Association of physical activity on changes in cognitive function: Boston Puerto Rican Health Study

Abstract

Objective:

To evaluate the association of habitual physical activity engagement on changes in cognitive function among Puerto Rican adults.

Methods:

Longitudinal data (2-year follow-up) from the Boston Puerto Rican Health Study were analyzed (n = 862; mean age = 56.5 year). A daily energy expenditure score was calculated using the number of hours over a 24-h period engaged in various activities, including sleeping, light activity, and moderate-to-vigorous exercise. Energy expenditure estimates were weighted based on the rate of oxygen consumption associated with each activity. Seven cognitive function outcomes were evaluated, including an assessment of general cognitive function, episodic memory, attention and working memory, cognitive flexibility, response inhibition, processing speed, and visuo-spatial organization. From these, overall executive function and memory capacity were derived using principal components analysis.

Results:

Physical activity was not associated with changes in overall executive function. However, compared to those with low baseline physical activity, those with moderate physical activity had 48% reduced odds of having ≥1 standard deviation decline in memory function (OR = 0.52; 95% CI: 0.32, 0.84; P = 0.008) in 2 years.

Conclusion:

Among Puerto Rican adults, physical activity may help attenuate memory decline.

Introduction

Emerging research from acute exercise experiments [1–5], chronic exercise training [6], and large-scale epidemiological studies [7] demonstrates that physical activity is favorably associated with cognitive function, including episodic memory, working memory, and executive function. We have previously detailed potential mechanisms through which physical activity may contribute to cognitive function [8–10], noting unique pathways depending on the cognitive outcome [11]. Such broad mechanisms include, for example, alterations in neuronal excitability, growth factor production, neurogenesis, gliogenesis, and angiogenesis. There may also be unique mechanisms through which exercise may influence various cognitive domains [11]. For example, exercise may help to subserve episodic memory primarily through alterations in long-term potentiation [8]. Executive function involves various higher order cognitions, such as planning, reasoning, and inhibition, and thus, may have additional and/or unique exercise-related mechanisms, including improved cerebral blood flow and cerebrovascular reactivity [12,13], or altered activity in neural structures related to planning (e.g. frontoparietal system) [14] and inhibition (e.g. lateral and orbitofrontal divisions of the prefrontal cortex) [15].

Although recent encouraging work suggests broad neurocognitive benefits of physical activity, few studies on this topic have comprehensively evaluated the association of physical activity on a multitude of cognitive outcomes. Thus, herein, we evaluate the association of physical activity on overall executive function (including aspects of cognitive flexibility and processing speed) and memory performance (including aspects of working memory, episodic memory, and visuo-spatial organization).

Another novel aspect of the present study is that we study these relationships among Puerto Rican adults, which has yet to be evaluated in this emerging line of inquiry. Evaluating this research question, within this population, is noteworthy, as Puerto Rican’s have health profile characteristics (e.g. greater depression [16–18], diabetes [19], and obesity [20] prevalence), which are also associated with cognitive decline [21 ,22]. As a result, the purpose of this study, written as a brief report, was to evaluate the association of physical activity on a variety of cognitive outcomes among an adult sample of Puerto Ricans.

Methods

Study design and participants

The present study uses data from the Boston Puerto Rican Health Study, which is an ongoing epidemiological study funded by the National Institutes of Health. Detailed information regarding the design of this study has been described elsewhere [23]. In brief, Puerto Rican adults (45–75 years) in the Boston, Massachusetts area, were randomly sampled within qualified households in selected blocks. In total, 2170 individuals were identified, and among these, 1811 agreed to be interviewed. Among these, nine participants were excluded due to a low (≤10) score on the Mini-Mental State Examination (MMSE). A total of 1500 individuals completed the baseline household interview. The present study included 1107 participants with complete data on the study variables for the baseline sample. Two-year follow-up assessments were evaluated to examine the association of physical activity on changes in cognition. The follow-up sample included 862 participants with cognition data at both time periods.

Measurement of physical activity

As described elsewhere [23], and tested within a Puerto Rican sample [24], physical activity was assessed using a modified version of the Paffenbarger questionnaire, administered as part of the Harvard Alumni Study [25,26]. A daily energy expenditure was calculated using the number of hours over a 24-h period engaged in various activities, including sleeping, light activity, and moderate-to-vigorous exercise. Energy expenditure estimates were weighted based on the rate of oxygen consumption associated with each activity (weights ranging from 1 to 5, following those used by Kannel and Sorlie [23,27]). Physical activity was then categorized as low (<30), moderate (30–39), or high (40+).

Measurement of cognitive function

As described elsewhere [28,29], seven cognitive function assessments were administered in the household interview. These included the Mini-Mental State Examination (MMSE; overall measure of general cognitive function), List Learning (episodic memory), Digit Span (attention and working memory), Stroop (cognitive flexibility, response inhibition and processing speed), Clock Drawing (visuo-spatial organization), Figure Copying (visuo-spatial organization), and Letter Fluency (verbal fluency) assessments. Specific details on the protocol for each of these cognitive assessments is accessible on the Boston Puerto Rican Health Study website [30]. In brief, the MMSE is a 30-item assessment to evaluate the severity of cognitive impairment, including functions such as attention, recall, language, and orientation. The List Learning task, similar to the Ray Auditory Verbal Learning Task, includes learning a list of 16 words (5 trials), followed by a distractor list (List B), 2 immediate recalls (free and cued), 2 long-term (20-min delay) recalls (free and cued), and a memory recognition task. The Digit Span included a digit forward and digit backward task. The Stroop task involved three separate tasks, including, over a 45-s period, reading as many words as possible, naming colors, and an incongruent task involving the identification of the ink a word is colored as opposed to the written word. The Clocking Drawing task involved drawing a clock demonstrating that the time is 10 after 11. The Figure Copying task involved copying various visuo-spatial illustrations. Lastly, the Letter Fluency task involved, within a 60-s period, coming up with as many words that start with a specific letter of the alphabet (three separate trials were completed, one for ‘C’, ‘F’, and ‘L’)Similar. to other reports, and to minimize multiple comparisons and to facilitate a parsimonious representation of the data, principal components analysis with varimax rotation was used as a data reduction technique to derive composite scores for separate cognitive domains. Based on loadings for individual baseline test scores, the two resulting components with eigenvalues greater than 1 were designated as measuring the cognitive domains of executive function and memory. Composite scores were also computed for 2-year follow-up testing data, using the means, standard deviations (SDs), and scoring coefficients from the baseline PCA. Difference scores for baseline-follow-up were calculated for each domain, and descriptive statistics were obtained for these difference scores. For each cognitive domain, ‘major decline’ was defined by a drop of ≥1 SD from the group mean difference score. By definition, each PCA-derived composite score had a mean of 0 and an SD of 1 [28,29].

Statistical analysis and co variates

Multivariable linear regression models, implemented in Stata (v. 12, College Station, TX, USA), were fitted to examine associations between physical activity at baseline and changes in executive function and memory performance. The outcome measure was the cognitive change score (Time 2-Time 1). Models were computed separately for the cognitive outcome measure, with physical activity serving as an independent variable (low physical activity as referent). Multivariable logistic regression analyses were also used to examine associations between baseline physical activity and odds of having a ≥1 SD decline in executive function and memory over the 2-year follow-up. These analyses were conducted to confirm the findings from the linear regression models as well as to aid in the clinical interpretation of the findings.

Primary models were adjusted for age, sex, education, income, measured body mass index (BMI), hypertension (>140/90 mmHg or on hypertension medication), diabetes (fasting glucose >126 mg/dL or on diabetes medication), depressive symptoms (Center for Epidemiology Studies Depression Scale), literacy, perceived health status, diet (healthy eating index, HEI-2005), smoking status (never, former, current), and alcohol use (never, in the past but not within 1 year, within past year, or within past 30 days) (Table 1). For sensitivity analyses, we also controlled for the baseline cognitive parameter in the model. Statistical significance was set at an a-priori nominal alpha of 0.05.

Table 1.

Demographic and behavioral characteristics of the sample (n = 862).

Variable	Point estimate	SD
Age, mean years	56.5	7.5
Sex, % female	74.3
Education, %
No schooling or less than fifth grade	16.2
Fifth-eighth grade	25.5
ninth-twelfth grade	41.6
Some college or bachelor’s degree	14.5
At least some graduate school	2.1
Income, mean annual household ($)	18,808	19,807
Body mass index, mean kg/m2	31.8	6.2
Hypertension (>140/90 mmHg), %	41.3
Diabetes (glucose 126 + mg/dL or medication), %	36.4
Depression Score, %
15 or less	41.3
16–21	14.8
22+	43.9
Literacy, % illiterate	0.8
Health status, %
Excellent	5.0
Very good	6.1
Good	20.1
Fair	56.3
Poor	12.5
Diet, mean healthy eating index (HEI-2005)	71.8	9.7
Smoking, %
Never	47.2
Smoked in past, but not currently	30.4
Current smoker	22.4
Alcohol Use, %
Never	28.1
In the past, but not within past year	28.8
Within past 30 days	26.3
Within past year	16.8
Physical Activity Score, %
<30	44.1
30–39	51.6
40+	4.3

Results

Tables 1 and 2, respectively, display the demographic and cognitive data. Participants had a mean SD age of 56.5 (7.5) years, with the sample predominately female (74.3%). The majority had a high school degree or less (83.3%), with a mean annual household income of $18,808. Most participants were obese or overweight (mean BMI = 31.8 kg/m²), and 41.3% and 36.4% had hypertension and diabetes, respectively. Similarly, the majority (68.8%) of the participants self-reported their health status as fair or poor. Lastly, the proportion of low, moderate, and high baseline physical activity, respectively, was 44.1, 51.6, and 4.3%.

Table 2.

Cognitive function scores.

Cognitive variable	Point estimate (SD)
Immediate recall (over five trials)	8.14 (3.3)
Recognition, mean	13.14 (2.5)
Digit span
Digit forward, mean	7.45 (1.9)
Digit backward, mean	3.53 (1.5)
Stroop
# words read in 45 s, mean	70.5 (20.1)
# of colors named in 45 s, mean	46.8 (14.9)
# correct ink but ignore word in 45 s, mean	23.8 (10.5)
Clock drawing, mean (higher = greater drawing accuracy)	2.20 (0.9)
Figure copying, mean (higher = greater ability to copy)	4.02 (2.4)

# = number

Baseline physical activity was not associated with 2-year change in executive function (Table 3). However, those engaging in moderate physical activity (vs. low) had improved preservation of memory function (β = 0.15; 95% CI: 0.01, 0.28; P = 0.03). Additional sensitivity analyses were conducted to evaluate this relationship while controlling for baseline memory function and, with this inclusion, results were unchanged. For example, those engaging in moderate physical activity (vs. low) continued to have improved preservation of memory function (β = 0.12; 95% CI: O.Q1, 0.23; P = 0.03).

Table 3.

Multivariable linear regression analyses examining the association between baseline physical activity and changes in executive function and memory (n = 862).

Physical activity	Executive function			Memory
	β	95% CI	p-value	β	95% CI	p-value
Low	Referent			Referent
Moderate	−0.01	−0.10, 0.27	0.69	0.15	0.01, 0.28	0.03
High	−0.003	−0.22, 0.21	0.97	0.08	−0.25, 0.41	0.64

Models controlled for age, sex, education, income, BMI, hypertension, diabetes, depression, literacy, perceived health status, diet (HEI-2005), smoking, and alcohol use.

Similar results emerged when examining the association between baseline physical activity and odds of having a ≥ 1 SD decline in executive function and memory (Table 4). Compared to those with low baseline physical activity, those with moderate physical activity had 48% reduced odds of having a 1 + SD decline in memory function (OR = 0.52; 95% CI: 0.32, 0.84; P = 0.008). Results were similar when also controlling for baseline memory function (OR = 0.50; 95% CI: 0.31, 0.82, P = 0.007). Further, for all analyses, reduced models (e.g. not including potential mediators, such as hypertension, diabetes, and BMI) were computed, but these results were similar to the fully adjusted models. Results were also unchanged when baseline MMSE was also controlled for in the model.

Table 4.

Multivariable logistic regression analyses examining the association between baseline physical activity and odds of having a 1 + standard deviation decline in executive function and memory (n = 862).

Physical activity	Executive function			Memory
	OR	95% CI	p-value	OR	95% CI	p-value
Low	Referent			Referent
Moderate	1.58	0.77, 3.27	0.21	0.52	0.32, 0.84	0.008
High	3.02	0.75, 12.1	0.12	1.06	0.37, 3.02	0.90

Models controlled for age, sex, education, income, BMI, hypertension, diabetes, epression, literacy, perceived health status, diet (HEI-2005), smoking, and alcohol use.

Discussion

The purpose of this study was to extend the emerging field of exercise neurophysiology by examining the association between physical activity and cognitive function among Puerto Rican adults. Our main findings were that physical activity was not significantly associated with change in executive function, but that moderate physical activity was associated with reduced decline in memory function over a 2-year period.

These findings align with other related studies, primarily in U.S. populations, showing that higher levels of physical activity are associated with greater function on tasks related to memory [7,31]. These benefits, in part, may relate to physical activity-related alterations in the production of growth factors (e.g. brain-derived neurotrophic factor[10], IGF-1 [9]), new neurons [32], and glial cells [33,34], as well as adaptations in cerebral angiogenesis [35] and hippocampal volume [36]. The null findings for executive function are surprising. Unlike memory function, the mean change in executive function was minimal (Table 1), and thus, this homogenous response, over a relatively short time period, may have prevented an association between physical activity and changes in executive function. Another interesting finding was that, although moderate amounts of physical activity were associated with reduced decline in memory function, higher amounts of physical activity did not show this effect. This aligns with our previous work in a nationally representative sample of older adult Americans, demonstrating an inverted U-shaped relationship between physical activity and cognition [7]. For the present study, it is possible that the null findings for higher volumes of physical activity may be attributed to the minimal number of individuals in this category (4.3%). Alternatively, as we have discussed elsewhere [7], this potential dose-response relationship may have biological explanations. For example, other work demonstrates that long duration exercise may be less beneficial than moderate-length exercise sessions in facilitating cognitive function [37] and may induce a less favorable oxidative stress-related response [38].

Limitations of this study include self-reported assessment of physical activity. Despite these limitations, strengths of this study include the comprehensive assessment of cognitive function, evaluating the association between physical activity and cognition in an under-investigation population at risk for cognitive decline, and employing a longitudinal study design.

In conclusion, in this brief report, we extend this emerging line of research by showing that, among Puerto Rican adults, moderate amounts of physical activity are associated with reduced odds of memory decline. Future experimental work on this topic, and among this population, is warranted. Such work should continue to evaluate these relationships among various demographic parameters. It will be important to determine whether such effects are more pronounced among, for example, select race-ethnicity or age groups. Our study was conducted among middle-age Puerto Ricans. The majority of work evaluating the potential neurocognitive effects of exercise on cognition have focused on older adults. Thus, specifically among this population, additional work in older (and younger) samples would contribute to the literature.