Disparities in Metabolic Syndrome and Neurocognitive Function Among Older Hispanics/Latinos with Human Immunodeficiency Virus

Abstract

Neurocognitive impairment and metabolic syndrome (MetS) are prevalent in persons with HIV (PWH). We examined disparities in HIV-associated neurocognitive function between Hispanic and non-Hispanic White older PWH, and the role of MetS in explaining these disparities. Participants included 116 community-dwelling PWH aged 50–75 years enrolled in a cohort study in southern California [58 Hispanic (53% Spanish speaking) and 58 age-comparable non-Hispanic White; overall group: age: M = 57.9, standard deviation (SD) = 5.7; education (years): M = 13, SD = 3.4; 83% male, 58% AIDS, 94% on antiretroviral therapy]. Global neurocognition was derived from T-scores adjusted for demographics (age, education, sex, ethnicity, language) on a battery of 10 cognitive tests. MetS was ascertained via standard criteria that considered central obesity, and fasting elevated triglycerides, low high-density lipoprotein cholesterol and elevated glucose, or medical treatment for these conditions. Covariates examined included sociodemographic, psychiatric, substance use and HIV disease characteristics. Compared with non-Hispanic Whites, Hispanics showed worse global neurocognitive function (Cohen's d = 0.56, p < 0.05) and had higher rates of MetS (38% vs. 56%, p < 0.05). A stepwise regression model including ethnicity and significant covariates showed Hispanic ethnicity was the sole significant predictor of worse global neurocognition (B = −3.82, SE = 1.27, p < 0.01). A model also including MetS showed that both Hispanic ethnicity (B = −3.39, SE = 1.31, p = 0.01) and MetS (B = −2.73, SE = 1.31, p = 0.04) were independently associated with worse neurocognition. In conclusion, findings indicate that increased MetS is associated with worse neurocognitive function in both Hispanic and non-Hispanic White older PWH, but does not explain neurocognitive disparities. MetS remains an important target for intervention efforts to ameliorate neurocognitive dysfunction among diverse older PWH.

Introduction

With the advent of modern combination antiretroviral therapy (ART), there have been dramatic reductions in HIV-related morbidity and mortality. Yet, the adverse consequences of HIV and related comorbidities on the central nervous system remain prevalent, with ∼40% of persons living with HIV showing neurocognitive impairment (NCI).¹ Although these impairments are most typically mild to moderate, they are important predictors of everyday functioning (e.g., management of medication regimen and driving).^2–4 Their prevalence also tends to be more common in older persons with HIV (PWH), particularly in the presence of comorbidities that become more prevalent as PWH age.^5–7

Hispanics/Latinos/as/x, hereafter referred to as Hispanics, are the largest minority group in the United States, and the ethnic group with the highest population growth within the older adult population.⁸ They are also disproportionally affected by HIV, and this is particularly salient among older persons, with Hispanics older than 50 years having three times the rate of infection of older non-Hispanic Whites.⁹ Hispanics are also often diagnosed with HIV and receive care for HIV later in the course of the disease compared with non-Hispanic Whites.^10,11

Hispanics with HIV tend to present with increased risk for NCI,^7,12–14 and neurocognitive decline^15,16 compared with other ethnic/racial groups in the United States. Yet only a few studies have examined disparities in NCI among older Hispanic and non-Hispanic White PWH,⁷ with most focused primarily on Caribbean Hispanics and none including Spanish speakers. Considering the heterogeneity within the Hispanic population,¹⁷ investigating NCI across older Hispanic PWH from different countries or origin/descent and language background is important to most accurately identify predictors that might differ among subgroups. The real-world impact of NCI, along with the confluence of risk for NCI in HIV infection, older age, and among Hispanics, underscore the significance of identifying mechanisms underlying disparities in NCI in Hispanics aging with HIV.

Little is known about factors that might be driving disparities in NCI among older Hispanic PWH. A cross-sectional multisite study, examining English-speaking adult PWH, showed that although historical HIV disease burden (i.e., nadir CD4) partially mediated ethnic differences in NCI, much of the disparity remained unexplained, even after considering substance use and psychiatric comorbidities in addition to a host of HIV disease characteristics.¹² A study based on longitudinal data from a similar cohort found that comorbidities, including cardiometabolic factors, partially accounted for elevated rate of neurocognitive decline among Hispanics compared with non-Hispanic Whites, but did not fully explain them.

Together, these results indicate that historical HIV disease characteristics and comorbidities might explain some of the ethnic differences in HIV-associated NCI. Adopting a culturally informed approach that considers factors particularly relevant to Hispanic PWH might help further identify important mechanisms underlying disparities in neurocognitive outcomes among Hispanics aging with HIV. This is a crucial step for the development of culturally relevant interventions aimed at reducing neurocognitive disparities.

Among the many potential mechanisms for increased risk for NCI among Hispanics aging with HIV, this study focused on metabolic syndrome (MetS), which is a cluster of abnormalities that increase risk for cardiovascular and cerebrovascular disease.^18–21 MetS is prominent among Hispanics,^22–24 with prevalence of 36% and 34% among US Hispanic women and men, respectively.²² MetS and associated cardiovascular risk factors are also prevalent among Hispanics with HIV,^25,26 are more common in advanced age,²³ and have been linked to worse HIV disease outcomes.^27–29 However, differences in MetS between older Hispanics and non-Hispanic Whites with HIV have not been thoroughly examined. Furthermore, MetS has been associated with NCI³⁰ and neurobehavioral disturbances (e.g., apathy and executive dysfunction)³¹ in HIV, but whether MetS might help explain disparities in NCI between older Hispanic and non-Hispanic White PWH is unknown.

The overall goal of this study was to examine MetS and its link to neurocognitive function in a diverse cohort of older PWH. Our first aim was to examine ethnic differences in neurocognition among older PWH living in southern California. Based on prior research from the northeastern United States, which included primarily English-speaking Caribbean older Hispanic PWH,⁷ we hypothesized that older Hispanics with HIV in southern California (primarily of Mexican origin and both English and Spanish speakers) would show worse neurocognitive function than their non-Hispanic White counterparts. Aim 2 was to examine rates of MetS among older Hispanic and non-Hispanic White PWH, and we hypothesized that Hispanics would have higher rates of MetS than non-Hispanic Whites. Aim 3 was to investigate whether MetS might explain any observed ethnic differences in neurocognitive performance among older PWH.

Methods

Participants

In this cross-sectional study, participants were 116 community-dwelling PWH aged 50–75 years who were enrolled in the NIH-funded HIV in Older Latinos (HOLA) study (58 Hispanic/Latino and 58 age-comparable non-Hispanic Whites). Inclusion criteria for the HOLA study were as follows: (1) presence of HIV infection, (2) age 50 years and older, (3) self-identification as Hispanic/Latino or non-Hispanic White, and (4) English or Spanish speaking. Exclusion criteria consisted of (1) history of comorbid neurological illness or injury that would likely impact cognition (e.g., seizure disorder, closed head injury with loss of consciousness for longer than 30 min); (2) history of psychotic disorder, (3) positive urine toxicology screen for substances known to affect neurocognition, and (4) sensory or physical problems that would interfere with neurocognitive testing. Inclusion criteria for the present analyses included having data available on MetS and at least 7 of the 10 tests included in the neurocognitive battery.

Participants were recruited from existing cohort studies at the University of California San Diego HIV Neurobehavioral Research Program (HNRP), where the study was housed, and via outreach efforts in the local Hispanic community. Study visits were completed between June 2015 and December 2018.

Materials and Procedures

Consistent with US Census Bureau methodology,³² ethnicity and race were determined via two separate questions. Ethnicity was ascertained by the question “Are you of Spanish/Hispanic/Latino origin?” with the following response options: “No,” “Yes, Mexican, Mexican American, Chicano,” “Yes, Puerto Rican”, “Yes, Cuban,” “Yes, Other Spanish/Hispanic/Latino.” Race was ascertained by asking participants to select all races that applied to them with response options including terms for American Indian or Alaskan Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, White, and other race.

Participants were tested in their primary language (English or Spanish) by trained bilingual study staff. Primary language was determined via the combination of a validated algorithm that uses information based on self-report about the participant's preferred language, fluency in English and Spanish, and language use in several specific contexts (e.g., conversing in and outside the home, listening to the radio or TV, reading),³³ together with performance-based verbal fluency assessments in English and Spanish.^34,35

The study was approved by the UCSD Institutional Review Board; all participants provided written informed consent, and research was completed in accordance with the Declaration of Helsinki.

Sociodemographic characteristics

Age, sex, education, and socioeconomic status (SES) were ascertained by self-report. SES was assessed via the 2018 US Census Poverty Thresholds.³⁶ Participants were categorized as being below the poverty threshold based on a combination of the number of individuals who were living in the same household and the level of family/household income in the last year (i.e., living alone and having income <$10,000; two to six individuals in the household and income <$20,000; seven individuals and income <$35,000).

Neurocognition was assessed via a comprehensive battery comprising 10 cognitive tests with available demographically adjusted normative data for English- and Spanish-speaking Hispanics, which assessed various cognitive domains including executive function/attention, verbal fluency, learning and memory, processing speed and visuospatial skills^33,37–43 (Table 1). Our primary outcome was a global cognition composite computed by averaging the Fluid Composite of the NIH Toolbox Cognition Battery (NIHTB-CB)³⁷ T-scores, and individual cognitive test T-scores for other tests in the battery. The resulting global cognition T-scores adjusted for age, sex, years of education, race/ethnicity, and language of tests (English and Spanish). In secondary analyses we examined performance by cognitive domain, based on average T-scores of tests in each domain, except for visuospatial skills, which was measured via T-scores on a single test (as listed in Table 1).

Table 1.

Neurocognitive Test Battery

Domain	Tests
Executive function	NIH-TB CB Dimensional Change Card Sort
Attention	NIH-TB CB Flanker Inhibitory Control and Attention
	NIH-TB CB List Sorting Working Memory
Verbal fluency	SENAS Category Fluency
	Letter Fluency (PMR or FAS)
Learning and memory	NIH-TB CB Picture Sequence Memory
	SENAS Word List Learning- I
Processing speed	NIH-TB CB Pattern Comparison
	WAIS-III Digit Symbol
Visuospatial skills	SENAS Spatial Localization

NIH-TB CB, National Institutes of Health Toolbox Cognition Battery; SENAS, Spanish and English Neuropsychological Assessment Scales; WAIS-III, Wechsler Adult Intelligence Scale, Third Edition.

HIV disease characteristics and MetS

Participants underwent a fasting blood draw, neuromedical interview, and physical examination to determine HIV disease characteristics and MetS. HIV infection was ascertained by enzyme-linked immunosorbent assay with Western blot confirmation. Routine clinical chemistry panels, complete blood counts, rapid plasma reagin, hepatitis C antibody, and CD4⁺ T cells (flow cytometry) were performed at a Clinical Laboratory Improvement Amendments (CLIA)-certified, or CLIA equivalent, laboratory. HIV RNA levels in plasma were measured by reverse transcriptase–polymerase chain reaction (Roche Amplicor, v. 1.5; lower limit of quantitation, 50 copies per milliliter). Nadir CD4, estimated duration of HIV disease, and current and past use of ART were ascertained by self-report and review of medical records when available. We computed the Veterans Aging Cohort Study (VACS) Index as in previous studies.^44,45 This index combines age, traditional HIV biomarkers (HIV-1 plasma RNA and current CD4 count), and non-HIV biomarkers (indicators of renal and liver function, anemia, and hepatitis C coinfection). It is predictive of mortality,^44–47 and NCI and neurocognitive decline in HIV, although it has been less predictive of NCI among young to middle-aged Hispanic PWH.^48–50 MetS was defined via established criteria by the presence of three or more of the following:²³ (1) waist circumference >102 cm in men and >88 cm in women; (2) triglycerides ≥150 mg/dL or treatment for lipid abnormality; (3) high-density lipoprotein (HDL) cholesterol <40 mg/dL in men and <50 mg/dL in women or treatment for lipid abnormality; (4) blood pressure ≥130/85 mm Hg or use of antihypertension medications; and (5) fasting glucose ≥110 mg/dL or diagnosis of diabetes or use of antidiabetic medications. Categorical diagnosis of MetS was the main measure and we investigated individual components of the syndrome in secondary analyses.

Psychiatric and substance use disorders

Diagnoses of lifetime (past and current) major depressive disorder, bipolar disorder, and 14 substance use disorders (including marijuana) were rendered using the Composite International Diagnostic Interview-2.1 (CIDI).⁵¹ This is a computer-assisted interview, which provides a cross-cultural assessment of psychiatric disorders using Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) criteria.⁵² The Beck Depression Inventory-II was used to assess current depressive symptoms.⁵³ Perceived stress, psychological well-being, and social satisfaction were assessed via the NIH Toolbox Emotion Battery (NIHTB-EB).^54,55 Perceived stress was assessed via a single scale and summary T-scores were computed for psychological well-being (including scales on meaning and purpose, life satisfaction, and positive affect) and social satisfaction (including scales on friendship, emotional support, instrumental support, and reverse-coded loneliness and perceived rejection) based on published methods.⁵⁴

Culturally relevant factors

Hispanic participants also completed measures assessing culturally relevant factors, including a self-report questionnaire on sociodemographic characteristics (e.g., place of birth, time in the United States, years of education in the United States and other countries) that was developed for this study based on earlier work.^33,34,56 Acculturation was measured via the Psychological Acculturation Scale, a 10-item measure that evaluates an individual's emotional connection to and knowledge of the US non-Hispanic White and Latino cultures.⁵⁷ Item scores range from 1 to 9 and higher scores suggest greater psychological acculturation into the US non-Hispanic White culture. Language use was assessed by self-report³³ and performance-based measure.^34,35 Participants were classified as monolingual Spanish- or English-speaking if they reported being able to speak only one of these languages.³³

Participants who reported speaking both languages to some extent completed tests of verbal fluency in English and Spanish (letter/phonemic fluency for words starting with P,M,R in Spanish, and F,A,S in English).³⁵ Consistent with prior procedures,³⁵ we estimated relative degree of Spanish–English bilingualism as the ratio of the verbal fluency score in Spanish to the sum of the average fluency scores in Spanish and English [PMR/(PMR + FAS)]. Based on this score participants were classified as Spanish dominant (≥0.67), English dominant (≤0.33), or balanced bilingual (0.34–0.66).³⁵

Statistical analyses

We examined differences between Hispanic and non-Hispanic White PWH on sociodemographic factors, HIV disease characteristics, and psychiatric comorbidities, via a series of independent sample t-tests, and chi-square tests (or Fisher's exact). To examine ethnic differences in neurocognitive function (Aim 1), we ran a series of independent sample t-tests on global neurocognitive T-scores (primary outcome) and domain neurocognitive T-scores (secondary outcome) and computed Cohen's d effect sizes for these comparisons. To address Aim 2, we examined differences in rates of MetS and its components by ethnic group via a series of logistic regression models. Given differences in the sex composition of ethnic groups in the current sample and prior findings showing notable sex differences in MetS,⁵⁸ in follow-up analyses, we ran comparable logistic regression models also including sex as a covariate.

To address Aim 3, we first ran univariable linear regression analyses with ethnicity as a predictor for neurocognitive outcomes that differed by ethnic group based on findings from Aim 1. We then ran multivariable regression models including MetS, ethnicity, and significant covariates. Covariates to be included in these models were selected by identifying sociodemographic, HIV disease, and psychiatric comorbidities that differed between ethnic groups at p < 0.10, and then entering these variables into backward stepwise regression models (minimum AIC criteria, no stopping rule). MetS and ethnicity were forced into the final models if not automatically selected via these procedures. Finally, we ran pathway analyses using bias-corrected and accelerated (BCa) method to test for mediation effect of MetS on association between Hispanic ethnicity and cognitive outcomes (“mediation” package in R).

Results

Characterization of the study cohort

Table 2 reports and compares participants' sociodemographic, HIV disease, and psychiatric characteristics by ethnic group. Hispanics were more likely to be women and have household incomes below the poverty line, and had significantly fewer years of education than non-Hispanic Whites, although there was substantial overlap in the range of years of education across groups (Hispanics: 0–20 years; non-Hispanic Whites: 6–20 years). Hispanics reported significantly shorter estimated duration of HIV disease and ART use despite similar ages of the groups, with no significant group differences on other HIV disease characteristics. There were no significant differences on rates of lifetime or current major depressive disorder, bipolar disorder, or any substance use disorder, but a smaller proportion of Hispanics reported a lifetime history of cannabis use disorder. There were also no statistically significant differences on mood symptoms, perceived stress or NIHTB Emotion Battery Summary T-scores (small ethnicity effect sizes: psychological well-being Cohen's d = 0.28; social satisfaction Cohen's d = 0.30).

Table 2.

Cohort Characteristics by Ethnic Group

	Hispanic (n = 58)	Non-Hispanic White (n = 58)	p a
Sociodemographic characteristics
Age, M (SD)	57.33 (5.48)	58.62 (5.94)	0.23
Sex (% male)	74.14%	91.38%	0.01
Education (years), M (SD)	12.24 (4.01)	14.31 (2.14)	<0.001
SES (% below poverty line)	36.84%	17.86%	0.02
HIV disease characteristics
% AIDS	58.62%	56.90%	0.85
Nadir CD4, median (IQR)	187 (40, 342)	171 (76, 300)	0.54
Duration of infection (years), median (IQR)	17.81 (11.97, 25.19)	26.49 (19.02, 30.05)	<0.001
Current CD4, median (IQR)	630 (486, 807)	666 (433, 758)	0.90
Months exposure ART, median (IQR)	156.62 (93.40, 210.99)	198.62 (116.42, 254.13)	0.02
VACS Index, M (SD)	26.57 (13.91)	26.79 (13.03)	0.93
% On ART	93.10%	94.83%	0.70
% Detectable plasma RNAb	3.77%	3.64%	0.97
Psychiatric comorbidities, %
Current Major Depressive Disorder	7.02%	13.79%	0.23
Current Bipolar Disorder	0.00%	0.00%	1.0
Current Any Substance Use Disorder	5.26%	1.72%	0.30
Lifetime Major Depressive Disorder	57.89%	70.69%	0.15
Lifetime Bipolar Disorder	8.77%	3.45%	0.27
Lifetime Any Substance Use Disorder	63.16%	77.59%	0.09
Lifetime Alcohol Use Disorder	47.37%	58.62%	0.23
Lifetime Cannabis Use Disorder	15.79%	36.21%	0.01
Lifetime Other Substance Use Disorder	38.60%	44.83%	0.50
Beck Depression Inventory, M (SD)	10.20 (9.37)	10.57 (9.67)	0.83
NIHTB-EB Perceived Stress T-scores, M (SD)	49.60 (10.91)	51.14 (11.03)	0.47
NIHTB-EB Psychological Well-Being Summary T-scores, M (SD)	49.93 (9.93)	47.07 (10.70)	0.17
NIHTB-EB Social Satisfaction Summary T-scores, M (SD)	46.18 (11.14)	42.61 (12.50)	0.14
Culturally relevant factors
Hispanic/Latino background, %	—	—	—
Mexican	86.21%	—	—
Cuban	1.72%	—	—
Central America	5.17%
More than one Latino background	6.89%
Place of Birth (United States), %	36.84%	—	—
Years living in the United States,c M (SD) [range]	30.06 (14.73) [1–57]	—	—
Years of education in the United States,c M (SD) [range]	9.33 (4.89) [0–16]	—	—
Acculturation, M (SD) [range]	4.41 (1.87) [1–8.6]	—	—
Language use, %	—	—	—
Monolingual Spanish speaking	31.03%	—	—
Spanish dominantd	6.89%	—	—
Balanced bilinguald	37.93%	—	—
English dominantd	8.62%	—	—
Monolingual English speaking	15.52%	—	—
Language of test, %			—
Spanish	53.45%	0.00%	—
English	46.55%	100.00%	—

^a Results of independent samples t-tests, Wilcoxon signed-rank tests or chi-square tests.

^b Among those on ART.

^c Among those born outside the United States.

^d Based on performance-based verbal fluency in English and Spanish, available for participants who reported some degree of bilingualism.

ART, antiretroviral therapy; IQR, interquartile range; NIHTB-EB, National Institutes of Health Toolbox Emotion Battery; RNA, ribonucleic acid; SD, standard deviation; SES, socioeconomic status; VACS, Veterans Aging Cohort Study, higher VACS Index scores indicate worse health.

As given in Table 2, most of the Hispanic sample was of Mexican origin/descent and born outside the United States. Degree of acculturation and Spanish/English language use varied widely within the Hispanic sample. Approximately half of the sample reported some degree of bilingualism, with a third of the sample speaking Spanish only and the remaining English only. A little over half of the Hispanic sample was tested in Spanish.

Differences in neurocognitive performance by ethnic group

Figure 1 provides global and domain neurocognitive T-scores by ethnic group. Independent sample t-tests showed Hispanics had significantly lower global neurocognition (p < 0.01, Cohen's d = 0.56), learning and memory (p < 0.01, Cohen's d = 0.51), and visual-spatial skills (p = 0.046, Cohen's d = 0.38) T-scores than non-Hispanic Whites, with no significant group differences on other domains (ps = 0.13–0.15, Cohen's d = 0.27–0.28).

FIG. 1.

Mean neurocognition T-scores (standard errors) among older Hispanic and non-Hispanic White PWH. *p < 0.05. **p < 0.01 (p values are based on independent sample t-tests by ethnic group). PWH, persons with HIV.

In follow-up analyses, we considered whether potentially different exposures to the neurocognitive tests used in this study might explain the observed ethnic differences in neurocognitive test performance. This was the first known exposure to the SENAS battery for all participants. An independent sample t-test comparing the number of prior NIHTB-CB administrations between Hispanic (M = 0.85, SD = 1.39) and non-Hispanic White PWH (M = 1.88, SD = 2.03) was significant (p < 01). Thus, we ran separate linear regression models on global cognition and learning and memory, including number of prior NIHTB-CB administrations as a covariate. Results from these models showed that there continued to be significant ethnic group differences for both global cognition (Estimate = −3.62, SE = 1.33, p < 0.01) and learning and memory (Estimate = −3.39, SE = 1.38, p < 0.02), and that number of prior NIHTB-CB administrations was not associated with neurocognition in either model (ps > 0.59). We did not run similar analyses on visual-spatial skills given that this domain did not include tests on the NIHTB-CB.

MetS and its components by ethnic group

Figure 2 provides the proportions of older PWH with MetS and its components by ethnic group. A series of logistic regression models showed that, compared with non-Hispanic Whites, Hispanics had significantly higher rates of MetS [odds ratio (OR) = 2.13, 95% confidence interval (CI) = 1.01–4.53, p = 0.049], and marginally higher rates of central obesity (OR = 1.92, 95% CI = 0.89–4.16, p = 0.09) and elevated triglycerides [OR = 1.87, 95% confidence interval (CI) = 0.90–3.92, p = 0.09], but with notable ORs, and no significant differences on other MetS components. ORs for ethnic differences in MetS among older PWH were generally comparable when sex was included as a covariate (OR = 1.98, 95% CI = 0.92–4.28, p = 0.08), and there was no significant effect of sex (p = 0.38). Among the MetS components, a model including sex as a covariate showed that Ors for ethnic differences in central obesity were reduced (OR = 1.40, 95% CI = 0.60–3.25, p = 0.44). There was also a significant effect of sex, indicating older women with HIV were more likely to show central obesity than men with HIV (OR = 8.96, 95% CI = 2.69–29.77, p < 0.001) in the overall sample. In contrast, ORs for ethnic differences in elevated triglycerides were increased (OR = 2.49, 95% CI = 1.12–5.51, p = 0.02) when adjusting for sex, with women showing significantly lower risk for elevated triglycerides than men (OR = 0.22, 95% CI = 0.07–0.70, p = 0.01).

FIG. 2.

Proportion of Hispanic and non-Hispanic White older PWH with metabolic syndrome and its components. *p < 0.05, ^p < 0.10 (p values are based on independent sample t-tests by ethnic group).

MetS and its role in the association of Hispanic ethnicity with neurocognition

To examine whether MetS might mediate the association between Hispanic ethnicity and neurocognition, we first ran linear regression models on neurocognitive T-scores that differed by ethnic group [i.e., global cognition (Table 3 Model 1a), learning and memory (Table 3 Model 2a), and visual-spatial skills (Table 3 Model 3a)] including only ethnicity as a predictor. We then ran separate backward stepwise linear regression models on global cognition, learning and memory, and visual-spatial skills, with predictors being ethnicity, MetS, and covariates that differed between groups at p < 0.10. The final model on global cognition did not select any covariates and showed that both Hispanic ethnicity (p = 0.01) and MetS (p = 0.04) were each significantly associated with worse global neurocognitive T-scores (Table 3, Model 1b).

Table 3.

Association of Metabolic Syndrome with Global and Domain Neurocognitive T-scores

	Global cognition		Learning and memory		Visual-spatial skills
	Model 1a	Model 1b	Model 2a	Model 2b	Model 3a	Model 3b
	B (SE)	B (SE)	B (SE)	B (SE)	B (SE)	B (SE)
Ethnicity (Hispanic)	−3.82 (1.27)**	−3.39 (1.31)*	−3.61 (1.31)**	−3.84 (1.40)**	−3.76 (1.32)*	−1.79 (1.87)
MetS	—	−2.73 (1.31)*	—	−0.35 (1.35)	—	−6.93 (1.85)***
Sex (female)	—	—	—	−3.48 (1.77)^	—	—
ART duration	—	—	—	−0.01 (0.00)	—	0.02 (0.01)^

^p < 0.06.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

Models a included ethnicity as the sole predictor. Models b included the following predictors: ethnicity, MetS and significant covariates (i.e., covariates that differed by ethnic group at p < 0.10 [sex, education, estimated duration of infection, months exposure ART, Lifetime Substance Use Disorder, and number of prior administrations of the NIHTB-CB] and were selected via backward stepwise regressions to be included in final models).

ART, antiretroviral therapy; MetS, metabolic syndrome; SE, standard error; reference group for ethnicity is non-Hispanic White; reference group for sex is male.

Furthermore, estimates of the association between Hispanic ethnicity and global neurocognition were comparable with those of the univariable linear regression model (Table 3, Model 1a). Similar analyses on learning and memory T-scores showed that Hispanic ethnicity continued to be significantly associated with learning and memory after considering MetS and significant covariates (Table 3 Model 2b), with comparable estimates for this association compared with the univariable model (Table 3, Model 2a). MetS was not significantly associated with learning and memory T-scores in this adjusted model. The final model on visual-spatial skills showed that Hispanic ethnicity was not significantly associated with visual-spatial skills (p = 0.34) when MetS and covariates were included (Table 3, Model 3b), with estimates for this association being notably reduced from models without MetS and covariates (Table 3, Model 3a). In this final model (Model 3b), MetS was significantly associated with worse visual-spatial skills (p < 0.01).

Pathway analyses on global neurocognition, learning and memory, and visual-spatial skills (including covariates selected via stepwise regressions on models presented previously) showed no significant mediation effect of MetS on the association between Hispanic ethnicity and neurocognition (Table 4).

Table 4.

Models Examining Whether Metabolic Syndrome Accounts for the Association Between Ethnicity and Neurocognitive T-Scores Among Older Persons Living with HIV

	Global neurocognitiona		Learning and memoryb		Visual-spatial skillsc
	Estimate [95% CI]	p	Estimate [95% CI]	p	Estimate [95% CI]	p
ACME	0.02 [−1.75 to 0.05]	0.118	−0.05 [−0.94 to 0.29]	0.812	−1.01 [−3.44 to 0.08]	0.094
Hispanic ethnicity	−3.39 [−5.90 to −0.66]	0.016	−3.84 [−6.46 to −1.19]	0.002	−1.79 [−5.46 to 2.10]	0.33
Total effect	−3.36 [−6.53 to −1.32]	0.002	−3.89 [−6.63 to −1.20]	<0.001	−2.80 [−7.06 to 1.05]	0.122

Models examining the association between Hispanic ethnicity and neurocognitive T-scores, including estimates of the indirect effects testing whether MetS might account for this association among older persons living with HIV.

^a Model on global neurocognitive T-scores with no covariates included.

^b Model on Learning and Memory T-scores and adjusted for sex and estimated duration of antiretroviral therapy.

^c Model on visual-spatial skills T-scores and adjusted for estimated duration of antiretroviral therapy.

ACME, average causal mediation effect; CI, confidence interval.

Given the lack of a mediation effect, in follow-up analyses we investigated whether the association between MetS and global neurocognitive function was comparable by ethnic group. A multivariable linear regression model on global neurocognitive T-scores with terms for MetS, ethnicity, and their interaction, showed no significant MetS and ethnicity interaction (Estimate = 2.68, SE = 2.62, p = 0.31).

Discussion

Consistent with our first and second hypotheses, older Hispanic PWH showed worse neurocognition and higher rates of MetS than an age-comparable group of non-Hispanic White PWH. Inconsistent with our third hypothesis, MetS did not explain neurocognitive disparities between Hispanic and non-Hispanic White older PWH. Yet, MetS was an independent significant predictor of neurocognition in addition to Hispanic ethnicity, and the presence of MetS was comparably associated with lower neurocognitive function in both Hispanic and non-Hispanic White older PWH.

Our findings showing that older Hispanic PWH with varying degrees of acculturation and language use had worse neurocognition than their non-Hispanic White counterparts, extend prior research^7,16 and indicate that ethnic disparities in HIV-associated NCI might be present across diverse subgroups of older Hispanics in the United States. The limited existing research on neurocognitive disparities among older Hispanic PWH showed that English-speaking Hispanics living in the northeastern United States and primarily of Caribbean descent had increased rates of NCI compared with non-Hispanic White PWH and younger cohorts of PWH.⁷ Present findings extend these prior results to show that older Hispanic PWH living in southern California and primarily of Mexican background also show worse neurocognitive performance than their non-Hispanic White counterparts. Furthermore, neuropsychological tests included in the earlier study of older Hispanic PWH in the northeastern United States¹³ did not have normative data available.

In this study we utilized neuropsychological normative data that accounted for the effect of demographic factors (age, years of education, sex, Hispanic ethnicity, and language) on “normal” variations in test performance. This strengthens the notion that ethnic differences observed in neurocognitive performance among older PWH are likely related to factors associated with aging with HIV, rather than normal variations on test performance associated with demographics.

As hypothesized, rates of MetS were higher among older Hispanic PWH (56%) compared with older non-Hispanic White PWH (38%). They were also higher than those reported in the general US population,^22,24 and a sample of Hispanic PWH in Puerto Rico (35%),²⁵ which was on average 13 years younger than participants in this study. These findings and the negative association between MetS and cognitive function underscore the need for prevention and treatment efforts to ameliorate MetS, especially in older Hispanic PWH. Low HDL cholesterol and high blood pressure were the two most prevalent components of MetS in both Hispanic and non-Hispanic White older PWH, which is consistent with prior studies in PWH.^25,31 Central obesity and elevated triglycerides showed marginally significant ethnic disparities in unadjusted models.

However, these ethnic differences were reduced for central obesity and increased for elevated triglycerides when sex was included in the models that considered the overall sample, with women showing increased central obesity and decreased elevated triglycerides compared with older men with HIV. These findings ought to be interpreted with caution given the relatively small number of women in this study. Future larger studies that consider sex differences along with social determinants of MetS in HIV⁵⁹ might shed further light onto mechanisms of disparities in MetS among older PWH.

Contrary to our hypothesis, we found that MetS did not explain ethnic disparities on neurocognitive performance among older PWH. Instead, MetS was a significant independent predictor of global neurocognition along with ethnicity. These results extend prior findings in HIV³⁰ and indicate that interventions targeting MetS may have added benefits to neurocognition among both Hispanics and non-Hispanic White older PWH. They also suggest that other factors associated with Hispanic ethnicity might be important for the development of HIV-associated NCI. Future work examining social determinants of neurocognitive disorders in HIV^4,60 and aging,⁶¹ along with HIV-related genetic factors associated with NCI in Hispanics⁶² and biomarkers of Alzheimer's disease might advance understanding of ethnic neurocognitive disparities in older PWH.

The cross-sectional design of this study limits our ability to ascribe directionality to our findings. Relatedly, a cross-sectional design is inherently limited in its ability to separate life-long group differences in cognitive test performance from acquired deficits. The utilization of normative data helps with identifying acquired cognitive impairment—defined as performance that is below expected levels. The normative data utilized in this study are from two different samples, (i.e., the NIHTB-CB was normed in a nationally representative sample in the United States, and the SENAS battery in a sample of older adults in California), which interjects potential sources of error and bias.

Yet, the fact that both samples included Hispanics (English and Spanish speaking) and non-Hispanic Whites, and that both sets of normative data considered the influence of ethnicity and language along with other demographic characteristics on cognitive test performance is a strength of our study. Our Hispanic sample was quite diverse in terms of their language use and degree of acculturation, and largely comprised persons of Mexican origin living in the US–Mexico borderland region. Although this region of the country houses approximately half of all Hispanics in the United States,⁶³ caution is warranted in generalizing present findings to Hispanics from other heritages or geographical regions.

Our assessment of MetS included a widely used composite based on established criteria. Yet, there are other markers of CVD risk burden, such as the Framingham Cardiovascular Risk Score and Global Vascular Risk Score,^64,65 that have been linked to neurocognition in Hispanics. Their consideration in future studies might advance understanding of the role of cardiovascular risk in explaining ethnic disparities in HIV-associated NCI, particularly given prior findings in English-speaking Hispanics.¹⁶

Future studies including larger diverse groups of Hispanic PWH would be important to determine whether present findings generalize to varied Hispanic subgroups, and if there are culturally relevant (e.g., Hispanic background, acculturation) and other demographic factors (e.g., sex) that moderate the associations observed in present analyses. Furthermore, the inclusion of Hispanics with and without HIV would help advance understanding of whether mechanisms underlying the association of MetS with neurocognitive function among Hispanics differ by HIV status. Longitudinal observational research, and intervention studies targeting MetS that examine changes in neurocognition over time, would help determine the predictive value of the observed associations.

In conclusion, the estimated large numbers of PWH reaching older age and the increased risk for HIV-associated NCI in older Hispanics underscores the importance of research into mechanisms driving NCI in this group. This study showed that MetS was associated with worse neurocognition in both Hispanic and non-Hispanic White older PWH. This suggests that MetS is a potentially important target to consider in the development of prevention and treatment efforts to ameliorate neurocognitive dysfunction among older PWH of ethnically diverse backgrounds.

Author Disclosure Statement

R.C.M. is a cofounder of KeyWise AI, Inc., and NeuroUX, Inc., No other authors have conflicts of interests to declare.

Funding Information

This work was funded by the following grants from the National Institutes of Health (M.J.M.: Grant Nos. K23MH105297, K24AG075240), (Z.Z.: Grant No. R01AG066657), (L.K.: Grant Nos. T32DA031098, T32AA013525), (R.K.: Grant No. P30MH062512).