Determinants of Exercise Adherence in Sedentary Middle-aged and Older Adults

Abstract

Regular exercise is known to positively impact neurocognitive health, particularly in aging individuals. However, low adherence, particularly among older adults, hinders the adoption of exercise routines. While brain plasticity mechanisms largely support the cognitive benefits of exercise, the link between physiological and behavioral factors influencing exercise adherence remains unclear. This study aimed to explore this association in sedentary middle-aged and older adults. Thirty-one participants underwent an evaluation of cortico-motor plasticity using transcranial magnetic stimulation (TMS) to measure changes in motor-evoked potentials following intermittent theta-burst stimulation (iTBS). Health history, cardiorespiratory fitness, and exercise-related behavioral factors were also assessed. The participants engaged in a 2-month supervised aerobic exercise program, attending sessions three times a week for 60 minutes each, totaling 24 sessions at a moderate-to-vigorous intensity. They were divided into Completers (n=19), who attended all sessions, and Dropouts (n=12), who withdrew early. Completers exhibited lower smoking rates, exercise barriers, and resting heart rates compared to Dropouts. For Completers, TMS/iTBS cortico-motor plasticity was associated with better exercise adherence (r= −0.53, corrected p= .019). Exploratory hypothesis-generating regression analysis suggested that post-iTBS changes (β= −7.78, p= .013) and self-efficacy (β= −.51, p= .019) may predict exercise adherence (adjusted-R²= 0.44). In conclusion, this study highlights the significance of TMS/iTBS cortico-motor plasticity, self-efficacy, and cardiovascular health in exercise adherence. Given the well-established cognitive benefits of exercise, addressing sedentary behavior and enhancing self-efficacy are crucial for promoting adherence and optimizing brain health. Clinicians and researchers should prioritize assessing these variables to improve the effectiveness of exercise programs.

ARTWORK GRAPHICAL ABSTRACT

Note. Numbers 1 through 5 refer to each result labeled in the artwork’s abstract figure. Following, we present the result highlights. 1. “Completers” and “Dropouts” differed in neither baseline TMS/iTBS cortico-motor excitability nor cortico-motor plasticity mechanisms (red color boxes). Despite no between-group differences, only “Completers” demonstrated a significant potentiation post-TMS/iTBS, but it did not hold after correction (green color arrow). Both groups demonstrated a larger effect size. 2. We demonstrated that the Dropout participants’ exercise barriers (e.g., lack of time, health, and budget) were significantly greater compared to the Completers group (green filled box). We also revealed that Dropout participants had significantly lower cardiorespiratory fitness, as demonstrated by lower RHR, exercise capacity, and estimated VO² peak compared to “Completers.” (green filled box) 3. In the Completers group, exercise adherence, defined as the average length of intervention in days to complete the 24 prescribed exercise sessions, was 64.4 (9.2) days. The average number of rescheduled sessions for this group was 4.3 (7.3). 4. The interval Post10–20_%Δ variable was the only selected TMS/iTBS cortico-motor plasticity predictor of exercise adherence and explained 22% of the variance (green-filled box). By including only exercise-related behavioral and fitness variables, exercise self-efficacy was the only predictor of exercise adherence in middle-aged and older adults and explained 21% of the variance. 5. Adjusted models were fitted and revealed that both TMS/iTBS cortico-motor plasticity and exercise self-efficacy were the best predictors of exercise adherence and explained 44% of the variance (green-filled box). This result indicates that for every one-unit increase in TMS/iTBS cortico-motor plasticity mechanisms, the estimated mean length of intervention in days (exercise adherence) decreases by approximately eight days when controlling for self-efficacy. Models adjusting for age, gender, and cardiorespiratory fitness were fitted but did not improve the model.

1. Introduction

Compelling evidence indicates that regular physical activity later in life increases longevity and may prevent or delay the onset of age-related cognitive decline and functional impairment (Falck et al., 2019; Morgan et al., 2019; Zubala et al., 2017). Despite this potential, approximately one-quarter of men and one-third of women worldwide are physically inactive (Guthold et al., 2018). Insufficiently active and sedentary behavior substantially increases healthcare costs and is a leading risk factor for cardiovascular events and all-cause mortality (Nguyen et al., 2022; Patterson et al., 2018). Specifically, inactive individuals have a 20-to-30% increased risk of premature death compared to sufficiently active people (Guthold et al., 2018; Piercy et al., 2018; World Health Organization, 2018). These findings show that sedentary individuals are not capitalizing on the potential health benefits of physical activity and highlight the urgent need to comprehend the motivating factors that can effectively engage middle-aged and older adults to adhere to an exercise program.

Despite extensive evidence highlighting the adverse impacts of sedentary behavior on health, social well-being, economic factors, and the numerous advantages of regular exercise, a substantial practical challenge remains – motivating sedentary individuals to adopt an exercise regimen. Furthermore, most sedentary older adults struggle to maintain exercise routines, and practical strategies to counter exercise attrition in this demographic are currently lacking. While most clinicians recommend that their patients increase physical activity levels and that older adults themselves recognize its health benefits, the overwhelming majority fail to participate in regular physical activity. The average adherence rates to exercise programs in clinical and research settings are also limited, and some studies report up to approximately 50% attrition (Picorelli et al., 2014). The causes of poor exercise adherence are likely multifactorial and have been suggested to include individual, social, environmental, and behavioral factors (Rivera-Torres et al., 2019).

Physical fitness, cardiovascular health, self-perceived well-being, self-efficacy, and prior exercise history are commonly cited factors associated with exercise adherence in older adults. Fitness levels and physical functioning capacity are crucial indicators of exercise adherence in this demographic. Put another way, aging adults with better physical fitness, fewer cardiovascular risk factors (e.g., obesity), and higher perceived health tend to adhere more consistently to exercise regimens (Picorelli et al., 2014; Rivera-Torres et al., 2019; Stineman et al., 2011). Within behavioral neuroscience, the social cognitive theory emphasizes self-efficacy, defined as an individual’s belief in their ability to control and execute specific tasks related to situational demands, as a key mediator of behavior change (Bandura, 1986; McAuley et al., 1993; Rhodes et al., 2020). Sedentary older adults with higher self-efficacy are more likely to maintain adherence to exercise programs once initiated (Neupert et al., 2009). Additionally, an individual’s lifetime exercise history may significantly influence exercise adherence later in life (Kuh & Cooper, 1992). Overcoming exercise barriers is another critical aspect of adherence among older adults. These barriers, distinct from those faced by younger adults, can include poor self-perception of health, age-related health issues, lack of knowledge about exercise guidelines, and environmental factors (e.g., living in a walkable area, neighborhood safety, and exercise facility affordability and accessibility (Bethancourt et al., 2014; Schutzer & Graves, 2004).

While the aforementioned factors offer valuable insights into the challenges of exercise adherence, an emerging frontier requires further exploration - the role of brain neurophysiology in exercise adherence. It is imperative to explore how neuroplastic mechanisms affect behavior change and exercise adherence, particularly if the assessment of the neuroplasticity mechanisms can predict effective and long-lasting adherence to an exercise program. Although evidence in this field is evolving, it promises to enhance our comprehension of exercise adherence. Thus, our study aims to contribute to the literature by highlighting the bidirectional relationship between physical exercise and neuroplasticity. Positive changes in synaptic neuroplasticity, like long-lasting changes in evoked postsynaptic potential or long-term potentiation (LTP) synaptic efficacy, are commonly cited as the underlying mechanisms for exercise benefits in cognitive health, such as motor learning and memory, in both animal and human studies (Christie et al., 2008; Lin et al., 2018). A common paradigm in the literature involves the evaluation of the modulation of motor evoked potentials (MEPs) induced by Transcranial Magnetic Stimulation (TMS)/intermittent Theta-Burst Stimulation (iTBS). This modulation exhibits significant parallels with the mechanisms of Long-Term Potentiation (LTP) synaptic plasticity, including a reciprocal time course of modulation post-stimulation and dependence on N-methyl-D-aspartate (NMDA) and Gamma-Aminobutyric Acid (GABA)ergic receptor activity (Huang et al., 2007; Nowak et al., 2017). The TMS/iTBS approach has been widely used to assess the role of LTP-like neuroplastic mechanisms in healthy individuals and those with metabolic, neurologic or neuropsychiatric disorders (Buss et al., 2020; Cabral et al., 2022; Freitas et al., 2011; Fried et al., 2016). Conversely, the literature currently lacks sufficient evidence to utilize TMS/iTBS plasticity as a basis for constructing a predictive model for exercise adherence. The effectiveness of brain plasticity mechanisms can potentially boost physical exercise performance by promoting behavior change and enhancing decision-making. This, in turn, may assist in initiating and sustaining exercise programs. For example, alterations in sedentary status due to a walking intervention could be influenced by baseline functional connectivity in brain regions related to executive control and effort-based decision-making (Morris et al., 2022). Additionally, sedentary behavior lacks inhibitory control and tends to reduce the attraction to a task requiring high effort and energy expenditure (e.g., physical exercise), aligning with the theory of effort minimization (Cheval et al., 2020; Cheval & Boisgontier, 2021). This highlights the need for a comprehensive physiological and behavioral assessment, especially for those at risk for cognitive decline and facing challenges with lifestyle interventions.

1.1. Rationale Summary for the Present Study and Research Hypotheses

Compelling evidence in behavior change has widely discussed intervening exercise adherence factors and how clinicians and researchers should address these factors in older adults. However, there is still an increasing need to explore what incentivizes older adults to fully adhere to an exercise program to avoid the consequences of physical inactivity and partake in the benefits of an active lifestyle. Moreover, there is insufficient evidence of an association between physiological and behavioral mechanisms influencing adherence to an exercise program in older adults.

Therefore, in this study, we compared individuals’ baseline demographic characteristics, health status, TMS/iTBS cortico-motor plasticity, cardiorespiratory fitness, and exercise-related behavioral measures between those who completed an 8-week aerobic exercise intervention “Completers” and those who withdrew early “Dropouts.” First, we hypothesized that the Dropouts would differ significantly in baseline health status, TMS/iTBS cortico-motor plasticity, cardiorespiratory fitness (CRF), and exercise-related behavioral parameters compared to those Completers. Thus, we predicted that when compared to the Completers, middle-aged and older adults in the Dropout group would present with 1) a greater number of cardiovascular health risk factors, 2) reduced TMS/iTBS cortico-motor plasticity (as assessed by TMS/iTBS cortico-motor plasticity protocol), 3) lower CRF, and 4) worse performance in the behavioral parameters as demonstrated by lower self-efficacy and exercise lifetime history and a greater number of exercise barriers. Second, within the Completers, we hypothesized that baseline TMS/iTBS cortico-motor plasticity would be associated with adherence to a 2-month exercise program. Consistent with the World Health Organization, we defined exercise adherence as the extent to which participants engaged in the prescribed dosing regimen (World Health Organization, 2003). We predicted that greater TMs/iTBS cortico-motor plasticity would be associated with greater adherence to the prescribed exercise regimen and thus may influence long-lasting exercise engagement. Third, in the Completers, we explored predictor variables that best explain response variability in exercise adherence to a 2-month aerobic exercise intervention by combining TMS/iTBS cortico-motor plasticity, CRF, and exercise-related behavioral parameters.

2. Method

2.1. Study Design

This study represents a secondary analysis of an interventional study in neurologically healthy but physically sedentary adults. Specifically, this study tested differences among those who completed or dropped out from the intervention and explored a hypothesis-generating analysis to search for predictor variables that best explain response variability in exercise adherence to a 2-month aerobic exercise intervention. We included pre-intervention data from 41 individuals who were recruited for an in-person option of an exercise intervention study. Nineteen participants completed the intervention, and twenty-two were lost to follow-up. Out of the twenty-two participants lost to follow-up, seventeen withdrew from the study voluntarily, while five were excluded due to pandemic-related restrictions and social distancing practices. These five participants were not included in the study analysis. Out of the seventeen that withdrew, five participants dropped out before performing any of the included assessments in this study, and thus, we did not have them in the analysis. The local institutional review board approved the study protocol and registered it in the Clinical Trials database (https://clinicaltrials.gov, #NCT03804528). We collected data from February 2019 to March 2020. Figure 1 shows the study design summary.

Figure 1.

Study design summary.

2.2. Participants

We recruited participants by 1) flyers throughout the University of Miami Miller School of Medicine and Coral Gables Campuses and the Miami-Dade community (e.g., public libraries), 2) online research database tools, and 3) via the University Research Informatics Data Environment Consent to Contact Initiative.

The inclusion criteria were: 1) age ≥ 55 years, 2) no clinically detectable cognitive impairment (MoCA score ≥ 24), 3) sedentary (defined as a ‘low’ category using the International Physical Activity Questionnaire [IPAQ] short form) (Craig et al., 2003), and 4) English as a primary language. Major exclusion criteria were: 1) any unstable medical condition, 2) medical contraindication to physical exercise, and 3) contraindications to undergo TMS assessment recommended by the International Federation of Clinical Neurophysiology (Rossi et al., 2021; Rossini et al., 2015). All participants provided written informed consent, and the University of Miami institutional review board (#20180926) approved all procedures.

2.3. Data Collection

The screening process was conducted through phone interviews followed by an in-person meeting to obtain written informed consent and collect physical measures (e.g., vital signs), demographics (e.g., age, sex, education status, body mass index), and medical and family history (e.g., comorbidities). Data collection for this study included 1) TMS/iTBS cortico-motor excitability and plasticity, 2) cardiorespiratory fitness, and 3) exercise-related behavioral parameters.

2.4. Outcomes Measures

2.4.1. TMS/iTBS cortico-motor excitability and plasticity

All study parameters followed the current guidelines for the safe application of TMS recommended by the International Federation of Clinical Neurophysiology (Rossi et al., 2021; Rossini et al., 2015), and all involved technicians and scientists met the recommended training criteria (Fried et al., 2021).

In our study, single-pulse TMS and iTBS were delivered using a biphasic static-cooled handheld MagPro MCF-B65 figure-of-eight coil connected to a MagPro X100 stimulator (MagVenture A/S, Farum, Denmark). Cortico-motor excitability and plasticity were assessed from the hand representation of the primary motor cortex in the hemisphere representing the dominant hand, and peak-to-peak MEPs were recorded using surface electromyography electrodes applied to the contralateral first dorsal interosseous muscle. The infrared-based frameless stereotaxic system Localite TMS Navigator (Localite GmbH, Bonn, Germany) was used to target TMS and maintain coil position within sessions.

The TMS/iTBS assessment followed the procedures: 1) Motor cortex “hotspot” searching, 2) Resting motor threshold (RMT), 3) Cortico-motor excitability single-pulse TMS baseline response, 4) Active Motor Threshold (AMT), 5) Theta burst stimulation, and 6) Cortico-motor excitability post-iTBS spTMS stimulation response. Following IFCN guidelines (Rossini et al., 2015), we defined RMT as the minimum stimulus intensity that produced a small motor-evoked potential (MEP; about 50 μV in 50% of 10 trials) during the relaxation of the FDI muscle, while AMT was defined as the minimum stimulus intensity that produces a small MEP (about 200 μV in 50% of 10 trials) during isometric contraction of the FDI, at minimal voluntary contraction. Cortico-motor excitability was measured as the average MEP amplitude from 90 spTMS trials delivered in 3 blocks of 30 pulses (randomly jittered at 5–7 seconds) at 120% of biphasic RMT before iTBS (baseline) and regular intervals (5, 10, 20, and 30 min) post iTBS protocol. RMT, AMT, and cortico-motor excitability were measured using a 1) TMS coil handle oriented 45^◦ relative to the participant’s mid-sagittal axis and delivering a biphasic pulse (anterior-posterior–posterior-anterior current in the brain). The iTBS protocol consisted of bursts of 3 pulses at 50 Hz repeated at intervals of 200 ms in a two-second-on, 8-second-off pattern for a total of 600 pulses delivered at 80% of AMT (Huang et al., 2005). To reduce the influence of extreme values, we log₁₀-transformed individual MEPs, averaged across each time point, and back-transformed them into geometric means. All subsequent analyses were performed using geometric means values. A TMS/iTBS cortico-motor plasticity index was calculated as the percent change in cortico-motor excitability from baseline to post-iTBS response (see statistical method for additional details).

2.4.2. Cardiorespiratory fitness

A trained physical therapist conducted the Incremental Shuttle Walk Test (ISWT) to measure functional exercise capacity and estimate cardiorespiratory fitness (Dourado et al., 2011; Singh et al., 1992). The ISWT is a valid, reliable, and safe cardiorespiratory fitness measure and correlates well with gold-standard measures of maximum oxygen consumption in cognitively healthy adults (Dourado et al., 2013; Lima et al., 2019; Neves et al., 2015). Before the test commenced, participants were fitted with a heart rate monitor (Polar H10, Polar Electro Inc) for continuous monitoring throughout the test. We also documented blood pressure (OMRON BP7350, Omron Healthcare Inc), oxygen saturation (Diagnostix 2100 fingertip pulse oximeter, American Diagnostic Corp), and rate of perceived exertion at rest (measured with the Borg scale) (Borg, 1982), during, upon completion, and 5 minutes after test cessation.

Participants were asked to walk 10 m around a marking between two traffic cones, maintaining the speed indicated by the beeps on the audio recording. The walking speed increased by 0.17 meters per second (m/s), with an initial speed of 0.5 m/s. The test was terminated when: 1) the participant was not able to maintain the required speed (defined as >0.5 m from the cone when the beep sounds on a second successive 10 meters length), 2) at the request of the participant, or for any incompatible physical and neurological signs or symptoms (e.g., extreme discomfort, pain, dyspnea, dizziness, vertigo, and angina), or 3) if the researcher determined that the participant was not fit to continue (e.g., participant reaches the age-predicted maximum heart rate (220 - age). Maximum walking distance and body mass was recorded and used to estimate the VO2 peak calculated by the equation (Dourado et al., 2013):

$$\mathit{\text{Estimated Vo}}2\mathit{\text{peak}}\left[\frac{\frac{ml}{kg}}{min}\right]=\frac{257+0.038*\mathit{\text{ISWT maximun distance }}\left[\mathit{\text{meters}}\right]*\mathit{\text{Body mass }}\left[kg\right]}{\mathit{\text{Body mass }}\left[kg\right]}$$

We also recorded each individual’s maximal heart rate and heart rate recovery (HRR). HRR was operationally defined as the change in the heart rate from the peak of exercise to the heart rate after 1-min and 2-min cessation (Lamberts et al., 2004).

2.4.3. Exercise-related behavioral factors

Exercise Lifetime history.

We assessed exercise lifetime history using a modified Lifetime Physical Activity Questionnaire (LPAQ), which evaluates hours spent in various physical activities across the lifespan (Chasan-Taber et al., 2002). We focused on moderate and vigorous activities to standardize the activities and meet the references of the minimum moderate to vigorous intensity for active individuals (Piercy et al., 2018). LAPQ-based estimates of hours spent in each physical activity were averaged in the number of hours/week and converted into units of energy expended by multiplying the time spent in each activity by the metabolic equivalent task (MET-hours/week/year) over the lifetime. The conversion to MET values was notably extracted from the Compendium of Physical Activity (Ainsworth et al., 2011).

Exercise self-efficacy.

We assessed exercise self-efficacy using the Exercise Self-efficacy Questionnaire (ESEQ). The ESEQ assessed how confident participants would perform physical exercise under different conditions or constraints (Neupert et al., 2009). The ESEQ included nine items and generated a maximum total score of 36 (a high score means better results – meaning better exercise self-efficacy).

Exercise barriers.

We assessed exercise barriers by selecting items reported in previous research that involved environmental and social influences in physical exercise practice (Booth et al., 2000). The Exercise Barriers Questionnaire assessed participants’ agreement with a list of statements of about 15 commonly cited barriers to the practice of physical activity and exercise. The questionnaire generated a maximum total score of 45 (a high score implies poorer results – meaning a greater number of reported exercise barriers).

2.4.4. Exercise Adherence (dependent variable)

Within the Completers, exercise adherence was measured to the extent to which the individual’s behavior was consistent with the exercise dosage regimen prescribed in the protocol (World Health Organization, 2003). Thus, based on the length of the intervention, we defined exercise adherence as the total time in days to complete the 24 prescribed exercise sessions. Exercise sessions were previously scheduled per protocol (3 sessions per week), and the time to complete the intervention varied on the participant’s compliance, availability, and accountability. We offered participants the opportunity to make up missed sessions, and the intervention was only fulfilled when the participant completed all 24 sessions.

2.5. Intervention

The physical exercise intervention was administered at the University of Miami Miller School of Medicine Wellness Center and supervised by a study team member during all times and sessions. Each participant engaged in 60-minute (5-minute warm-up, 50 minutes of continuous exercise in the target zone, and 5 minutes of cool-down) sessions delivered three times/week for eight consecutive weeks (a total of 24 sessions). Aerobic exercise modalities were offered for participants’ selection at each section and included treadmill, elliptical, stationary bike, or stationary recumbent bike. Participants were fitted with a heart rate monitor to facilitate data collection of the cardiac signals. They were instructed to maintain a moderate steady-state intensity at 55–64% of maximal heart rate (determined via the exercise test) for the first 4-weeks and a vigorous intensity of 65–90% of maximal heart rate for the second half of the exercise program. At each session, we monitored heart rate and participant’s exertion (measured with the Borg scale) before, every 5 min of the 50-minute session, and 5 min upon completion of the exercise session. Blood pressure was assessed before and after each exercise session.

2.6. Statistical and Power Analysis

We performed all statistical analyses using JMP Pro (v15.0, The SAS Institute Inc., Cary, North Carolina, USA) and set a two-tailed 95% confidence interval (ɑ = .05). Data on sociodemographic characteristics and health status, TMS/iTBS cortico-motor excitability and plasticity, and cardiorespiratory fitness were represented as means ± SD and percentage (%) of the total. Data were tested for normality of distribution using the Shapiro-Wilk test, homogeneity of variances using Levene’s test, homoscedasticity by plotting residuals and predictors values, and variance inflation factor to avoid multicollinearity (mean VIF < 10). We calculated effect sizes, and were reported in partial eta-squared (η²) and interpreted as follows: small effect (.01 −.05), medium effect (.06 −.13), and large effect (> .14) (Cohen, 1988). The Holm-Bonferroni method was applied to correct p-values and reduce the familywise error rate for multiple comparisons.

Sensitivity power analysis was conducted on G*Power 3.1 software (Faul et al., 2007). Our sample of 31 participants (19 Completers and 12 Dropouts) provided a two-tailed 80% power to detect a large eta-squared between-group effect size (η² = .22). Within the Completers group; the sample provided 80% power to detect a correlation of |r|≥ .56 and an adjusted r-squared of 0.61 when accounting for TMS/iTBS cortico-motor plasticity and self-efficacy predictors.

To test our primary hypothesis, which states significant differences in baseline health status, TMS/iTBS cortico-motor excitability and plasticity, CRF, and exercise-related behavioral factors between Dropouts and Completers, and to analyze the determinants influencing the initiation of an exercise program, we initiated by comparing individual characteristics. This included sociodemographic, health status, cardiorespiratory fitness, and behavioral variables at baseline among participants. We employed either pooled or unpooled T-test and Chi-square test, depending on the variable category or distribution. We performed t-tests to compare differences by Group (Completers vs. Dropouts) in TMS/iTBS cortico-motor excitability response by assessing RMT, AMT, and baseline MEPs. We evaluated post-iTBS modulation of MEPs within-group by entering the MEP geometric mean as a dependent variable in random-effects linear models assessed within-group factor Time (Baseline, Post0-10, Post10–20, Post20–30) differences. Following a well-established approach (Fried et al., 2017; Wischnewski & Schutter, 2015), to compare iTBS-induced changes in cortico-motor plasticity, post-iTBS MEP geometric means were expressed as the percent change (%Δ) from baseline and were entered into mixed-effects linear models between-Group factor, the within-subject factor Time, and the higher order interaction of Group*Time. Interval time response aimed to decrease variability and, thus, improve the reproducibility of the measure. This method has been reported (Cabral et al., 2022), and the time window within the post-iTBS time, specifically measuring the iTBS response at 10–20 minutes post-iTBS, corresponds to the peak effect of iTBS in healthy adults (Wischnewski & Schutter, 2015).

To test our secondary hypothesis, which states that baseline TMS/iTBS cortico-motor plasticity would be associated with exercise adherence, we fitted Pearson correlation (r) and scatterplots to represent the association visually.

To test our exploratory analysis, we fitted regression models to assess the predictors of exercise adherence in the participants who completed the exercise intervention. In these models, exercise adherence was defined as the number of days taken to complete the 24 exercise sessions prescribed, accounting for any rescheduled sessions. The independent variables were 1) TMS/iTBS cortico-motor plasticity measures (Post0-10%Δ, Post10–20%Δ, and Post20–30%Δ), 2) exercise-related behavioral measures (exercise lifetime history, exercise self-efficacy, and exercise barriers), and 3) cardiorespiratory fitness (estimated VO₂ peak). First, we performed two multivariable stepwise regression models to analyze 1) the effects of TMS/iTBS cortico-motor plasticity measures and 2) the effects of exercise-related behavioral and cardiorespiratory measures on exercise adherence using data from all participants in the Completers group (p-value threshold, put-in criteria, ≤ 0.10; put-out criteria, ≤ 0.10). Second, we fitted a multiple linear regression combining the variables selected in the previous stepwise regressions. Third, we adjusted the models to control for age and gender. In light of the exploratory nature of these analyses, it is important to note that individual p-values were not adjusted for multiple comparisons and, thus, should be interpreted accordingly.

3. Results

Table 1 presents detailed baseline demographic and clinical characteristics of the 31 participants who started the intervention and were included in the analysis. From those, nineteen participants (61%) completed the intervention, and twelve (39%) dropped out. The smoking history was the only significant variable that differed between participants’ Completers (11% smokers) and Dropouts (45% smokers, uncorrected p = .03).

Table 1.

Baseline demographic characteristics and health status.

Demographics and health status mean ± SD		Completers (n = 19)	Dropouts (n = 12)	p
Demographics
	Age, years	61.8 ± 7.1	62.1 ± 5.3	0.91
	Gender, female n (%)	12 (63)	4 (33)	0.10
	Ethnicity, n (%)			0.12
	Asian	1 (5)	1 (8)
	Black	2 (11)	6 (50)
	Hispanic	10 (53)	2 (17)
	Hispanic/White	2 (11)	1 (8)
	White	4 (21)	2 (17)
	Education level, n (%)			0.81
	High School/Technical/Associate	5 (27)	5 (42)
	Complete College	9 (47)	4 (33)
	Postgraduate	5 (26)	3 (25)
Health status
	BMI, n (%)			0.18
	Normal (18,5 – 24,9)	3 (16)	4 (33)
	Overweight (25 – 29,9)	7 (37)	3 (25)
	Obese (≥ 30)	9 (48)	5 (42)
	Mean ± SD, kg/m2	29.5 ± 4.9	30.3± 7.5	0.74
	Physical Activity Level
	IPAQ total, METs, mean± SD	372 ± 374	330 ± 377	0.79
	IPAQ weekday sitting, hours, mean± SD	6.3 ± 2.8	7.3 ± 4.2	0.50
	Global Cognition Status
	MoCA total, mean±SD	25.8 ± 1.9	25.3 ± 1.6	0.56
	Diseases and Comorbidities, mean±SD	3.7 ± 2.8	3.6 ± 2.8	0.96
	History of smoking, yes, mean±SD	2 (11)	5 (45)	0.03a

Abbreviations. BMI = Body Mass Index; IPAQ = International Physical Activity Questionnaire; MET = Metabolic Equivalent; MoCA = Montreal Cognitive Assessment.

^{a =}This finding did not hold statistical significance following Holm-Bonferroni correction.

3.1. “Completers” vs. “Dropouts”

Figure and Table 2 detail results for comparing the baseline between the Completers and Dropouts groups. T-tests found no significant differences between groups for RMT (t₂₉ = 1.9, uncorrected p = .07, η² = .11), AMT (t₂₉ = 1.25, uncorrected p = .22, η² = .05), or baseline MEPs (t₂₈ = −1.15, uncorrected p = .26, η² = .05). These results indicate that the Completers and Dropouts groups did not differ in baseline cortico-motor excitability response to TMS. A random-effects analysis indicated a significant effect of Time for post-iTBS MEPs in the Completers group (F_3,51= 2.83, uncorrected p = .04, η² = .14), see Figure 2) but not in the Dropouts, despite the larger effect size (F_3,21 = 1.36, uncorrected p = .28, η² = .16). These results indicate a significant effect of iTBS on MEP amplitudes in the Completers group along with a non-significant effect in the Dropouts, which was likely due to insufficient power given an equivalent effect size was observed with fewer participants. This finding did not hold statistical significance following the Holm-Bonferroni correction (corrected p-values > .05). To analyze a between-group comparison of post-iTBS MEPs%Δ responses, a mixed-effect linear model showed no effect of Group (F_1,23 = 2.1, uncorrected p = 0.16, η² = .08) and neither with Time (F_3,67 = 1.2, uncorrected p = 0.30, η² = .05) nor the Group*Time interaction (F_3,67 = 2.6, uncorrected p = 0.06, η² = .10). These findings indicate that the iTBS-induced change in MEP amplitudes did not differ significantly between Completers and Dropouts at the p=.05 level.

Table 2.

Data on length of intervention and baseline predictors outcome measures.

Variable, mean ± SD	Completers (n = 19)	Dropouts (n = 12)	95%CI mean diff.	t (df)	p	|η 2 |
Exercise adherence
Length of intervention, days	64.4 ± 9.2	-	-	-	-	-
TMS/iTBS Cortico-motor excitability
RMT, %	56.0 ± 12.7	64.9 ± 12.7	−0.67, 18.51	1.90 (29)	.07	.11
AMT, %	46.0 ± 11.9	51.2 ± 9.8	−3.25, 13.58	1.25 (29)	.22	.05
Baseline MEPs, mV, mean (SE)	0.86 ± .53	0.64 ± .34	−0.56, 0.16	−1.15 (28)	.26	.05
Exercise-related behavioral parameters
Exercise Lifetime History, total METs/week/year	188.3 ± 195.2	165.8 ± 155.0	−193.6, 148.5	−0.27 (23)	.79	.01
Exercise Self-Efficacy, total score	11.6 ± 8.4	12.7 ± 8.4	−5.2, 7.5	0.37 (28)	.71	.01
Exercise Barriers, total score	3.6 ± 2.1	7.0 ± 6.1	0.02, 6.8	2.08 (22)	.048a	.18
Cardiorespiratory fitness (ISWT)
Resting HR, beats/min	69.1 ± 9.9	78.2 ± 12.2	0.8, 17.4	2.24 (28)	.032a	.15
HR reserve, beats/min	62.6 ± 17.1	57.2 ± 16.8	−19.5, 8.7	−0.79 (24)	.43	.03
HRR1, beats/min	27.8 ± 11.4	24.1 ± 8.7	−12.8, 5.4	−0.84 (23)	.41	.03
HRR2, beats/min	38.1 ± 12.6	35.5 ± 6.9	−12.5, 7.2	−0.55 (23)	.58	.02
Exercise capacity, ISWT distance, m	538.7 ± 120.3	441.8 ± 105.2	−190.5, −3.2	−2.13 (24)	.043a	.15
Estimated VO2 peak, ml/kg/min	23.2 ± 5.1	19.0 ± 3.8	−8.2, −1.0	−2.12 (23)	.045a	.16

Notes. MET = Metabolic Equivalent; ISWT = Incremental Shuttle Walking Test; HRR = Heart Rate Recovery; HR = Heart Rate.

^{a =}This finding did not hold statistical significance following Holm-Bonferroni correction.

Corrected p-values > .05.

Figure 2.

Group post-iTBS MEPs (mV) response comparison of “Completers” in blue and “Dropouts” in red.

Note. Mean (Standard Error). TMS/iTBS cortico-motor induced plasticity of Post0-30 MEPs was significantly modulated in the “Completers” individuals but not in the “Dropouts” individuals. MEPs = motor evoked potentials; %Δ = percent change; * = this finding did not hold statistical significance following Holm-Bonferroni correction.

3.2. Cardiorespiratory Fitness and Exercise Behavior

Table 2 details the results for comparing the baseline between the Completers and Dropouts groups. T-tests found that the Dropout participants’ exercise barriers were significantly greater than the Completers group (t₂₂ = 2.08, uncorrected p = .048, η² = .18). Overall, the most common exercise barriers reported by the participants, respectively, were: 1) Lack of company and would be more active with a partner or in a group (53% agreed or partially agreed), 2) Getting tired easily (44%), 3) Cannot afford a membership (40%), and 4) Do not have any time for exercise (33%). We also discovered that Dropout participants had significantly lower cardiorespiratory fitness than Completers which was demonstrated by the significant differences in resting heart rate (t₂₈ = 2.24, uncorrected p = .032, η² = .15), exercise capacity (t₂₄ = −2.13, uncorrected p = .043, η² = .15), and estimated VO₂peak (t₂₃ = −2.12, uncorrected p = .045, η² = .16). Despite these findings revealing a larger effect size, this finding did not hold statistical significance following Holm-Bonferroni correction. Exercise lifetime history and self-efficacy did not show a statistical difference between groups (p > .05; see Table 2 for detailed statistics).

3.3. Determinants of Exercise Adherence in the “Completers” and Exercise Modality

Preference

The average length of intervention in days to complete the prescribed 24 exercise sessions was 64.4 (9.2) days. The average number of rescheduled sessions for this group was 4.3 (7.3). The preferred exercise modality throughout the sessions was the stationary bike (47.4%), followed by the treadmill (43.4%) and the elliptical (8.6%).

3.4. TMS/iTBS cortico-motor plasticity and Exercise Adherence Correlation Analysis

For Completers, the Pearson coefficient revealed a large negative significant correlation between TMS/iTBS cortico-motor plasticity (Post10–20%Δ) and days to complete the intervention (r = −0.53, corrected p = .019, Figure 3), indicating that individuals who had greater TMS/iTBS plasticity mechanisms also demonstrated greater adherence to the intervention.

Figure 3.

Correlation between TMS/iTBS plasticity and exercise adherence.

Note. Post10–20%Δ represents TMS/iTBS cortico-motor plasticity, and it is defined as the percentage change (%Δ) in peak-to-peak MEP amplitude from baseline to 10–20 minutes post-iTBS. We presented the values in decimal form, where −1.0 represents a decrement in TMS/iTBS cortico-motor plasticity of 100% and +1.0 represents an increase of 100%.

3.5. TMS-iTBS Neuroplastic Determinants of Exercise Adherence

A multivariable stepwise linear regression was fitted and selected the time interval of Post10–20%Δ (Post-iTBS MEPs percent change from baseline to 10–20 minutes) as a predictor of exercise adherence (F_1,18 = 6.7, uncorrected p = 0.019, adj. R² = .22, Table 3). This finding aligns with prior research indicating Post10–20 as the peak effects and highest test-retest reliability of iTBS in older adults (Fried et al., 2017; Wischnewski & Schutter, 2015). It also revealed a significant association between TMS-iTBS cortico-motor plasticity (Post10–20%Δ) and exercise adherence (β = −8.3, uncorrected p = .024). This suggests that for every unit increase in Post10–20%Δ, the length of intervention decreases by approximately eight days, demonstrating greater exercise adherence.

Table 3.

Fitted models of the association between TMS-iTBS cortico-motor plasticity and exercise adherence.

Model 1a	β Coefficients (95% CI)	SE	Adj R-squared	p-value
Constant	66.2	1.95	.22	<.0001b
Post10–20%Δ	−8.3 (−15.5, −1.3)	3.39		.024b

Abbreviations. Post10–20%Δ= percent change in the TMS-iTBS cortico-motor plasticity index measure from baseline to interval 10–20 minutes; SE= Standard Error; CI Confidence Interval.

Note.

^{a =}Prediction equation for model 1: Exercise adherence (days)= 66.2 +(−8.3 * Post10–20%Δ)

^{b =}significant uncorrected p-value

3.6. Exercise Behavior Determinants of Exercise Adherence

A second stepwise regression model with only behavioral and fitness variables was fitted. Exercise self-efficacy was selected as the only behavioral and fitness predictor of exercise adherence (F_1,18 = 5.76, uncorrected p = .028, adj. R² = .21, Table 4). This result revealed a significant association between exercise self-efficacy and adherence (β = −.55, uncorrected p = .028), indicating that individuals with higher self-efficacy tend to complete the intervention in a shorter duration, demonstrating greater exercise adherence.

Table 4.

Fitted models of the association between behavioral variables and exercise adherence.

Model 2a	β Coefficients (95% CI)	SE	Adj R-squared	p-value
Constant	70.8	3.2	.21	<.0001b
Self-efficacy	−0.55 (−1.03, −.06)	.23		.028b

Abbreviations. SE = Standard Error; CI = Confidence Interval.

Note.

^{a =}Prediction equation for model 3: Exercise adherence (days)= 70.8 +(−0.55 * Self-efficacy)

^{b =}significant uncorrected p-value

3.6. TMS-iTBS Cortico-motor Plasticity and Self-efficacy Predict Exercise Adherence

A multiple linear regression revealed both TMS-iTBS cortico-motor plasticity (β = −7.78, p = 0.013) and exercise self-efficacy (β = −.51, p = 0.019) as predictors of exercise adherence and explained 44% of the variance (F_2,18 = 7.9, uncorrected p = .0041, adj. R² = 0.44, Table 5). This result indicates that for every one-unit increase in Post10–20%Δ, the estimated mean length of intervention in days (exercise adherence) decreases by 7.78 when controlling for self-efficacy. Models adjusting for age, gender, and cardiorespiratory fitness were fitted but did not improve the adjusted model 1 (Table 5).

Table 5.

Adjusted fitted models of the association among TMS-iTBS cortico-motor plasticity, self-efficacy, and exercise adherence.

Adjusted Model	β Coefficients (95% CI)	SE	Adj R-squared	p-value
Adjusted Model 1 a
Constant	71.9	2.7	.43	<.0001b
Post10–20%Δ	−7.78 (−13.7, −1.9)	2.8		.013b
Self-efficacy	−0.51 (−.92, −.09)	0.19		.019b
Adjusted Model 2 (F5,18 = 2.96, p = .053)
Constant	71.53	23.1	.35	.0084b
Self-efficacy	−0.46 (−0.9, 0.02)	0.22		.058
Post10–20%Δ	−8.30 (−14.9, −1.7)	3.05		.017b
Estimated VO2 peak	0.25 (−0.5, 1.1)	0.37		.50
Age	−0.10 (−0.7, 0.5)	0.30		.75
Gender (Male)	−0.72 (−4.8, 3.4)	1.90		.71

Abbreviations. Post10–20%Δ= percent change in the TMS-iTBS cortico-motor plasticity index measure from baseline to interval 10–20 minutes; Estimated VO2 peak estimated oxygen consumption at exercise peak; SE= Standard Error; CI= Confidence Interval.

Note.

^{a =}Prediction equation for adjusted model 1: Exercise adherence (days)= 71.9 +(−7.78 * T10–20%Δ) + (−0.51 * Self-efficacy)

^{b =}significant uncorrected p-value

4. Discussion

4.1. Summary, relevance, and discussion of main results

This study addressed two primary questions regarding exercise adherence. Firstly, we aimed to investigate the physiological focusing on TMS/iTBS cortico-motor plasticity and exercise-related behavioral determinants influencing individuals’ completion of an exercise program. Secondly, we observed which of these physiological and behavioral factors hinder or facilitate individuals’ consistent adherence to the prescribed exercise regimen outlined in the protocol. The findings offered partial support for our primary hypothesis. Specifically, the presence of risk factors such as smoking history, low cardiorespiratory fitness, and a higher number of exercise barriers underscored their significant role in determining whether participants completed the exercise program. These results underscore the importance of considering these factors when designing interventions aimed at improving exercise adherence among middle-aged and older adults. Despite the study’s limitations, including the lack of statistical power, our analysis did not find evidence supporting the role of TMS/iTBS cortico-motor plasticity in influencing exercise completion when comparing “completers” and “dropouts”. However, we observed a strong correlation between baseline TMS/iTBS cortico-motor plasticity and exercise adherence, supporting our secondary hypothesis. This suggests that while TMS/iTBS cortico-motor plasticity may not directly impact exercise completion, which still requires further investigation, it could play a role in predicting individuals’ adherence to exercise regimens and facilitating the process of long-term behavior change in the long term.

We also performed a novel hypothesis-generating analysis to enhance our comprehension and identify predictor variables that explain the variability in exercise adherence to a planned aerobic exercise intervention. This approach, hitherto under-examined in the literature, aimed to refine our understanding of the factors influencing adherence to such interventions. The results from this exploratory analysis suggested a potential role of TMS/iTBS cortico-motor plasticity mechanisms and exercise self-efficacy as predictors of exercise adherence. Together, these factors accounted for approximately 50% of the variation in consistently completing the prescribed exercise regimen with the prescribed dose. Notably, the modulation of MEPs in cortico-motor plasticity response to TMS following iTBS has been interpreted as indicative of neuroplastic mechanisms resembling LTP-like response rather than directly associated with executive control areas, such asthe dorsolateral prefrontal cortex region (Friedman & Robbins, 2022).

In addition, we have identified various environmental and social barriers that hinder exercise and warrant further investigation. These include the absence of a workout partner or group, experiencing quick exhaustion, financial constraints preventing membership affordability, and a lack of available time for physical activity. These findings contribute to our understanding of the complex interplay between physiological and behavioral factors in determining exercise adherence among middle-aged and older adults. By identifying specific risk factors and potential predictors of exercise adherence, our study offers valuable insights for the development of targeted interventions and future studies aimed at improving long-term adherence to exercise programs in this population.

4.2. Understanding the Impact of brain plasticity and self-efficacy on behavior change and exercise adherence

The divergent results observed between TMS/iTBS cortico-motor plasticity mechanisms and exercise adherence, as evidenced in our findings, prompt further investigation within the transtheoretical model of behavior change (TTM) framework. Firstly, comparing individuals who completed the program to those who did not offer insights into the factors influencing participant initiation of exercise. This can be understood through the TTM, which conceptualizes behavior change as a dynamic process categorized into five stages: pre-contemplation, contemplation, preparation, action, and maintenance (Gourlan et al., 2016; Selçuk-Tosun & Zincir, 2019). Given the gradual and deliberative nature of behavior change, our comparison between “completers” and “dropouts” aimed to explore the effectiveness of participants transitioning from the “preparation” to the “action” stage. This transition signifies a crucial shift from deciding and preparing to actually incorporating the behavior change into their daily lives. Furthermore, during an exercise program, instances of relapse may signal a regression to earlier stages, highlighting the need for additional support. This may have occurred with those who dropped out of the study. In our analysis, factors that likely contributed most to participant withdrawal, potentially causing relapses, included smoking history, exercise barriers, and poorer fitness, as evidenced by higher resting heart rate and lower exercise capacity. In fact, our study showed no statistical group differences in TMS/iTBS cortico-plasticity between completers and dropouts. However, we highlight that this would reflect a lack of statistical power and requires further exploration using TMS and electroencephalography. Specifically, exploring higher-order cognitive regions like the dorsolateral prefrontal cortex is important. This area involves crucial inhibitory control over subcortical regions such as the amygdala and the hypothalamic-pituitary-adrenal axis, potentially affecting impulse control and adherence to complex health behaviors (Bigliassi & Filho, 2022; Friedman & Robbins, 2022).

Completing an exercise program is essential, but most importantly, it is appropriately engaging by achieving the prescribed exercise dose and parameters for full benefits and long-term maintenance of the behavior (Gomes-Osman et al., 2018). Along those lines, brain activity assessment-modulation, exemplified in this study by TMS/iTBS cortico-motor plasticity assessment, holds the potential to offer crucial insights into comprehending the readiness for behavioral change and understanding factors related to gaining exercise habits. One key aspect of readiness to change is the ability of an individual brain network to transition effectively between a “resting state” and a “task state.” This capacity has been linked to measures of general intelligence, indicating that individuals with higher-performing cognitive abilities exhibit a greater state of readiness for engaging in behaviors than those with lower-performing cognitive skills (Schultz & Cole, 2016). Therefore, analyzing factors contributing to exercise adherence solely among “completers” gives us the opportunity to understand readiness to change and what drives maintenance in the “action stage”, potentially progressing to the “maintenance stage” of behavior change. Achieving this transition typically requires a longer intervention phase (≥ 6 months), for individuals to fully acquire behavior and enter the maintenance stage, reducing the likelihood of relapses (Gourlan et al., 2016; Prochaska & Di Clemente, 1982). Although our intervention lasted only two months, it was still sufficient to offer insights into factors influencing individuals’ maintenance in the action stage and their progression toward acquiring behaviors such as regular exercise.

In our study, two primary factors emerged as significant contributors to long-term adherence to behaviors: the efficacy of TMS/iTBS cortico-motor plasticity mechanisms and exercise self-efficacy. First, our study revealed a large correlation between baseline TMS/iTBS cortico-motor plasticity and exercise adherence. This finding suggests that individuals with greater efficacy of TMS/iTBS cortico-motor plasticity mechanisms may be more prone to initiating behavior changes, potentially explaining their ability to sustain adequate participation in the exercise program. Second, we demonstrated that exercise adherence (length of days to complete the intervention) is independently associated with TMS/iTBS cortico-motor plasticity and self-efficacy. By combining both factors, we improved our model and revealed that greater efficacy of TMS/iTBS cortico-motor plasticity mechanisms and greater self-efficacy predict about half of exercise adherence in middle-aged and older adults. By going deep into the definitions of neuroplastic changes and self-efficacy, we can further understand these results. Neuroplastic mechanisms broadly refer to the brain’s ability to reorganize itself by forming new neural connections throughout life (Mateos-Aparicio & Rodríguez-Moreno, 2019). This process plays a crucial role in learning new skills, adapting to environmental changes, and recovering from injury (Pascual-Leone et al., 2005). On the other hand, exercise self-efficacy pertains to an individual’s belief in their ability to engage in and maintain an exercise regimen successfully. It reflects confidence in overcoming barriers, practicing a workout routine, and achieving fitness goals by achieving expected outcomes and inhibiting undesired tasks. These two factors, neuroplasticity and exercise self-efficacy, are connected in their influence on long-term behavior adherence (McAuley et al., 1993; Rhodes et al., 2020). Individuals with higher levels of neuroplasticity may find it easier to learn and adopt new exercise habits, while those with strong exercise self-efficacy are more likely to persist in their routines despite challenges or setbacks. For example, older adults who demonstrate greater self-efficacy and self-control and exhibit self-regulatory strategies (e.g., goal setting) may overcome automatic low-effort tasks (e.g., sedentary behaviors) and favor behavior change (Rhodes et al., 2020). Therefore, understanding and fostering these factors can significantly enhance individuals’ engagement and adherence to long-term behavioral changes, including regular exercise (Neupert et al., 2009). It is noteworthy that the length of days to complete the intervention is directly related to the participant’s accountability, which was affected by the number of missed and rescheduled sessions. Therefore, it is important to acknowledge that sedentariness by itself, in combination with a lack of self-efficacy, as demonstrated in this study, can potentially alter the neuroplasticity process, posing challenges to the beneficial effects of exercise on cognitive health. Consequently, it becomes increasingly reasonable to postulate that exercise programs may encounter limitations or exhibit reduced effectiveness in sedentary aging adults due to predisposed challenges when attempting to initiate or fully engage in the prescribed exercise regimen. These challenges may include a lack of intrinsic motivation, difficulty overcoming sedentary habits ingrained over time, insufficient confidence in their physical abilities, concerns about potential discomfort or injury, unfamiliarity with exercise techniques or equipment, apprehensions related to social settings or public scrutiny, perceived time constraints, or financial limitations inhibiting access to appropriate exercise facilities or professional guidance. It is essential to proactively acknowledge and address these individual challenges to optimize the success and impact of exercise interventions among sedentary aging adults, ensuring they can reap the full cognitive and physical benefits of regular exercise.

4.3. The behavioral relevance of CRF, cardiovascular and brain risk factors, and exercise barriers on exercise completion

It is well-known that sedentary behavior is influenced by multifarious biopsychosocial factors that directly and indirectly affect brain health. Similarly, our study suggests that smoking history and poorer fitness, as demonstrated by higher resting heart rate and lower exercise capacity, may not only affect brain health and increase dementia risk (Livingston et al., 2017) but also compound increased sedentary behavior by acting as determinants of exercise completion. These findings are not surprising when considering the inverse relationship between smoking rate and physical fitness. Studies in young and older adults have shown lower physical endurance and decreased exercise frequency in smokers than in non-smokers (Conway & Niles, 2017; Mesquita et al., 2015; Swan et al., 2018). Over time, physical inactivity and chronic smoking often lead to all-cause mortality and comorbidities such as type-2 diabetes, cardiovascular disease, stroke, and cancer or chronic lung disease (Jackson et al., 2019; Ng et al., 2020). Despite regular physical activity reducing up to 30% of the risk of all-cause mortality in both smokers and non-smokers, the more significant benefits are generally observed when exercising at goal levels and quitting smoking (O’Donovan et al., 2017). In addition to tobacco use and poorer fitness, the social and health aspects linked to the significant exercise barriers reported (e.g., lack of company and easy fatigue) may have worsened the completion of the exercise. Socialization and optimal overall self-health perception are essential for exercise adherence in older adults. Social engagement during training in a group may be beneficial to increase exercise adherence, improve cognitive functions (e.g., inhibitory control), and reduce the risk of dementia (Lindsay Smith et al., 2017; Penninkilampi et al., 2018; Sommerlad et al., 2018). Thus, screening for and guiding modifiable lifestyle behavior changes, such as smoking cessation, social engagement, and regular exercise, appear essential to facilitate behavior change in this population. In summary, many exercise barriers are linked to social determinants of health, and it is crucial to identify and increase awareness of those factors that may be easily targeted through low-cost and accessible exercise programs.

4.4. Study limitations and directions for future research

Despite the promising results, it is important to acknowledge and address this study’s limitations. First, an informal qualitative insight suggested that the extensive battery of assessments and exercise design (e.g., progressive exercise parameters instead of fixed high-level parameters) may have contributed to the early withdrawal of many participants. As stated in the methods, this is a secondary analysis of an interventional study that includes a comprehensive assessment battery that was not fully included in this study (Cabral et al., 2021). For future studies, we suggest adding additional characteristics to the exercise program to increase exercise adherence, including gradually increasing exercise duration, goal setting, exploring individual needs, and many others (Collado-Mateo et al., 2021). Future studies should address these elements by comparing isolated exercise interventions and an individualized multicomponent exercise intervention exploring behavioral, educational, social, personal, and physiological aspects. It is relevant to emphasize the importance of implementing supervised remote-based programs, as they can yield additional benefits in exercise engagement (Hinchman et al., 2022). This approach offers several advantages, including increased accessibility, convenience, and flexibility for participants, which can enhance exercise adherence and motivation. Moreover, the use of technology, such as video conferencing or mobile applications, enables real-time feedback, personalized instruction, and progress tracking, further optimizing the effectiveness and safety of the exercise program (Muellmann et al., 2018). Second, we assessed cortical excitability and plasticity from the motor cortex using EMG as the output measure. While this approach provided valuable insights into the neurophysiological effects of exercise adherence on motor-related processes, it is important to acknowledge that assessing similar measures from higher-order cognitive areas (i.e., the dorsolateral prefrontal cortex) by coupling TMS and electroencephalography would also be relevant to understanding the neurophysiological mechanisms underlying exercise adherence. Additionally, a clinical implication of the current findings is called ‘pragmatic neuroscience,’ which implies that the use of neuromodulatory and neuro screening techniques, particularly TMS (isolated or in combination with other neuroimaging), to identify neuro markers, predict individuals’ behavior, and optimize interventions by manipulating exercise program characteristics would favor the individuals’ characteristics and behaviors (Gabrieli et al., 2015).

Third, we acknowledged that our study was constrained by a relatively small sample size, which limited statistical power. The high dropout rates, exceeding 50%, also potentially introduced bias into our results. This includes participants who withdrew during the assessment phase and individuals whose participation was interrupted due to the restrictive measures imposed by the COVID-19 pandemic. Regarding the exploratory regression analysis, we acknowledge that some results may be susceptible to type II errors. In future studies with larger sample sizes, we recommend implementing p-value corrections. This will help validate our generated research hypothesis, specifically examining whether cortico-motor plasticity assessed through the TMS-iTBS approach is associated with exercise adherence in middle-aged individuals while accounting for exercise self-efficacy.

We acknowledged the potential biases and limitations associated with stepwise regression and the benefits and considerations of alternative model selection methods such as the Akaike information criterion (AIC) (Zhang, 2016). This limitation was minimized by 1) acknowledging the suitability of the method for hypothesis-generating, 2) the comprehensive rationale provided, aiming to mitigate the risk of stepwise error selection, and 3) the benefits rely on the advantages of adjusted R² in explaining the model’s clinical relevance and overall interpretability. Stepwise regression serves as a valuable tool for exploratory analysis and hypothesis generation by systematically including and excluding variables based on their contribution to the model fit. While AIC-based methods, for instance, may offer advantages in model selection, the complexity of variable relationships and dataset characteristics may favor stepwise regression in certain cases. Additionally, our approach aimed to mitigate stepwise error selection by providing robust background and rationale for variable selection, guided by literature, theoretical frameworks, and empirical evidence supporting each predictor variable’s relevance, significance, and clinical implication in relation to the outcome of interest. We conducted two stepwise regression models, each focusing on distinct variable natures to minimize model inflation and potential biases. Furthermore, our comparison of adjusted R² and standard error against AIC revealed that adjusted R2 may offer a superior ability to explain the model’s clinical relevance and interpretability. Adjusted R² offers a clear measure of predictive power while considering the number of predictors, providing valuable insights into model explanatory capabilities. In contrast, while AIC balances goodness of fit and model complexity, its interpretation may pose challenges in clinical settings due to its complexity. Thus, understanding these methodological nuances is crucial for robust statistical analysis and informed decision-making in research and clinical practice.

Researchers should also explore additional confounding variables, such as intrinsic and extrinsic motivation, preference and tolerance profiles, and an individual’s affective response to exercise (Hutchinson et al., 2023). Future studies should also address a follow-up assessment that would include insights into long-term adherence to the behavior, health outcomes, number of relapses, and challenges.

4.5. Conclusion

These findings offer meaningful physiological and behavioral insights and clinical applications to increase adherence to an exercise intervention. We have provided further data to support the association and behavioral relevance of cardiovascular and brain health risk factors – such as smoking and sedentariness - and the impact of TMS/iTBS cortico-motor plasticity and self-efficacy mechanisms on exercise adherence. Moreover, as existing literature highlights the critical role of neuroplasticity in facilitating post-exercise cognitive gains and promoting brain health, we have emphasized how sedentariness and low self-efficacy can potentially disrupt these mechanisms and hinder adherence to exercise programs, evidencing a potential bidirectional relationship between physical exercise and neuroplasticity. Therefore, considering that exercise is currently one of the most effective interventions available for mitigating cognitive decline and enhancing brain health, it is suggestive for therapists, clinicians, and scientists to prioritize the assessment of these variables. By that means, we can identify individuals who may face challenges in adhering to exercise programs and interventions, allowing for targeted strategies to improve adherence and optimize the benefits for brain health. Further understanding additional variables influencing exercise adherence will enable therapists to refine, optimize, and individualize exercise interventions for brain health.

Data availability:

The data supporting the findings described in this article will be available from the authors upon reasonable request.