Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Am J Geriatr Psychiatry. Author manuscript; available in PMC 2013 August 1.
Published in final edited form as:
PMCID: PMC3337345

Neural Correlates Associated with Cognitive Decline in Late-life Depression

Lihong Wang, M.D., Ph.D.,1,2 Guy G. Potter, Ph.D.,2 R.K. Ranga Krishnan, M.D.,3 Florin Dolcos, Ph.D.,4 Gwenn S. Smith, Ph.D.,5 and David C. Steffens, M.D.2



Persistent cognitive impairment (PCI) after remission of depressive symptoms is a major adverse outcome of late-life depression (LLD). The purpose of this study was to examine neural substrates associated with PCI in LLD.


Longitudinal study


Outpatient depression treatment study at Duke University


Twenty-three patients with LLD completed a two-year follow-up study, and were in a remitted or partially remitted state at Year 2.


At first entry to the study (Year 0), all participants had a functional magnetic resonance imaging (fMRI) scan while performing an emotional oddball task. For the purpose of this report, the primary fMRI outcome was brain activation during target detection, which is a measure of executive function. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropsychological battery was used to assess cognitive status yearly, and the Montgomery-Asberg Depression Rating Scale was used to assess severity of depression at Year 0 and every 6 months thereafter for two years. We investigated changes in brain activation at Year 0 associated with PCI over two years.


Patients with PCI at the two-year follow-up date had significantly decreased activation at Year 0 in the dorsal anterior cingulate cortex, hippocampus, inferior frontal cortex, and insula compared to non-PCI patients.


Our results suggest individuals who have LLD with PCI have decreased activation in the similar neural networks associated with the development of Alzheimer’s disease among nondepressed individuals. Measuring neural activity in these regions in individuals with LLD may help identify patients at-risk for cognitive impairment.

Depression in late adulthood is associated with profound changes in cognitive and emotional functioning. Cognitive impairment in some patients may persist despite remission of depressive symptoms 12. Continuous impairment or further decline in cognition after remission of depression symptoms for is often described as persistent cognitive impairment (PCI)34. Patients with PCI are of particular clinical concern because they have a higher risk of developing of dementia including Alzheimer’s disease (AD) 45. Therefore, understanding the neural mechanisms associated with PCI in late-life depression (LLD) is important for early recognition and clinical intervention among these individuals. Although PCI has been widely recognized as an adverse outcome of LLD, little is known about its neural substrates beyond findings of reduced hippocampal volume6.

Extensive neuroimaging studies on Alzheimer’s disease and mild cognitive impairment (MCI, prodromal AD) have been conducted in recent years4, 716, and provide some information on the potential neural substrates of PCI in LLD. Structural MRI has confirmed the importance of volumetric changes in the medial temporal lobe (MTL) memory system, and has specifically identified reduced volume in the entorhinal cortex as a biomarker for prodromal AD11. Recent functional resonance imaging (fMRI) studies have also revealed prodromal AD affects a large-scale episodic memory network, including the anterior cingulate, insula, inferior frontal and parietal cortices, in addition to regions typically associated with the MTL memory system.8, 1718.

Although there are reports showing alterations in the hippocampus, frontal and parietal cortices, and cingulate in neuroimaging studies in LLD 1922, little is known about which regions are specifically associated with PCI and cognitive decline. We previously studied neural substrates of depression in acute and remitted patients compared to control participants 21. We found decreased activation in the dorsolateral prefrontal cortex (dlPFC), which seemed to be depressive-state related. In addition, both the acute and remitted groups showed reduced activation in the anterior-middle cingulate (BA 24) and anterior portion of the posterior cortex (aPCC) compared to the control group during a target detection task. In that study, it was not clear whether the decreased activation in the cingulate cortex in the remitted patients was a residual effect of depression or whether it was related to possible cognitive impairment. Clarifying whether neural circuits involved in LLD with PCI are similar to those conventionally found among individuals with MCI (but never depressed) is important to better understanding the etiologies of both LLD and MCI/AD.

LLD is a heterogeneous disorder, and so are the potential causes of cognitive impairment during an acute depressive episode in LLD. There are two major types of depression-associated pathology that have been hypothesized to contribute to LLD as a risk factor for cognitive impairment. One is the progression of vascular lesions (subcortical gray matter and frontal deep white matter hyperintensities [DWMH]), and the other is neurotoxicity from the stress-glucocorticoid cascade. Accumulated evidence has suggested that the vascular pathology results in prefrontal-striatal dysfunction2326, whereas neurotoxicity from the chronic stress-glucocoticoid cascade leads to hippocampal damage6, 27. Butters and colleagues4 have proposed pathological pathways that might link the presence of cognitive impairment in LLD to AD. Individuals in this group are presumed to have accumulated AD pathology concurrent with depression-associated pathology. Concurrence of vascular pathology in LLD with accumulated AD pathology has been postulated as the most frequently occurring pathway leading to or accelerating the development of amnestic-multidomain MCI and AD along with cerebrovascular disease4. Importantly, this type of cognitive impairment results in cognitive decline over time and progression to AD. Another pathway proposed by Butters et al. is AD pathology concurrent with depression, which may reflect mood alteration due to the primary disease, and which has a diagnostic endpoint of AD. Therefore, it is very likely there is AD pathology among individuals with PCI in LLD, and one can predict that PCI in LLD should have deficits in the same regions that are essential for the development of AD in nondepressed individuals, such as the hippocampus and the cingulate cortex.

In the current study, we performed an fMRI scan during an emotional oddball task in LLD patients, and followed up the outcomes of cognition and depressive symptoms in the patients for two years. The goal of the study was to examine neural correlates of PCI occurring after a late-life episode of Major Depressive Disorder (MDD). We hypothesized that individuals with PCI have deficits in the same brain regions (such as the hippocampus) reported in nondepressed individuals who have MCI with AD pathology 8.



All participants were recruited from a depression treatment study at the NIMH-sponsored Conte Center for the Neuroscience of Depression in Late-Life at Duke University Medical Center (DUMC). In our initial fMRI study, we recruited 32 individuals who met DSM-IV criteria for MDD at the time of enrollment in the treatment study. All participants were over 60 years of age, and were either in an active depressive episode or in a remitted or a partially remitted state at the time they participated in the fMRI study. All participants completed a two-year clinical follow-up of their depressive symptoms and cognitive status. Among them, 23 patients were in remission or in a partially remitted state at the last observation (Year 2). This study focused on the 23 patients who were in at least partial remission at Year 2.

The exclusion criteria included (1) another major psychiatric illness; (2) alcohol or drug abuse or dependence; (3) neurological illness, such as dementia, stroke, and epilepsy; (4) medical illness, medication use, or disability that would prevent the participant from completing neurocognitive testing; and (5) contraindications to MRI. Two depressed individuals meeting criteria for co-morbid generalized anxiety disorder were not excluded since MDD was the primary diagnosis. A complete description of the study enrollment protocol is described elsewhere28. Detailed demographic and clinical assessments are listed in Table 1 for the participants who were included in our final analysis. The study was approved by the Institutional Review Board at Duke University, and all participants provided written informed consent after the procedures had been fully explained.

Table 1
Clinical Profile of Participants with Persistent Cognitive Impairment (PCI) Compared to Those without Persistent Cognitive Impairment (non-PCI)

Longitudinal Assessments

Depressed participants were enrolled in a longitudinal study where they received pharmacologic treatment based on a standardized algorithm 28. Among the 23 participants, 15 were receiving monotherapy (9 on an SSRI, 2 on venlafaxine, and 4 on bupropion), 2 were receiving combination treatment (SSRI and bupropion), and 6 were receiving no antidepressants at the time of the fMRI scan. The Montgomery-Åsberg Depression Rating Scale (MADRS)29 was used to assess symptom severity. At initial entry to the study (Year 0), the MADRS was completed by a geriatric psychiatrist, and by trained study personnel before the fMRI scan. The MADRS was administered at least once every six months during the two-year study period. All raters were trained on the MADRS, and high inter-rater reliability (kappa > .9) was established30. The mean (SD) MADRS for all participants at Year 0 was 12.9 (10.5), and at Year 2 was 5.5 (5.0). In the current study, we used the MADRS score to define remission. Participants with MADRS < 8 for at least 6 months were considered as remitted and participants with 8 < MADRS < 15 were considered in partial remission.

Screening to rule out prevalent dementia was conducted by a geriatric psychiatrist and included a clinical evaluation, medical record review, and consultation with referring physicians when applicable. All participants also completed the Mini Mental State Examination (MMSE). Any participants with a MMSE score below 25 were not included in the current study.

Individuals participating in the study received yearly neuropsychological assessment 28, which included the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropsychological battery. The CERAD battery has demonstrated accurate case ascertainment of MCI and dementia in large scale epidemiological studies, and is effective in tracking longitudinal changes in these conditions 3133. We used the CERAD Total Score (TS) as our index of cognitive impairment. The CERAD TS is a summary of all the individual neuropsychological test scores that constitute the CERAD battery, excluding the MMSE. This metric has demonstrated high accuracy in differentiating independent samples of normal adults, MCI patients, and AD34 patients. The CERAD TS has also demonstrated utility for tracking AD progression35. As described below, we used Chandler et al.’s 34 cutoff value between normal cognition and MCI to define our PCI group, which includes demographic adjustment of the raw CERAD TS score for age, education, and gender. The mean (SD) of the adjusted CERAD TS across all participants was 89.8(7.2) for Year 0 and 91.0 (9.1) for Year 2. Participants were considered as having PCI if they: 1) had an adjusted CERAD TS <85 at both the Year 0 and Year 2; 2) had no improvement in CERAD TS at Year 2 relative to Year 0; and 3) were not in an active depressive state at Year 2. Using these criteria, 7 out of the 23 participants were classified as PCI. Detailed demographic information for PCI and the non-PCI group is listed in Table 1.

Functional MRI Task

All participants completed a visual emotional oddball task at Year 0. Detailed description of stimuli and task design have been published previously 21. Briefly, there were four types of grayscale pictures presented randomly: 1) circles of varying sizes and luminance, serving as attentional targets, 2) sad distractors, 3) neutral distractors, and 4) phase-scrambled images of the distractors. The imaging session consisted of 10 runs, each containing 150 stimuli (stimulus duration = 1500 ms, inter-stimulus interval = 2000 ms). The participants’ task was to press a response button using their right index finger upon detection of a target oddball stimulus. The attentional target, sad distractors, and neutral distractors appeared with a presentation frequency of 3.33% each, with scrambled pictures comprising the remaining 90% of stimuli. Our previous studies have consistently shown that the dorsal executive function regions are activated during target detection, whereas the emotional system is activated by sad distractors36. Given that we were interested in brain regions associated with PCI, we focused our analysis on brain activation during target detection as the relevant brain regions were activated under this condition in our previous studies 21, 3637.

Image Acquisition and Analysis

Functional images were acquired on a 4.0 Tesla GE scanner with identical acquisition parameters as Wang et al.36. Oblique spoiled gradient-recalled acquisition images (three-dimensional, whole-brain) were acquired parallel to the anterior commissure (AC) -posterior commissure (PC) plane thickness of 1.9 mm. Inward spiral gradient images were acquired with the following parameters: TR = 2000 ms, TE = 31 ms, FOV = 24 cm, flip angle = 90°, matrix = 64×64×34, slice thickness = 3.75 mm with 3.75mm3 isotropic voxels. FEAT (FMRI Expert Analysis Tool) Version 5.98, part of the FSL analysis package (FMRIB’s Software Library,, was used to conduct image pre-processing procedures. These procedures include slice-timing alignment, motion correction, coregistration, non-brain voxel extraction, normalization, smoothing (5 mm3 kernel) and high-pass temporal filtering (1/60Hz).

The general linear model was used for the first-level FSL FEAT analysis, and contained three types of events: negative distractors, neutral distractors, and targets. However, we only focused on activation to target events in the analysis at the group level. Results for targets were combined across runs for each participant, using fixed-effects models. At the group analysis level, we used a mixed effects analysis for controlling within- and between-subjects variation. Given that only seven out of 23 participants had PCI over the two year period, we used the ‘outlier deweighting function’ (a mathematic tool in the FSL FEAT software) to ensure the findings were not biased by outliers. The outlier deweighting method models errors as being both from normal population and from the outliers. Furthermore, it uses variance information from lower levels in the hierarchy38. Statistical results were thresholded at a voxel significance level of Z > 2.3 and a whole-brain-corrected cluster significance threshold of p < 0.05. First, we used voxelwise ANOVA analysis to compare the activation difference in response to targets between participants with PCI and participants without PCI. Then we conducted a regression analysis to examine brain regions that were linearly associated with cognitive performance at year 0 and with cognitive decline over the two years by using the CERAD TS at Year 0 and the change in CERAD TS (Year 2 – Year 0) as independent regressors (i.e., orthogonalized to each other). We controlled for the potential confounding effects of age, education, and severity of depression in all of these analyses.

Finally, we conducted a conjunction analysis to obtain regions that 1) indicated significant differences between PCI and with non-PCI individuals and 2) were correlated with cognitive decline. Only the regions that overlapped in the conjunction analysis were considered important for predicting subsequent PCI. To illustrate the correlation, we conducted ROI (region of interest) analyses to produce regression plots by extracting the significant clusters that differed between PCI and non-PCI groups. Confirmatory regression analyses were conducted to correlate mean peak signal changes in each ROI with unadjusted CERAD TS at Year 0 and the change in unadjusted CERAD TS (Year 2 – Year 0) using a statistical threshold of r > 0.4 and p < 0.05 (two-tailed). The ROI analysis was used to confirm that the results of the voxelwise correlation analyses were not driven by outliers.


Comparison of the clinical profile between the PCI and non-PCI groups

Seven out of 23 participants had PCI, which resulted in a 30% prevalence among our sample of older adults with LLD. A detailed comparison of the clinical profile between the two groups is presented in Table 1. The PCI group was significantly older than the non-PCI group (two sample t-test, t21=2.16, p<0.02), and was less depressed than the non-PCI group at the time they participated in the fMRI scan at Year 0 (two sample t-test on MADRS score, t21=4.68, p=0.001). Only one out of the 7 patients with PCI (14 %) was in an active depressive state at Year 0, whereas the number of non-PCI individuals with active depression was 11 out of 16 (69%). Given the difference in age and severity of depression (at year 0) between the PCI and non-PCI groups, we controlled for age, education, and MADRS in our whole-brain voxelwise fMRI analyses.

Cognitive performance on the CERAD TS was significantly different between the two groups at both Year 0 and Year 2. While the PCI group had significantly lower cognitive performance from Year 0 to Year 2 (paired t-test on adjusted CERAD TS score, t6=2.20, p=0.04), the non-CI group had significantly higher cognitive performance (paired t-test, t15=2.78, p=0.02) over the two year period. Since the majority of the participants in the non-PCI group were acutely depressed at baseline, improved CERAD TS scores at Year 2 for this group may suggest a mood-congruent improvement in cognition.

Subdividing the CERAD TS into memory and non-memory items revealed that the PCI group had lower scores than the non-PCI group at both Year 0 and Year 2 across both groups of items, indicating that the lower CERAD TS of PCI participants included both memory and non-memory domains (Table 1).

Activation associated with PCI

Our previous research showed that target stimuli in the oddball task activated the dorsal executive system, including the dlPFC, dorsal anterior cingulate (dACC), anterior portion of the posterior cingulate, inferior parietal cortex, insula, striatum, and cerebellum 21, 3637. Here, we found that individuals with PCI over the two years had significantly decreased activation in both dorsal anterior cingulate (dACC, BA24) and posterior cingulate cortices at Year 0 compared with non-PCI patients. The activation in the dACC was extended to the dorsal supplementary motor cortex (or so-called dorsomedial prefrontal cortex, BA 6). We refer this region as dACC/dmPFC in the text below. Decreased activations in PCI relative to non-PCI group were also found in bilateral hippocamal formation (entorhinal cortex), right inferior frontal cortex (IFC), left inferior parietal cortex, left insula, left thalamus, right caudate, and bilateral cerebellum (Figure 1, Table 2). To confirm the results from the voxel-based analysis, ROI analysis on the significant clusters including the dACC/dmPFC and right hippocampus were conducted. Significant differences between the PCI and non-PCI groups in these ROIs were tested using a two-sample Kruskal-Wallis test. The results confirmed the significant difference between the PCI and non-PCI groups (dACC/dmPFC, χ2(1, n = 23) =5.47, p=0.02; hippocampus, χ2(1, n =23) =10.72, p=0.001).

Figure 1
Brain regions showing significantly reduced activation in the PCI group compared with the non-PCI group. dACC=dorsal anterior cingulate, dmPFC=dorsomedial prefrontal cortex, HC=hippocampal complex, IFC=inferior frontal cortex, IPC=inferior parietal cortex. ...
Table 2
Brain Regions Showing Significantly Decreased Activation during Targets Detection at Year 0 in Participants with Persistent Cognitive Impairment (PCI) Compared to Participants without Persistent Cognitive Impairment (non-PCI).

Because six participants in the non-PCI group were not taking antidepressant medication, we examined whether medication status influenced our results. By removing these 6 participants from the non-PCI group, the comparison groups consisted of 7 PCI and 10 non-PCI participants. Removing these individuals did not change the results significantly.

Furthermore, regression analysis across the 23 participants (controlling for age, education, and MADRS) revealed that activation in many regions at Year 0 was correlated with the degree of subsequent cognitive decline (Year 2 CERAD TS - Year 0 CERAD TS). These regions included the dACC (BA24, extended to middle cingulate), aPCC (BA31), bilateral hippocampus, bilateral middle temporal cortex, bilateral orbitofrontal cortex (OFC), bilateral insula, and left inferior parietal cortex. (Figure 2, Table 3). As shown in Figure 3 from the conjunction analysis, right dACC, hippocampus, bilateral insula, right IFC, and fusiform cortex were the regions that overlapped across the two analyses, indicating that activation in these regions at Year 0 was associated with persistent cognitive impairment.

Figure 2
Brain regions revealing a linear correlation between brain activation to targets and cognitive decline measured by the change in CERAD TS (Year 2- Year 0). dACC= dorsal anterior cingulate cortex, HC=hippocampal complex, IFC=inferior frontal cortex, IPC=inferior ...
Figure 3
The Brain maps illustrating the overlapped brain regions showing significantly reduced activation in the PCI group compared with the non-PCI group and meanwhile the activations significantly correlated with the degree of cognitive decline based on the ...
Table 3
Brain Regions Showing Linear Correlation of Activation during Targets Detection with the Degree of Cognitive Decline Measured by the Change of CERAD Total Score (Year 2 – Year 0)

Activation associated with cognitive function at Year 0

We found that the activation in the dACC/dmPFC (BA 24 and BA 6) and the left inferior parietal cortex cortices were significantly correlated with the CERAD TS at Year 0 (Figure 4, Table 4). Although the ROI plot in Figure 3 also showed a correlation between hippocampus activation and TS at Year 0 (Figure 3, lower left), it did not pass the threshold in the whole brain analysis.

Figure 4
Brain regions revealing a linear correlation between brain activation to targets with cognitive function as measured by CERAD TS at Year 0. The image inside the white box (lower right) illustrates the overlapping voxels from three analyses: 1) activations ...
Table 4
Brain Regions Showing Significant Correlation of Activation during Targets Detection with the CERAD Total Score at Year 0

Activation associated with both cognitive function at year 0 and PCI

Since the majority of our patients had depressive symptoms in varying degrees at the time of their fMRI scan, it is very likely that cognitive performance in some participants at Year 0 was related to depression status and depression severity. In addition to using MADRS as a covariate to control the influence of depression severity (note that the MADRS was not correlated with CEREAD TS at baseline, r22=0.38, p=0.08), we also did a conjunction analysis to identify the overlapping regions where activation was associated with 1) cognitive function at year 0, 2) PCI status, and 3) cognitive decline over the two years. We found that the dACC (BA24, the voxels in white color in Figure 4) showed an overlap in the conjunction analysis, whereas activation in the left inferior parietal cortex (Figure 4) was only associated with cognitive function at year 0, and did not overlap with the PCI-related region.


The purpose of the study was to investigate neural substrates in depressed patients that can predict PCI. We found reduced activation in the dACC, hippocampus, IFC, and insula in PCI patients compared with non-PCI patients. In particular, low activation in the dACC, hippocampus, insula, and fusiform cortex were also associated with greater cognitive decline over the two-year period. Therefore, reduced activation in these regions in patients with LLD suggests an increased likelihood of PCI over a two-year period.

Decreased activation in the medial temporal lobe (hippocampus and fusiform gyrus) and cingulate cortex has been frequently found in patients with MCI and AD15, 39. Consistent with our prediction, depressed patients with PCI indeed had deficits in the same neural circuits as non-depressed patients with MCI. Although the results do not directly imply that PCI in LLD necessarily share the same neuropathology as MCI/AD without depression, the results suggest that PCI in LLD both demonstrate alterations in the same neural circuits, which explains commonalities in cognitive deficits. A large sample size and a direct comparison between PCI in depression and never-depressed MCI are needed to further confirm the results. Confirming the current results is important since it will not only allow us to predict cognitive impairment in patients with LLD, but it will also deepen our understanding of neural mechanisms linking depression with MCI

Another core finding of the study is that decreased activation of the dACC/dmPFC was not only correlated with cognitive function at Year 0, but also associated with PCI. Involvement of ACC in the pathology of LLD has been well documented in the literature, 2122, 4041 although the reported change in directionality was different depending on the type of task employed in different studies. We have previously shown reduced activation in the ACC regardless of depression state 21. Our current results further suggest that deficits in dACC function could be a leading cause of cognitive impairment and cognitive decline in LLD. Our results are consistent with positron emission tomography (PET) imaging studies that have shown beta-amyloid deposition in both geriatric depressed patients and MCI patients in many of these anterior and posterior cortical regions 4243. Therefore, decreased activation to targets in this region warrants closer clinical follow-up. Decreased activation in the left inferior parietal cortex was only associated with cognitive function at Year 0 but not with PCI, which suggests that the decreased activation in the left inferior parietal cortex can be normalized after remission of depression.

Activation in the hippocampus was associated with cognitive decline and correlated with cognitive function at Year 0 in the ROI analysis but not in the whole-brain corrected analysis, suggesting that individual variance plays an important role. In our sample, individuals who were in remission had lower cognitive function, reduced activation in the hippocampus, and greater cognitive decline 2 years later. Therefore, our results were not confounded by depressive symptoms. Hippocampal atrophy and reduced hippocampal activation are hallmarks of AD. Reduced hippocampal volume is also the most consistent finding in studies on cognitive impairments in patients with LLD 6, 4445. Here, we extended previous findings and showed that low activation in the hippocampus in remitted depression could predict cognitive decline.

Thirty percent of patients with LLD in our sample had PCI, which is consistent with previous findings1. One limitation of the current study is the small sample size. Another limitation of the study is that for those remitted patients who had PCI, they also had cognitive impairment at Year 0. As a result, we cannot determine whether the cognitive impairment appeared before or after depression and whether it was independent of depression. Further follow-up study on remitted patients without cognitive impairment would be necessary to clarify the causal relationship of cognitive impairment and depression.

In summary, our study showed that PCI in LLD shares deficits in a large-scale brain network similar to MCI/AD. Reduced activation in the dACC was associated with cognitive performance at Year 0 and cognitive decline over the two year period. Therefore, reduced activation in the dACC could be an indicator of acute cognitive dysfunction, longitudinal cognitive dysfunction, or both. Reduced activation in the hippocampus seemed to be a clearer indicator of PCI. Future studies with larger sample sizes, direct comparison of PCI and never-depressed MCI, and better tracking of the onset of cognitive impairment are important for further improving our understanding of the pathophysiology underlying PCI in depression, and the relationship between the circuitry in MCI and PCI with depression. An understanding of the neural circuitry of PCI in depression would inform the design of future mechanistic studies to understand the role of neurodegenerative (atrophy, beta-amyloid) and cerebrovascular processes relative to the functional neuroanatomic abnormalities.


This research was supported by the Duke Silvio O. Conte Center for the Neuroscience of Depression (P50-MH60451), the Paul B. Beeson Career Developmental Awards (K23-AG028982), and a National Alliance for Research in Schizophrenia and Depression Young Investigator Award (LW). GGP is supported by NIA Career Developmental Awards (K23MH87741), FD was supported by a Young Investigator Award from the National Alliance for Research on Schizophrenia and Depression and a CPRF Award from the Canadian Psychiatric Research Foundation, and DC is supported by a NIMH Mid-Career Development Award (K24 MH70027).


The authors have no conflict interests to disclose


1. Bhalla RK, Butters MA, Becker JT, et al. Patterns of mild cognitive impairment after treatment of depression in the elderly. Am J Geriatr Psychiatry. 2009;17:308–16. [PMC free article] [PubMed]
2. Lee JS, Potter GG, Wagner HR, Welsh-Bohmer KA, Steffens DC. Persistent mild cognitive impairment in geriatric depression. Int Psychogeriatr. 2007;19:125–35. [PubMed]
3. Bhalla RK, Butters MA, Mulsant BH, et al. Persistence of neuropsychologic deficits in the remitted state of late-life depression. Am J Geriatr Psychiatry. 2006;14:419–27. [PubMed]
4. Butters MA, Young JB, Lopez O, et al. Pathways linking late-life depression to persistent cognitive impairment and dementia. Dialogues Clin Neurosci. 2008;10:345–57. [PMC free article] [PubMed]
5. Alexopoulos GS, Meyers BS, Young RC, Mattis S, Kakuma T. The course of geriatric depression with “reversible dementia”: a controlled study. Am J Psychiatry. 1993;150:1693–9. [PubMed]
6. O’Brien JT, Lloyd A, McKeith I, Gholkar A, Ferrier N. A longitudinal study of hippocampal volume, cortisol levels, and cognition in older depressed subjects. Am J Psychiatry. 2004;161:2081–90. [PubMed]
7. Dickerson BC, Sperling RA. Functional abnormalities of the medial temporal lobe memory system in mild cognitive impairment and Alzheimer’s disease: insights from functional MRI studies. Neuropsychologia. 2008;46:1624–35. [PMC free article] [PubMed]
8. Dickerson BC, Sperling RA. Large-scale functional brain network abnormalities in Alzheimer’s disease: insights from functional neuroimaging. Behav Neurol. 2009;21:63–75. [PMC free article] [PubMed]
9. Dierckx E, Engelborghs S, De Raedt R, De Deyn PP, Ponjaert-Kristoffersen I. Differentiation between mild cognitive impairment, Alzheimer’s disease and depression by means of cued recall. Psychol Med. 2007;37:747–55. [PubMed]
10. Greicius MD, Srivastava G, Reiss AL, Menon V. Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: evidence from functional MRI. Proc Natl Acad Sci U S A. 2004;101:4637–42. [PubMed]
11. Holland D, Brewer JB, Hagler DJ, et al. Subregional neuroanatomical change as a biomarker for Alzheimer’s disease. Proc Natl Acad Sci U S A. 2009;106:20954–9. [PubMed]
12. Kloppel S, Stonnington CM, Chu C, et al. Automatic classification of MR scans in Alzheimer’s disease. Brain. 2008;131:681–9. [PMC free article] [PubMed]
13. Petrella JR, Wang L, Krishnan S, et al. Cortical deactivation in mild cognitive impairment: high-field-strength functional MR imaging. Radiology. 2007;245:224–35. [PubMed]
14. Sorg C, Riedl V, Muhlau M, et al. Selective changes of resting-state networks in individuals at risk for Alzheimer’s disease. Proc Natl Acad Sci U S A. 2007;104:18760–5. [PubMed]
15. Sperling R. Functional MRI studies of associative encoding in normal aging, mild cognitive impairment, and Alzheimer’s disease. Ann N Y Acad Sci. 2007;1097:146–55. [PubMed]
16. Zhou J, Greicius MD, Gennatas ED, et al. Divergent network connectivity changes in behavioural variant frontotemporal dementia and Alzheimer’s disease. Brain. 2010;133:1352–67. [PMC free article] [PubMed]
17. Bokde AL, Ewers M, Hampel H. Assessing neuronal networks: understanding Alzheimer’s disease. Prog Neurobiol. 2009;89:125–33. [PubMed]
18. Bokde AL, Lopez-Bayo P, Meindl T, et al. Functional connectivity of the fusiform gyrus during a face-matching task in subjects with mild cognitive impairment. Brain. 2006;129:1113–24. [PubMed]
19. Aizenstein HJ, Butters MA, Wu M, et al. Altered functioning of the executive control circuit in late-life depression: episodic and persistent phenomena. Am J Geriatr Psychiatry. 2009;17:30–42. [PMC free article] [PubMed]
20. de Asis JM, Stern E, Alexopoulos GS, et al. Hippocampal and anterior cingulate activation deficits in patients with geriatric depression. Am J Psychiatry. 2001;158:1321–3. [PubMed]
21. Wang L, Krishnan KR, Steffens DC, Potter GG, Dolcos F, McCarthy G. Depressive state- and disease-related alterations in neural responses to affective and executive challenges in geriatric depression. Am J Psychiatry. 2008;165:863–71. [PubMed]
22. Smith GS, Kramer E, Ma Y, et al. The functional neuroanatomy of geriatric depression. Int J Geriatr Psychiatry. 2009;24:798–808. [PMC free article] [PubMed]
23. Alexopoulos GS, Meyers BS, Young RC, Campbell S, Silbersweig D, Charlson M. ‘Vascular depression’ hypothesis. Arch Gen Psychiatry. 1997;54:915–22. [PubMed]
24. Alexopoulos GS, Meyers BS, Young RC, et al. Executive dysfunction and long-term outcomes of geriatric depression. Arch Gen Psychiatry. 2000;57:285–90. [PubMed]
25. Alexopoulos GS, Young RC, Meyers BS. Geriatric depression: age of onset and dementia. Biol Psychiatry. 1993;34:141–5. [PubMed]
26. Sheline YI, Pieper CF, Barch DM, et al. Support for the vascular depression hypothesis in late-life depression: results of a 2-site, prospective, antidepressant treatment trial. Arch Gen Psychiatry. 2010;67:277–85. [PMC free article] [PubMed]
27. Sheline YI, Wang PW, Gado MH, Csernansky JG, Vannier MW. Hippocampal atrophy in recurrent major depression. Proc Natl Acad Sci U S A. 1996;93:3908–13. [PubMed]
28. Steffens DC, Welsh-Bohmer KA, Burke JR, et al. Methodology and preliminary results from the neurocognitive outcomes of depression in the elderly study. J Geriatr Psychiatry Neurol. 2004;17:202–11. [PubMed]
29. Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry. 1979;134:382–9. [PubMed]
30. Steffens DC. A multiplicity of approaches to characterize geriatric depression and its outcomes. Curr Opin Psychiatry. 2009 [PMC free article] [PubMed]
31. Breitner JC, Welsh KA, Robinette CD, Gau BA, Folstein MF, Brandt J. Alzheimer’s disease in the NAS-NRC registry of aging twin veterans. II. Longitudinal findings in a pilot series. National Academy of Sciences. National Research Council Registry Dementia. 1994;5:99–105. [PubMed]
32. Plassman BL, Langa KM, Fisher GG, et al. Prevalence of cognitive impairment without dementia in the United States. Ann Intern Med. 2008;148:427–34. [PMC free article] [PubMed]
33. Welsh KA, Butters N, Hughes JP, Mohs RC, Heyman A. Detection and staging of dementia in Alzheimer’s disease. Findings from the British Women’s Heart and Health Study and the British Regional Heart StudyUse of the neuropsychological measures developed for the Consortium to Establish a Registry for Alzheimer’s Disease. Arch Neurol. 1992;49:448–52. [PubMed]
34. Chandler MJ, Lacritz LH, Hynan LS, et al. A total score for the CERAD neuropsychological battery. Neurology. 2005;65:102–6. [PubMed]
35. Rossetti HC, Munro Cullum C, Hynan LS, Lacritz LH. The CERAD Neuropsychologic Battery Total Score and the progression of Alzheimer disease. Alzheimer Dis Assoc Disord. 2010;24:138–42. [PMC free article] [PubMed]
36. Wang L, McCarthy G, Song AW, Labar KS. Amygdala activation to sad pictures during high-field (4 tesla) functional magnetic resonance imaging. Emotion. 2005;5:12–22. [PubMed]
37. Wang L, LaBar KS, Smoski M, et al. Prefrontal mechanisms for executive control over emotional distraction are altered in major depression. Psychiatry Res. 2008;163:143–55. [PMC free article] [PubMed]
38. Woolrich M. Robust group analysis using outlier inference. Neuroimage. 2008;41:286–301. [PubMed]
39. Petrella JR, Krishnan S, Slavin MJ, Tran TT, Murty L, Doraiswamy PM. Mild cognitive impairment: evaluation with 4-T functional MR imaging. Radiology. 2006;240:177–86. [PubMed]
40. Aizenstein HJ, Butters MA, Figurski JL, Stenger VA, Reynolds CF, 3rd, Carter CS. Prefrontal and striatal activation during sequence learning in geriatric depression. Biol Psychiatry. 2005;58:290–6. [PubMed]
41. Murphy CF, Gunning-Dixon FM, Hoptman MJ, et al. White-matter integrity predicts stroop performance in patients with geriatric depression. Biol Psychiatry. 2007;61:1007–10. [PMC free article] [PubMed]
42. Butters MA, Klunk WE, Mathis CA, et al. Imaging Alzheimer pathology in late-life depression with PET and Pittsburgh Compound-B. Alzheimer Dis Assoc Disord. 2008;22:261–8. [PMC free article] [PubMed]
43. Wolk DA, Price JC, Saxton JA, et al. Amyloid imaging in mild cognitive impairment subtypes. Ann Neurol. 2009;65:557–68. [PMC free article] [PubMed]
44. Sheline YI. Neuroimaging studies of mood disorder effects on the brain. Biol Psychiatry. 2003;54:338–52. [PubMed]
45. Steffens DC, Byrum CE, McQuoid DR, et al. Hippocampal volume in geriatric depression. Biol Psychiatry. 2000;48:301–9. [PubMed]