Depression and the Diagnosis of MCI in a Culturally Diverse Sample in the United States

Abstract

Objective

To analyze (1) whether there are ethnic differences in the severity of depressive symptoms between groups of elders classified as cognitively normal (CN) or amnestic mild cognitive impairment (aMCI) and (2) the influence of depressive symptoms on specific cognitive performance by ethnicity across diagnoses, controlling for covariates.

Methods

164 Hispanics residing in the United States (HAs) and European Americans (EAs) (100 women; M_age = 72.1, SD = 8.0) were diagnosed as either CN or aMCI. Depressive symptoms were measured with the Geriatric Depression Scale (GDS-15). Cognition was assessed using the Loewenstein-Acevedo Scales for Semantic Interference and Learning (semantic memory), Multilingual Naming Test (confrontation naming), and the Stroop Test (Color–Word condition; executive function). A 2 × 2 univariate ANCOVA as well as linear and logistic regressions explored differences in depressive symptoms among diagnostic and ethnic groups.

Results

Higher depression was seen in aMCI compared to the CN group for both ethnicities, after controlling for age, education, gender, and Mini-Mental State Examination score. Greater levels of depression also predicted lower scores in confrontation naming and semantic memory for only the EA group and marginally in scores of executive function for HA participants. GDS-15 scores of ≤ 4 also predicted less likelihood of aMCI diagnosis.

Conclusions

Severity of depressive symptoms was associated with greater cognitive impairment, independent of ethnicity. Significant results suggest detrimental effects of depression on clinical diagnoses most evidently for subjects from the EA group.

The contribution of depression to the diagnosis of MCI in a culturally diverse sample in the United States

Depression is one of the most pervasive disorders observed in old age, with a prevalence in the United States that ranges between 1% and 16% for major depression, 2%–19% for minor depression, and 7.2%–49% for clinically relevant depressive symptoms of elderly individuals living within the community or in nursing homes (Djernes, 2006; Skoog, 2011). The World Health Organization estimated that the rate of depressive disorders in the elderly population varies between 10% and 20% (Rangaswamy, 2001). According to Ismail and colleagues (2017), the overall pooled prevalence of reported depression in patients with mild cognitive impairment (MCI) was 32%.

Depressive symptoms in late life are now recognized as a “prodrome” or a major predictor of amnestic MCI (Geda et al., 2014), as well as Alzheimer’s disease (AD; Enache, Winblad, & Aarsland, 2011; Mosoiu, 2016). Depression has been consistently associated with increased incidence of dementia and a higher rate of progression from MCI to AD (Byers & Yaffe, 2011; Gallagher, Kiss, Lanctot, & Herrmann, 2018; Geerlings et al., 2000; Mourao, Mansur, Malloy-Diniz, Castro Costa, & Diniz, 2016; Ownby, Crocco, Acevedo, John, & Loewenstein, 2006). In a systematic review and meta-analysis of 23 studies observing the risk of incident all-cause dementia (due to an unspecified disease), AD, and vascular dementia, Diniz, Butters, Albert, Dew, and Reynolds (2013) demonstrated that individuals with late-life depression (LLD) had a significantly higher risk for occurrence of all three neurological disorders (with the most elevated risk for incident vascular dementia). Depression and additional neuropsychiatric symptoms of anxiety, psychosis, agitation, and apathy have been associated with an increased likelihood of progression from MCI to AD/dementia (Donovan et al., 2015; Geda et al., 2014; Pink et al., 2015; Rosenberg, Mielke, Appleby, Oh, & Yonas, 2013; Wilson, Begeny, Boyle, Schneider, & Bennet, 2011).

Most studies investigating depression and cognition in the elderly have analyzed late-onset depression (LOD; occurring after age 60), as opposed to early-onset depression (EOD; occurring before age 60). LOD presents higher medical comorbidity and demonstrates more frontal–subcortical damage and white matter lesions, producing increased cognitive impairment and less response to treatment (Espinoza & Kaufman, 2014). Koenig and coworkers (2015) examined the profile of cognitive impairment in individuals with a history of depression (classified either as currently or previously depressed) and found that compared to age- and education-matched controls, depressed individuals performed worse than never-depressed participants on tasks of global cognition in addition to tasks of episodic memory, attention, processing speed, as well as verbal and visuospatial ability, with no overall differences observed in executive function. Previously depressed participants performed similarly to the nondepressed group on the Boston Naming Test (Kaplan, Goodglass, & Weintraub, 1983), a measure of confrontation naming. Therefore, this demonstrates mixed findings in determining the influence of depression on cognitive impairment among elderly individuals.

Few studies have observed which cognitive deficits may be best predicted by depressive symptoms in abnormal aging, despite the established relationship between depression and MCI as well as AD/dementia (i.e., Diniz et al., 2013; Mourao et al., 2016). Research evaluating cognitive function and depression in MCI has revealed lower delayed free recall in depressed individuals with MCI compared to nondepressed MCI and cognitively normal (CN) adults (Hudon, Belleville, & Gauthier, 2008). Furthermore, studies have found an association between depression and lower performance on other cognitive tasks that involve processing speed, memory, and executive function (Gonzales et al., 2017; Steenland et al., 2012; Thomas & O’Brien, 2008). Still, it remains unknown which neuropsychological domain is most sensitive to depression in normal elders and in those diagnosed with MCI.

Some reports in the United States refer to disparities of depression across ethnic groups. For example, depression is often presented more frequently in Hispanics compared to white non-Hispanics, with similar or higher reported levels compared to African Americans in preretirement age (Dunlop, Song, Lyons, Manheim, & Chang, 2003; González, Tarraf, Whitfield, & Vega, 2010; Rodriguez-Galan & Falcón, 2009; Russell & Taylor, 2009). Contributing factors to depression in Hispanics (O’Bryant et al., 2013) include less years of education and lower socioeconomic status described in this ethnic group compared to non-Hispanic Americans (Mungas, Reed, Haan, & González, 2005; Williams, Mohammed, Leavell, & Collins, 2010). Depression and diabetes as well as age and level of education have been shown to increase the risk of MCI for Mexican Americans, whereas this has not been seen for non-Hispanic whites (Johnson et al., 2015). It has also been suggested that levels of acculturation (a process of adaptation or assimilation by an ethnic or racial group to a host culture; Berry, 2003) may influence immigrants’ performance on neuropsychological tests in addition to levels of depression, cultural barriers, and poorer health status (Arnold, Montgomery, Castañeda, & Longoria, 1994; González, Haan, & Hinton, 2001; Torres, 2010).

In summary, depression has been shown to influence levels of neuropsychological functioning in CN subjects and in those diagnosed with MCI (Gonzales et al., 2017; Koenig et al., 2015; Steenland et al., 2012), but it is still unclear whether certain cognitive domains are differentially affected by depressive symptoms and if this relationship is influenced by the individual’s ethnicity and level of acculturation to the host culture. This study aimed to address this gap by examining (a) differences in levels of depression for an ethnically diverse sample of CN individuals and those with amnestic mild cognitive impairment (aMCI) and (b) the impact of depression on certain neuropsychological domains comparing Hispanic Americans (HAs) and European Americans (EAs) across both diagnostic groups.

It was hypothesized that elevated levels of depressive symptoms as measured by the Geriatric Depression Scale (GDS)-15 would be reported in the HA sample compared to EA participants across aMCI and CN, with the highest scores observed in the aMCI group (based on previous findings from O’Bryant et al., 2013 and Ortiz, Fitten, Cummings, Hwang, & Fonseca, 2006). We also expected a main effect of ethnicity, with more depressive symptoms reported in the HA group. In addition, this study explored the association between scores of depression and performance on specific neuropsychological tasks, which included the Stroop Color–Word (SCW; inhibitory control), Loewenstein-Acevedo Scales for Semantic Interference and Learning (LASSI-L) 20-min delayed recall (semantic memory; Curiel et al., 2013) and the Multilingual Naming Test (MINT, confrontation naming; Gollan, Weissberger, Runnqvist, Montoya, & Cera, 2012) for the two ethnicities, HA and EA, across both diagnoses of CN and aMCI. Scores on the SCW and LASSI-L delayed recall were predicted to be negatively associated with levels of depression in the CN and aMCI groups. This prediction is in line with previous findings reporting reduced executive function and memory performance in individuals with elevated depressive symptoms for both CN and MCI diagnoses (Butters et al., 2004; Gonzales et al., 2017; Thomas & O’Brien, 2008; Steenland et al., 2012).

Finally, exploratory analyses were used to determine if absence of depression (as measured by a dichotomous variable for the GDS-15 of ≤ 4, and >4 representing depression) predicted less likelihood of being diagnosed with aMCI (González et al., 2001; Kørner et al., 2006; Martínez de la Iglesia et al., 2002), while controlling for age, education, and ethnicity. For only the HA sample, the association between level of acculturation to the U.S. American culture and number of depressive symptoms was also investigated, after controlling for relevant demographic variables. It was anticipated that participants with a higher level of acculturation would present lower depressive symptoms.

Method

Participants

The analysis of the current study was based on a subsample of participants from the 1Florida Alzheimer's Disease Research Center, part of a 5-year longitudinal study with data collected between September 2015 and October 2018. We included baseline data (Year 1) from participants who received a clinical diagnosis of CN or aMCI (N = 210). Sixteen participants were removed due to comorbid diagnoses of schizophrenia, bipolar disorder, and posttraumatic stress disorder. One participant was removed because of a diagnosis of non-amnestic MCI. Additionally, 29 participants were removed who were immigrants from countries other than Spanish-speaking Latin American countries, participants whose first language was not Spanish or English, and those whose ethnicities were not HA or EA, resulting in a total sample of 164 (100 women; HA n = 104; EA n = 60). The characteristics of the sample are included on Table 1. The means for age, years of education, and Mini-Mental State Examination (MMSE) total were 72.13 (SD = 7.97), 15.02 (SD = 3.39), and 27.88 (SD = 2.29), respectively. The sample was 63.4% Hispanic (n = 104), with 79.8% of the HA sample (n = 83) tested in Spanish. The remaining 21 HA participants were fluent Spanish–English bilinguals who preferred to be tested in English. Sixty-seven percent (n = 110) received a consensus diagnosis of aMCI.

Table 1.

Characteristics of the sample using percentages by ethnic groups

	Hispanic American		European American
Variable	Normal n = 31	aMCI n = 73	Normal n = 23	aMCI n = 37
Gender (female) Born in US Raised in US > 12 years of education Handedness (right)	77.4% (24) 0.0% (0) 16.1% (5) 77.4% (24) 90.3% (28)	58.9% (43) 4.1% (3) 13.7% (10) 60.3% (44) 97.3% (71)	69.6% (16) 100.0% (23) 100.0% (23) 95.7% (22) 65.2% (15)	45.9% (17) 100.0% (37) 100.0% (37) 83.8% (31) 83.8% (31)
Antidepressant use	38.7% (12)	52.1% (38)	34.8% (8)	54.1% (20)
Active depression	35.5% (11)	43.8% (32)	34.8% (8)	43.2% (16)
Active anxiety	45.2% (14)	31.5% (23)	34.8% (8)	37.8% (14)
Diabetes	16.1% (5)	32.9% (24)	17.4% (4)	13.5% (5)
Hypertension	58.1% (18)	63.0% (46)	30.4% (7)	43.2% (16)
Hypercholesterolemia	67.7% (21)	76.7% (56)	43.5% (10)	70.3% (26)

aMCI = amnestic mild cognitive impairment.

The Functional Activities Questionnaire (FAQ; Pfeffer, Kurosaki, Harrah, Chance, & Filos, 1982) was also administered by the clinician to the participant’s study partner in order to examine daily function. The FAQ is a questionnaire that assesses the participant’s level of functionality in aspects of daily living which include preparing meals and coffee, writing checks/assembling tax records, shopping and traveling, and remembering appointments. The mean FAQ score was 2.38 (SD = 2.87) (reported on a 4-point Likert scale of 0 = normal, 1 = has difficulty but does by self, 2 = requires assistance, and 3 = dependent). There were no significant differences observed between HA and EA groups in the level of functionality for both diagnoses as measured by the FAQ.

Within the sample, 47.6% (n = 78) were taking antidepressants, including selective serotonin reuptake inhibitors and selective serotonin and norepinephrine reuptake inhibitors. Also, a few other cases (4.9%; n = 8) were taking other anxiolytic, antipsychotic, or opioid medications. The prevalence of these medications, active anxiety (determined by dichotomous values of absent and recent/active) and active depression (based on dichotomous values of yes and no), did not significantly differ by ethnic or diagnostic groups (see Table 1).

Additionally, other comorbid disorders that were treated in this sample included hypertension, cardiovascular disease/high cholesterol, and diabetes. There were no significant differences in the GDS scores between those who had diabetes, t(18.71) = 1.65, p = .12, hypertension, t(135) = −.15, p = .88, and cardiovascular/cholesterol diseases, t(166) = −.02, p = .98, with and without treatment. There were no significant differences between ethnicities in the percentage of participants from the CN and aMCI groups suffering from vascular or neurological diseases, hypertension, or hypercholesterolemia. However, chi-square tests revealed significant differences in the prevalence of diabetes for the aMCI diagnosis across ethnic groups, Pearson χ²(1) = 4.74, p = .03, with the HA group more likely to be diagnosed with diabetes compared to the EA participants. No differences were observed in the normal group.

Demographic variables between the two ethnic groups are presented on Table 2. The EA group had significantly higher years of education. Table 3 shows the demographic characteristics by ethnic and diagnostic group. In the aMCI group, the mean years of education was significantly higher for EA compared to HA, and this difference was marginal for normal participants. It is also important to note that there was unequal variance (indicated by a significant Levene’s statistic for age, p = .03) in the aMCI group. However, after using a Mann–Whitney U test (a robust and reliable test for unequal variances), there were no significant differences between ethnicities in age.

Table 2.

Means and standard deviations of demographic variables and the MMSE scores by ethnicity

	Hispanic American Mean (SD) n = 104	European American Mean (SD) n = 60	F	p	ηp2
Age	71.52 (7.37)	73.18 (8.88)	1.66	.20	.07
Education (years)	14.37 (3.39)	16.17 (3.11)	11.41	.001	.02
MMSE	27.69 (2.34)	28.22 (2.17)	2.01	.16	.01

Notes. η_p² = partial eta squared; MMSE = Mini-Mental State Examination.

Table 3.

Means (SD) and ANOVAs by ethnicity and diagnostic groups

	Hispanic American Mean (SD) n = 104		European American Mean (SD) n = 60
	Normal n = 31	aMCI n = 73	Normal n = 23		aMCI n = 37
Age	69.90 (6.96)	72.21 (7.48)	70.30 (6.05)		74.97 (9.92)
Education	15.42 (3.04)	13.92 (3.45)	17.00 (2.80)		15.65 (3.22)
MMSE	29.35 (0.84)	26.99 (2.42)	29.17 (0.98)		27.62 (2.49)
	Normal n = 54			aMCI n = 110
	F	p	ηp2	F	p	ηp2
Age	0.05	.83	.00	2.68	.10	.02
Education	3.82	.06	.07	6.47	.01	.06
MMSE	0.53	.47	.01	1.66	.20	.02

Notes. MMSE = Mini-Mental State Examination; aMCI = amnestic Mild Cognitive Impairment.

Diagnostic determination

The determination of CN or aMCI was made by a consensus diagnosis between a neurologist, a neuropsychologist, and a psychiatrist, after analyzing psychometric neuropsychological tests, performing a physical examination, and interviewing the patient and study partner to assess levels of daily functioning using the FAQ. The diagnosis was in accordance with the criteria for aMCI, which required the presence of memory complaints, particularly in episodic memory, with preserved performance in global cognition and daily life activities (Petersen et al., 1999). To reduce the number of common false-negative diagnostic errors related to Petersen and coworkers’ conventional diagnostic criteria for MCI (Edmonds et al., 2016), we used neuropsychological criteria for MCI of 1.5 SD below the mean on at least one cognitive task, as well as a score on the FAQ of ≥9.

A neuropsychological test battery was administered in English to all EA participants and in the preferred language—English or Spanish—to HA participants. The diagnosis of aMCI was based on the following criteria: (a) subjective memory complaints made by the participant and/or informant; (b) evidence by clinical evaluation/history of memory or other cognitive decline; (c) a Global Clinical Dementia Rating (CDR) scale score of 0.5; (d) performance of 1.5 SD below the mean of age, education, and language-matched norms on one or more memory measures: the total immediate and delayed Hopkins Verbal Learning Test—Revised (HVLT-R) recall (Benedict, Schretlen, Groninger, & Brandt, 1998) or the delayed recall of the National Alzheimer’s Coordinating Center Logical Memory story passage (Beekly et al., 2007), which were used to differentiate MCI from CN elderly individuals (de Jager, Schrijnemaekers, Honey, & Budge, 2009); and (e) scores within normal range on tasks assessing non-memory cognitive functioning such as naming (MINT), visual attention (Trail Making Test [TMT] A and B), and inhibition (SCW). These tasks are part of the AD initiative for English and Spanish-speaking cohorts and have been applied for determination of diagnosis in several studies with English and Spanish-speaking participants (Loewenstein et al., 2016, 2017a, 2017b).

Measures

Neuropsychological tests

Participants completed a comprehensive neuropsychological evaluation which assessed various cognitive domains. Verbal memory was measured using the LASSI-L (Curiel et al., 2013), HVLT-R (Brandt, 1991), the Craft Story (Craft et al., 1996), and Logical Memory (Abikoff et al., 1987); confrontation naming was tested with the MINT (Gollan et al., 2012); visuospatial cognitive functioning was evaluated with the Benson Figure Drawing (Possin, Laluz, Alcantar, Miller, & Kramer, 2011) and Block Design (Wechsler, 2014a); overall cognition was estimated using the MMSE (Folstein, Folstein, & McHugh, 1975); executive function was appraised with the SCW (Stroop, 1935; Trenerry et al., 1989), as well as TMT B (Corrigan & Hinkeldey, 1987; Reitan, 1958; Reitan & Wolfson, 1993); and finally, verbal fluency was assessed using category (Benton, 1968) and phonemic fluency (Benton, Hamsher, & Sivan, 1994).

Spanish language evaluations were completed with equivalent standardized neuropsychological tests. Tasks administered to primary Spanish speakers had appropriate age, education, and cultural/language normative data for the translated versions (Arango-Lasprilla, Rivera, Aguayo et al., 2015; Arango-Lasprilla, Rivera, Garza et al., 2015; Benson, de Felipe, Xiadong, & Sano, 2014; Gollan et al., 2012; Golden, 1999; Ostrosky-Solís, López-Arango, & Ardila, 2000; Peña-Casanova et al., 2009a; Peña-Casanova et al., 2009b; Wechsler, 2014b). Testing was performed by proficient Spanish/English psychometricians.

The neuropsychological domains that were analyzed in this study included the following:

Semantic memory

The LASSI-L delayed recall trial examined the ability to recall two lists of 15 words that belong to three shared semantic categories (fruits, musical instruments, and articles of clothing) after a 20-min delay in order to assess semantic memory. This task has established concurrent and discriminative validity in English (Curiel et al., 2013). The 20-min delay trial has been shown to be a statistically significant predictor of AD diagnosis for Spanish-speaking participants (Matías-Guiu et al., 2017; see full description of task in Loewenstein et al., 2017b) and has even been effective in detecting differences between CN and pre-MCI participants (Crocco et al., 2018). The LASSI-L has high sensitivity (87.9%) and specificity (91.5%) in determining aMCI (Crocco, Curiel, Acevedo, Czaia, & Loewenstein, 2014) in addition to high test–retest reliability, r = 0.60–0.89 (Loewenstein & Acevedo, 2005).

Confrontation naming

The MINT measured confrontation naming. This task includes 32 pictures of objects in which the participant is prompted to produce the name of each object, and upon failure to retrieve the name, a semantic (description of the object) or phonemic (the beginning sound of the word) cue is given (Gollan et al., 2012). Total scores were calculated by adding the number of pictures named correctly, including retrieval with a semantic cue. Scoring was based on the 32-item MINT used in Ivanova, Salmon, and Gollan (2013) and normative data from Gollan et al. (2012).

Inhibitory control

This component of executive function was measured with the raw score of the SCW (Strauss, Allen, Jorgensen, & Cramer, 2005; Stroop, 1935; Trenerry, Crosson, DeBoe, & Leber, 1989). The SCW inhibition trial from the Stroop Test: Adult Version in English (Golden & Freshwater, 2002) and Spanish (Golden, 1999) measures processing speed, inhibitory abilities, and resistance to interference. The subject must refrain from reading a word that is a color and, instead, is instructed to identify the contrasting ink color of the text. Participants are corrected by the examiner if words are read instead of naming the color, incurring a time penalty. The number of correctly read words in 45 seconds is the total score, with higher scores reflecting better performance. The incongruent SCW condition has been indicated as a measure of inhibition of competing stimuli as well as maintenance of attentional set (MacLeod & MacDonald, 2000). Normative data of this task was included from Rivera and coworkers (2015) and Ivnik, Malec, Smith, Tangalos, and Petersen (1996).

Depression

To assess depression, the Geriatric Depression Scale-Short Form (GDS-15 in English: Sheikh & Yesavage, 1986; Yesavage et al., 1982; GDS-VE in Spanish: Martínez de la Iglesia et al., 2002) was employed. It includes 15 items such as “Are you basically satisfied with your life?” and “Are you afraid something bad is going to happen to you?” This questionnaire was administered by the psychiatrist to the participant in English or Spanish, whichever language the participant felt most comfortable to respond in. This scale has 15 statements with “yes” or “no” responses coded 0 or 1, depending on whether the item represented depressive symptoms (1) or not (0). The GDS-15 and GDS-30 in English have good validity among elderly populations (Kørner et al., 2006; Marc, Raue, & Bruce, 2008) and with MCI (Debruyne et al., 2009), in addition to demonstrating high internal consistency (Cronbach’s α = .80; D’ath, Katona, Mullan, Evans, & Katona, 1994). Also, the Spanish GDS-15 and GDS-30 versions have been validated in subjects with dementia and have high reliability (Cronbach’s α = .81; Fernández-San Martín et al., 2002; Lucas-Carrasco, 2012; Martínez de la Iglesia et al., 2002). For the current study, as a continuous variable, the total GDS scores represented prevalence of depressive symptoms, and as a dichotomous variable, these scores represented incidence of depression (present or not present). The cutoff score for GDS as a dichotomous variable in the logistic regression analyses was 4/5 (≤4: no depression; > 4: depression), which represented sensitivity and specificity for significant depressive symptomology according to previous research in English and Spanish (D’ath et al., 1994; de Craen, Heeren, & Gussekloo, 2003; Kørner et al., 2006; Martínez de la Iglesia et al., 2002).

Acculturation

The Bidimensional Acculturation Scale (BAS; Marín & Gamba, 1996) was used to observe acculturation to the U.S. American culture and respective native country cultures by examining levels of English and Spanish language use. This questionnaire uses a 1–4 Likert scale measuring two major cultural dimensions, “Hispanic” and “non-Hispanic,” with 12 items per culture. It includes three language-related questions such as language use (“How often do you speak English?” with a 4-point Likert scale, 1: Almost Never, 2: Sometimes, 3: Often, and 4: Almost Always), linguistic proficiency (“How well do you speak English?” with a 4-point Likert scale, 1: Very Poorly, 2: Poorly, 3: Well, and 4: Very Well), and electronic media (“How often do you watch television programs in English?” with a 4-point Likert scale, 1: Almost Never, 2: Sometimes, 3: Often, and 4: Almost Always). The 12 items for each cultural domain (Hispanic and non-Hispanic) were averaged across all items for each respondent, and the level of acculturation for the non-Hispanic domain (or acculturation to the U.S. American culture) was used in the analyses (Marín & Gamba, 1996).

Procedure

This study was approved by the Mount Sinai Medical Center Institutional Review Board. After written informed consent was obtained, the neuropsychological battery was administered to each participant in an approximately 3–4-hr session. There was a break for lunch halfway through testing, with a $10 voucher provided to both the study partner and the participant, as well as $70 given upon completion of the study session. Participants were also interviewed and evaluated by a psychiatrist to assess the presence of depression, additional neuropsychiatric symptoms such as anxiety, functional daily activities, and overall physical health.

Statistical analyses

All analyses were conducted using SPSS Version 25. One-way ANOVAs as well as chi-square tests were performed to examine HA and EA differences in age, years of education, MMSE, and comorbid diseases as well as differences in neuropsychological tests scores (i.e., SCW raw score, LASSI-L delayed recall, and MINT total).

A 2 × 2 univariate analysis of covariance (ANCOVA) was used to explore differences in levels of depression across two diagnostic groups: CN and aMCI in both ethnicities, HA and EA. Because age, years of education, gender, and global cognition (MMSE) may cause discrepancies in the scores, these variables were included as covariates for the analyses. Due to the high number of participants in the HA group with diabetes compared to the EA group, an additional ANCOVA was performed including diabetes as a covariate.

In order to determine which neuropsychological scores may be best predicted by depression in each ethnic group after adding the other covariates, a series of three stepwise regressions were performed using the total GDS score as a continuous independent variable in Step 1, age and education were added at Step 2, and the MMSE was entered at Step 3 with the neuropsychological test scores (SCW, LASSI-L delayed recall, and MINT total) included as dependent or outcome variables. Because diabetes significantly differed across ethnicity, it was also a covariate predictor for this analysis at Step 4 of the regression model for each neuropsychological test.

Exploratory logistic regression analyses were performed using the binomial dependent variable of diagnostic category (CN and aMCI) with predictors of GDS grouped by scores of ≤4 representing no depression and >4 indicating significant reported depression, age, years of education, and ethnicity. To measure the contribution of diabetes to this model, an additional logistic regression was conducted with this variable included as a predictor of diagnostic category. Finally, a bivariate partial Pearson correlation analyzed level of acculturation to the U.S. American culture (BAS score) and GDS only for the HA sample. All tests assumed a p value of <.05 as statistically significant.

Results

Ethnic group differences

Table 4 contains the means and standard deviations for the cognitive tests by diagnostic and ethnic group. For the normal and aMCI groups, the MINT total was significantly higher for the EA compared to the HA group.

Table 4.

Means (SD) and ANOVAs by ethnicity for test variables

	Hispanic American n = 104		European American n = 60
	Normal n = 31	aMCI n = 73	Normal n = 23		aMCI n = 37
Stroop CW	33.77 (7.23)	27.81 (9.46)	34.13 (8.20)		25.32 (9.23)
LASSI delay	20.00 (4.13)	13.05 (5.78)	20.65 (3.52)		14.27 (5.53)
MINT total	27.03 (2.11)	25.44 (4.47)	29.78 (2.56)		28.19 (4.01)
	Normal n = 54			aMCI n = 110
	F	p	ηp2	F	p	ηp2
Stroop CW	0.03	.87	.00	1.72	.19	.02
LASSI delay	0.37	.55	.01	1.12	.29	.01
MINT total	18.76	<.001	.27	9.96	.002	.08

Notes. Stroop CW = Stroop Color–Word Raw score, LASSI delay = Loewenstein-Acevedo Scales of Semantic Interference and Learning—20-min delay, MINT total = Multilingual Naming Test total; aMCI = amnestic mild cognitive impairment.

Because the HA participants were tested in one of two languages, additional analyses were conducted to assess performance differences between HA tested in English or Spanish on the neuropsychological tests. Comparing HA participants in each evaluation language, none of the cognitive tests significantly differed across diagnostic group.

Depression results

The results of the 2 × 2 univariate ANCOVA analyzing diagnostic group (CN and aMCI) and ethnicity (HA and EA) as between factors on GDS scores are shown on Table 5. The main effect of ethnicity was not significant, whereas the effect for diagnostic group was significant with greater depressive symptoms found in the aMCI group compared to controls. None of the other covariates were significant (i.e., MMSE, age, education, gender). There was no significant interaction between ethnicity and diagnostic group in differences of total GDS score. When adding diabetes as a covariate to the ANCOVA model, the effect of this variable on depression was significant, F(1, 155) = 6.56, p = .01, η_p² = .04, with greater depressive symptoms reported in those diagnosed with diabetes. With diabetes included in the model, years of education became marginally significant, F(1, 155) = 3.41, p = .07, η_p² = .02, whereas diagnosis lost significance, F(1, 155) = 3.54, p = .06, η_p² = .02).

Table 5.

Means, standard deviations and univariate analysis of covariance for differences in the Geriatric Depression Scale (GDS-15) by ethnicity and diagnostic group (covariates: age, education and Mini-Mental State Examination [MMSE])

Variable	Hispanic American n = 104	European American n = 60
	Mean (SD)	Mean (SD)	Total mean (SD)	Variable	F	p	ηp2
Normal (n = 54)	1.94 (2.08)	1.26 (1.60)	1.65 (1.91)	Age Years of education	1.38 2.81	.24 .10	.01 .02
aMCI (n = 110)	2.70 (2.53)	2.19 (2.91)	2.53 (2.66)	Gender	0.82	.37	.01
Total Mean	2.47 (2.42)	1.83 (2.52)	2.24 (2.47)	MMSE Ethnicity Diagnosis Ethnicity*diagnosis	1.76 2.29 3.83 0.17	.19 13 .05 .68	.01 .01 .02 .00

Notes. Hispanic American: cognitively normal (CN) = 31, amnestic mild cognitive impairment (aMCI) = 73; European American: cognitively normal (CN) = 23, amnestic mild cognitive impairment (aMCI) = 37.

The three stepwise regressions are presented on Table 6 with predictors of GDS, age, years of education, MMSE, and diabetes on outcomes of neuropsychological task scores for each ethnic group in the combined CN and aMCI diagnoses. For the EA group, depression by itself was a significant predictor for the MINT total, explaining 13% of the variance. Age, years of education, MMSE, and diabetes were not significant predictors of MINT performance for EA participants when added to the model. GDS remained significant at Steps 2, 3, and 4, even after the covariates were introduced. Interestingly, the best regression model to predict MINT performance was at Step 1 with depression as the only predictor. For the EA group in the LASSI-L model, GDS alone explained a significant 8% of the variance for semantic memory delayed recall. Years of education and age explained an additional 18% of this score’s variance, and the MMSE contributed an extra 11%. After adding diabetes in Step 4 of the model, there was no significant change in predicted variance. For the SCW model, the only significant predictors for the EA group were age and MMSE scores at Steps 2, 3, and 4, explaining 22% of the variance. This suggests that younger participants with higher MMSE scores have better performance on the SCW.

Table 6.

Summary of stepwise regression analyses for ethnic groups of depression (GDS-15 total) predicting neuropsychological outcomes in normal and aMCI groups

Hispanic American						European American
MINT model
Predictors	B	SE B	β	t	p	B	SE B	β	t	p
Step 1
GDS	−.08	.16	−.05	−0.50	.620	−.53	.17	−.38	−3.08	.003
Adjusted R2	−.01					.13
F value	0.3				.620	9.50				.003
Step 2
GDS	−.10	.16	−.06	−0.63	.533	−.57	.18	−.40	−3.21	.002
Age	.01	.06	.02	0.15	.878	−.08	.05	−.19	−1.51	.138
Education	.30	.12	.25	2.47	.015	−.07	.14	−.06	−0.47	.640
Adjusted R2	.04					.13
F value	2.24				.088	3.98				.012
Step 3
GDS	−.02	.16	−.01	−0.11	.911	−.54	.18	−.38	−2.97	.004
Age	.04	.05	.08	0.76	.449	−.06	.06	−.14	−1.04	.303
Education	.19	.12	.16	1.61	.110	−.06	.14	−.05	−0.42	.674
MMSE	.56	.17	.33	3.30	.001	.19	.22	.11	0.86	.396
Adjusted R2	.12					.13
F value	4.57				.002	3.15				.021
Step 4
GDS	−.03	.16	−.02	−0.19	.848	−.55	.19	−.38	−2.91	.005
Age	.04	.05	.07	0.73	.467	−.06	.06	−.14	−1.00	.321
Education	.20	.12	.17	1.64	.105	−.06	.15	−.05	−0.41	.680
MMSE	.56	.17	.33	3.28	.001	.19	.22	.11	0.85	.401
Diabetes	.34	.85	.04	0.41	.686	.16	1.28	.02	0.13	.899
Adjusted R2	.12					.11
F value	3.66				.004	2.48				.043
LASSI-L model
Step 1
GDS	−.18	.26	−.07	−0.72	.474	−.71	.29	−.31	−2.47	.016
Adjusted R2	−.01					.08
F value	0.52				.474	6.11				.016
Step 2
GDS	−.18	.24	−.07	−0.75	.457	−.92	.27	−.40	−3.45	.001
Age	−.23	.08	−.27	−2.74	.007	−.30	.07	−.46	−3.96	<.001
Education	.25	.18	.13	1.35	.179	.01	.21	.00	0.03	.980
Adjusted R2	.09					.26
F value	4.42				.006	7.79				<.001
Step 3
GDS	.02	.21	.01	0.11	.913	−.76	.25	−.33	−3.03	.004
Age	−.15	.07	−.18	−2.04	.044	−.20	.08	−.31	−2.65	.011
Education	−.01	.16	−.01	−0.07	.942	.04	.20	.02	0.20	.840
MMSE	1.39	.23	.53	5.93	<.001	.98	.30	.37	3.25	.002
Adjusted R2	.33					.37
F value	13.26				<.001	9.47				<.001
Step 4
GDS	.07	.22	.03	0.30	.763	−.68	.25	−.30	−2.71	.009
Age	−.15	.07	−.17	−1.98	.051	−.21	.08	−.33	−2.86	.006
Education	−.03	.16	−.02	−0.16	.870	.03	.19	.02	0.16	.877
MMSE	1.40	.23	.53	5.94	<.001	.98	.30	.37	3.29	.002
Diabetes	−1.10	1.16	−.08	−0.95	.345	−2.49	1.71	−.16	−1.45	.152
Adjusted R2	.32					.38
F value	10.78				<.001	8.15				<.001
Step 1
GDS	−.67	.38	−.17	−1.78	.079	−.32	.51	−.08	−0.62	.537
Adjusted R2		.02					−.01
F value	3.16				.079	0.39				.537
Step 2
GDS	−.68	.32	−.18	−2.15	.034	−.73	.48	−.19	−1.51	.136
Age	−.51	.11	−.41	−4.72	<.001	−.45	.14	−.41	−3.32	.002
Education	.76	.23	.28	3.22	.002	.43	.39	.14	1.12	.266
Adjusted R2	.32					.15
F value	16.79				<.001	4.50				.007
Step 3
GDS	−.50	.30	−.13	−1.67	.098	−.50	.47	−.13	−1.07	.288
Age	−.44	.10	−.35	−4.25	<.001	−.31	.14	−.28	−2.20	.032
Education	.54	.23	.20	2.34	.022	.48	.37	.15	1.31	.197
MMSE	1.18	.33	.30	3.54	.001	1.41	.57	.31	2.48	.016
Adjusted R2	.39					.22
F value	17.20				<.001	5.22				.001
Step 4
GDS	−.33	.30	−.09	−1.10	.274	−.46	−.48	−.12	−0.95	.346
Age	−.42	.10	−.34	−4.22	<.001	−.32	.14	−.29	−2.23	.030
Education	.48	.22	.17	2.14	.035	.48	.37	.15	1.28	.206
MMSE	1.18	.32	.30	3.70	<.001	1.41	.57	.31	2.47	.017
Diabetes	−4.55	1.58	−.22	−2.88	.005	−1.47	3.28	−.05	−0.45	.656
Adjusted R2	.43					.21
F value	16.43				<.001	4.15				.003

Notes. Covariates included age, years of education, Mini-Mental State Examination (MMSE), and diabetes, which were entered in four steps. First step: Geriatric Depression Scale (GDS), second step: age and education, third step: MMSE; and fourth step: diabetes.

For the regression models of the HA group (presented on Table 6), severity of depressive symptoms did not significantly predict performance on the MINT, LASSI-L, and SCW scores. GDS was a significant contributor only for SCW when age and education were entered into the model (Step 2). However, this significance was lost when the MMSE and diabetes were included as predictors (Steps 3 and 4). Diabetes was a significant predictor of SCW scores in the HA group at Step 4, indicating that those with lower SCW scores were diagnosed with diabetes, were older, had fewer years of education, and scored lower on the MMSE. The HA participants with better MINT, LASSI-L, and SCW performance had higher MMSE scores independent of GDS. For the LASSI-L model in the HA group, the predictive value of the MMSE was significant in explaining an additional 24% of the variance to the 10% already explained by age and education together (Table 6). Education was also a significant predictor before entering the MMSE into the MINT and SCW models for HA participants, remaining significant only in the SCW regression model even after controlling for MMSE. It is important to note that the MMSE scores and years of education were significantly correlated for the HA group (Table 7), indicating that higher level of education was associated with better performance on the MMSE scores only for HA.

Table 7.

Neuropsychological task correlations

		Age	Education	MMSE	Total GDS
Hispanic American	Age	1.00	−.30**	−.27**	.04
	Education		1.00	.32**	.04
	MMSE			1.00	−.15
	Total GDS				1.00
European American	Age	1.00	−.12	.36**	−.20
	Education		1.00	−.04	.18
	MMSE			1.00	−.11
	Total GDS				1.00

Note. ^*p < .05. ^**p < .01. MMSE = Mini-Mental State Examination; GDS = Geriatric Depression Scale.

A binomial logistic regression model determined the outcome of the diagnostic group based on the predictor variables of age, the GDS grouped by total scores of ≤ 4 and >4 (representing no depression and depression, respectively), ethnic group, and level of education (see Table 8). The overall model was significant, χ²(4) = 16.63, p = .002. GDS marginally predicted aMCI diagnosis. Age and education were significant predictors of diagnosis, suggesting that older participants with a lower level of education and reporting more than four depressive symptoms had an increased likelihood of being diagnosed with aMCI. Ethnicity did not significantly influence the probability of being diagnosed with aMCI for the current study. When adding diabetes as a predictor to the model, the significance of the model decreased, χ²(5) = 16.91, p = .005, and diabetes was not a significant individual predictor of diagnostic group (likelihood ratio test χ²(1) = 0.27, p = .60, with the following parameter estimate coefficients: B = −.23, SE B = .44, Wald χ² = .27, p = .61).

Table 8.

Summary of binomial logistic regression analyses of depression in predicting diagnostic category of aMCI with normal as reference after controlling for age, education and ethnicity

MCI diagnosis	B	SE B	Wald	p	OR	95% CI OR
Age	.051	0.02	4.74	.030	1.05	[1.01, 1.10]
Education	−.133	0.06	5.44	.020	0.88	[0.78, 0.98]
Hispanic ethnicity	.223	0.37	0.36	.547	1.25	[0.61, 2.58]
GDS score ≤4	−1.14	0.61	3.72	.061	0.32	[0.10, 1.05]

Notes. Normal was used as the reference category. Two GDS categories were computed comparing scores of 0–4 or 5–15: GDS category 0–4: n = 142, and GDS category 5–15: n = 22. aMCI = amnestic Mild Cognitive Impairment.

A bivariate Pearson correlation revealed a significant association between GDS total and mean of acculturation to the non-Hispanic domain, r(78) = −.273, p = .02. When conducting a partial correlation of these two variables controlling for age, years of education, MMSE, and diabetes diagnosis, this association lost significance, r(72) = −.205, p = .08. These correlations indicated that lower levels of acculturation in the non-Hispanic domain were related to higher depressive symptoms, but this association was additionally influenced by demographic, health, and cognitive variables.

Discussion

In the present study, we investigated whether symptoms of depression differed as a function of ethnicity (EA and HA) by comparing GDS scores in a sample of elderly participants that belonged to diagnostic categories of CN or aMCI. We also analyzed the predictive value of depression on neuropsychological test scores of semantic memory, inhibitory control, and confrontation naming in addition to other covariates across HA and EA ethnic groups in both diagnoses. Results showed that reported depressive symptoms, as measured by the GDS-15, did not significantly differ between HA and EA samples after controlling for general cognition (MMSE), age, education, gender, and comorbid diabetes. Also, increased depressive symptoms were found in aMCI compared to CN, independent of ethnicity. GDS scores, age, and education predicted diagnosis, with individuals from both ethnic groups more likely to be diagnosed with aMCI if they were less educated, older and presented more depressive symptoms.

These results are consistent with previous research indicating that depression in older age significantly influences the progression from normal cognition to MCI (Steenland et al., 2012) and is associated with higher incidence of both MCI and dementia diagnoses (Buntinx, Kester, Bergers, & Knottnerus, 1996; Donovan et al., 2015; Geda et al., 2014; Peters et al., 2015; Rosenberg et al., 2013; Sosa-Ortiz, Acosta-Castillo, & Prince, 2012; Steenland et al., 2012; Tervo et al., 2004; Wilson et al., 2011). Surprisingly, ethnicity was not a contributing factor to depressive symptoms. Previous studies have found that more depressive symptoms were reported in Hispanic participants compared to non-Hispanic whites (Dunlop et al., 2003; González et al., 2010; Rodriguez-Galan & Falcón, 2009; Russell & Taylor, 2009). However, the Composite International Diagnostic Interview Short form was used in Dunlop et al. (2003) to classify and evaluate major depressive disorders. Other studies with ethnic discrepancies in depression have also used the Center for Epidemiological Studies Depression (CES-D) scale (González, Tarraf, Whitfield, & Vega, 2010; Rodriguez-Galan & Falcón, 2009; Russell & Taylor, 2009). In this study, the GDS-15 measured depression and included items that referred specifically to emotional and motivational symptoms of depression; however, there was no evaluation of physical symptoms, which are also common in older adults. Despite being equivalent valid screening measures for LLD (Lyness et al., 1997), psychometric differences of the GDS and the CES-D have been reported (van de Rest, van der Zwaluw, Beekman, de Groot, & Geleijnse, 2010). Another possible contributing factor to the absence of cross-cultural differences in depression is our inclusion of an aMCI clinical sample rather than a community-based nonclinical population (used in the previously mentioned studies).

The significant association between depression and specific cognitive test performance in the EA group is also consistent with preceding research. Prior studies have shown that decreased performance on tasks of semantic and episodic memory, and the risk of conversion to MCI have been predicted by neuropsychiatric symptoms such as depression, apathy, irritability, disinhibition, anxiety, and agitation (Geda et al., 2014; Donovan et al., 2015). Sabbagh et al. (2010) also found correlations between GDS scores and brain biomarkers of abnormal aging such as plaques and tangle counts, suggesting that symptoms of depression increase with illness severity.

For the current study, older EA participants with lower global cognition and higher GDS-15 scores had significantly lower performance on the MINT (confrontation naming) and on the LASSI-L delayed recall trial (semantic memory), with the GDS remaining a significant predictor even after controlling for demographic variables and diabetes diagnosis. The only significant variables in predicting HA neuropsychological outcomes were MMSE scores and age, independent of levels of depression. Younger HA participants with higher MMSE performed better on the delayed semantic memory task (LASSI-L) and the executive function inhibitory task (SCW), with an additional influence of diabetes on lower inhibitory control for only the HA group.

Previous studies have examined group differences on neuropsychological tasks between depressed and nondepressed individuals. Gonzales and colleagues (2017) found a similar result in which subsyndromal depressive symptoms were associated with decreased cognitive function as measured by tasks of global cognition, information processing speed, memory, and semantic fluency for an MCI sample of an unspecified ethnicity. Similarly, depressed CN and MCI individuals expressed differential impairment on cognitive tasks of working memory, executive function, and category fluency (Steenland et al., 2012). To our knowledge, there have not yet been any studies observing the effects of confrontation naming, as measured by the MINT, on predicting levels of depression. Also, few studies have addressed differences of depressive symptoms and their impact on neuropsychological test outcomes between EA and HA in the context of abnormal aging.

Discrepant findings of depression influencing cognition between HA and EA could be related to other confounding variables or cultural differences in coping with memory disorders. Early and coworkers (2013) examined the influence of demographic variables such as race, ethnicity, and education on baseline levels and longitudinal trajectories of cognitive decline for an ethnically diverse sample of African Americans, Hispanics, and Caucasians. Hispanics showed significantly less decline than Caucasians in semantic and episodic memory, as well as executive function, but this relationship lost significance after controlling for language of evaluation and education. In our study, depressive symptoms influenced neuropsychological task scores of confrontation naming and semantic memory exclusively for the EA group, which remained significant even after controlling for demographic variables. Additionally, upon comparing the neuropsychological outcomes of HA tested in English and Spanish, we found no significant differences between these subgroups. Therefore, language of evaluation did not seem to be a confounding variable on cognitive performance for the current sample. Besides these cultural differences in our sample, results showed no significant association between acculturation and symptoms of depression when controlling for demographic variables, which is in line with the findings of previous research (Muruthi, Zalla, & Lewis, 2019). Although most studies in which non-Hispanic Whites were included, ethnicity was not always specified, leading to a possible mix of cultures and languages contained within one group (Donovan et al., 2015; Geda et al., 2014; Gonzales et al., 2017). This may lead to missing the effects of cultural influence on the association between cognitive function and depression. Because the current study data were collected in Miami, a city in which the majority of the population identifies as Latino (United States Census, 2010), the results could indicate that the HA cohort was in a culturally appropriate environment due to the widespread use of the Spanish language and, therefore, was able to maintain a sense of societal support and identity with little acculturative stress. Increased social support (familial support) has been shown to relate to positive self-rated health (Mulvaney-Day, Alegria, & Sribney, 2007), and research has indicated that frequent social interactions can also be protective against dementia (Fratiglioni, Wang, Ericsson, Maytan, & Winblad, 2000). Therefore, one strength of the current study is that the ethnic groups were well-defined; all of the EA participants were born in the United States with origins from European countries, and all HA participants spoke Spanish with origins from Latin American countries.

Biological variables such as metabolic and vascular health conditions that contribute to cognition and comorbid diseases may differ across cultures and ethnic groups, possibly influencing the variance explained by depression (Goldstein, Levey, & Steenland, 2013; Luchsinger et al., 2007; Sparks et al., 2010). Cardiovascular conditions, such as high blood pressure, hypercholesterolemia, elevated central arterial stiffness, and heart disease, have been found to be more prominent in Hispanic populations when compared to EAs (Kirk et al., 2005; Pandey, Labarthe, Goff Jr., Chan, & Nichaman, 2001). These conditions are also associated with lower education and global cognition (Wright et al., 2006), as well as contribute to cognitive decline and dementia (Pasha, Haley, & Tanaka, 2014) for the Hispanic ethnicity. Additional research is needed to explore the influence of comorbid diseases on cognition and their possible interaction with depressive disorders and symptoms for a cross-cultural sample of Hispanic and European Americans.

Some limitations of the current study that should be addressed in future research include unequal sample sizes for the HA and EA groups and an over-representation of aMCI in the sample relative to CN. Additionally, this study did not analyze the distinction between EOD and LOD (Espinoza & Kaufman, 2014) and the responsiveness of the depressed patients to treatment. Moreover, the use of the conventional criteria for the diagnosis of aMCI, which requires a less stringent classification of the MCI diagnosis compared to the actuarial diagnostic method proposed by Jak and colleagues (2009) and Bondi and colleagues (2014), could have influenced our clinical groups. The Jak and Bondi criterion requires impairment of 1.5 SD below the mean on two or more measures within a cognitive domain and has been shown to be more sensitive than the conventional criterion for aMCI that we employed, requiring low performance on only one memory test (Bondi et al., 2014). Bondi and coworkers showed that those classified with MCI using the actuarial neuropsychological diagnostic method remained as MCI or progressed to dementia with increased biomarkers, and rarely reverted to a CN status. Therefore, for their sample, the traditional criteria were highly susceptible to false-positive diagnostic errors. This could raise the question that some of our participants under the MCI category could have been incorrectly identified as MCI due to less stringent neuropsychological diagnostic requirements. However, the traditional clinical criteria (i.e., evidence of concern about a change in cognition, lower performance in one or more cognitive domains, preservation of independence in functional abilities, and absence of dementia) used in the current study followed the diagnostic guidelines recommended by the National Institute on Aging-Alzheimer’s Association (Albert et al., 2011).

Despite these limitations, findings from the current study contribute to the literature on depression and aging by showing the significant predictive influence of depression on the clinical diagnosis of aMCI, independent of ethnicity. Also, a diverse interaction was demonstrated between cognitive performance and depression, demographic variables, and chronic comorbid disorders such as diabetes across ethnicity. The EA participants of this sample were more susceptible to the effects of depression on naming and memory test performance compared to HA participants, whose cognitive test performance was more influenced by demographic variables. Acculturation and number of depressive symptoms also did not predict aMCI diagnosis when age and years of education were taken into account.

Future research could analyze this cohort longitudinally to investigate if CN participants with higher GDS scores will ultimately progress to aMCI and eventually dementia during this study. Because we did not evaluate dementia here, we can continue to examine the development and interaction of depression and cognitive dysfunction during the progression from aMCI to dementia, for these data are from the first year of a 5-year longitudinal study. A prospective analysis of the current sample may help to elucidate the interaction between depression and cognitive impairment throughout the illness trajectory, as well as determine if there are cultural differences in this relationship over time. The differences in the manifestation of depressive symptoms between HA and EA participants diagnosed with aMCI should be further investigated, accounting for other relevant variables such as country of birth and years spent residing in the United States. In addition, depressive symptoms as a predictor for scores in other neuropsychological domains such as episodic memory, working memory, nonverbal memory, and visuospatial/constructional tasks should be analyzed to further clarify possible ethnic differences in the association of depression and cognitive function for both normal and abnormal aging.

Funding

Supported by the National Institute of Aging Grant numbers 5 P50 AG047726602 and 1 R01 AG047649-01A1.

Conflict of Interest

None declared.