Profiles of lifestyle health behaviors and cognitive decline in older adults

Abstract

INTRODUCTION

We aimed to identify profiles of modifiable, late‐life lifestyle health behaviors related to subsequent maintenance of cognition and explore sociodemographics and health characteristics as effect modifiers.

METHODS

Analyses used data from 715 older adults without baseline dementia from the Rush Memory and Aging Project and with lifestyle health behaviors (physical activity, cognitive activity, healthy diet, social activity) at baseline and ≥ 2 annual assessments of cognition. We used latent profile analysis to group participants based on behavior patterns and assessed change in cognition by group.

RESULTS

Three latent profiles were identified: high (n = 183), moderate (n = 441), and low (n = 91) engagement in health behaviors. Compared to high engagement, the moderate (mean difference [MD] = ‐0.02, 95% CI = [‐0.03;‐0.0002], p = 0.048) and low (MD = ‐0.06, 95% CI = [‐0.08;‐0.03], p < 0.0001) groups had faster annual rates of decline in global cognition, with no significant effects modifiers (vascular risk factors, apolipoprotein E [APOE] ε4, motor function).

DISCUSSION

Avoiding low levels of lifestyle health behaviors may help maintain cognition.

Highlights

• Latent profile analysis (LPA) captures lifestyle health behaviors associated with cognitive function.

• Such behavior include physical activity, cognitive activity, healthy diet, social activity.

• We used LPA to examine associations of behaviors and cognitive function over time.

• Older adults with low lifestyle health behaviors showed more rapid decline.

• To a lesser degree, so did those with moderate lifestyle health behaviors.

• Vascular conditions and risks, APOEε4, or motor function did not modify the effect.

1. BACKGROUND

Preventing Alzheimer's disease (AD) and other related dementias is an international public health priority because of increasing prevalence (75 million adults 65 years of age and older affected by 2030) and negative implications for health. ¹ Improving lifestyle health behaviors such as physical activity, ² , ³ , ⁴ , ⁵ cognitive activity, ⁶ , ⁷ , ⁸ healthy diet, ⁹ , ¹⁰ , ¹¹ and social activity ¹² , ¹³ , ¹⁴ , ¹⁵ are each separately associated with improved cognitive function in longitudinal and intervention investigations. These four behaviors can be integrated into daily life practices and habits in community and home settings. This is advantageous because it may be more effective to promote increasing or adding behaviors versus discouraging or eliminating unhealthy behaviors. ¹⁶

Building on this, clinical trials such as studies from World‐Wide FINGERS ¹⁷ , ¹⁸ , ¹⁹ are testing multidomain interventions that combine evidence‐based strategies to target several lifestyle health behaviors simultaneously. In existing multidomain lifestyle health behavior interventions in older adults, up to four lifestyle health behaviors ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ have been targeted in standardized interventions. However, this was done without considering variation in individual behavior profiles. An individualized, tailored approach that targets deficient behaviors only may lead to more sustainable interventions with greater adherence. ²⁸ , ²⁹ , ³⁰ , ³¹ Unfortunately, it has not been empirically established what specific combinations of lifestyle health behaviors would be optimal for individual cognitive function over time.

Previously, examinations of combinations of lifestyle health behaviors in association with cognitive function relied on summary measures of behaviors (e.g., counts). Though appealing because of simplicity, summary measures do not differentiate unique patterns across behaviors. ³² , ³³ , ³⁴ Instead, latent profile analysis (LPA) can be used to identify individuals with a similar pattern of lifestyle health behaviors associated with optimal maintenance of cognitive function. ³⁵ This method can enhance precise selection of specific evidence‐based interventions that will target only those lifestyle health behaviors that are associated with optimal cognitive function and are deficient for the individual participant.

RESEARCH IN CONTEXT

1. Systematic review: Few recent studies have used a latent profile approach to examine lifestyle health behaviors and their association with cognitive function in older adults. Latent profile analyses examining multidomain health behaviors with longitudinal cognitive outcomes and testing for potential effect modifiers are limited.

2. Interpretation: Our findings indicated that older persons who engaged in low levels of lifestyle health behaviors, and in moderate levels to a lesser degree, showed a faster rate of cognitive decline compared to those engaged in high levels of such behaviors. Vascular disease risk factors/conditions, apolipoprotein E (APOE) ε4, or motor function were not found to modify the effect.

3. Future directions: Recommendations for future research include: (a) adaptation and tailoring of existing evidence‐based interventions that promote achieving moderate levels of lifestyle health behaviors and (b) additional research to investigate lifestyle health behavior profiles in diverse populations, considering social determinants of health and a lifespan approach.

Recently, studies have used a latent profile or class approach to examine combinations of lifestyle health behaviors and examined relation of profiles or classes to cognitive function in cross‐sectional designs ³⁶ , ³⁷ or dementia incidence ³⁸ in older adults. These three studies included behaviors that could be promoted or increased to support a healthy lifestyle, though one also included smoking and alcohol consumption. ³⁶ Overall, studies found that higher engagement in cognitive, physical, and “going out” activities were related to better cognitive function, ³⁷ while not engaging in unhealthy behaviors like smoking. ³⁶ These studies suggested usefulness of employing a latent profile approach to examine engagement in multiple lifestyle health behaviors. It is still unknown how profiles of lifestyle health behaviors inform cognitive function in a longitudinal design. Also, though not examined in earlier studies, it is possible that the lifestyle health behavior profile associated with optimal cognitive function differs based on sociodemographic factors (e.g., sex, education, socioeconomic), genetic differences (i.e., apolipoprotein E [APOE]‐ε4), and health characteristics (e.g., vascular risk factors and disease, motor function). ³⁹ , ⁴⁰ , ⁴¹

In the current study, we conducted LPA using data from a longitudinal cohort study of community‐dwelling older adults without baseline dementia. We focused on lifestyle behaviors that can be integrated into daily life and maintained over long periods of time, with a goal of informing subsequent practical, lifestyle‐focused interventions. First, we identified latent profiles of baseline lifestyle health behaviors (i.e., combinations of physical activity, cognitive activity, healthy diet, and social activity), and examined profile differences in cognitive function (baseline levels of, and changes in, global cognition and five cognitive systems, including episodic memory, semantic memory, working memory, perceptual speed, visuospatial memory). Second, we explored sociodemographics (sex, education, socioeconomic status), APOEε4, and health characteristics (vascular risk factors and diseases, motor function) as effect modifiers. This approach can allow for identification of combinations of lifestyle health behaviors for cognitive function best suited for each sub‐group, which can encourage the future development of tailored interventions that are more acceptable and effective when implemented. ³⁵

2. METHODS

2.1. Study participants

The Rush Memory and Aging Project (MAP) is an ongoing longitudinal clinical‐pathologic study of aging and dementia, which started in 1997 and enrolls women and men from more than 40 retirement communities and housing facilities in the metropolitan Chicago area. ⁴² Eligibility for enrollment in MAP requires no known dementia and agreement to annual clinical evaluations and brain donation at death. Some participants may have impaired cognitive function at enrollment, including a small number who meet criteria for a dementia. These persons were excluded from analyses. This a common practice in prospective community‐bases cohort studies of aging and incident dementia. At study entry and annually thereafter, each participant underwent a uniform clinical evaluation, which included a structured medical history, complete neurologic examination, and cognitive testing. To date, the follow‐up rate exceeds 90%. All participants signed an informed consent and an Anatomical Gift Act for organ donation, and the study protocol was approved by an Institutional Review Board of Rush University Medical Center.

2.2. Eligible participants for analysis

Of the 2297 participants enrolled in MAP who had completed the initial clinical evaluation at the time of these analyses, we first identified the 1059 individuals involved in the dietary sub‐study (launched in 2004) and with Food Frequency Questionnaire data (see Section 2.3.3) validated, completed, and processed. We then excluded three participants who did not have concurrent data on the four lifestyle health behaviors of interest and excluded another 51 participants who were identified as having dementia at enrollment or between enrollment and the first assessment of all four lifestyle health behaviors (analytic baseline). Last, we excluded 290 participants without both a valid, completed, cognitive evaluation at analytic baseline (when assessment of all four lifestyle health behaviors occur) and follow‐up evaluation, resulting in a study sample of 715 participants for these analyses.

2.3. Lifestyle health behaviors

Lifestyle health behaviors, including physical activity, cognitive activity, healthy diet, and social activity, were assessed at each annual clinical evaluation; we focused on each participant's first concurrent report of the four relevant scales (i.e., hereafter, analytical baseline).

2.3.1. Physical activity

Physical activity was self‐reported using a modified version of the 1985 National Health Interview Survey. ⁴³ Activities included walking for exercise, gardening or yardwork, calisthenics or general exercise, bicycle riding, and swimming or water exercise. Participants were asked if they had engaged in any of those activities within the past 2 weeks and, if so, the number of occasions and average minutes per occasion. Minutes spent engaged in each activity were summed and expressed as hours of activity per week, as previously described. ⁴⁴

2.3.2. Cognitive activity

Frequency of participation in cognitively stimulating activities during the past year was self‐reported using a structured questionnaire. Seven activities were chosen that involved information processing or retention with minimal physical or social demands, including reading, visiting the library, reading newspapers, reading magazines, reading books, writing letters, and playing games (e.g., checkers, cards, puzzles). Participants rated each activity on a 5‐point scale (1 = once a year or less, 2 = several times a year, 3 = several times a month, 4 = several times a week, 5 = every day or nearly every day), and item scores were averaged to yield a total score ranging from 1 to 5, with higher scores indicating greater social activity. As described previously, ⁴⁴ scores on individual items were averaged to yield a composite measure of frequency of participation in cognitively stimulating activities.

2.3.3. Healthy diet

A healthy diet was assessed using a 144‐item semiquantitative Food Frequency Questionnaire (FFQ), modified from the Willett FFQ for use in older individuals. Validity and reliability of this FFQ was previously established. ⁴⁵ For each food item, participants were asked to report the usual frequency of intake during the past year. Nutrient levels and total energy were based either on natural portion sizes (e.g., one apple) or according to age‐ and sex‐specific portion sizes from national dietary surveys. ⁴⁵ In this study, we focused on the Mediterranean‐DASH Intervention for Neurodegenerative Delay (MIND) diet, developed at Rush University and based on the same cohort as studied here (MAP). The MIND diet is a hybrid of the Mediterranean diet and Dietary Approaches to Stop Hypertension diet patterns, with additional foods associated with cognitive health in later life. ⁹ The MIND diet score is based on achieving target amounts of 10 healthy components (green leafy vegetables, other vegetables, berries, fish, poultry, beans, whole grains, nuts, olive oil, and wine) and reducing intake of five unhealthy components (red and processed meat, pastries and sweets, fast and fried foods, full‐fat cheese, and butter/margarine). For each component, values of 0, 0.5, or 1 point are assigned based on a predefined number of servings, with higher scores representing healthier levels of intake; exceptions were olive oil consumption, which was scored 1 if identified by the participant as the primary oil usually used at home and 0 otherwise, and wine, where a value of 1 was assigned to those consuming one glass of wine per day, 0.5 to those consuming wine once per month to 6 times per week, and 0 to those who reported no intake or more than once per day. Thus, the possible MIND score ranged from 0 to 15 (with 15 as highest alignment to all components). ⁹ Dietary intake data collection began in February 2004 and has continued annually.

2.3.4. Social activity

The frequency of participation in social activities was assessed by asking how often during the past year participants engaged in six common activities involving social interaction. Activities included going to restaurants/sporting events/bingo, trips, community/volunteer work, visiting relatives or friends, group participation, or attending church/religious services. ⁴⁶ Participants rated each activity on a 5‐point scale (1 = once a year or less, 2 = several times a year, 3 = several times a month, 4 = several times a week, 5 = every day or nearly every day), and item scores were averaged to yield a total score ranging from 1 to 5, with higher scores indicating greater social activity.

2.4. Effect modifiers

Effect modifiers examined included sociodemographics (sex, years of formal education), presence of APOEε4 genotype (at least one allele), number of vascular disease risk factors (hypertension, diabetes, smoking history), number of vascular disease conditions (claudication, stroke, heart conditions, congestive heart failure), and a dichotomized global motor function score (low vs. high level, dichotomized using the median of our analytic sample), assessed at analytical baseline, except sex, education, and APOEε4, which were reported at enrollment. The composite measure of vascular disease risk factors was determined based on three self‐reported medical conditions: hypertension, diabetes, smoking history. For vascular disease conditions, the composite measure was determined based on four self‐reported medical conditions: claudication, stroke, heart conditions, and congestive heart failure. Initially, the global motor function score was determined based on 10 motor abilities related to muscle strength and motor performance, with higher scores indicating better performance. ⁴⁷

2.5. Cognitive function outcomes

Each annual clinical evaluation included a uniform medical history, neurologic examination, and detailed neuropsychological performance testing. ⁴² , ⁴⁸ A battery of 19 neuropsychological tests was administered annually and used to create a global composite measure of cognitive function in both cohorts. The tests evaluated five cognitive systems: episodic memory (Logical Memory Immediate Recall, Logical Memory Delayed Recall, Word List Memory, Word List Recall, Word List Recognition, East Boston Memory Test, East Boston Delayed Recall), semantic memory (Verbal Fluency, 15‐item Boston Naming Test, 15‐item Reading Test), working memory (Digit Span Forward, Digit Span Backward, Digit Ordering), perceptual speed (Number Comparison, Symbol Digit Test Modality, Stroop Word Reading, Stroop Color Naming), and visuospatial ability (Line Orientation, Progressive Matrices). To create a global cognition composite score, we converted raw scores on the 19 component tests across five cognitive systems to z‐scores using the baseline mean and standard deviation (SD) from the parent study. In secondary analyses, we examined the five cognitive systems separately (i.e., episodic memory, semantic memory, working memory, perceptual speed, visuospatial ability). Composite scores for each were calculated similarly as for the global composite. For all cognitive outcomes, higher scores indicated better cognitive function. In this study, analyses were based on cognitive measures collected from analytical baseline to the last follow‐up visit.

2.6. Statistical analyses

First, to assess lifestyle health behavior groups at analytical baseline, we conducted LPA, ⁴⁹ a statistical method for identifying unmeasured profile membership from a set of continuous variables, using the four continuous lifestyle health behaviors: weekly physical activity level, frequency of cognitively stimulating activity, MIND diet score, and social activity score. For a clear interpretation of which indicator values were above or below the sample means, we used the z‐standardized mean scale scores. Of note, the physical activity levels were log‐transformed to reduce skewness of the distribution. We fitted a one‐profile LPA model to the data, and then increased the number of profiles one at a time, until a predetermined maximum of seven profiles. Following the decision steps provided by Ram and Grimm, ⁵⁰ the final best resulting groups were selected based on LPA fit statistics including: the Bayesian Information Criterion (BIC) and the sample size‐adjusted BIC (SABIC), with lower values indicating improved fit; entropy to evaluate the confidence with which participants have been classified as belonging to one group or another; ⁵¹ and bootstrap likelihood ratio test to quantify specific comparisons between the model of interest and a model with one fewer class. ⁵² The decision on the final number of lifestyle behavior groups was determined by evaluating the number of participants assigned to each group and the theoretical coherence.

Second, to test the hypothesis that healthier lifestyle behaviors at analytical baseline were associated with better mean initial level of global cognition and slower annual rate of decline over time, we applied linear mixed‐effects models ⁵³ with the ability to handle intermittent missing data using the maximum likelihood methods and to consider measurement times and a total number of measures that can differ from one individual to another. The models included a linear function of time (in years since baseline) and a correlated individual random intercept and slope to capture inter‐individual variabilities—considering a linear function of time was reasonable because the follow‐up time was 4.3 years in this study.

The demographic‐adjusted model was adjusted for sex, education (continuous, years), and age (continuous, years) at analytical baseline. The fully adjusted model included additional terms for vascular disease risk factors (continuous), vascular disease conditions (continuous), and motor function (continuous) at analytical baseline. In secondary analyses, we explored associations of the lifestyle health behavior groups with each of the five cognitive systems.

Last, in sensitivity analyses, we examined whether the associations between the lifestyle groups and cognitive decline of global cognition were moderated by the modifiers of interest: sex, education, vascular disease risk factors, vascular disease conditions, motor function, and APOEε4 genotype. For each modifier, we applied a separate fully adjusted model, as specified above, further including a three‐way interaction between lifestyle groups, the modifier, and time.

Model assumptions were examined graphically and analytically and found to be adequately met. All statistical analyses were conducted using R software version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). We used the estimate_profiles function of tidyLPA R package version 1.0.8. for the LPAs ⁵⁴ and the hlme function of lcmm R package version 1.7.8. for the linear mixed models. ⁵⁵

3. RESULTS

3.1. Characteristics of study participants

The analytic sample included 715 MAP participants, predominantly women (75%), with a mean of 15 (SD = 3) years of education. The mean age was 81 (SD = 7) years at the analytical baseline. At the analytic baseline, half of the participants reported 2.8 (inter‐quartile range = [0.8‐4.5]) hours of physical activity per week, the mean cognitive activity score was 3.3 (SD = 0.6; possible scores range from 1 to 5), the mean MIND diet score was 7.9 (SD = 1.7; possible score 0 to 15), and the mean social activity score was 2.7 (SD = 0.6; possible score 1 to 5). Annual cognitive assessments spanned a mean of 4.3 (SD = 2.3, range = [1;8]) years during the follow‐up period, with a mean of four cognitive assessments (SD = 2, range = [2;8]).

As indicated in Figure 1, we identified three groups of engagement in lifestyle health behaviors at analytical baseline: generally high (n = 183), generally moderate (n = 441), and generally low (n = 91; details on the LPA fit statistics are provided in Table S1). When examining characteristics of participants according to their level of engagement (Table 1), those with higher engagement were more likely to be younger and to have more years of education, fewer vascular disease risk factors, higher scores of motor functioning, and higher levels of cognitive functions.

FIGURE 1.

Mean standardized values and 95% confidence intervals of lifestyle health behaviors in the three selected profiles of engagement at baseline (A) and individual observed trajectories of global cognition during follow‐up for 25 participants randomly selected in each profile (B), Memory and Aging Project.

TABLE 1.

Baseline characteristics according to the profiles of engagement in lifestyle health behaviors, Memory and Aging Project (N = 715).

	High engagement n = 183	Moderate engagement n = 441	Low engagement n = 91	p Value *
Demographics
Female, %	76	76	67	0.211
Age at baseline, years	79 (7)	81 (7)	81 (8)	0.002
Education, years	16 (3)	15 (3)	14 (3)	<0.0001
Follow‐up, years	4.3 (2.3)	4.4 (2.3)	3.8 (2.0)	0.054
No. of vascular disease risk factors †	0.9 (0.8)	1.0 (0.8)	1.2 (0.8)	0.029
No. of vascular disease conditions ‡	0.2 (0.5)	0.3 (0.6)	0.3 (0.6)	0.217
Motor functioning score	1.1 (0.19)	1.0 (0.2)	0.9 (0.2)	<0.0001
APOEε4 carriers (at least 1 allele), %	19	21	19	0.742
Lifestyle health behaviors
Physical activity (log‐transformed) §	1.5 (0.8)	0.8 (0.8)	0.1 (0.9)	<0.0001
Cognitive activity score (possible score 1–5)	3.5 (0.5)	3.4 (0.5)	2.3 (0.5)	<0.0001
MIND diet score (possible score 0–15)	10.0 (1.0)	7.3 (1.3)	6.5 (1.3)	<0.0001
Social activity score (possible score 1–5)	3.0 (0.5)	2.7 (0.5)	2.1 (0.5)	<0.0001
Cognitive function, z‐scores
Global cognition	0.42 (0.47)	0.28 (0.51)	−0.19 (0.71)	<0.0001
Episodic memory	0.42 (0.59)	0.26 (0.69)	−0.13 (0.95)	<0.0001
Semantic memory	0.50 (0.61)	0.30 (0.63)	−0.22 (1.02)	<0.0001
Perceptual orientation	0.47 (0.69)	0.33 (0.73)	−0.09 (0.89)	<0.0001
Perceptual speed	0.34 (0.67)	0.26 (0.67)	−0.41 (0.83)	<0.0001
Working memory	0.42 (0.63)	0.31 (0.76)	−0.13 (0.74)	<0.0001

Note. Values are expressed as mean (SD), unless otherwise specified. Boldface data indicate statistical significance.

Abbreviations: ANOVA, analysis of variance; APOEε4, apolipoprotein Eε4; MIND, Mediterranean‐DASH intervention for neurodegenerative delay; SD, standard deviation.

^*ANOVA test for quantitative variables, chi‐test for binary variables.

^†Hypertension, diabetes, smoking history.

^‡Claudication, stroke, heart conditions, congestive heart failure.

^§Physical activity levels were log‐transformed to reduce skewness of the distribution.

3.2. Association of engagement in lifestyle behaviors with global cognition

At analytical baseline, compared to participants with high engagement in healthy lifestyle behaviors, those with low engagement had worse initial mean levels of global cognition after adjusting for number of vascular disease risk factors, number of vascular disease conditions, and motor function score (estimated mean difference (MD) = −0.40 standard units, 95 percent confidence interval (95% CI) = [−0.52;−0.28], p < 0.0001) (Table 2). However, we did not find differences in the analytic baseline level of global cognition between the groups with moderate versus high engagement (all p > 0.423; Table 2).

TABLE 2.

Multivariable‐adjusted mean differences in initial level of global cognition and in annual rate of cognitive decline, according to the profiles of engagement in lifestyle health behaviors at baseline, Memory and Aging Project.

	Model 1 *		Model 2 †
Model terms	Mean difference (95% CI)	p Value	Mean difference (95% CI)	p Value
Mean difference in initial cognitive level
High engagement	Ref.		Ref.
Moderate engagement	‐0.03 (−0.12;0.05)	0.423	‐0.01 (−0.09;0.07)	0.797
Low engagement	‐0.45 (−0.58;−0.33)	<0.0001	‐0.40 (−0.52;−0.28)	<0.0001
Mean difference in annual rate of cognitive decline
High engagement	Ref.		Ref.
Moderate engagement	‐0.02 (−0.03;0.001)	0.065	‐0.02 (−0.03;−0.0002)	0.048
Low engagement	‐0.06 (−0.08;−0.03)	<0.0001	‐0.06 (−0.08;−0.03)	<0.0001

Note: Boldface data indicate statistical significance.

Abbreviation: 95% CI, 95 percent confidence interval.

^*Adjusted for sex, age at analytical baseline (continuous, in years), and education (continuous, in years).

^†Additionally adjusted for number of vascular disease risk factors (hypertension, diabetes, smoking history), number of vascular disease conditions (claudication, stroke, heart conditions, congestive heart failure), and motor function score (continuous).

During the follow‐up, we found that, compared to participants with high engagement in healthy lifestyle behaviors (mean estimated slope = −0.049), those with moderate and low engagement in healthy lifestyle behaviors had a steeper annual rate of decline in global cognition after full adjustment (moderate: estimated MD = −0.02 standard units per year, 95% CI = [−0.03;−0.0002], p = 0.048; low: estimated MD = −0.06 standard units per year, 95% CI = [−0.08;−0.03], p < 0.0001; (Table 2). As demonstrated in Figure 2, those with higher engagement in lifestyle health behaviors appeared to have consistently higher levels of global cognition at repeated timepoints during follow‐up.

FIGURE 2.

Estimated mean trajectories of global cognition among participants with high (n = 183), moderate (n = 441), and low (n = 91) engagement in healthy lifestyle behaviors at baseline, Memory and Aging Project. Trajectories were plotted for the most common profile of covariates in the study sample (i.e., female, 81 years of age at analytical baseline, 15 years of education, a score of 1 for vascular disease risk factors [average of 3 items], a score of 0.3 for vascular disease burden [average of 4 items], and a score of 1 for motor function). Shading represents the 95% confidence intervals.

When examining whether the associations between the lifestyle behavior groups and cognitive decline were moderated by sex, education, vascular risk factors, vascular conditions, motor function, or APOEε4 genotype, we found no evidence of effect modification (results not shown).

3.3. Association of engagement in lifestyle behaviors with different cognitive systems

When examining associations between the profiles of engagement in lifestyle behaviors with each cognitive system separately, we found that, compared to participants with high engagement, only those with low engagement had worse analytic baseline mean levels of cognition in each system after full adjustment (Table S2). In addition, we found that moderate engagement was associated with steeper cognitive decline in episodic memory and perceptual speed, and that low engagement was associated with steeper cognitive decline in episodic memory, semantic memory, and perceptual speed (Table S2).

4. DISCUSSION

In this longitudinal study using LPA, we identified profiles comprised of four lifestyle health behaviors (i.e., physical activity, cognitive activity, healthy diet, and social activity) in a cohort of older adults without dementia at analytical baseline from whom cognitive data were annually collected. The optimal solution from LPA yielded three lifestyle health behavior profiles: high engagement, moderate engagement, and low engagement in all four lifestyle behaviors. Contrary to expectations, there was little variation across the lifestyle health behaviors in each profile. Those who had high (or low) engagement in one lifestyle health behavior showed similar levels in the other behaviors.

Our three‐profile solution supports the notion that behaviors cumulatively contribute to a general healthy lifestyle, with potential benefits to cognitive function over time. Decades of literature have established that low engagement in one lifestyle health behavior often co‐occurs with low engagement with other lifestyle health behaviors. ⁵⁶ , ⁵⁷ This clustering of health behaviors has been shown to occur in multiple diverse populations, in the context of chronic disease prevention (e.g., cardiovascular disease), and across the lifespan, including in older populations. ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ In the earlier descriptive studies of lifestyle health behaviors and cognitive function that used summary scores of behaviors (i.e., index scores or “counts”), a greater number of lifestyle health behaviors was related to better cognitive function. ³² , ³³ , ³⁴ Likewise, a greater number of lifestyle health behaviors has been shown to increase the likelihood of healthy aging, including fewer chronic diseases and higher functional abilities. ⁶⁰ Clustering of lifestyle health behaviors can also occur in the context of behavioral intervention trials that target a single behavior. For instance, promoting a healthy diet in an intervention can have positive effects on other lifestyle health behaviors that were not targeted, such as physical activity. ⁵⁸ Yet, this “spillover effect” is inconsistent ⁶¹ and may not apply to all types of lifestyle health behaviors. Conversely, the relatively homogenous sample could also contribute to this lack of variation in levels of health behaviors across profiles.

We then examined the association of our three‐profile solution with the initial level of cognitive function and cognitive decline over time. Compared to participants with high engagement in lifestyle health behaviors, those with moderate and (separately) low engagement had steeper annual rates of cognitive decline over time, a finding consistent with the earlier studies. ³⁶ , ³⁷ However, the findings for the moderate group should be interpreted cautiously due to the borderline significant result. As such, the difference in cognitive trajectories between the moderate and high groups was much less pronounced compared to the difference between the low and high groups. Our results suggested that, instead of promoting high levels of lifestyle health behaviors only, promoting moderate levels in those with low engagement may have a greater clinical impact. This finding is particularly important to consider in the clinical setting because most older adults do not reach the accepted recommended levels of lifestyle health behaviors, especially within the higher age groups, when frailty and other challenges may be present. ⁶² , ⁶³ , ⁶⁴ Indeed, the majority of older adults are not able to engage in high levels of health behaviors for a variety of health, social, and environmental barriers. ⁶² , ⁶³ , ⁶⁵ , ⁶⁶ , ⁶⁷ Thus, it is important to consider alternative strategies that may still offer clinical benefits. A more approachable and feasible strategy may be to emphasize that small changes to daily engagement of lifestyle health behaviors that can reach the moderate level may be of benefit to cognition. ⁶⁸ , ⁶⁹ , ⁷⁰ , ⁷¹ This is highly preferable and realistic for most older adult populations who experience barriers to engagement and fosters a more long‐term, sustainable behavior change and thus greater long‐term effects on cognitive health. ²⁸ , ²⁹ , ³⁰ , ³¹ , ⁷²

Our analysis did not yield any significant effect modifiers in our study, including vascular disease risk factors, vascular disease conditions, APOEε4, and motor function, despite the known influence of these factors on cognitive decline. ³⁹ , ⁴⁰ , ⁴¹ This unexpected finding may be due to limitations with our study sample, which primarily represented those who are of white racial background, higher levels of education, and generally healthy. Additionally, there may be other important effect modifiers that were not examined, such as other health problems, environmental or neighborhood characteristics, and other social determinants of health. Future analyses are needed in diverse populations to explore additional factors that may influence how lifestyle health behavior profiles impact rates of cognitive decline in aging.

Our study had limitations. First, this analysis used data from older adult research volunteers from the MAP, who were largely female, of white racial background, and had an education level higher than the average in the United States, which was similar to the sample of earlier longitudinal studies that used latent profile/class analytic methods. ³² , ³³ , ³⁴ Moreover, our mean study sample was 81 years at analytic baseline, representing an older age group. This limitation in sample representation reduces our ability to generalize to other populations. To address this limitation in diversity, equity, and inclusion, additional studies are ongoing using similar methods and measures in diverse cohorts including at the Rush Alzheimer's Disease Center (e.g., the Minority Aging Research Study; the Latino Core). ⁷³ , ⁷⁴ Second, self‐report questionnaires were used to assess the four lifestyle health behaviors (i.e., physical activity, cognitive activity, healthy diet, and social activity). Though they have been previously validated in this sample, the self‐reported measures of lifestyle health behaviors may not have provided accurate representations of engagement and may result in bias toward reporting generally similar levels across multiple behaviors. It is possible that the general consistency across levels of all four behaviors may be an artifact of the self‐report measures. Additionally, the frequencies of some of the behaviors may be overestimated compared to data derived from more objective assessments. For instance, physical activity is often overestimated when assessed by self‐report questionnaires compared to objective measures of physical activity (e.g., accelerometer). ⁷⁵ We also examined engagement in the lifestyle health behaviors in the late‐life period only, which does not necessarily reflect exposure to and engagement in lifestyle health behaviors earlier in life (e.g., young adult or midlife periods) that can impact cognitive function over time. Third, our lifestyle health behaviors examined were not exhaustive, and we did not include smoking and alcohol consumption. In this sample, few participants reported being a smoker or overusing alcohol, so these questions could not be adequately addressed in our study. We focused on physical activity, cognitive activity, healthy diet, and social activity because they were assessed annually in the parent study and could be promoted or increased in intervention or public health settings, which may be more effective for promoting an overall healthy lifestyle. ¹⁶ Finally, we did not include confounding effects of income, socioeconomic status, and other important social determinants of health known to impact health.

Despite these limitations, our study has notable strengths, including that this study examined the association of lifestyle health behavior profiles with cognitive function over time using LPA. Unlike earlier studies, our study focused on behaviors can be easily integrated into daily life and be promoted in the context of an accessible and scalable behavioral intervention, including physical activity, cognitive activity, healthy diet, and social activity. An additional strength was our use of LPA methods to focus on individual variation in behaviors, versus a “one‐size‐fits‐all” approach that is more common. Our findings may suggest that moderate level of engagement in the four lifestyle health behaviors over time could result in less cognitive decline, similar to the rates that older adults with high levels of engagement showed. Future research should adapt, enhance, and tailor existing evidence‐based interventions to promote lifestyle health behaviors in older adults who experience low engagement. Moreover, there is an evident need to conduct additional research to investigate lifestyle health behavior profiles in diverse populations, considering the important effects of social determinants of health and engagement in lifestyle health behaviors over the lifespan.

AUTHOR CONTRIBUTIONS

The authors assume full responsibility for analyses and interpretation of these data. Drs. Halloway, Wagner, Schoeny, Arvanitakis, Tangney, and Lange‐Maia were involved in study conception and design, analysis, and interpretation of the data, drafting of the manuscript, and critically revised the manuscript. Dr Wagner conducted the analysis. Dr Bennett critically revised the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts to disclose. Author disclosures are available in the supporting information.

CONSENT STATEMENT

The Institutional Review Board of the Rush University Medical Center approved the study protocols, and all participants provided written consent for evaluations.

Supporting information

Supporting information

Supporting information

Halloway S, Wagner M, Tangney C, et al. Profiles of lifestyle health behaviors and cognitive decline in older adults. Alzheimer's Dement. 2024;20:472–482. 10.1002/alz.13459