Abstract
Physical activity is a crucial modifiable risk factor for preventing cognitive decline, stroke, and dementia. A new cardiovascular disease diagnosis offers an opportunity to improve health outcomes through lifestyle changes. However, sustaining behavior change, even with effective exercise interventions, remains challenging. Understanding the neurobehavioral mechanisms behind successful long-term behavior change can help tailor therapeutic strategies to enhance physical activity and reduce dementia and stroke risks in older adults. This study followed 295 cognitively unimpaired older adults from the UK-Biobank cohort, all newly diagnosed with cardiovascular disease, over five years. We aimed to predict future changes in moderate-to-vigorous physical activity (MVPA) engagement using baseline resting-state functional connectivity, along with sociodemographic factors (e.g., age, sex, socioeconomic status), behavioral characteristics (e.g., general health, pain, depression), cognitive function (e.g., attention, fluid intelligence), social factors (e.g., networks and support), and environmental context (e.g., access to green spaces). Our findings revealed that greenspace percentage, social support, and retirement status significantly predicted increased MVPA at follow-up. A support vector machine (SVM) model, utilizing baseline functional connectivity matrices, achieved 84% accuracy in predicting successful MVPA engagement at follow-up. Key features contributing to the model’s accuracy included functional connectivity between brain regions located in distinct large-scale networks, particularly the heteromodal cortex. Notably, the default mode network, frontoparietal control network, and salience/ventral attention network were prominently involved. These results underscore the importance of social and structural determinants of health, such as greenspace and social support, in promoting physical activity behavior change. Furthermore, they provide preliminary evidence of neural mechanisms that may underlie successful lifestyle modifications, informing future interventions to prevent vascular contributions to late-life cognitive decline.
1. Introduction
1.1. Preventing Vascular Contributions to Dementia
Cardiovascular diseases (CVD) substantially elevate the risk of both dementia and stroke due to shared pathophysiological mechanisms in the heart–brain axis (Saeed et al., 2023). Indeed, mixed dementia comprised of combined vascular and Alzheimer pathological changes is the most prevalent dementia etiology (Korczyn, 2002). With the global dementia burden projected to rise to 132 million by 2050 (World Health Organization, 2017), there is an urgent need for targeted strategies to mitigate the vascular contributions to late-life cognitive decline. Physical activity is highly effective in lowering dementia risk among individuals with cardiovascular risk factors (Dao et al., 2024; Gorelick et al., 2011), reducing cognitive impairment, vascular dementia, and all-cause mortality. Thus, integrating physical activity into treatment and prevention strategies for vascular-related cognitive decline, especially following a CVD diagnosis, is essential (Guadagni et al., 2020; Aarsland et al., 2010; Grodstein, 2007; Verdelho et al., 2012; Landman et al., 2021).
1.2. The Challenge of Successful Physical Activity Behaviour Change
Despite the well-established benefits of physical activity, physical inactivity remains prevalent, with approximately 27.5% of the global population not meeting recommended activity levels (Guthold et al., 2018). In the United States, inactivity increases with age, affecting 59.8% of middle-aged adults (ages 45–64) and 65.7% of older adults (65 and older) (Xu et al., 2023). Alarmingly, physical inactivity has escalated since the COVID-19 pandemic, particularly among older adults with chronic conditions who face significant barriers, including mobility limitations and socioeconomic challenges (Peçanha et al., 2020; Ng et al., 2022). Moreover, maintaining long-term behavior change is difficult. Observational studies report that only 25–40% of individuals sustain lifestyle modifications six months after a cardiovascular event, with adherence rates dropping to 3–11% after five years. Understanding why and when individuals engage in initiation or cessation of physical activity is crucial for designing effective interventions.
1.3. Critical Windows for Behaviour Change
The critical windows theory suggests successful behaviour change may be facilitated by an external threat from a major life event or circumstance (e.g., receiving a diagnosis of a new chronic illness such as a cardiovascular disease, retirement, pregnancy, or menopause), which might catalyze the reassessment of goals and increase motivation for change (Lane-Cordova et al., 2022). For example, individuals with chronic conditions—whether newly diagnosed, recovering, or managing ongoing illnesses—are often more likely to maintain or increase their leisure-time PA levels (Dai et al., 2014). Thus, life transitions can serve as critical moments for intervention, offering a window to promote long-term physical activity engagement.
1.4. Potential Neurobehavioral Mechanisms of Physical Activity Engagement
Factors such as self-efficacy (Jones et al., 2020), self-regulation (Williams et al., 2018), perceived benefits and biological sex—where males generally show higher adherence rates than females (Johnson and Lee, 2017)—play significant roles in regulating physical activity behaviors. Psychological factors, including depression, fatigue, and health perceptions, also influence adherence (Brown et al., 2016). Additionally, cognitive functions, particularly executive functions linked to cortical structural integrity, have been tied to exercise behaviors (Garcia et al., 2016; Patel et al., 2017).
Beyond individual-level factors, social and structural determinants of health significantly influence physical activity behavior (Carroll-Scott et al., 2013; Wen et al., 2007). Access to greenspace and neighborhood walkability (Richardson et al., 2013), social support(Ho et al., 2018), socioeconomic status (Wen et al., 2007) is strongly associated with physical activity levels. The environmental conditions in which individuals are born, live, learn, work, play, and age— referred to as structural and social determinants of health—have a cumulative impact on physical, mental, and brain health (Gómez et al., 2021a). These determinants not only shape physical activity behavior but also contribute to disparities in health outcomes, including Alzheimer’s disease and related dementias.
Despite these insights, many open questions remain regarding the mechanisms driving successful physical activity engagement. As noted by Cohen and Sherman (2014), a unified pathway that explains successful behaviour change has yet to be clearly identified. Neuroimaging provides insights into individual differences in brain organization and highlights neurodiversity— how brain functions vary across individuals based on age, sex, personality, culture, and genetics (Gabrieli et al., 2015). Baseline functional connectivity biomarkers have been found to predict successful engagement in healthy lifestyle habits (Ai et al., 2023), offering a promising avenue for personalized intervention strategies.
1.5. The Present Study
To better understand the drivers of sustained physical activity, this current study adopts a precision medicine framework combined with a whole-brain, data-driven machine learning approach. Specifically, we will examine the roles of sociodemographic factors (e.g., age, sex, socioeconomic status), behavioral characteristics (e.g., general health, pain, depression), cognitive function (e.g., attention, fluid intelligence), social factors (e.g., networks and support), environmental context (e.g., access to green spaces), and resting-state functional connectivity biomarkers. This comprehensive approach is designed to uncover tangible targets for future interventions tailored to individual needs, especially for those at heightened risk of cognitive decline. By employing a rigorous data-driven machine-learning (ML) approach, the current study aims to move beyond group-level differences to uncover the neurobehavioral mechanisms driving successful physical activity at the individual level (Gabrieli et al., 2015a).
2. Methods
2.1 Participants
295 (mean age = 63.13 years ± 7.5, 188 women) cognitively unimpaired and physically inactive older adults from the UK Biobank, a large-scale population-based longitudinal cohort were included in this study. Inclusion criteria were: 1) cognitively unimpaired at enrollment; 2) reported a new cardiovascular diagnosis (i.e., hypertension, type II diabetes, dyslipidemia, cardiac angina or myocardial infarction) between baseline (T1; 2014) and follow-up five years later (T2; 2019) (mean duration 5.2 years, SD 1.1); 3) did not meet the World Health Organization recommendation of 150 minutes/week of moderate-to-vigorous physical activity (MVPA) at baseline (Bull et al., 2020); and 4) age >= 60. These criteria yielded a final sample size of 295 after removing four participants for having poor quality brain imaging data. Unimpaired cognition was defined as follows: performance scores on each of the cognitive test were converted into percentile rank, and the raw score corresponding to the 5th percentile (or 95th, on tests where higher scores represented worse performance) was identified as the cut-off for impairment (Lyall et al., 2016). An illustration of the study timeline can be seen in Fig. 1. The brain imaging visit (Instance 2 of the UK Biobank) was considered baseline timepoint, and the first repeat imaging visit (Instance 3 of the UK Biobank) was considered the follow-up timepoint. Self-reported MVPA was recorded using the Lifetime Total Physical Activity Questionnaire (Friedenreich et al., 1998). Demographic variables including age, sex, years of education, and socioeconomic status (as measured through Townsend deprivation index)(Townsend et al., 2023) were included as covariates of non-interest. MRI data was obtained at baseline, before participants had received a new cardiovascular diagnosis, and MVPA data and cognitive function data were obtained for two time-points: at baseline and in follow up after 5 years (mean duration 4.2 years, SD 2.1; ranging from 8 months to 4.8 years). Below is the specific information about the physical activity, psychosocial, cognitive, and environmental measures that were used in the current study.
Instance 1 (Imaging) of the UK Biobank represents the baseline timepoint, while Instance 3 (repeat imaging) marks the follow-up. Baseline assessments included self-reported MVPA, cognitive function, psychosocial and behavioral factors (such as depression, anxiety, general pain, retirement), social support, and greenspace exposure, all of which were evaluated at both baseline and follow-up. rs-fMRI was assessed only at baseline. New cardiovascular diagnoses was reported between baseline and follow-up, with no cardiovascular diagnoses reported prior to baseline. MVPA: moderate-to-vigorous physical activity; rs-fMRI: resting-state functional Magnetic Resonance Imaging.
2.2 Behavioral measures and data analysis
Participants completed a comprehensive battery of psychosocial, behavioral, cognitive, and environmental assessments at both baseline and follow-up (Fig. 1). These assessments are briefly outlined below. To investigate the relationship between physical activity and cognition, we first examined whether baseline cognitive function predicted future physical activity behavior. We then assessed whether increases in physical activity at follow-up were linked to cognitive gains. We also evaluated whether social and structural health determinants at baseline predicted successful engagement in physical activity. Resting-state fMRI (rs-fMRI) was assessed at baseline only, prior to any cardiovascular diagnoses, reducing the likelihood of confounding effects related to blood flow alterations, given that BOLD signals are susceptible to cardiovascular disease (CVD)-related changes (Makedonov et al., 2013; Tsvetanov et al., 2021). Outliers were identified using Cook’s distance with a threshold of 0.5, resulting in the detection of 30 outliers.
2.2.1 Townsend deprivation index
The Townsend Deprivation Score is an area-based score of social deprivation that was determined immediately prior to the participant joining the Biobank, based on data from the preceding national census (Woodward et al., 2021). Each participant was assigned a score corresponding to their postcode area.
2.2.2 The lifetime total physical activity questionnaire
The questionnaire developed and validated by Friedenreich et al. (1998) was utilized to record self-reported MVPA. Participants reported both the frequency and duration of each type of physical activity they engaged in per week. The total time spent on moderate and vigorous activities was then calculated to derive the overall MVPA in minutes per week. This total score served as an indicator of each individual’s physical activity engagement. The scale was administered at both baseline and follow-up timepoints to assess changes over time.
2.2.3 Cognitive assessments
A computerized cognitive battery was administered using a touchscreen tablet. The tests were specifically developed for the UK Biobank and have been validated (Fawns-Ritchie & Deary, 2020; Lyall et al., 2016), while sharing features with established cognitive assessments. The battery included the following tasks: Reaction time, Numeric memory, Fluid intelligence, Matrix pattern completion, Tower rearranging, and Trail making. A detailed description about these tasks can be found in Supplementary materials.
2.2.4 Social support
The measures available in the UK-Biobank for social support come from the items “How often do you visit friends or family or have them visit you?” and “How often are you able to confide in someone close to you?”. Participants rated each item on a Likert scale from 0 (Never or almost never) to 6 (Almost daily). For the frequency of visits, the categories “never or almost never” and “no friends or family outside the household” were combined into a single category, “never.” This adjustment was made because these responses were similar, and there were only a few participants with no friends or family outside the household (n = 16). Scores ranged from 0 to 6 and were treated as a continuous measure.
2.2.5 Greenspace and coastal proximity assessment
environmental indicators included in this study were the proportion of green space and water within 300 m of residential addresses, using the 2005 Generalised Land Use Database for England and Centre for Ecology and Hydrology 2007 Land Cover Map data for Great Britain(Smucny et al., 2022). The buffer size of 300m was decided based on relevant health evidence and public policy on both density and accessibility. Coastal proximity was estimated using Euclidean distance raster (Wheeler et al., 2012).
2.2.6 Psychosocial factors
Psychosocial factors were assessed through self-reported experiences, including depression, anxiety, general pain, and lifestyle factors such as retirement status. Depression was evaluated using two items: “Feeling down, depressed, or hopeless” and “Little interest or pleasure in doing things.” Participants rated their experiences on a four-point scale, ranging from 0 (Not at all) to 4 (Nearly every day). Anxiety was assessed similarly, with two items: “Feeling nervous, anxious, or on edge” and “Not being able to stop or control worrying.” General pain was measured using a single item: “Have you had pains all over your body for more than 3 months?” Responses were recorded as yes (1) or no (0). Participants also rated their overall health perception on a scale from 1 (Excellent) to 4 (Poor). Additionally, participants indicated their retirement status with a simple yes (1) or no (0) response. These assessments were conducted at both baseline and follow-up timepoints.
2.2 MRI Data Acquisition
Details of image acquisition and processing are available in the UK Biobank Protocol (http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=2367), and Brain Imaging Documentation (http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=1977). Briefly, all brain MRI data were acquired on a Siemens Skyra 3T scanner with a standard Siemens 32-channel RF receiver head coil, using the following parameters: TR = 2000 ms; TI = 800 ms; R = 2; FOV = 208 × 256 × 256 mm; voxel size = 1 × 1 × 1 mm. For resting-state fMRI scans, two consecutive functional T2*-weighted runs were collected with eyes closed using a blood oxygen level dependent (BOLD) sensitive, single-shot echo planar imaging (EPI) sequence with the following parameters: TR = 735 ms; TE = 39 ms; flip angle = 52°; FOV 88 x 88 x 64 matrix; resolution = 2.4 × 2.4 × 2.4 mm; 490 volumes; and acquisition time = 6 minutes per run.
2.3 Resting-State Functional MRI Data Preprocessing
Preprocessing of raw functional images from the UK Biobank was done using the fMRIprep pipeline (version 20.2.4) (Esteban et al., 2019). For each of the BOLD runs per participant, the following preprocessing was performed: First, the T1w reference was skull-stripped using a Nipype implementation of the antsBrainExtraction.sh (ANTs) tool. A B0-nonuniformity map (or fieldmap) was estimated based on a phase-difference map calculated with a dual-echo gradient-recall echo (GRE) sequence, which was then co-registered to the target EPI reference run and converted to a displacements field map. A distortion-corrected BOLD EPI reference image was constructed and registered to the T1-weighted reference using a boundary-based approach (using bbregister, Freesurfer). Rigid-body head-motion parameters with respect to the BOLD EPI reference were estimated (using mcflirt, FSL 5.0.9) (Jenkinson et al., 2002) before spatiotemporal filtering was performed. BOLD runs belonging to the single band acquisition sessions were slice-time corrected (using 3dTshift, AFNI 20160207). The BOLD time series were resampled into their original, native space by applying a single, composite transform to correct for scan-to-scan head motion and susceptibility distortions. Functional scans were spatially smoothed using a 6 mm full width at half maximum (FWHM) Gaussian smoothing kernel.
Additional preprocessing steps were undertaken to remove physiological, subject motion, and outlier-related artifacts, which were implemented using the nilearn package. Non-neuronal sources of noise from white matter and CSF were estimated and removed using the anatomical CompCor method (aCompCor) (Behzadi et al., 2007) to allow for valid identification of correlated and anticorrelated networks (Chai et al., 2012; Murphy et al., 2009). Temporal band-pass filtering (0.008–0.09 Hz) was then applied. Additionally, scan-to-scan mean head motion (framewise displacement) was used as a covariate of non-interest in all second-level analyses (mean head motion = 0.2 mm, SD = 0.1 mm). Head motion is a known important potential confound as it produces systematic and spurious patterns in connectivity and is accentuated in Alzheimer’s disease (AD) and cognitively typical aging populations (Power et al., 2012). Critically, we did not identify a relationship between the mean head motion parameter and the primary behavioral variable of interest, physical activity change (all p > 0.05). The framewise displacement timeseries was determined by calculating the maximum shift in the position of six control points situated at the center of a bounding box around the brain, computed independently for each scan. Four participants were removed from the UK-Biobank sample final analysis for having >30 scan volumes flagged, leading to the final sample size of 295 participants. This cut off was determined based on preserving at least 5 minutes of scanning time (Van Dijk et al., 2010).
2.4 Machine Learning Modelling
To predict future successful physical activity behavior change following a new cardiovascular diagnosis with baseline RSFC, we used the support vector machine (SVM) algorithm from the scikit-learn (v0.21.3) library, utilizing the pydra-ml (v0.3.1) toolbox. This algorithm aims to find the optimal hyperplane that maximizes the margin between classes. It can handle non-linear relationships by using kernel functions. We used the default parameters, which include the Radial Basis Function kernel to capture non-linearities in the data.
Amid ongoing debates about the most accurate machine learning algorithms for prediction, recent comparisons suggest that a suite of approaches may be clinically beneficial. To assess the robustness of our findings, we repeated our analysis using additional machine learning algorithms of increasing complexity, defined by the computational resources required for model simulation. Specifically, we examined linear regression, random forest, and multi-layer perceptron algorithms, using default parameters unless otherwise specified. Further details on these algorithms are available in the supplementary material.
To generate independent test and train data splits, we used a bootstrapped group shuffle split sampling scheme. For each iteration of bootstrapping, a random selection of 20% of the participants, balanced between the two groups, was designated as the held-out test set. The remaining 80% of participants were used for training. This process was repeated 50 times, fitting and testing the four classifiers for each test/train split. To reduce computational time, we used the default of 50 bootstrapping splits from pydra-ml toolbox. The Area Under the Receiver Operating Characteristic Curve (ROC AUC; perfect classification = 1; chance = 0.5) was calculated to evaluate model performance on each bootstrapping iteration, resulting in a distribution of 50 ROC AUC scores for each classifier. The model’s performance was statistically compared to its null model’s distribution using an empirical p-value, which is a common and effective measure for evaluating classifier performance. The significance level was set to alpha = 0.05. We employed Kernel SHAP (SHapley Additive exPlanations) (Goodwin et al., 2022) to assess the significance of baseline RSFC features in predicting successful engagement in physical activity We computed the average absolute SHAP values across all predictions, weighted by the model’s median performance, and calculated mean SHAP values across splits for each model. This entire pipeline, encompassing machine learning models, bootstrapping, and SHAP analysis, was implemented using pydra-ml toolbox.
2.6 Reducing collinearity using Independence Factor to enhance model interpretability
Collinearity among features can significantly affect model generation and interpretation, particularly in RSFC analyses. To address this, we employed the Independence Factor method (Low et al., 2024), which iteratively removes features with strong dependence above a set threshold, ensuring a consistent set of features across models. Using distance correlation, which accommodates non-monotonic relationships, we systematically increased the threshold to eliminate redundant features while preserving model performance within a narrow margin. We applied thresholds ranging from 1.0 (keeping all features) to 0.2 (removing features with distance correlation above 0.2). Our goal was to identify a feature set that maintained model performance within three percentage points of using all features, resulting in a more parsimonious and interpretable model without compromising accuracy, essential for clinical applicability.
2.7 Performance using most important and least important features
To address the question of why certain features are important, we evaluated model performance under two scenarios: one using only the top 20 features and another excluding these features. This method mitigates the common pitfall in brain-behavior prediction analyses, where the significance of the top features may not reflect their true impact on model performance. By comparing performance metrics in both scenarios, we can gain a more nuanced understanding of the highlighted features’ contributions and derive mechanistic insights into the neural correlates of successful behavior change.
3. Results
3.1 Behavioral Results
There was a positive behavior change observed, with the sample maintaining increased physical activity in long-term follow-up after a new cardiovascular diagnosis (mean change in MVPA = 4.02 min/week ± 1.23). Notably, 160 individuals exceeded the threshold of 150 min/week of MVPA, indicating successful behavior change at follow-up (mean change in MVPA among those who exceeded the threshold = 7.10 min/week ± 2.1). However, no significant association was found between baseline cognition in any domain (i.e., reaction time, numeric memory, fluid intelligence, trail making, symbol digit substitution, tower rearranging, prospective memory, pairs matching) and change in MVPA. Similarly, change in cognition (i.e., the difference between cognitive function scores at follow-up and baseline) was not associated with physical activity gains.
Next, we investigated the relationship between behavioral factors and their predictive value for future engagement in physical activity, specifically changes in moderate-to-vigorous physical activity (MVPA). Spearman correlation analyses were conducted to assess associations between each behavioral variable and change in MVPA. Our analysis revealed a significant positive correlation between the percentage of greenspace in participants’ environments and increased MVPA at follow-up relative to baseline (r = 0.4, p = 0.01, FDR-corrected). In contrast, coastal proximity did not demonstrate a significant relationship with MVPA. Social support, operationalized as the frequency of visits with friends or family, was also significantly positively correlated with increased MVPA (r = 0.3, p = 0.01, FDR-corrected). However, no significant association was observed between MVPA and the ability to confide in others, another measure of social support. Retirement status was significantly negatively associated with MVPA, indicating that retired individuals exhibited a decrease in MVPA at follow-up compared to baseline (r = - 0.32, p = 0.01, FDR-corrected). No significant correlations were observed between changes in MVPA and other variables, including depression, anxiety, general pain, or overall health ratings.
Importantly, separate analyses for males and females revealed no significant sex differences in the relationship with MVPA. Likewise, the Townsend Deprivation Index was not significantly associated with changes in MVPA.
3.2 Neuroimaging machine learning results
In Table 2, we present the performance results of various feature selection scenarios for the SVM model: using all features, removing redundant features, selecting only the top 20 features to assess their unique contributions, and excluding the top 20 features to evaluate their necessity for high performance. The SVM model consistently demonstrated strong performance across all scenarios. To mitigate the influence of dependent features that can provide similar information and distort feature importance analyses, we employed the Independence Factor method to eliminate redundant features. We determined that a feature set size of 250 yielded the best performance with the fewest features for subsequent analyses. We assessed model performance after removing highly collinear features from the original 400. The bootstrapped ROC AUC distributions and permutation tests for the model using the non-redundant 250 feature set are presented in Supplemental Figure 2. After applying the Benjamini-Hochberg procedure to correct for multiple comparisons, we found that the distributions of the SVM model was significantly different from its respective null distributions. We conducted similar analyses with other machine learning algorithms, and their performance results are displayed in the supplemental figure 1.
SD = Standard Deviation.
Median ROC AUC score (90% confidence interval). SVM: Support Vector Machine.
3.3 Top functional connections predicting behavioral change
Figure 2 shows the most informative functional connections in the prediction model (see also Table 3 in supplementary table 2). These are predominantly between large scale networks and within the left hemisphere. Notably, nodes with higher importance are primarily located within the default mode network, the frontoparietal control network, and the salience ventral attention network, as shown in Figure 2b.
(a) Anatomical depiction of significant features and their corresponding importance values. Nodes are defined as functionally distinct brain regions, while edges are the functional connections between the nodes. Node size depicts the frequency of that brain region on the list of predictive features, while edge thickness represents connection importance. Purple signifies positive RSFC connectivity whereas grey signifies negative RSFC connectivity associated with enhance physical activity gains at follow up compared to baseline. (b) Distribution of nodes from each intrinsic resting-state network (i.e., Yeo 7 networks) in the roster of important features.
Enhanced positive RSFC within the DMN was linked to increased physical activity engagement at follow-up compared to baseline. Moreover, successful behavior change at follow-up was associated with increased positive RSFC between the DMN and the frontoparietal control network. Additionally, increased antagonistic RSFC between frontoparietal control network and limbic network was also correlated with successful behavior change.
4. Discussion
In this study, we systematically evaluated whether RSFC, psychosocial factors, and social and structural determinants of health could predict successful long-term engagement in physical activity among older adults newly diagnosed with cardiovascular disease. Our behavioral analysis underscored the importance of structural and social determinants, particularly the percentage of greenspace in one’s environment and levels of social support, as key predictors of positive changes in physical activity. Additionally, lifestyle factors such as retirement status were also found to significantly influence physical activity outcomes. Moreover, we identified functional connectivity features that were strongly predictive of sustained increases in MVPA activity following a cardiovascular diagnosis. These predictive features primarily reflected between-network functional connectivity, specifically involving the heteromodal cortex, highlighting the role of higher-order brain networks in driving behavior change.
Our study builds on the evidence underlying the impact social determinants of health might have on age-related brain health outcomes; for example, the influence of upstream factors on downstream protective behaviours such as physical activity engagement. A growing body of work suggests that greenspace and social support are promising avenues to enhance physical activity. Proximity to greenspace was associated with higher levels of walking and lower rates of circulatory and all-cause mortality rates (Giles-Corti et al., 2005; Mitchell and Popham, 2008). Similarly, high social support from friends and family was significantly associated with increased and sustained MVPA (Kouvonen et al, 2005; Eyler et al., 2002). Our findings corroborate these studies, linking greenspace percentage and social support to increased MVPA. However, our results suggest that not all components of social support have the same influence; specifically, the frequency of visits from family and friends was a strong predictor of behavior change, while the ability to confide in others was not. This suggests a complex relationship between social support and physical activity that requires further exploration. Contrary to previous studies suggesting that retirement leads to increased leisure-time physical activity (Evenson et al., 2002), our results indicated that retirement was associated with decreased MVPA over the five-year follow-up. This discrepancy highlights the importance of assessing physical activity across various domains beyond leisure time and suggests that retirement may not uniformly lead to increased physical activity.
We identified neural markers that predict successful engagement in physical activity among older adults after being diagnosed with a cardiovascular risk factor. Notably, the most frequent features involved heteromodal cortex. Nodes in the temporal and prefrontal regions of the DMN and FPN were most important in predicting successful long-term physical activity engagement. Our findings support the default-executive coupling hypothesis of aging, which suggests that as goal-directed cognition relies less on declining control resources and more on accumulated knowledge (semanticization of cognition), the DMN becomes more closely coupled with the FPN (Spreng & Turner, 2019). This coupling increases as prior knowledge becomes more relevant to goals, providing an advantage to older adults. The shift from exploration to exploitation in cognitive strategies with age is associated with enhanced functional connectivity between the FPN and the DMN. A possible reason why a new CVD diagnosis could serve as a critical moment for behavior change is that it may prompt older adults to rely more on semantic knowledge, shifting their goal structures toward emotionally meaningful objectives that offer immediate personal satisfaction.
The present findings replicate and substantially extend previous findings linking prefrontal cortex and behaviour change. Prefrontal cortex, including ventromedial and dorsomedial prefrontal cortex, play crucial roles in self-referential processing and perceived social support, respectively (Jenkinson et al., 2002; Macrae, 2004; D’Argembeau et al., 2010; Bechara & Damasio, 2002; Eslinger & Damasio, 1985; Amodio & Frith, 2006; Waytz et al., 2012; Bzdok et al., 2012; Schuwerk et al., 2014) Our findings underscore the importance of the prefrontal cortex, showing its engagement across multiple networks, including the default mode network (DMN), salience network, and control networks.
We observed that baseline RSFC features predictive of successful long-term physical activity behavior change are primarily left-lateralized. One potential explanation could be that hemispheric specialization of the left lateralization contribution to resistance to interference, a core aspect of cognitive control. Past studies have found that ability to resolve Stroop interference localized to the left prefrontal cortex (Geddes et al., 2014; Gläscher et al., 2012; Tsuchida and Fellows, 2013).
There are several limitations to this study. First, we relied on self-reported measures of MVPA. These self-reports should be interpreted with caution due to potential reverse causation effects. Second, objective measures such as accelerometers can differentiate between sedentary behavior, light activity, and moderate/vigorous activity, and can also provide physiological metrics for estimating cardiorespiratory fitness. Differentiating between physical activity and cardiorespiratory fitness could be important for understanding their distinct impacts on cognitive and brain health. Finally, the correlational nature of functional connectivity analyses prevents us from determining causality in the brain behavior relationship identified (Siddiqi et al., 2021; Silvanto & Pascual-Leone, 2012; Vaidya et al., 2019).
Despite these limitations, our study has several notable strengths. It represents the most comprehensive assessment of neurobehavioral factors predicting successful longer-term physical behavior change following a new cardiovascular diagnosis in aging. This study highlights the importance of going beyond individual-level factors and including structural factors such as greenspace and social support to promote physical activity behaviour change. Future research could also investigate differences related to aging and apply similar analyses to younger adults to enhance understanding of how these factors vary across age groups.
4. Conclusion
This study demonstrated that individual differences in the baseline RSFC, social and structural determinants of health successfully predict future engagement in physical activity among older adults that are newly diagnosed with cardiovascular diagnosis. Leveraging mechanistic predictors of future physical activity and adopting a precision medicine framework will potentially lead to targeted interventions that result in sustained behavioural change.
Data Availability
All data produced in the present study are available upon reasonable request to the authors
Acknowledgements
This research was undertaken thanks in part to funding from a National Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant (DGECR-2022-00299), an NSERC Early Career Researcher Supplement (RGPIN-2022-04496), a Fonds de Recherche Santé Québec (FRSQ) Salary Award, the Canada Brain Research Fund (CBRF), an innovative arrangement between the Government of Canada (through Health Canada) and Brain Canada Foundation, a Brain Canada Future Leaders Award, an Alzheimer Society Research Program (ASRP) New Investigator Grant, the Canadian Institutes of Health Research, the Canada First Research Excellence Fund, awarded through the Healthy Brains, Healthy Lives initiative at McGill University, and the National Institutes of Health (P30 AG048785) to MRG, and the Consortium pour l’Identification précoce de la Maladie d’Alzheimer-Québec (CIMA-Q) awarded to NT.
References
