Cognitive Function as a Predictor of Incident Hypertension in Hispanic/Latino Adults

Abstract

Background:

Studies showed that hypertension predicts cognitive performance. It remains uncertain whether a bidirectional relationship exists. We investigated whether cognitive function predicts incident hypertension.

Methods:

Data from Hispanic Community Health Study/Study of Latinos were analyzed. Global cognitive score (GC) of non-hypertensive participants at baseline was derived by averaging z scores across four neurocognitive tests (Brief Spanish-English Verbal Learning Test Sum and Recall; Word Fluency; and Digit Symbol Substitution Test). Incident hypertension was investigated, on average, 6 years later. Sociodemographic characteristics and unhealthy behaviors (obesity, physical activity not at goal, low diet quality, smoking) were ascertained at baseline. Association of GC at baseline with incident hypertension (defined as BP ≥130/80 mmHg or on treatment) was investigated using logistic regression analyses adjusted for sociodemographic characteristics and time between visits. Multimodal analyses assessed whether unhealthy behaviors were pathway variables between GC and incident hypertension.

Results:

A total of 6,755 participants (mean [IQR] age, 53[48–59] years; 62% female) were included. At follow-up, 57% of individuals developed hypertension. In our final model, higher GC was associated with lower odds of incident hypertension (OR=0.85; 95% CI 0.74– 0.98). In pathway analysis, indirect effects through unhealthy behaviors were small. Obesity was the only significant mediator (indirect OR=0.95, 95% CI 0.90–0.99). Smoking and physical activity not at goal were in the pathway of GC and obesity, but did not mediate mediate GC-incident hypertension.

Conclusions:

Higher GC was associated with lower incidence of hypertension six years later. Unhealthy behaviors may influence the association of cognitive function and incident hypertension.

Graphical Abstract

Introduction

A wealth of evidence demonstrates that midlife hypertension is a predictor of late life cognitive dysfunction and dementia¹. Chronic hypertension leads to territorial and microvascular brain injury, white matter disease, microstructural abnormalities, and cerebral atrophy. These processes impair synaptic transmission, compromise neuronal survival, and ultimately lead to cognitive impairment². In addition, there is compelling evidence indicating that the trajectories of cardiac and brain health are intricately intertwined³. From a hemodynamic standpoint, neuroanatomical areas such as the rostral ventrolateral medulla and the paraventricular nucleus regulate blood pressure in normal and pathologic conditions⁴. Thus, from a physio-pathological perspective, dysfunction of neural pathways that regulate autonomic responses could, in theory, lead to hypertension. Interestingly, a bidirectional association has been observed between cognitive impairment and precursors of hypertension such as obesity, smoking, and unhealthy diet^5–8. This suggests that features commonly observed in individuals with cognitive impairment, including lack of insight, cognitive flexibility, and self-control could result in the adoption of unhealthy behaviors and lead to hypertension. However, the role of cognitive performance in the development of hypertension has not been investigated in prospective cohorts.

The Hispanic/Latino community, the second largest ethnoracial group in the US, faces a high burden of hypertension as well as Alzheimer’s disease and related dementias, and lower rates of hypertension control^9,10. In this context, examining how cognitive performance, cardiovascular risk trajectories and social and socioeconomic factors interact across the lifespan of Hispanic/Latino adults is crucial to understanding the bidirectional links between brain and vascular health in this population.

In this study we investigated the association between cognitive scores and incident hypertension in middle and older aged Hispanic/Latino adults who participated in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Our primary hypothesis was that higher global cognition scores at baseline were associated with lower risk of incident hypertension at follow up, independent of demographic and socioeconomic variables. In addition, in an exploratory analysis, we studied if unhealthy behaviors were pathway variables in this association.

Subject/Materials and Methods

Data Availability Statement

Because of the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols may be sent to the HCHS/SOL study, available at https://sites9.cscc.unc.edu/hchs/new-investigator.

Study Sample

The HCHS/SOL study (n=16,415) is an ongoing prospective cohort of Hispanic/Latino adults from four US metropolitan areas (Bronx, NY; Chicago, IL; Miami, FL; and San Diego, CA)¹¹. Institutional review boards at each participating site approved the study protocol and participants provided informed consent. Research complied with the 1975 Helsinki Declaration. Briefly, HCHS/SOL participants aged 18–74 years underwent a comprehensive examination at Visit 1 (2008–2011) which included anthropometric and blood pressure measurements; neurocognitive assessment (among participants ages ≥45 years); and the collection of medical history, lifestyle factors, medication usage, and sociocultural aspects. Participants underwent a follow-up examination at Visit 2 (2014–2017) when blood pressure measurements were repeated. This study included non-hypertensive participants who underwent neurocognitive assessments at Visit 1.

Outcome

The definition of hypertension at Visit 1 and 2 was based on the American College of Cardiology and American Heart Association recommendations¹². Hypertension was defined as a blood pressure of ≥130 mm Hg systolic or ≥80 mm diastolic, or on treatment with antihypertensive medications. Three seated blood pressure measurements were obtained one minute apart following a 5-minute rest period using an automatic sphygmomanometer. The average of these three blood pressure values was used in this analysis.

Exposure

The main exposure, global cognition(GC) at Visit 1, was ascertained using a battery of neurocognitive tests which consistent of (1) Brief Spanish-English Verbal Learning Test (B-SEVLT) Sum for verbal episodic learning; (2) B-SEVLT Recall for delayed verbal episodic memory; (3) word fluency (WF) for phonemic verbal fluency and executive function; and (4) Digit Symbol Substitution Test (DSST) for sustained attention and psychomotor speed¹³. These tests were administered in-person by trained bilingual psychometrists in the participants’ preferred language. The scores obtained in each test were z-transformed and averaged to create a GC score. These HCHS/SOL cognitive tests, scoring procedures, and application to the HCHS/SOL data have been previously reported¹⁴.

Covariates and pathway variables

All the variables were ascertained at Visit 1. Age (years), sex, field center, household income (<$10,000; $10,001-$20,000; $20,001-$40,000; $40,001-$75,000; and >$75,000), nativity (born in US 50 states/DC versus born outside states/DC including US territories, hereafter “foreign born”), heritage group (Dominican, Central American, Cuban, Mexican, Puerto-Rican, South American, more than one heritage/other), education (less than high school, high school or equivalent and more than high school), and health insurance status (yes vs. no) were self-reported. Unhealthy behaviors were obesity, physical activity (PA) not at goal, current smoker, and low diet quality. Weight and height were measured with participants wearing light clothing, and body mass index (BMI) was calculated by dividing weight (kg) by height (m²). Obesity was defined as a BMI ≥30 kg/m². PA was assessed using a modified version of the World Health Organization Global PA Questionnaire (GPAQ)¹⁵. Based on the goals recommended by the Office of Disease Prevention and Health Promotion, PA not at goal was defined as ≤75 minutes/week of total vigorous physical, ≤150 minutes/week of moderate PA, or ≤150 minutes/week of combined vigorous and moderate PA¹⁶. Participants who had smoked at least one hundred cigarettes in their lifetime and reported smoking daily or some days at the time of Visit 1 were classified as current smokers. Dietary quality was ascertained by determining adherence to the Mediterranean diet¹⁷. Participants completed two 24-hour dietary recalls of predefined food and nutritional categories, 30 days apart. Adherence to a Mediterranean diet was determined from the average of both recall where possible (or a single recall for those who complete only one; n=805 (12%)), using the Mediterranean Diet Score (MeDiS), which has been described previously¹⁷. MeDiS ranges from 0–9, with higher scores indicating better adherence. In some Hispanic or Latino populations, the intake of refined grains and potatoes, which may not be fully beneficial, may be high¹⁸. A modified MeDi, which replaces the whole grains group with total-grains and excludes potato products from the vegetable group, has been proposed for this population¹⁹. In this study, MeDi refers to the modified MeDi. Participants with a MeDiS in the lowest quartile were considered to have low diet quality; quartiles were selected to ensure adequate distribution and comparability across heritage groups.

Statistical analysis

Among individuals with data on neurocognitive tests (n=9,223), we excluded 2,468 with prevalent hypertension at Visit 1 or with missing data on hypertension at visit 1. Our final analytic sample consisted of 6,755 cases with complete data. Descriptive characteristics were presented as medians and interquartile ranges (IQR) for continuous variables, and numbers and unweighted percentages for categorical variables. All cognitive outcomes were z-scored to facilitate comparison of the estimated associations across tests. Comparisons between groups were conducted using the Wilcoxon rank-sum test for continuous variables and the Chi-square test for categorical variables.

For our primary outcome, we used regression analysis to investigate the association of GC at Visit 1 with incident hypertension at Visit 2. In secondary analysis, we used the score of each neurocognitive test as the predictor. Results are expressed as Odds Ratios (OR) with corresponding 95% CI and p values. In addition to OR, we estimated absolute risk differences (ARD) and population attributable fraction (PAF) with 95% confidence intervals using a model-based counterfactual framework that incorporated complex survey design weights, where GC was increased by one standard deviation in the counterfactual scenario.

We constructed three incremental models. Model 1 was adjusted for age at Visit 1 and time elapsed between visits. Model 2 was adjusted for age, sex, center, heritage group, household income, years in the US, nativity, health insurance status, and time elapsed between visits. Since childhood education and higher educational attainment predict hypertension²⁰ and cognitive performance later in life²¹, model 3 was further adjusted for education. In addition, we performed regression analyses between baseline GC and incident hypertension adjusting for unhealthy behaviors. Collinearity between cognitive outcomes and unhealthy behaviors was assessed using the Variance Inflation Factor, Tolerance and Condition Index. Analyses accounted for the complex sample design, and sampling weights were used to adjust for the unequal sampling probability and nonresponse. Sampling weights were trimmed and calibrated by age, sex, and heritage group to the characteristics of the four sampled communities using 2010 US Census data. Since high-normal systolic blood pressure values have been associated with both incident hypertension and reduced cognitive performance, we performed stratified analysis excluding participants with elevated systolic blood pressure, defined as 120 and 130 mmHg.

In an exploratory analysis, we investigated the potential mediating effect of unhealthy behaviors in the association between GC and incident hypertension. We tested indirect effects using a structural equation modeling framework (SEM), incorporating unhealthy behaviors first as single variables and then jointly in a multimodal model to assess pathways linking GC to incident hypertension, while adjusting for model 3 covariates. Interactions between mediators and cross-mediation pathways were included to account for the interrelationships between mediators. ORs, 95% CIs, and p values were reported for all pathways. The maximum likelihood with robust standard error estimation was used as the primary analytic approach, with Monte Carlo integration applied to handle binary mediators and categorical covariates. To further evaluate the robustness of indirect effects, we conducted a sensitivity analysis using bias-corrected bootstrap confidence intervals, using a simplified covariate specification.

For the SEM analyses, unhealthy behaviors (obesity, smoking, physical activity not at goal and poor diet) were modeled as parallel mediators with freely estimated covariance. Binary mediators and the binary outcome were specified using a logit link and estimated with maximum likelihood with robust standard errors (MLR) and Monte Carlo numerical integration (Mplus default integration points). Cross-mediational pathways among mediators (e.g, GC → smoking → obesity → hypertension) were explicitly modeled to capture shared variance and sequential behavioral effects. Indirect effects were computed on the log-odds scale and transformed to odds ratios. Robustness of indirect pathways was evaluated using bias-corrected bootstrap confidence intervals (5,000 draws).

Additionally, to ascertain the directionality of the effect between the exposure and mediators, we examined the independent effect of GC on incident hypertension by re-running our regression model adjusting for Model 3 covariates plus the pathway variables identified in our multimodal analysis. Furthermore, we re-estimated an alternative SEM specifying obesity as exposure, incident hypertension as the outcome, and GC as a mediator using the same MLR/logit specification with the full covariate set, deriving direct and indirect effects with 95% CIs.

Analyses were conducted in SAS 9.4 and Mplus 8.11 software. Statistical significance was set at p<0.05.

Results

The median age at Visit 1 was 53 years (IQR 48–59) and 62% of the participants were female. The median time between the two visits was 5.8 years (IQR 5.5–6.2), with 71% of the analytic sample having resided in the U.S. for at least 10 years. Approximately 41% met criteria for obesity, 20% were current smokers, 34% had low diet quality, and 39% did not meet the recommended goals for PA. In contrast, participants with baseline hypertension were older, had higher BMI, lower cognitive test performance, and a greater burden of adverse sociodemographic and health characteristics. By Visit 2, 57% of the sample developed hypertension (Table 1 and supplemental table S1).

Table 1.

Baseline characteristics of the HCHS/SOL target population

Characteristics	Participants n=6,755 (unweighted data)
Age, median (IQR), years	53.0 (48.0–59.0)
Elapsed time between visits, median (IQR), years	5.8 (5.5–6.2)
Sex Female-no. (%)	4175 (62)
Body Mass Index, median (IQR), (kg/m2)	28.8 (25.9–32.4)
Obesity (≥ 30 kg/m2), no. (%)	2734 (41)
Field center, no. (%)
Bronx	1480 (22)
Chicago	1666 (25)
Miami	1857 (27)
San Diego	1752 (26)
Heritage group-no. (%)
Dominican	541 (8)
Central American	709 (10)
Cuban	1096 (16)
Mexican	2620 (39)
Puerto-Rican	1105 (17)
South American	518 (8)
More than one heritage/Other	144 (2)
Income-no. (%)
<$10,000	1003 (16)
$10,001−$20,000	2009 (32)
$20,001−$40,000	2143 (34)
$40,001−$75,000	792 (13)
>$75,000	280 (5)
Education level-no (%)
Less Than High School	2731 (41)
High School or Equivalent	1459 (22)
Greater than High School or Equivalent	2544 (38)
No health insurance-no. (%)	3338 (50)
Nativity and years in the US, no. (%)
Foreign born, in the US for <10 years	1275 (19)
Foreign born, in the US for ≥10 years	4772 (71)
US born	682 (10)
Moderate and vigorous physical activity, median (IQR), min/week	260.0 (20.0–1050.0)
Physical activity not at goal –no. (%)	2631 (39)
Current Smoker-no. (%)	1381 (20)
Mediterranean Diet Score, Median (IQR)	5.0 (4.0–6.0)
Low diet quality- no. (%)	2304 (44)
Incident Hypertension (130/80 cutoff)- no. (%)	2989 (57)
Cognitive performance (z-scores), median (IQR)
Brief- Spanish-English Verbal Learning Sum	−0.02 (−0.74 to 0.71)
Brief- Spanish-English Verbal Learning Recall	0.18 (−0.52 to 0.53)
Word fluency	−0.08 (−0.77 to 0.62)
Digit Symbol Substitution Test	−0.07 (−0.67 to 0.60)
Global Cognition	0.002 (−0.53 to 0.53)

All the variables were determined at Visit 1 except for incident hypertension which was determined at Visit 2.

In the age- and time between visits-adjusted model, higher scores in GC, as well as each sub-component (B-SEVLT Sum, B-SEVLT Recall, WF, and DSST) were associated with lower odds of incident hypertension. In the fully adjusted model, GC remained as significant predictors of incident hypertension (OR=0.85; 95% CI, 0.74–0.98) (Table 2), corresponding to an absolute risk difference of 1.9% (95% CI 0.4–3.5%) and a population attributable fraction of 5.3% (95% CI 1.2–9.5%). In secondary analysis by neurocognitive test, only WF was associated with incident hypertension (OR=0.90; 95% CI, 0.82–0.99). We also observed that the addition of unhealthy behaviors (obesity, current smoking, and PA not at goal) to our final model, did not change the results (supplemental table S2). Furthermore, in adjusted stratified analysis done in individuals with blood pressure <120/80 mmHg, the association between GC and incident hypertension was retained (OR = 0.81; 95% CI: 0.68–0.97).

Table 2.

Association of incident hypertension with baseline cognitive function

Variable (Z-scored)	OR (95% CI)
	Model 1	Model 2	Model 3
Global Cognition	0.76 (0.67–0.86)***	0.81 (0.71–0.92)**	0.85 (0.74–0.98)*
Brief- Spanish-English Verbal Learning Sum	0.88 (0.81–0.97)*	0.91 (0.82–0.99)*	0.93 (0.85–1.03)
Brief- Spanish-English Verbal Learning Recall	0.86 (0.79–0.94)**	0.90 (0.82–0.99)*	0.93 (0.85–1.02)
Word fluency	0.83 (0.76–0.91)***	0.87 (0.79–0.96)*	0.90 (0.82–0.99)*
Digit Symbol Substitution Test	0.85 (0.77–0.93)**	0.87 (0.80–0.98)*	0.94 (0.84–1.05)

Model 1: age at baseline visit and time between visits; Model 2: age at baseline visit, sex, center, heritage group, income, years in US, nativity, insurance status, and time between visits; Model 3: age at baseline visit, sex, center, heritage group, income, years in US, nativity, insurance status, time between visits, and education.

^*p<0.05

^**p<0.005

^***p<0.0001.

We then performed adjusted single-pathway mediation analyses between GC, incident hypertension, and unhealthy behaviors. No collinearity was observed between the variables included in the analysis. Obesity was a significant pathway variable between GC and incident hypertension (indirect effect for obesity, OR=0.95; 95% CI, 0.90–0.99). Low quality diet was associated with incident hypertension (OR=1.39; 95% CI, 1.19–1.64) but was not a pathway variable between GC and incident hypertension (supplemental figure S1). PA not at goal and smoking were pathway variables in the association between GC and obesity (Table 3, supplemental figure S2A). In addition, current smoking was in the pathway between GC and low diet quality (Table 3, supplemental figure S2D).

Table 3.

Pathway analysis between baseline global cognition and unhealthy behaviors

Pathway variable	Outcome, OR (95% CI)
	Obesity	Physical activity not at goal	Current smoker	Low diet quality
Obesity
Direct pathway	-	0.88 (0.81–0.95)*	0.78 (0.70–0.86)***	0.94 (0.85–1.04)
Indirect pathway		0.98 (0.95–1.00)	1.03 (0.99–1.07)	0.98 (0.97–1.00)
Physical activity not at goal
Direct pathway	0.93 (0.86–1.01)	-	0.78 (0.70–0.86)***	0.95 (0.85–1.05)
Indirect pathway	0.96 (0.94–0.99)**		1.00 (0.98–1.02)	0.98 (0.95–0.99)*
Current smoker
Direct pathway	0.91 (0.84–0.98)*	0.87 (0.80–0.95)**	-	0.96 (0.86–1.06)
Indirect pathway	1.11 (1.05–1.16)*	1.00 (0.97–1.04)		0.89 (0.83–0.94)***
Low diet quality
Direct pathway	0.92 (0.85–1.00)*	0.87 (0.81–0.95)**	0.78(0.71–0.87)	-
Indirect pathway	0.99 (0.97–1.01)	0.99 (0.97–1.01)	0.97 (0.92–1.02)

Results adjusted by age at baseline, sex, center, heritage group, income, years in US, nativity, insurance status, and education.

^*p<0.05

^**p<0.005

^***p<0.0001.

Using this information, we created a multiple pathway model which showed that, after adjusting for Model 3 covariates, obesity partially mediated the association between GC and incident hypertension at Visit 2 (OR_{indirect effect through obesity}=0.95; 95% CI, 0.90–0.99, Figure). Higher GC was associated with lower odds of obesity (OR=0.89; 95% CI, 0.81–0.98), current smoking (OR=0.77; 95% CI, 0.68–0.87) and PA not at goal (OR=0.87; 95% CI, 0.79–0.95). PA not at goal and current smoking were pathway variables between GC on obesity. Importantly, neither of these variables were mediators in the GC-incident hypertension relationship. Low diet quality was associated with higher odds of both obesity and incident hypertension; however, because GC was not associated with MeDi adherence, diet did not mediate the GC- incident hypertension relationship (Figure).

Figure. Pathway analyses between baseline global cognition, unhealthy behaviors, and incident hypertension.

The multimodal model was designed to examine whether unhealthy behaviors are pathway variables in the relationship between global cognition and incident hypertension. The model controls for age at baseline visit, sex, center, heritage group, income, years in US, nativity, insurance status, time between visits, and education. The association between hypertension and global cognition is further adjusted by time between visits. Only significant associations are shown. Estimates (OR, 95% CI) represent positive direct effects (solid lines), negative direct effects (dashed lines), or indirect effects through behavioral mediators (dotted line)

Although all behaviors were evaluated as potential mediators, obesity was the only variable that was associated with both GC and incident hypertension—while smoking and PA were associated with GC but not with incident hypertension. The combined indirect effects across all behaviors were small and not statistically significant (overall indirect OR=0.96, 95% CI 0.89–1.03); obesity was the only behavior that significantly mediated the association between GC and incident hypertension (indirect OR=0.95, 95% CI 0.90–0.99).

In the alternative model using obesity as the exposure, GC as the mediator, and incident hypertension at V2 as the outcome, higher GC remained inversely associated with incident hypertension (OR = 0.88, 95% CI 0.80–0.96). In addition, obesity showed a strong direct association with incident hypertension (OR = 1.62, 95% CI 1.43–1.84). However, the indirect pathway through cognition was negligible (OR indirect = 1.01, 95% CI 1.00–1.01), leaving our conclusions unchanged.

In sensitivity analyses using bias-corrected bootstrap confidence intervals, the indirect effect of GC on hypertension through obesity remained significant (OR=0.95; 95% CI 0.90–0.99), consistent with the primary MLR model. Point estimates across pathways were highly similar between models, indicating that the mediation findings were robust to estimation approach (supplemental methods).

Discussion

Our study suggests that GC is associated with the development of hypertension in middle- to older-aged Hispanic/Latino individuals. In particular, each standard deviation increase in GC score at Visit 1 was linked to 15% lower odds of developing hypertension by Visit 2, corresponding to a model-based absolute risk difference of 1.9% and a population-attributable fraction of approximately 5%. These findings indicate that while the individual-level effect of cognition on hypertension risk is modest, improving cognitive health could have a meaningful impact on the population level.

In addition, the results of our exploratory analysis suggest that modifiable unhealthy behaviors could be pathway variables in the relationship of GC and incident hypertension. However, obesity emerged as the sole significant mediator among the included unhealthy behaviors, because it was the only factors simultaneously associated with both GC and hypertension risk. Other behaviors influenced obesity (smoking, physical activity) but did not meet the statistical criteria for mediation GC-hypertension relationship. These findings align with the established biological link between adipose tissue, vascular dysfunction and blood pressure regulation.

A wealth of evidence obtained in clinical and preclinical models confirms that midlife hypertension can compromise brain health by causing stroke, microstructural changes, and cognitive impairment later in life^3,22,23. Data obtained in rodent models have led to the identification of different pathogenic mechanisms that explain these associations, such as neurovascular uncoupling, blood brain barrier dysfunction, neuroinflammation, and oxidative stress^22,24.

Maintaining appropriate cerebral blood flow is essential for preserving brain health. To achieve this, the brain exerts a delicate control of physiological variables that regulate tissue perfusion, including blood pressure. Thus, we hypothesized a bidirectional association between optimal brain health and hemodynamic control. To date, there is evidence supporting the notion that cognitive dysfunction can precede cardiac disease. In the Atherosclerosis Risk in Communities study, lower adjusted scores on tests of cognitive function predicted a greater risk of incident cardiovascular events in the average follow up of 6.4 years²⁵. Additionally, in the Prospective Study of Pravastatin in the Elderly at Risk, participants in the lowest third of executive function had a 1.85-fold higher risk of coronary heart disease at 3 years compared to participants in the highest third²⁶. However, the predictive value of cognitive function for incident hypertension has not been described in prospective cohorts. In our study, we found an inverse relationship between GC and the odds of developing hypertension. In addition, our results suggest that unhealthy behaviors could be pathway variables in this association. Specifically, improved GC, either directly or through optimal PA or decreased smoking, was associated with decreased obesity which, in turn, led to a lower risk of hypertension (Figure). However, neither optimal PA nor smoking were direct mediators in the relationship between GC and incident hypertension.

Among the different neurocognitive tests included in the analysis, WF was associated with incident hypertension. WF assesses phonemic fluency and relies on generating words within 1 min for the letters F and A. The WF primarily assesses executive function, though it can also be influenced by attention, semantic memory, and processing speed²⁷. Dysfunction in neuroanatomical areas associated with these domains, such as prefrontal cortex and hippocampus can impair impulse control and result in the adoption of unhealthy behaviors, some of which are associated with hypertension. In a systematic review of 28 longitudinal and 63 cross-sectional studies, worse executive function was associated with obesity⁵. Dietary habits are determined by the interplay of different neuroanatomical areas, including those involved in hedonic circuitry, such as the nucleus accumbens, ventral pallidum, and brain stem, and in Pavlovian learning, such as the amygdala²⁸. In addition, it was proposed that, through reduced neurogenesis and enhanced inflammation, the elevated intake of an obesity-promoting Western diet leads to hippocampal dysfunction which weakens regulation of food intake and results in an increased intake of harmful foods and obesity²⁹. Thus, pathology in these neuroanatomical areas can change dietary self-control. In addition, the mesolimbic system and prefrontal cortex, both involved in impulsivity and reward mechanisms, have been shown to predict the early initiation of smoking³⁰.

In our model, obesity appeared as a key pathway variable between GC and incident hypertension. Obesity is an established predictor of incident hypertension³¹. This association stems from an increased production of cytokines in adipocytes, such as leptin, which lead to the activation of the renin–angiotensin–aldosterone system and the sympathetic nervous system³² with the consequent adverse effect on blood pressure control. GC was not associated with low diet quality; however, poor diet was associated with obesity and incident hypertension. These findings correlate with evidence showing that higher adherence to the Mediterranean Diet (which focuses on the consumption of fresh unprocessed foods, healthy fats, and low sodium) promotes cardiovascular health by decreasing arterial blood pressure, cholesterol levels, and the risk of overweight and obesity³³.

Importantly, while our analysis used data from selected midlife points, many behavioral and clinical risk factors, such as obesity, smoking or poor diet, likely accumulate over decades. In our sample, 41% of participants were already obese at baseline, suggesting that these exposures may have been long-standing and could themselves have contributed to early cognitive decline. Lifelong exposure beginning in childhood or early adulthood are increasingly recognized as shaping both cardiometabolic and cognitive trajectories^34,35. Therefore, the associations observed here probably reflect not only concurrent interactions, but also cumulative effects of sustained health behaviors and physiological dysregulation throughout the course of life.

Our study, together with results previously reported in HCHS/SOL^36,37, provides evidence that cognition and incident hypertension have a bidirectional association and raises the possibility of a cycle model whereby decreased cognition, through the adoption of unhealthy behaviors, results in hypertension in midlife, which further exacerbates cognitive dysfunction later in life. This dynamic cycle may begin even earlier during life: individuals with poorer cognitive abilities in childhood and adolescence may be less likely to adopt or maintain healthy lifestyles, predisposing them to obesity and hypertension, which in turn impair cognition in later years³⁸. In our study, lower GC was associated with later hypertension, partially through obesity and health behaviors, supporting the notion that cognitive capacity influences self-regulation and adherence to healthy choices. Conversely, hypertension related brain injury could later worsen executive control, reinforcing unhealthy behaviors. Several neuroanatomical areas associated with the adoption of detrimental lifestyles, such as the prefrontal cortex, hippocampus, and amygdala, participate in risk assessment, emotion processing, and decision making^30,39,40. Emerging data from behavioral economics, a field that seeks to understand how the interplay of cognitive, emotional, and social factors shape choices, show that gratification, inertia, and loss of aversion may outweigh self-interest, leading individuals to deviate from rational action when they make decisions⁴¹. Thus, from a practical perspective, it is possible that programs that combine behavioral incentives and education may be necessary to change behaviors and break the cognitive function-hypertension cycle.

Our study has strengths and limitations. First, several covariates were self-reported introducing the possibility of recall bias. Second, we used four cognitive tests to assess cognitive function and cannot exclude the possibility that including other neurocognitive tests could yield different results. Third, the association between GC and incident hypertension was retained after adjustment for different social determinants of health, including household income, years in the US, nativity, health insurance status, and education; in future studies we will assess the modulating effect of acculturation, psychological stress, and social support in such a relationship. Fourth, since our study included only Hispanic/Latino participants, our results cannot be directly extrapolated to other populations. Fifth, while our findings indicate an association between cognitive function and the subsequent development of hypertension, they should not be interpreted as evidence of causality. Moreover, because unhealthy behaviors and GC were measured concurrently, the direction of effect between these variables cannot be conclusively determined. Therefore, the results of the multimodal analysis should be regarded as hypothesis-generating rather than as proof of causation. Despite these limitations, our study analyzed one of the largest and most diverse cohort of middle-aged and older Hispanic or Latino individuals living in US, to describe the role of cognition performance on incident hypertension six years later. In addition, owing to the longitudinal design of HCHS/SOL, our analysis allows us to establish temporal precedence between cognitive function and incident hypertension.

Finally, sex-associated variations have been observed in both cognition and hypertension. Thus, in subsequent studies we will explore sex and cognition interactions and sex-stratified analyses, as these may reveal important effect modifications and provide a more nuanced understanding of how cognition relates to hypertension risk.

In conclusion, we found that better GC function at baseline was associated with lower odds of developing hypertension six years later. Steering individuals’ behavior consistent with their goals or preferences in the management of cardiovascular disease may reduce the incidence of hypertension, one of the most important determinants of brain health.

Pathophysiological Novelty and Relevance.

What is New?

This study demonstrates that higher cognitive performance, assessed through a composite global cognitive score, is associated with lower odds of incident hypertension in a large community-based cohort of Hispanic/Latino adults. These findings extend the traditional model that views vascular risk factors as a precursor of cognitive decline, showing instead that cognition itself may influence the development of hypertension. Our multimodal analysis further identified modifiable factors -obesity, physical inactivity, and smoking- as potential pathway variables linking cognitive function to incident hypertension.

What is Relevant?

These findings suggest an interrelated association between cognitive function and hypertension, with cognition potentially serving as an underrecognized risk factor for incident hypertension. Individuals with lower self-regulatory capacity may be less likely to adopt or maintain healthy lifestyles, thereby predisposing them to obesity and subsequent hypertension.

Integrating cognitive assessment into cardiovascular prevention strategies may help identify individuals who would benefit from behavioral support or cognitive-behavioral interventions.

Clinical /Pathophysiological Implications

Our results support a bidirectional model of brain-vascular interaction in which higher cognitive scores are associated with better health behaviors and lower odds of incident hypertension, which in turn reduces neurovascular injury and cognitive decline. Targeting both cognitive and behavioral domains, through education, cognitive training and lifestyle modification, may be essential to breaking the cognition-hypertension cycle and may ultimately reduce both vascular and neurocognitive disease burden.

Perspective:

In this study, higher cognitive function was associated with lower incidence of hypertension over six years among middle and older Hispanic/Latino adults. These results extend the traditional view that hypertension leads to cognitive decline by suggesting that a potential reverse pathway in which reduced cognitive capacity, particularly in executive domains, may predispose individuals to unhealthy behaviors such as smoking, physical inactivity and obesity, which are risk factors for hypertension.

Accumulating evidence indicates that cognitive and cardiometabolic health influence one other across life span. Our findings support the concept of a cognition-hypertension cycle, in which impaired self-regulation and decision making contribute to the adoption of adverse heath behaviors, that, in turn, worsen vascular function and further compromise cognition.

These observations highlight the importance of integrating cognitive screening and behavioral support into cardiovascular prevention programs. Interventions that enhance executive function, improve health literacy, or provide behavioral reinforcement may help mitigate the risk of hypertension in cognitively vulnerable populations. Future research should test multidomain interventions combining cognitive training, lifestyle counseling and blood pressure control strategies to break this reciprocal cycle and promote both brain and vascular health.

List of Non-standard Abbreviations and Acronyms

HCHS/SOL

Hispanic Community Health Study/Study of Latinos

BMI

Body Mass Index

PA

Physical Activity

BP

Blood Pressure

GC

Global Cognitive Score

HTN

Hypertension

B-SEVLT

Brief Spanish-English Verbal Learning Test

WF

Word Fluency

DSST

Digit Symbol Substitution Test