5.7 (2.8) years, died, and underwent a postmortem neuropathological examination that provided estimates of the densities of Lewy bodies, neurofibrillary tangles, and aminergic neurons in the locus ceruleus, dorsal raphe nucleus, substantia nigra, and ventral tegmental area.

Main Outcomes and Measures  The number of depressive symptoms (mean [SD], 1.61 [1.48]; range, 0-6; skewness, 0.94) on the Center for Epidemiological Studies Depression Scale averaged across annual evaluations.

Results  Brainstem Lewy bodies were associated with depressive symptoms, and the association was attenuated in those taking antidepressant medication. Brainstem tangles were associated with more depressive symptoms in those without cognitive impairment but with fewer symptoms in those with mild cognitive impairment. Lower density of tyrosine hydroxylase–immunoreactive neurons in the ventral tegmental area was robustly associated with a higher level of depressive symptoms (mean [SE] estimate, −0.014 [0.003]; P < .001; 16.3% increase in adjusted R²). The association was not modified by medication use or cognitive impairment. Neither tyrosine hydroxlyase–immunoreactive neurons in the locus ceruleus nor tryptophan hydroxlyase–immunoreactive neurons in the dorsal raphe nucleus were related to depressive symptoms.

Conclusions and Relevance  The results suggest that the mesolimbic dopamine system, especially the ventral tegmental area, has an important role in late-life depressive symptoms.

Late-life depressive symptoms are common¹ and are associated with increased risk of disability,² dementia,³ and death.⁴ Because brainstem aminergic nuclei have an important role in the pathogenesis of depression⁵ and are early sites of age-related neurodegeneration,^6,7 it has long been hypothesized that late-life depressive symptoms are partly due to neurodegenerative changes in these nuclei.^8-11 Support for the hypothesis has been inconsistent, likely because of several factors. Few studies have had more than 50 participants^8,10,12 or used continuous measures of depressive symptoms,^9,10 limiting statistical power. Also, much research has been conducted among individuals with dementia.^6,8-10,13-15 However, the presence of dementia introduces substantial error into the measurement of depressive symptoms,^16,17 in addition to extensive pathological conditions elsewhere in the brain, and the neurodegenerative changes associated with dementia are robustly related to cognitive,¹⁸ sensory,¹⁹ and motor²⁰ symptoms years before dementia onset, suggesting that associations with depressive symptoms might be easier to identify during this period.¹¹ Finally, despite evidence that brainstem dopamine neurons have an important role in regulating depression-related behavior,^21-23 previous investigations have primarily focused on the locus ceruleus and dorsal raphe nucleus.

The present study examines the association of neurodegenerative lesions and aminergic neurons in 4 brainstem nuclei with the level of depressive symptoms in late life. At annual intervals for a mean of 5.7 years, older persons without dementia had structured evaluations that included a standard self-report measure of depressive symptoms. At death, a uniform neuropathological examination was performed that yielded estimates of the densities of Lewy bodies, neurofibrillary tangles, and aminergic neurons in the locus ceruleus, dorsal raphe nucleus, substantia nigra, and ventral tegmental area. We tested the hypothesis that higher pathological burden and lower neuronal density are associated with more depressive symptoms.

The Rush Memory and Aging Project is an ongoing clinicopathological study that began September 1, 1997, and includes annual clinical evaluations and brain autopsy at death.²⁴ Persons were recruited from social service agencies, subsidized housing facilities, retirement communities, and churches in the Chicago, Illinois, metropolitan region. After a presentation on the project, interested individuals met for further discussions with staff, who obtained written informed consent. The project was approved by the institutional review board of Rush University Medical Center, Chicago, Illinois.

There had been 550 deaths (among 1533 participants) at the time of these analyses. A brain autopsy had been performed in 439 (79.8%), and a uniform neuropathological examination had been completed in the first consecutive 417 individuals. From these, 170 individuals were selected to provide a wide range of cognitive, motor, and affective functioning proximate to death for a clinicopathological substudy of brainstem aminergic nuclei as described elsewhere.^25,26 We excluded 45 individuals with a diagnosis of dementia (see the Clinical Evaluation subsection), and 1 person was missing data about depressive symptoms. Analyses are based on the remaining 124 individuals, who had a mean (SD) age at death of 87.7 (5.7) years. They had completed a mean (SD) of 14.3 (2.5) years of education, 69.4% were women, and 43.5% had mild cognitive impairment.

A 10-item version²⁷ of the Center for Epidemiological Studies Depression Scale²⁸ was used to assess depressive symptoms at each annual clinical evaluation. For each item, the interviewer read a brief statement (eg, “I felt depressed”), and the participant indicated whether or not he or she had felt that way much of the past week. The score is the number of symptoms reported (range, 0-10). This 10-item scale is less burdensome than the original 20-item scale but has similar psychometric properties,²⁷ and its measures have been associated with adverse health outcomes, including dementia²⁹ and death.⁴ Supported by prior factor analyses,^27,30 subscores were created for negative affect (3 items, including “I felt depressed,” “I felt lonely,” and “I felt sad”), positive affect (2 items, including “I was happy” and “I enjoyed life”), somatic symptoms (3 items, including “I felt like everything I did was an effort,” “My sleep was restless,” and “I could not get going”), and interpersonal problems (2 items, including “People were unfriendly” and “I felt that people disliked me”), as previously described.³¹

The annual evaluations included a structured medical history, neuropsychological testing, and neurological examination.²⁴ Prescription and over-the-counter medications were inspected, identified, and coded using the Medi-Span Master Drug Database system.³² The neurological examination included a modified version³³ of the motor portion of the Unified Parkinson’s Disease Rating Scale,³⁴ as previously described.³⁵ After each evaluation, an experienced clinician diagnosed dementia and mild cognitive impairment. Classification of dementia was based on the criteria of the joint working group of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer Disease and Related Disorders Association, which require a history of cognitive decline and impairment in at least 2 cognitive domains.³⁶ The diagnosis of mild cognitive impairment required impairment in 1 or more cognitive domains in the absence of dementia.³⁷ Upon death, all clinical data were reviewed by a board-certified neurologist blinded to all pathological data, and final diagnoses of mild cognitive impairment and dementia were made.

The brain was removed a median of 6.0 hours (interquartile range, 4.8-8.6 hours) after death, which occurred a median of 9.3 months (interquartile range, 5.5-12.9 months) after the last complete clinical evaluation. Examiners blinded to all clinical data followed a standard protocol for tissue preservation, tissue sectioning, and quantification of pathological findings.²⁰

To examine the nuclei of interest, consecutive transverse blocks of fixed tissue were obtained at 4 brainstem levels, as previously described.^25,26 The first block included the substantia nigra and paranigral nucleus of the ventral tegmental area (Figure 1A); the second block included the rostral level of the dorsal raphe nucleus, the trochlear nucleus, and the decussation of the superior cerebellar peduncles; the third block included the caudal level of the dorsal raphe nucleus, rostral levels of the locus ceruleus nuclei, and the mesencephalic nucleus of the trigeminal nerve; and the fourth block included the main body of the locus ceruleus. Following routine processing, blocks were embedded in paraffin, and 6-μm sections stained with hematoxylin-eosin were used to survey the regions of interest. Blocks without anatomically matching levels of the regions of interest were excluded. Immunohistochemistry was performed using an indirect immunoenzyme horseradish peroxidase method with 3,3′diaminobenzidine as the chromogen in an autostainer (Leica Bond Max; Leica Microsystems) using a kit (Bond Polymer Refine Detection Kit; Leica Microsystems). Tyrosine hydroxylase–positive and tyrosine hydroxylase–negative neurons in the substantia nigra, ventral tegmental area, and locus ceruleus were identified using a monoclonal anti–tyrosine hydroxylase antibody (1:750; ImmunoStar). Nickel chloride (5%) added to the 3,3′diaminobenzidine substrate (Vector Laboratories) enhanced the signal of the tyrosine hydroxylase–positive neurons and their processes to a brown-black color. The count of ventral tegmental area neurons was based on the paranigral nucleus. The figures show examples of high and low densities of tyrosine hydroxylase–positive neurons in the paranigral nucleus of the ventral tegmental area (Figure 1B and C), substantia nigra (Figure 2A and B), and locus ceruleus (Figure 2C and D). Serotonergic neurons in the dorsal raphe nucleus were identified using an anti–tryptophan hydroxylase monoclonal antibody (1:1000; Sigma Chemical Company). Examples of high and low densities of tryptophan hydroxylase–positive neurons are shown in Figure 2E and Figure 2F, respectively.

Neuronal density, neurofibrillary tangles, and Lewy bodies were quantitated at 1 level of the substantia nigra and the ventral tegmental area, at 2 levels of the dorsal raphe nucleus, and at 2 levels of the right and left locus ceruleus. The immunohistochemical preparations were used to outline the regions of interest using software (Stereo Investigator Program; MBF Biosciences) attached to a microscope (BX60; Olympus) with a motorized stage. These outlines were superimposed on slides used to determine the neuron, neurofibrillary tangle, and Lewy body counts in the locus ceruleus, dorsal raphe nucleus, and ventral tegmental area. The counting frame and motorized stage facilitated counting in contiguous fields in the outlined regions. In these fields, unstained and immunostained neurons with nuclei were manually selected at a magnification of ×400. Densities of neurons, tangles, and Lewy bodies in the 4 quadrants of the substantia nigra were determined, as described previously.²⁵

Analyses were based on the density per millimeter squared of immunostained neurons. The interrater reliability of these neuronal density measures, estimated among a subset of 12 participants, was high for the locus ceruleus (r = 0.97), dorsal raphe nucleus (r = 0.96), substantia nigra (r = 0.97), and ventral tegmental area (r = 0.99) (P < .001 for all). Their discriminant validity is supported by previous research linking neuronal density in the substantia nigra to subclinical parkinsonism²⁵ and linking neuronal density in the locus ceruleus to cognitive decline.²⁶

Neurofibrillary tangles were identified using a monoclonal anti–paired helical filament tau phosphorylated antibody (clone AT8, 1:2000; Thermo Scientific), and Lewy bodies were identified using a monoclonal phosphorylated antibody to α-synuclein (1:20 000; Wako Chemical USA Inc).²⁰ The total numbers of neurofibrillary tangles and Lewy bodies in each brainstem nucleus within the outlined area referred to above were counted manually in the regions of interest, and results were expressed as density per millimeter squared.

To assess pH, 250 to 300 mg of frozen cerebellum was homogenized (Polytron PT10-35GT; Kinematica) in 5 volumes of nuclease-free water adjusted to pH 7.2. The pH of each homogenate was measured twice at room temperature (sympHony SB70 pH meter; VWR). Analyses are based on the mean of the 2 readings, which were highly correlated (r = 0.998, P < .001).

All analyses included terms to adjust for the effects of sex, age at death, and education. A mixed-effects model was used to determine whether depressive symptoms changed over time. The mean depressive symptom score was regressed on postmortem variables in a series of linear regression models. The initial model had a term for the composite measure of brainstem Lewy bodies. The model was repeated using the following variables: controlling for parkinsonism; with depression subscores as outcomes; separately analyzing Lewy bodies within each nucleus; with an indicator for antidepressant use and its interaction with Lewy bodies; with an indicator for mild cognitive impairment and its interaction with Lewy bodies; and controlling for pH, postmortem interval, and time from the last clinical evaluation to death. Similar analyses were conducted with the measures of brainstem tangle density. The 4 neuronal density scores were modeled separately, ventral tegmental area neurons and nigral neurons were modeled together, and ventral tegmental area neurons were modeled controlling for pH, postmortem interval, and time from the last clinical evaluation to death. We also assessed whether ventral tegmental area neurons were related to depression subscores, accounted for the association of Lewy bodies with depressive symptoms, or interacted with antidepressant medication use or mild cognitive impairment.

At the initial evaluation, depressive symptom scores ranged from 0 to 5 (mean [SD], 1.48 [1.66]; skewness, 1.04). During a mean (SD) of 5.7 (2.8) annual follow-up years, depressive symptoms did not seem to change (eFigure 1 in the Supplement). In a mixed-effects model adjusted for age, sex, and education, there was no evidence that the number of depressive symptoms changed over time (mean [SD] estimate for time, 0.005 [0.03]; P = .85) or that residual variation not captured by the model was related to the brainstem postmortem variables. Therefore, we averaged scores across evaluations to create a composite index of late-life depressive symptoms (mean [SD], 1.61 [1.48]; range, 0-6; skewness, 0.94), as done in previous clinicopathological research.^38,39 The composite score was not related to sex (t₁₂₂ = 1.42, P = .16), age at death (r = −0.00, P = .97), education (r = −0.09, P = .30), pH (r = −0.12, P = .19), postmortem interval (r = 0.11, P = .22), or time from the last clinical evaluation to death (r = 0.17, P = .05).

Because the measures of tangles and Lewy bodies in the nuclei clustered in a factor analysis (Table), we used composite measures of tangle density (n = 124; mean [SD], 1.66 [1.51]; range, 0-5.44; skewness, 1.80) and Lewy body density (n = 124; mean [SD], 0.09 [ 0.29]; range, 0-0.88; skewness, 3.52) based on all 4 nuclei in primary analyses. The composite measures were not correlated (r = 0.13, P = .16). Higher density of tangles (r = 0.24, P = .006) but not Lewy bodies (r = 0.03, P = .75) was related to older age at death; neither was related to sex, education, pH, postmortem interval, or time from the last clinical evaluation to death.

In a linear regression model adjusted for sex, age at death, and education, higher density of brainstem Lewy bodies was associated with a higher level of depressive symptoms (mean [SE] estimate, 1.133 [0.443]; P = .01; 4.0% increase in adjusted R²). Results were similar after controlling for pH, postmortem interval, and time from the last clinical evaluation to death (mean [SE] estimate, 1.159 [0.437]; P = .009) or parkinsonism proximate to death (mean [SE] estimate, 1.168 [0.450]; P = .01). The association was observed in 3 of 4 nuclei, with mean (SE) estimates of 0.363 (0.175) for locus ceruleus (P = .04), 5.302 (2.000) for dorsal raphe nucleus (P = .009), 1.301 (0.609) for substantia nigra (P = .04), and 0.420 (0.368) for ventral tegmental area (P = .26). In analyses of subscores, higher Lewy body density was marginally related to negative affect (mean [SE] estimate, 0.495 [0.237]; P = .04), robustly related to somatic symptoms (mean [SE] estimate, 0.634 [0.204]; P = .002), and unrelated to positive affect (mean [SE] estimate, 0.007 [0.107]; P = .95) or interpersonal problems (mean [SE] estimate, −0.005 [0.033]; P = .88).

We conducted additional analyses to determine whether antidepressant medication use (32.3% during at least part of the observation period) (eTable 1 in the Supplement) or mild cognitive impairment (43.5%) was obscuring the association of Lewy bodies with depressive symptoms. There was no interaction with mild cognitive impairment (mean [SE] estimate, 0.859 [0.991]; P = .39), but an interaction was observed with medication use (mean [SE] estimate, −2.476 [0.913]; P = .008): the association of Lewy bodies with depressive symptoms was attenuated in those taking antidepressant medication compared with those not taking antidepressant medication.

The composite measure of brainstem tangle density was not related to depressive symptoms (mean [SE] estimate, −0.038 [0.090]; P = .68); adjustment for pH, postmortem interval, and time from the last clinical evaluation to death yielded comparable results (mean [SE] estimate, 0.087 [0.093]; P = .35), and the results of a subsequent analysis with a quadratic term for tangles (mean [SE] estimate, −0.061 [0.039]; P = .12) did not suggest a nonlinear association. No association with tangles was present in the individual nuclei, with mean (SE) estimates of 0.031 (0.071) for locus ceruleus (P = .66), −0.007 (0.045) for dorsal raphe nucleus (P = .88), 0.482 (0.332) for substantia nigra (P = .15), and −0.105 (0.087) for ventral tegmental area (P = .23). There was no interaction with antidepressant medication use (mean [SE] estimate, 0.027 [0.217]; P = .90), but an interaction was observed with mild cognitive impairment (mean [SE] estimate, −0.491 [0.190]; P = .01): higher tangle density was associated with higher level of depressive symptoms in those without cognitive impairment but with lower level of depressive symptoms in those with mild cognitive impairment (eFigure 2 in the Supplement).

Because of missing or insufficient tissue, 6 to 13 participants per nucleus had missing neuronal data. eFigure 3 in the Supplement shows the distributions of aminergic neuronal density in the locus ceruleus (n = 118; mean [SD], 44.1 [17.8]; skewness, 0.39), dorsal raphe nucleus (n = 117; mean [SD], 105.4 [28.3]; skewness, 0.54), substantia nigra (n = 113; mean [SD], 31.1 [11.9]; skewness, 0.01), and ventral tegmental area (n = 111; mean [SD], 82.8 [47.9]; skewness, 0.33). None of the density measures were related to sex, age at death, or education; their crude correlations with pH, postmortem interval, time from the last clinical evaluation to death, and brainstem pathological conditions are summarized in eTable 2 in the Supplement.

We constructed a series of linear regression models adjusted for sex, age at death, and education to assess the hypothesized association between brainstem aminergic neurons and depressive symptoms. Lower density of tyrosine hydroxylase–immunoreactive neurons in the ventral tegmental area was related to higher level of depressive symptoms in the ventral tegmental area (Figure 3) (mean [SE] estimate, −0.014 [0.003]; P < .001; 16.3% increase in adjusted R²) and substantia nigra (mean [SE], estimate, −0.025 [0.012]; P = .03; 1.5% increase in adjusted R²) but not the locus ceruleus (mean [SE] estimate, −0.011 [0.008]; P = .18). There was no association for tryptophan hydroxylase–immunoreactive neurons in the dorsal raphe nucleus (mean [SE] estimate, 0.000 [0.005]; P > .99). When the ventral tegmental area and substantia nigra were included in the same model, only neurons in the ventral tegmental area were related to depressive symptoms (mean [SE] estimate, −0.012 [0.003]; P < .001). Adjustment for pH, postmortem interval, and time from the last clinical evaluation to death did not affect the association of ventral tegmental area neurons with depressive symptoms (mean [SE] estimate, −0.013 [0.003]; P < .001). The use of antidepressant medication was not related to the density of ventral tegmental area neurons (t₁₀₉ = −1.21, P = .23). Neither antidepressant use (mean [SE] estimate, 0.006 [0.005]; P = .26) nor the presence of mild cognitive impairment (mean [SE] estimate, 0.001 [0.006]; P = .88) modified the association of ventral tegmental area neurons with depressive symptoms.

Because lower density of ventral tegmental area neurons was associated with higher density of brainstem Lewy bodies (r = −0.19, P = .046), we modeled them together. In this analysis, the association of Lewy bodies with depressive symptoms was attenuated (mean [SE] estimate, 0.615 [0.434]; P = .16), but the association of neurons with symptoms persisted (mean [SE] estimate, −0.013 [0.003]; P < .001), suggesting that the association of brainstem Lewy bodies with depressive symptoms was in part due to the density of ventral tegmental area neurons.

To further investigate how ventral tegmental area neurons were related to depressive symptoms, we analyzed different symptom domains. Lower density of tyrosine hydroxylase–immunoreactive neurons in the ventral tegmental area was associated with negative affect (mean [SE] estimate, −0.007 [0.002]; P < .001), lack of positive affect (mean [SE] estimate, −0.002 [0.001]; P = .003), and somatic symptoms (mean [SE] estimate, −0.004 [0.001]; P = .007) but not with interpersonal problems (mean [SE] estimate, −0.0003 [0.0001]; P = .17).

During a mean of approximately 5 years of observation, depressive symptoms were assessed annually in 124 older persons who subsequently died and had a postmortem neuropathological examination. Lower density of tyrosine hydroxylase–immunoreactive neurons in the ventral tegmental area was associated with higher level of depressive symptoms, with 16.3% of the variability in symptoms explained by individual differences in density. The results suggest that the mesolimbic dopamine system has an important role in late-life depressive symptoms.

Animal models^21-23 and human neuroimaging studies^40,41 have highlighted the potential importance of mesolimbic dopaminergic mechanisms in depression, particularly for aspects of apathy and anhedonia. However, we are unaware of previous clinicopathological studies of ventral tegmental area neuronal density and depressive symptoms. Higher neuronal ratings in the substantia nigra were associated with lower likelihood of depressive symptoms in a study¹⁴ of older persons without dementia. The association in the present study between the density of nigral dopaminergic neurons and depressive symptoms is consistent with this finding and extends it by more precisely quantifying nigral neurons and by showing that the association is largely confounded by the density of dopaminergic neurons in the ventral tegmental area. We did not find the density of noradrenergic neurons in the locus ceruleus or serotonergic neurons in the dorsal raphe nucleus to be related to depressive symptoms. These null results agree with some studies^42,43 but not another study⁴⁴ of the locus ceruleus and are consistent with the weight of the evidence from studies of dorsal raphe nucleus neurons in depression, which have reported levels to be lower,^11,45 no different,^12,46 and higher⁴⁷ than controls.

The factors responsible for the observed individual differences in the density of dopaminergic neurons in the ventral tegmental area are unknown, and because the present study did not include persons who died before old age, it is uncertain whether the association between dopaminergic neurons in the ventral tegmental area and depressive symptoms is specific to late life. That the association was observed after excluding individuals with dementia and controlling for postmortem markers linked to dementia suggests that some of the variability in dopamine neuron density was long-standing. Therefore, it is possible that a less populated ventral tegmental area due to neurodevelopmental or other causes occurring earlier in adulthood somehow increases vulnerability to depressive symptoms. However, a substantial subgroup had few dopamine neurons in the ventral tegmental area (eFigure 3 in the Supplement), and older age had an almost significant correlation with lower neuronal density (eTable 2 in the Supplement), suggesting that age-related pathological processes are also involved. Because Parkinson disease also involves a loss of brainstem dopaminergic neurons for uncertain reasons, it provides a useful point of comparison. Depressive symptoms are common in the disease⁴⁸ and are reduced by pramipexole hydrochloride use,^49,50 a dopamine agonist, suggesting that late-life depressive symptoms in some persons without Parkinson disease might also be alleviated by pharmacologic enhancement of dopamine transmission with agonists or with reuptake inhibitors such as bupropion hydrochloride⁵¹ or nomifensine maleate.⁵²

Neurodegenerative lesions in brainstem aminergic nuclei were related to depressive symptoms, but the associations were weak and conditional on other factors. Therefore, the presence of Lewy bodies in brainstem aminergic nuclei was related to depressive symptoms, consistent with a previous clinicopathological study¹² and the prominent place of depressive symptoms in synucleinopathies, such as Parkinson disease and Lewy body disease. However, the association was only seen in those not taking antidepressants, and controlling for ventral tegmental area neuronal density eliminated it.

There was no overall association between brainstem tangles and depressive symptoms, consistent with previous research in persons without dementia^12,44 and with some studies^9,44 but not another study¹³ of persons with dementia. In contrast with previous research, the present study included clinical classification of mild cognitive impairment, and higher brainstem tangle density was associated with a higher level of depressive symptoms in persons without cognitive impairment but with a lower level of depressive symptoms in those with mild cognitive impairment. This suggests that tangles make a modest contribution to depressive symptoms before the onset of cognitive impairment but that, as cognitive problems develop, depressive symptoms may be less consistently reported, possibly because they are harder to remember or acknowledge.^16,17 In addition, incipient cognitive decline may disrupt the continuity of mood states and their ability to regulate behavior. These observations agree with longitudinal data that suggest little increase in depressive symptoms during the development of mild cognitive impairment and dementia.^31,53,54

Study strengths and limitations should be noted. There was a high rate of participation in the annual clinical evaluations and brain autopsy, minimizing bias due to selective attrition. Uniform annual evaluations and the application of widely used diagnostic criteria allowed us to reliably classify mild cognitive impairment and dementia. The measure of depressive symptoms was based on a standard scale administered at regular intervals for several years, which probably enhanced assessment of the enduring tendency to be depressed while reducing measurement error and the influence of episodic fluctuations in mood. An important limitation is that the cohort was selected, so replication of these results will be important. In addition, the findings for the ventral tegmental area are based on the paranigral nucleus, and it is uncertain whether they will generalize to other subpopulations of ventral tegmental area dopaminergic neurons.⁵⁵

Submitted for Publication: November 13, 2012; final revision received February 18, 2013; accepted March 29, 2013.

Corresponding Author: Robert S. Wilson, PhD, Rush Alzheimer’s Disease Center, Rush University Medical Center, 600 S Paulina St, Ste 1038, Chicago, IL 60612 (rwilson@rush.edu).

Published Online: October 16, 2013. doi:10.1001/jamapsychiatry.2013.2224.

Author Contributions: Dr Wilson had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Wilson, Nag, Boyle, Arnold, Bennett.

Acquisition of data: Nag, Hizel, Schneider, Bennett.

Analysis and interpretation of data: Wilson, Nag, Boyle, Yu, Buchman, Shah, Schneider, Arnold, Bennett.

Drafting of the manuscript: Wilson, Nag, Hizel, Schneider, Arnold.

Critical revision of the manuscript for important intellectual content: Wilson, Nag, Boyle, Yu, Buchman, Shah, Schneider, Arnold, Bennett.

Statistical analysis: Yu.

Obtained funding: Wilson, Buchman, Bennett.

Administrative, technical, and material support: Nag, Boyle, Hizel, Schneider, Arnold, Bennett.

Study supervision: Schneider, Bennett.

Conflict of Interest Disclosures: None reported.

Funding/Support: This research was supported by grants R01AG17917, R01AG15819, R01AG24871, R01AG34374, R01AG33678, R01AG43379, and R01NS78009 from the National Institutes of Health and by the Illinois Department of Public Health.

Role of the Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Traci Colvin, MPH, and Karen Skish, MS, assisted with study coordination; John Gibbons, MS, and Greg Klein, MS, contributed to data management; and Wenqing Fan, MS, provided statistical programming. We thank the neuropathologic laboratory of the Rush Memory and Aging Project and the many Illinois residents for participating in the project.