The association of late-life depressive symptoms with brain amyloid-β deposition: the ARIC-PET study

Abstract

Late-life depression is associated with an increased risk of developing dementia related to Alzheimer’s disease, yet the mechanism underlying this relationship remains poorly understood. This study investigated the association of late-life depressive symptoms (LLDS) with brain amyloid deposition by PET. 334 dementia-free individuals from the Atherosclerosis Risk in Communities Study had Florbetapir-PET scans in late-life (ages 67–89). Elevated global and regional brain amyloid deposition was defined as a Florbetapir standardized uptake value ratio (SUVR) > 1.2. LLDS was assessed using the 11-item Center for Epidemiological Studies Depression Scale (CES-D) and defined as late-life CES-D ≥ 9. Several secondary depression measures (e.g., antidepressant use) were also explored. Stratified analyses across race, sex, and cognitive status groups were conducted. Participants (median age 76 years; 57.2 % female; 42.5 % Black; 26.9 % mild cognitive impairment) showed no association of LLDS with global amyloid deposition. Although antidepressant use was not associated with global amyloid overall, antidepressant use was associated with elevated global amyloid in individuals with normal cognition but not mild cognitive impairment. There was no effect modification by race and sex. Overall, our findings suggest that LLDS are generally not associated with global amyloid deposition, but further investigation of this relationship in larger sample sizes is warranted.

1. Introduction

Late-life depression (LLD) is a common health concern among older adults, linked to executive dysfunction, cardiovascular comorbidities, and mortality [1,2,3]. Moreover, a history of depression and LLD have been identified to be associated with a significantly increased risk of dementia, including Alzheimer’s disease (AD)-related dementia, highlighting LLD as a neuropsychiatric condition with significant implications for cognition and quality of life among older adults [4,5,6,7,8].

The relationship between LLD and AD is multifaceted, with its underlying mechanisms remaining inadequately understood. Prior research has suggested that depressive symptoms in late-adulthood represent an early manifestation of AD during the preclinical stage [9,10] or can occur as a subsequent consequence of AD [11]. Conversely, other evidence has supported that LLD may serve as a risk factor for incident dementia, potentially playing a causal role in its development [12,13,14]. The connection between depression and dementia, but lack of understanding regarding the pathway that links the two, emphasizes the value of studying AD-specific biomarkers in association with LLD.

Deposition of amyloid protein plaques in the brain is linked to cognitive decline and neurodegenerative changes and is pathognomonic of AD [15,16]. Recent advances in our ability to measure brain amyloid via positron emission tomography (PET) imaging have allowed the assessment of amyloid deposition in vivo [17]. Previous cross-sectional studies have investigated the association of brain-amyloid deposition with depression or depressive symptoms, primarily focusing on whether amyloid deposition contributes to the development of these conditions while controlling for potential confounders [18,19,20]. Fewer studies have explored this association in the other direction, examining depression or depressive symptoms as a predictor of increased brain-amyloid deposition. Several studies have shown associations between depression measures and global [21] or regional amyloid deposition [22,23,24]. On the other hand, other studies have suggested that LLD may be associated with lower global amyloid deposition [25] or found no association between depression measures and global or regional amyloid deposition [26,27,28]. Varied findings across studies may be a result of methodological differences in imaging techniques, depression assessments, and participant characteristics.

Existing research indicates that LLD and elevated amyloid deposition are independently more prevalent in specific sociodemographic groups, suggesting that the association of LLD with amyloid deposition could vary across sex, race, and cognitive status [29,30,31,32,33]. However, previous studies investigating the association between LLD and brain amyloid deposition have had limited racial diversity, with findings reflecting a single racial group, emphasizing the importance of evaluating this association in diverse populations. Careful consideration of potential confounders is also essential to clarify the relationship between depression and amyloid.

We aimed to comprehensively investigate the association between late-life depressive symptoms (LLDS) and brain amyloid deposition in older adults without dementia in the ARIC-PET cohort. We hypothesized that individuals with LLDS exhibit elevated global amyloid as assessed by PET imaging during late-life. Furthermore, we explored variations in amyloid deposition across distinct brain regions, given prior research indicating heightened amyloid burden in specific brain regions associated with depression. Lastly, this study investigated if the association between LLDS and amyloid deposition differs across race, sex, and cognitive status. Specifically, we hypothesized that associations between LLDS and elevated global amyloid deposition would be particularly strong among individuals of Black race, women, and individuals with MCI.

2. Methods

2.1. Study population

The Atherosclerosis Risk in Communities (ARIC) Study is an ongoing prospective study that included 15,792 adults ages 45–64 from 4 US communities at baseline [34]. The ARIC study was designed to conduct population surveillance for risk factors and incidence for various cardiovascular conditions. Participants attended five in-person study visits from baseline through visit 5, including visit 1 (1987–1989), visit 2 (1990–1992), visit 3 (1993–1995), visit 4 (1996–1998), and visit 5 (2011–2013). At visit 5, as part of the ARIC-Neurocognitive Study (ARIC-NCS), a full neurocognitive battery was administered to 6,538 surviving participants who attended the in-person visit. In addition to an extensive neuropsychological battery, an informant interview was conducted for a subset of participants, and cognitive status (normal/MCI/dementia) was determined by expert classification [35]. A subset of ARIC-NCS participants was selected for a brain MRI if they did not have a contraindication to MRI and met one of the following criteria: (1) prior MRI within the ARIC study, (2) low or declining scores on the neuropsychological battery, or (3) a random subset of cognitively normal participants stratified by age (n = 1968)[36].

A subset of those who received a brain MRI at visit 5, and were from one of three ARIC sites (Jackson, MS; Washington County, MD; Forsyth County, NC) was recruited to receive a Florbetapir-PET scan as part of the ancillary ARIC-PET Amyloid Imaging Study [31]. Only participants without dementia were selected based on the following exclusionary criteria: (1) expert committee adjudication of dementia, or if adjudication not complete: (2) Clinical Dementia Rating sum of boxes score > 3, Functional Activities Questionnaire > 5, or Mini-Mental State Examination score < 19 (Black participants) or < 21 (White participants). Participants with heavy alcohol use, renal dysfunction (creatinine > 2 mg/dL), or prolonged QT-c interval (>450 ms) were further excluded. Full ARIC-PET ancillary study details and protocol have previously been published [31].

Of 346 participants who had PET imaging between 2011–2014 as part of the ARIC-PET Study, two participants were excluded due to non-White or non-Black race. These exclusions, standard in ARIC, were necessary as the small number of participants in certain racial groups precluded important adjustment for race in analytical models. One participant was excluded due to dementia diagnosis by the end of visit 5. Nine participants were excluded due to missing APOE genotype information. Thus, our analytic sample encompassed a total of 334 participants, aged 67–89 at the time of PET scans.

2.2. Brain MRI and PET

Participants received 3 Tesla brain MRI scans at each site between 2011 and 2013, and MRI analysis was completed at the Mayo Clinic MRI Reading Center. Participants received 18-Florbetapir PET scans between 2011 and 2014, within one year of their brain MRI scan. PET images were reviewed for image quality and incidental findings, coregistered to MRI scans, and quantified for 18-Florbetapir standardized uptake value ratio (SUVR) at the ARIC-PET reading center (DFW Lab − Johns Hopkins University; Baltimore, MD). SUVRs for 34 individual regions of interest were quantified using the cerebellum as reference. PET processing details and protocol of utilizing global SUVR as a measure of Florbetapir-PET are further described (Supplementary Methods 1).

Our primary analysis used a global measure of amyloid, defined as the weighted average of the SUVR of the following cortical regions: orbitofrontal lobe, prefrontal lobe, superior frontal cortices, lateral temporal lobe, parietal lobe, occipital lobe, precuneus, anterior cingulate, and posterior cingulate [31]. Secondary analyses explored regional SUVR values of other cortical (medial temporal lobe) and subcortical regions (amygdala, thalamus, putamen, and caudate) as well, based on prior studies [37,38,23,24]. Consistent with prior ARIC-PET studies, SUVRs were dichotomized at the sample median of 1.2, with elevated global and regional amyloid defined as SUVR > 1.2[31].

2.3. Late life depressive symptoms

LLDS were assessed at visit 5 (2011–2013; ages 67–89) using the 11-item Center for Epidemiologic Studies Depression Scale (CES-D). The 11-item CES-D, a shortened form of the original 20-item CES-D, is advantageous for older populations as it is less burdensome [39]. Furthermore, the shortened CES-D strongly correlates with the full version of the CES-D (r = 0.95) [39].

The CES-D is designed to assess depressive symptoms within the past week in the general population [40]. The items on the scale represent statements consistent with depression symptoms, such as “I felt lonely” and “My sleep was restless.” Responses to the items were self-reported, with participants responding with the following options: hardly ever or never (<1 day in the past week), some of the time (1–2 days in the past week), or much or most of the time (3–7 days in the past week). Scores range from 0 to 22, with higher scores indicative of more severe depressive symptomology [40].

We defined LLDS, our primary independent variable, with a CES-D score ≥ 9, a threshold previously established as clinically indicative of depression [41].

2.4. Midlife vital exhaustion

Midlife vital exhaustion (MVE) was assessed at visit 2 (1990–1992; ages 46–69) using the Maastricht Questionnaire. Vital exhaustion captures feelings of excessive fatigue, dejection, and irritability [42]. Vital exhaustion is not a formal depression measure but has a moderate correlation with depressive symptoms (r = 0.67) [43]. Since the ARIC study did not assess depressive symptomatology earlier in life, we utilized MVE within our secondary analyses to capture aspects of midlife mental wellbeing, as an imperfect proxy for depression, and to assess how duration of poor wellbeing through the life course may be associated with amyloid deposition.

The Maastricht questionnaire is a 21-item questionnaire in which responses were totaled to obtain an overall vital exhaustion score. Responses to the items were self-reported, and each response was scored 0 for “No”, 1 for “Don’t know,” and 2 for “Yes”. Scores range from 0 to 42, with higher scores indicative of more severe vital exhaustion. MVE included individuals with MQ score ≥ 13[44].

2.5. Antidepressant use

Data on antidepressant use in late life were obtained by identifying generic product codes from ARIC medication inventory lists, which were compiled by ARIC study staff based on medication containers brought by participants during visit 5. Antidepressant use was defined as an individual using any of the following medications: MAO inhibitors, tricyclic agents, or other miscellaneous antidepressants (i.e., SSRI’s, SNRI’s, serotonin modulators, and atypical antidepressants). Antidepressant collection details and protocol have previously been published [45].

2.6. Primary and secondary definitions of depression

Primary analyses evaluated LLDS (CES-D score ≥ 9) as the independent variable.

Other measures of depressive symptoms, as part of our secondary analyses, included the following: (1) continuous CES-D scores, (2) antidepressant use, (3) composite LLDS (antidepressant use or CES-D ≥ 9), (4) midlife vital exhaustion (MQ ≥ 13), and (5) composite MVE/LLDS across the life course. We characterized composite MVE/LLDS using four categories: Level 1 included individuals with neither MVE nor LLDS (MQ < 12 and CES-D < 9), Level 2 included those with MVE but no LLDS (MQ ≥ 13 and CES-D < 9), Level 3 included those with no MVE and LLDS (MQ < 12 and CES-D ≥ 9), and lastly, Level 4 encompassed those with both MVE and LLDS (MQ ≥ 13 and CES-D ≥ 9). Level 1 was used as the reference group, while level 4 was considered high composite MVE/LLDS, and represents individuals with a likely greater duration of poor mental wellbeing (Supplementary Table 1).

2.7. Demographic and covariate information

Primary demographic variables such as age at time of scan (in years), race (White/Black), sex (man/woman), and education level (less than completed high school; high school, general educational development, or vocational school; any college, graduate, or professional school) were self-reported at visit 1. APOE genotype was assessed using TaqMan assay (Applied Biosystems; Foster City, CA), and individuals were categorized based on APOE ε4 carrier status (carrier/non-carrier).

The following covariates were assessed at both visit 2 (concurrent with MDS assessment) and visit 5 (concurrent with LLDS assessment): hypertension (systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or use of antihypertensive medication), diabetes (fasting glucose ≥126 mg/dl, non-fasting glucose ≥200 mg/dl, diabetes medication use, or self-reported diabetes diagnosis by a physician), body mass index (BMI) (kg/m²), and current smoking status. Cognitive status (MCI/normal) at visit 5 was determined based on computer algorithmic diagnoses incorporating cognitive data along with informant interviews, with further adjudication by expert review.

2.8. Main analyses

Descriptive statistics were calculated for all groups (LLDS vs. no LLDS), including t-tests for continuous data or χ² tests for categorical data. We performed multivariable logistic regression analysis, with dichotomized global cortical SUVR as the primary outcome and LLDS (CES-D ≥ 9) as the primary exposure. The primary model was adjusted for covariates previously identified as potential confounders to the association between LLDS and global cortical amyloid deposition [[31], [46], [47]]. Covariates were adjusted for in a stepwise fashion. Model 1 was adjusted for age, sex, race, and education level. Model 2 was additionally adjusted for hypertension and diabetes. Model 3 was adjusted for all aforementioned covariates in addition to smoking status and BMI. Due to the only one participant with LLDS who was also a carrier of APOE ε4, analyses with LLDS (CES-D ≥ 9) as the primary exposure were not adjusted for APOE ε4 carrier status.

2.9. Secondary analyses

For secondary analyses, we repeated the above analyses but with dichotomized regional SUVR as the primary outcome (still using the 1.2 SUVR threshold) and LLDS (CES-D ≥ 9) as the exposure. Additional logistic regression models were used to assess the association of secondary definitions of depression, as described above, with elevated amyloid deposition. For MVE/LLDS across the life course, we performed multivariable logistic regression, with dichotomized global cortical SUVR as the primary outcome and level of composite MVE/LLDS (level 4, level 3, level 2) as the exposure, with level 1 as the reference group. To account for multiple comparisons across 14 regions of interest and 8 depression measures, we applied Bonferroni’s corrections, resulting in corrected p-values of p = 0.003 and p = 0.006, respectively.

We performed stratified analyses to compare women vs. men, Black vs. White individuals, MCI vs. normal cognition, and in some models, APOE ε4 carriers vs. APOE ε4 non-carriers. In separate models, we also tested formal interaction terms for each of these effect modifiers. Stratified analyses were minimally adjusted to Model 1 covariates, with the exclusion of APOE ε4 carrier status due to insufficient numbers.

2.10. Sensitivity analyses

Sensitivity analyses were performed by restricting our primary analysis to those with normal cognition and using alternative SUVR cutoffs (SUVR ≥ 1.11)[48]. To examine the effects of participant selection from visit 5 into the ARIC-PET ancillary study, a sensitivity analysis was conducted using inverse probability weighting (IPW). IPW uses regression weighting to correct for sampling bias by assigning a higher weight to participants who were less likely to participate in ARIC-PET. Stabilized weights were assigned to individuals who attended visit 5 based on predicted probabilities derived from two logistic regressions with demographic and clinical covariates, and with the outcome as participation in ARIC-PET. Specifically, we first estimated the predicted probability of participating in ARIC-PET in a reduced logistic regression model, and then in an extended logistic regression model. Stabilized weights were calculated as the predicted probability of participating in ARIC PET based on the full model, divided by the predicted probability based on the reduced model. The primary analysis was repeated applying the stabilized weights representing the inverse of the probability of participating in ARIC-PET.

3. Results

3.1. Participant characteristics

Among all participants, the median age was 76 years, 57.2 % were women, and 57.5 % were of Black race. 21 (6 %) individuals had LLDS in late life. Compared to individuals without LLDS, individuals with LLDS were more likely to have lower education, be smokers, be non-carriers of APOE ε4 alleles, and use antidepressants. There was a similar proportion of individuals with MCI in the LLDS and no LLDS group (Table 1).

Table 1.

Participant characteristics by late-life depressive symptom status.

	No Late-life Depressive Symptoms	Late-life Depressive Symptoms	p-value a
	(N = 313)	(N = 21)
Age, y, median (min–max)	75 (67–87)	78 (70–89)	0.08
Women, n, %	179 (57.2 %)	12 (57.1 %)	1.00
Black Race, n, %	132 (42.2 %)	10 (47.6 %)	0.79
Education Level, n, %			<0.001 *
< High School	45 (14.4 %)	10 (47.6 %)
High School, GED, or Vocational School	136 (43.5 %)	8 (38.1 %)
College, Graduate, or Professional School	132 (42.2 %)	3 (14.3 %)
APOE ε4 Genotype, n, %			0.03 *
2/2, 2/3, or 3/3	211 (67.4 %)	20 (95.2 %)
2/4 or 3/4	94 (30.0 %)	1 (4.8 %)
4/4	8 (2.6 %)	0 (0 %)
MCI at Visit 5, n, %	84 (26.8 %)	6 (28.6 %)	1.00
Hypertension, n, %	221 (70.6 %)	17 (81.0 %)	0.28
Body Mass Index (kg/m2), median (min–max)	28.6 (16.6–51.5)	27.7 (19.1–40.3)	0.49
Diabetes, n, %	108 (34.5 %)	12 (57.1 %)	0.07
Current Smoking, n, %	13 (4.2 %)	4 (19.0 %)	0.009 *
Antidepressant Use at Visit 5, n, %	40 (12.8 %)	10 (47.6 %)	<0.001 *
Global Cortex SUVR, median (min–max)	1.20 (0.9–2.3)	1.26 (1.0–2.3)	0.28
Abbreviations: SUVR = Standardized Uptake Value Ratio; GED = General Education Development; APOE = Apolipoprotein E; MCI = Mild Cognitive Impairment. * p < 0.05 a p-values calculated using t-tests (continuous data) or χ2 tests (categorical data).

3.2. LLDS and Global/Regional amyloid deposition

LLDS (CES-D ≥ 9) was not significantly associated with elevated amyloid deposition in fully adjusted models (Fig. 1; Fig. 2). Among all regions evaluated, LLDS was independently associated only with elevated amyloid deposition in the amygdala (OR: 4.24, 95 % CI: 1.47–13.14) (Supplementary Table 2). However, after applying the Bonferroni correction to account for multiple comparisons (corrected p-value < 0.003), the associations of LLDS with amyloid deposition in the amygdala were no longer statistically significant.

Fig. 1.

Adjusted Odds Ratios of Elevated Global Amyloid Deposition Associated with Each Depression Definition (Model 3) ^a. Abbreviations: MVE = Midlife Vital Exhaustion; LLDS = Late-life Depressive Symptoms; CES-D = Center for Epidemiologic Studies Depression Scale. ^a All models adjusted for following Model 3 covariates: age, sex, race, education level, APOE e4 carrier status, hypertension, diabetes, BMI, and smoking. Elevated amyloid deposition defined as SUVR > 1.2. ^† Models were not adjusted for APOE e4 carrier status due to insufficient numbers.

Fig. 2.

Unadjusted Box Plots of Global Amyloid Deposition in Individuals with LLDS vs Individuals without LLDS ^a. Abbreviations: LLDS = Late-life Depressive Symptoms; SUVR = Standardized Uptake Value Ratio; APOE = Apolipoprotein E. ^a Dashed line at SUVR cutoff value of 1.2.

3.3. Secondary depression measures and global amyloid deposition

CES-D defined continuously was not associated with elevated global amyloid. MVE, antidepressant use, and composite LLDS (antidepressant use or CES-D ≥ 9) were also not associated with elevated global amyloid deposition. The highest level of composite MVE/LLDS (level 4), which included individuals with both MVE and LLDS, was not associated with elevated amyloid deposition, when compared to level 1. Composite LLDS (antidepressant use or CES-D ≥ 9) was associated with elevated global amyloid in the whole sample, when adjusted for Model 1 covariates (OR: 1.93, 95 % CI: 1.04–3.63, p-value = 0.038). However, this association was not statistically significant when adjusted for Model 3 covariates (OR: 1.74, 95 % CI: 0.91–3.38, p-value = 0.09) (Fig. 1; Supplementary Table 3) or after applying the Bonferroni correction to account for multiple comparisons.

3.4. Stratified analyses by sex, race, and cognitive status

We did not find significant interactions between sex or race with LLDS for either elevated global or regional amygdala amyloid deposition (Supplementary Tables 4–5; Supplementary Figs. 1–3). Regarding our secondary definitions, we also did not find significant interactions between sex or race and secondary definitions of LLDS for elevated global (Supplementary Table 4). Stratified analyses by cognitive status (and formal testing for interaction) showed no evidence of effect modification by cognitive status for LLDS with elevated global or regional amyloid, with no association between LLDS and amyloid in those either with normal cognition or with MCI (Supplementary Table 4). However, antidepressant use was associated with elevated global amyloid deposition in those with normal cognition (OR: 2.65, 95 % CI: 1.23–5.94) but not MCI (OR: 0.52, 95 % CI: 0.12–2.24; p-interaction = 0.049) (Fig. 3; Supplementary Table 4; Supplementary Fig. 4), with a similar pattern when the composite LLDS measure was evaluated (antidepressant use or elevated CES-D).

Fig. 3.

Adjusted Odds Ratios for Elevated Global Amyloid Deposition Associated with Antidepressant Use ^a. Abbreviations: MCI = Mild Cognitive Impairment; APOE = Apolipoprotein E. * p-interaction < 0.05, showing effect modification by cognitive status of the association between antidepressant use and SUVR. ^a All models adjusted for the following Model 1 covariates: age, sex, race, education, and APOE e4 carrier status. Elevated amyloid deposition defined as SUVR > 1.2.

3.6. Sensitivity analyses and inverse probability weighting

Utilizing an SUVR cutoff of 1.11 or restricting our sample to only those with normal cognition yielded no significant association between LLDS and global amyloid deposition (Supplementary Table 6; Supplementary Table 7). Results of the inverse probability weighting analysis (Supplementary Table 8), where the cohort was weighted back to the broader visit 5 sample in order to account for potential selection bias, were similar to the main results.

4. Discussion

In our investigation involving a cohort of 334 Black and White participants without dementia, we observed that LLDS were not associated with global amyloid deposition. We evaluated several other secondary definitions of depression, and although a composite definition incorporating either a high CES-D score or antidepressant use, allowing for inclusion of a larger sample size meeting this criterion of depression, was significantly associated with elevated brain amyloid in minimally adjusted models, it was no longer significant in further adjusted models, nor did it survive adjustment for multiple comparisons.

Our findings are consistent with several prior studies suggesting that midlife depressive symptoms [26], lifetime history of major depression [24], and late-onset depression after age 50 [24] are not associated with global amyloid deposition on Florbetapir-PET scans. Our results further correspond with investigations utilizing C-PiB PET scans, or harmonized data of various types of amyloid-PET scans, suggesting that prior depressive symptoms experienced in midlife or late life do not have any association with global amyloid deposition [27,28]. However, the literature on this topic remains inconclusive as other investigations have identified significant associations between LLDS and amyloid deposition. For example, one study found that individuals with LLD, both with MCI and normal cognition, exhibited significantly less global cortical amyloid deposition than non-depressed cognitively normal controls [25], suggesting the possibility that LLD and dementia is linked through non-amyloid-associated mechanisms. Differences in the extent of underlying cerebrovascular disease or vascular risk factors may also account for this finding. On the other hand, another study discovered each one-point increase on the Beck Depression Inventory II (BDI-II) to be associated with increased odds for elevated global amyloid deposition [21]. While our findings do not support the notion that LLDS are associated with global amyloid deposition, further investigation in larger cohorts is warranted to clarify this relationship.

In understanding how LLDS and amyloid relate to each other, it is important to consider that this relationship may operate in reverse, in which amyloid deposition may contribute to the development of LLDS. Since cerebral accumulation of amyloid precedes the clinical diagnosis of AD, LLDS may represent pre-clinical symptoms of AD arising from early neurodegenerative changes [49]. While a recent meta-analysis of primarily cross-sectional studies has suggested no association between amyloid deposition and depressive symptoms [50], several longitudinal cohort studies have reported greater amyloid burden to be associated with increased incident depressive symptoms in cognitively normal individuals [51,52,53,54], suggesting that the temporal and dynamic nature of depressive symptoms may be important in elucidating the relationship between depression and AD. Alternatively, our findings bring up the possibility that the previously observed link between LLDS and AD may not be explained by increased cerebral amyloid deposition, but rather by another pathologic mechanism, such as cerebral tau accumulation. Several studies have found associations between cerebral tau accumulation on PET, but not amyloid, and depression [18,20]. Similarly, other studies have found that depressive symptoms are associated with greater tau accumulation in specific regions of interest among individuals with normal cognition [55] and APOE ε4 carriers [26].

In our study, we observed an association of LLDS with elevated amyloid deposition in the amygdala, though this finding did not survive correction for multiple comparisons. A study by Krell-Roesch et al. revealed an association between elevated amyloid in the amygdala and depressive symptoms (β₁: 0.80; SE: 0.26; p-value = 0.002) within a group of cognitively normal individuals and individuals with MCI [38]. Notably, Krell-Roesch et al. also identified that elevated amyloid in the striatum and thalamus was associated with depressive symptoms in adults without dementia [38]. The traditional neuropathologic paradigm of AD delineates a sequential progression of amyloid deposition, commencing from neocortical regions during the preclinical phase and involving other subcortical regions in later phases [56]. Hence, amyloidosis within the amygdala typically signifies progression beyond the preclinical phase and is indicative of heightened cognitive impairment [57]. However, our findings, alongside findings from previous studies, suggest the possibility for LLDS to alter amyloid dynamics in subcortical areas, such as the amygdala, in individuals without dementia. Further investigation is merited to clarify the connection between depressive symptoms and amyloid deposition in subcortical regions which have established connections to emotional processing and depression, such as the amygdala [58].

Existing research has highlighted differences in amyloid deposition across race and sex groups, particularly with Black individuals and women having greater amyloid deposition compared to White individuals and men, respectively [31]. Prior studies have indicated that Black individuals have greater depression severity than White individuals, and that women are more likely to experience LLD than men, as are individuals with MCI compared to individuals with normal cognition [29,30,32,33]. Contrary to what the existing literature may suggest, in our study, we did not observe the association of LLDS or persistent depressive symptoms with amyloid deposition to be modified by race, sex, or cognitive status. We speculate that these findings could be attributed to the small number of individuals with LLDS or the possibility that individuals who opt to receive a PET scan in late-life may be healthier than their age-matched counterparts within the general population. Furthermore, the participants in our study may reflect the distinct characteristics of the communities they are from, as our study recruited from U.S. communities with the goal of representing those communities, whereas many other studies relied on clinical populations.

We considered antidepressant use as an additional definition for depression. Across our entire sample, we did not observe any association of antidepressant use with global amyloid deposition. This result aligns with previous research indicating that the use of selective serotonin reuptake inhibitors (SSRIs) [27] and duration of SSRI usage [25] do not correlate with amyloid deposition. However, our analysis revealed evidence of effect modification by cognitive status, in which antidepressant use was associated with increased amyloid deposition in those with normal cognition but not in those with MCI. Antidepressant use may serve as a proxy for depressive symptoms that are being treated, and significant findings could suggest a link between depressive symptoms and amyloid deposition. However, since we observed no significant association when utilizing a composite LLDS definition (CES-D ≥ 9 or antidepressant use), this explanation appears unlikely. It is more likely this finding can be attributed to a potential ceiling effect in participants with MCI, as SUVR values for participants with MCI were already statistically higher than those of cognitively normal participants (p-value = 0.0001). It is also important to note that individuals take antidepressants for a variety of reasons, including anxiety, chronic pain, migraines, or sleep disturbances [59]. Therefore, the use of antidepressants could also reflect other underlying chronic conditions or comorbidities which may increase risk of elevated amyloid deposition [60,37]. Further, antidepressants may also be more likely to be prescribed to individuals from certain demographic groups [61,62]. Ultimately, these subgroup findings should be interpreted with caution given the lack of consistent results by cognitive status for all of our depression measures.

Our study was mainly limited by the low number of individuals who met criteria for LLDS (n = 21), which greatly reduced our statistical power. Additionally, our study sample likely exhibited a healthier profile compared to an age-matched cohort from the general population due to survivorship bias and the exclusion criteria associated with participation in the ARIC-PET study. The fact that our results remained similar, however, in our IPW-weighted analyses, suggests that selection into ARIC-PET is not driving the lack of an association between LLDS with amyloid. Furthermore, given the limited number of participants meeting criteria for LLDS (CES-D ≥ 9), we employed several measures to enhance robustness of our findings, including multiple secondary definitions of depression, which generally remained not associated with amyloid. We acknowledge that our measure for assessing LLDS, the CES-D, is not equivalent to a clinical diagnosis of LLDS. Further, our definition of LLDS (CES-D ≥ 9) may not precisely capture intended clinical neuropsychological symptoms and instead may represent symptoms found frequently in association with aging. Similarly, a formal measure of depressive symptoms, such as the CES-D, was not used to assess depressive symptoms during midlife in our cohort, so we utilized the Maastricht Questionnaire for vital exhaustion to capture aspects of midlife mental wellbeing. Though there is considerable overlap between the three defining features of vital exhaustion (pervasive fatigue, irritability, and demoralization) and depressive symptomatology, the Maastricht Questionnaire is limited in capturing other features of depression such as sad mood, physical activity changes, weight/appetite changes, or trouble concentrating [63]. Analytically, we adjusted for many potential confounders, such as sociodemographic characteristics, vascular risk factors, and APOE ε4 genotype, but it is possible that there are additional confounders that were not controlled for, including consideration of other medical or psychiatric illness or substance abuse. Furthermore, because this is a community-based sample, our participants had higher rates of underlying vascular disease or vascular risk factors than is what typically seen in a clinical sample, which makes it likely other factors contribute to LLDS in this population. Finally, there is likely residual confounding of a potential association between any of our depressive measures and amyloid, given the observational nature of the study.

Despite these limitations, our study offers numerous strengths that contribute to the existing body of literature on depression and AD. Our study sample was composed of both Black and White individuals and included comprehensive data on potential confounding variables, such as APOE genotype, vascular risk factors, and demographic characteristics. Moreover, our stratified analyses allowed us to explore variations in the association of LLDS with amyloid deposition across different subgroups, including Black vs. White individuals, women vs. men, and cognitively normal individuals vs. individuals with MCI. Lastly, we investigated amyloid deposition both globally and across various cortical and subcortical regions, allowing us to assess the relationship between LLDS and amyloid deposition in specific regions related to emotion processing.

5. Conclusion

Our study demonstrates that LLDS are not associated with global amyloid deposition when accounting for potential confounders. Further, our investigation revealed an association between antidepressant use and global amyloid deposition only in those with normal cognition but not in those with MCI. While our findings do not support LLDS as a risk factor for the pathological progression of AD, they suggest a potential for LLDS to alter amyloid dynamics in subcortical brain regions linked to depression. Moreover, our results highlight the significance of considering cognition in the interplay between LLDS and AD, as well as considering the possibility that LLDS and AD may be linked through alternate pathways, such as cerebrovascular disease. Future research employing longitudinal design is warranted to enhance our understanding of potential pathways linking mood disorders and AD.

CRediT authorship contribution statement

Sylee Kanetkar: Writing – review & editing, Writing – original draft, Visualization, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Valerie N. Morrill: Writing – review & editing, Supervision, Methodology, Formal analysis, Data curation, Conceptualization. Marco T. Egle: Writing – review & editing. Keenan A. Walker: Writing – review & editing. Dean F. Wong: Writing – review & editing, Resources. Rebecca F. Gottesman: Writing – review & editing, Supervision, Methodology, Formal analysis, Data curation, Conceptualization.

Funding

Open access funding provided by the National Institutes of Health. The ARIC study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (NHLBI) (75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, 75N92022D00005). The ARIC Neurocognitive Study is supported by U01HL096812, U01HL096814, U01HL096899, U01HL096902, and U01HL096917 from the National Institutes of Health (NHLBI, NINDS, National Institute on Aging [NIA], and National Institute on Deafness and Communications Disorders [NIDCD]). The ARIC-Positron Emission Tomography (PET) ancillary study is funded by the NIA (R01AG040282). Avid Radiopharmaceuticals provided the Florbetapir isotope for the study. SK, VNM, MTE, and RFG were supported by the NINDS Intramural Research Program. KAW was supported by the NIA Intramural Research Program.

This research was supported by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH authors are considered Works of the United States Government. The findings and conclusions presented in this paper are those of the authors and do not necsssarily reflect the views of the NIH or the US Department of Health and Human Services.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: DFW is a consultant for Engrail Therapeutics and receives contract funds from Eisai, Anavex, Intracellular Technologies, LB Pharma, and Roche. These potential conflicts of interest have been reviewed and are managed by Washington University School of Medicine. Other authors have no conflicts of interest.

Appendix A. Supplementary data

The following are the Supplementary data to this article: