Accelerometer-measured and self-reported physical activity and sedentary time and their relationships with the P300 in a Go/No-Go task in older adults

Abstract

Older adults who experience cognitive decline are more likely to have a reduced quality of life. Identifying lifestyle factors that may influence cognitive processing and in turn improve quality of life during older adulthood is an important area of interest. Cognitive function, as measured by the P300 event-related potential (ERP), has been noted to be modified by physical activity; however, no study to date has examined relationships between this neurophysiological measure and physical activity and sedentary time in older adults. Furthermore, there is a gap in understanding as to whether physical activity and sedentary time assessed using self-reported and accelerometer-based methods similarly relate to the P300. This study aimed to assess the P300 during a Go/No-Go task in relation to self-reported and accelerometer-based physical activity and sedentary time in a community sample of 75 older adults. Results indicated that participants engaging in more moderate-to-vigorous physical activity had larger P300 amplitudes across self-reported and accelerometer-based measurements; however, no relationships between sedentary time and P300 amplitude were observed. Notably, accelerometer-based moderate-to-vigorous physical activity explained P300 amplitudes over and above self-reported moderate-to-vigorous physical activity—an effect that remained significant even after accounting for age. Although these results highlight the importance of moderate-to-vigorous physical activity in relation to cognitive function, as measured via the P300 in older adults, a secondary analysis indicated that engaging in lifestyle activity may have similar effects on the P300 as moderate-to-vigorous physical activity. In sum, the present study highlights the role of habitual engagement in physical activity as a possible means for supporting cognitive function during the aging process.

1. Introduction:

Cognitive decline, which occurs on a continuum, is one of the most feared side effects of aging (Deary et al., 2009), can lead to substantial economic burden, reduced social interactions, and decreased health-related quality of life in older adults (Deary et al., 2009; James et al., 2011; Kazazi et al., 2018). Therefore, maintaining cognitive functioning is an important element for sustained daily functioning for older adults (Depp & Jeste, 2006; Beaton et al., 2015). Given the health-related consequences of cognitive decline, understanding relationships between lifestyle factors that account for individual differences in cognitive functioning, which may impact cognitive decline, has become crucial.

1.1. Physical Activity and Sedentary Time in Older Adults

Physical activity (PA) has shown efficacy as a protective factor and intervention strategy to address cognitive decline in older adults. For example, previous studies have shown a link between greater PA levels and reduced cognitive decline in older adults (Sofi et al., 2011; Nagamatsu et al., 2013; Hu et al., 2014; Svantesson et al., 2015; Lautenschlager, 2019; Cunningham et al., 2020; Lim et al., 2020). Moreover, data from 216,593 adults in the Behavioral Risk Factor Surveillance (BRFSS) survey, an ongoing national survey study collecting information about health-related behaviors among U.S. adults, demonstrated that the participants who were physically active reported lower subjective cognitive decline compared to those who were inactive (Xu et al., 2023). On the other hand, evidence on the influence of sedentary time (ST), which is defined as sitting or reclining with low energy expenditure (Tremblay et al., 2010; Sedentary Behaviour Research Network, 2012), on cognitive functioning in older adults is more conflicting. Some studies have shown sedentary behavior to be positively related with cognitive functioning (e.g., Moreira et al., 2022), while a multi-site analysis found no association between ST and cognitive function in older adults (e.g., Maasakkers et al., 2020).

To date, the relationship between PA, ST, and cognitive function among older adults has depended primarily on self-reported measures of PA and ST (Sofi et al., 2011; Lim et al., 2020), which has some limitations. For example, self-reports have been used to categorize older adults as being physically active or inactive/sedentary (Coelho et al., 2020; Eisenstein et al., 2021). This limits the ability to parse out the variability in movements and activity throughout the day due to classifying PA and ST patterns based on pre-defined thresholds. Self-reports are also subject to recall bias (e.g., Lee & Shiroma, 2014), which is particularly relevant given work demonstrating that older adult may unintentionally over- or under-report their PA and ST (VandeBunte et al., 2022; Meh et al., 2023). In contrast, accelerometer-based methods used to quantify PA and ST can offer refined insight in terms of duration, frequency, and intensity (Prince et al., 2008). Moreover, these methods do not require older adults to remember minute details that may be difficult to recall. Although there is some support regarding the relationship between higher daily accelerometer-measured PA levels and a slower rate of cognitive decline among older adults (Buchman et al., 2012), research using accelerometers to better understand patterns of PA and ST in older adults is nascent. Therefore, a goal for the current study was to add to the relatively sparse literature investigating the connection between accelerometer-based PA, ST, and cognitive functioning in older adults.

1.2. Event-Related Potentials (ERPs)

ERPs, which are derived from a continuous electroencephalogram (EEG), are time-locked measures of nearly instantaneous changes in neural activity that can be utilized in the quantification of cognitive functioning. Use of ERPs is cost-effective and due to their limited contraindications, more accessible and well-tolerated for aging populations compared to other imaging methods such as functional magnetic resonance imaging (fMRI). In addition, ERPs reflect non-invasive biomarkers that can be used to assess individual differences in cognitive function and outcomes to interventions (Proudfit, 2015; Kappenman & Luck, 2016; Alderman et al., 2019; Hajcak et al., 2019). The P300 is a highly investigated ERP component (Donoghue & Voytek, 2022) that manifests as positive-going ERP approximately 300 ms after stimulus presentation. It is commonly linked to cognitive processes involved in attention, processing speed, and context updating (Novak & Foti, 2015; Hajcak & Foti, 2020); however, the specific functional role of the P300 is related to the context in which is it elicited (Polich, 2007). For example, the P300s elicited during tasks such as the Go/No-Go (Pfefferbaum et al., 1985), which have stimuli varying in frequency and response demand, are thought to reflect inhibitory control and target/novelty detection (Brush et al., 2022; Gajewski & Falkenstein, 2013; Polich, 2007; Wessel, 2018). Both inhibitory control and target detection are crucial elements in cognitive processing that have been found to be reduced in aging populations with more cognitive decline (Hasher & Zacks, 1988; Daffner et al., 2006). Translationally, inhibitory control enables an individual to overcome mental, behavioral, and emotional impulses—processes that are essential for day-to-day functioning, goal-directed behavior, and increased quality of life (Miyake & Friedman, 2012; Diamond, 2013). Target detection is thought to reflect processes sensitive to attentional allocation and stimulus salience (Hajcak & Foti, 2020).

1.3. P300 and PA in Older Adults

Previous work has explored the P300 and its association with PA in older adults. Consistent with evidence on the relationship in younger adults (e.g., Hillman et al., 2012; Tsai et al., 2017; Wu et al., 2022), a systematic review detailed evidence that older adults with higher levels of PA had larger P300 amplitudes (Pedroso et al., 2017). In three separate studies that assessed self-reported PA, active older adults had greater P300 amplitude than their sedentary or less active counterparts (Eriksen flankers, Hillman et al., 2003; task-switching task, Fong et al., 2014; Sternberg working memory task, Chang et al., 2013). Even though early results seem to point to a positive impact of greater PA on cognitive function, a recent systematic review by Kao et al. (2020) examining the relationship between PA and the P300 showed that all evidence supporting benefits of PA on P300 in older adults has been based on self-reported data, which is affected by the limitations outlined above. To our knowledge, no previous work has been conducted to assess the influence of accelerometer-based PA and ST on the P300 during cognitive inhibition and target detection, as elicited by the Go/No-Go task in older adults.

The purpose of the current study was to examine cross-sectional associations between P300 amplitude and self-reported and accelerometer-based measures of PA and ST in a community sample of older adults. We were also interested in assessing the convergent validity of the self-reported and accelerometer-based measures of PA and ST to inform our secondary aim, which was to determine whether they explained unique or overlapping variance in P300 amplitude. Given previous work investigating the P300, PA, and ST in older adults, we hypothesized that greater time spent sedentary would be associated with a reduced P300 amplitude, whereas greater time spent engaged in moderate-to-vigorous physical activity (MVPA) would be correlated with larger P300 amplitudes during the Go/No-Go task. Given the limitations of self-report methods, we also expected that accelerometer-based measurements would explain relationships between PA and ST over and above self-reported PA and ST.

2. Method

2.1. Participants

The data utilized in this study were collected during a parent study which investigated the role of a brief exercise (i.e., single session) intervention on cognition; the primary aims for the parent study will be reported elsewhere. Participants aged 60 to 85 years with mild-to-no cognitive impairment were recruited² from the greater Tallahassee, FL area. Those with a current self-reported diagnosis of a major neurocognitive disorder, evidence of greater than mild cognitive impairment as indicated by a score of < 14 on the Montreal Cognitive Assessment Blind (MoCA Blind; Pendlebury et al. 2013; Wittich et al., 2010), or inability to engage in exercise by verbal report or lack of physician approval were excluded from the study. A total of 94 older adults were recruited and consented for the study; however, data from 19 participants were excluded for the following reasons: 3 participants only completed the baseline visit, 11 participants from the parent study opted to not wear the accelerometer; 3 participants had unusable EEG data; and 2 participants did not wear the accelerometer for at least four days. Data from 75 participants were used for the analyses of accelerometer-based activity in the present study, whereas 68 participants were used for analyses of self-reported activity due to 7 participants not fully completing their self-report data. All study procedures were approved by the Institutional Review Board at Florida State University. Additional details regarding participant compensation are below.

2.2. Procedures

The parent study asked participants to partake in 3 separate sessions and were compensated in cash based on completion of procedures per visit. At the initial session, participants conducted a semi-structured clinical interview, neuropsychological assessments, self-report surveys and the NIH Toolbox Emotion Battery (Gershon et al., 2013). Additionally, during the initial visit, participants were invited to participate in an optional portion of the study, which involved wearing a wrist-based accelerometer device on their non-dominant wrist for one week and tracking their wear time using a log diary prior to their first EEG visit (i.e., second visit). At the start of this visit, the watch and the log diary were collected.

The duration between the initial visit and the first EEG session (in which the Go/No-Go task was completed) was approximately one-to-two weeks apart (M = 11.66, SD = 5.19 days). The present study uses data collected during the initial visit (i.e., self-report data), accelerometer data collected during the week after the initial visit, and the baseline Go/No-Go EEG recording from the second visit, prior to exercise intervention.

2.3. Accelerometer Data

Participants were given a tri-axial Actigraph GT9X Link accelerometer (Actigraph Corp., Pensacola, FL, USA; weight, 14 g; dynamic range [primary accelerometer], ±8g; dynamic range [secondary accelerometer] = ±16g) and asked to wear the watch on their non-dominant wrist for a minimum of four days and up to one week between their first (initial) and second study visits. In addition, participants were asked to maintain a watch log diary to record sleep time, wake time, moments in which they removed the watch, the reason for its removal, and to list activities they completed that day (e.g., water aerobics, gardening, grocery shopping). To accurately reflect an individual’s typical activity levels, participants’ data were included if there was at least four days of watch wear for at least 480 minutes or longer during waking hours (e.g., Gough et al., 2021; Rich et al., 2013). Additional non-wear time was screened with the use of the wear time validation by Troiano (2007). Raw signal was sampled at 30 Hz and then re-integrated into 15-s epochs for data analysis.

Using ActiLife Software (version 6.13.4; Actigraph Corp., Pensacola, FL, USA) and in accordance with cut-points established for the non-dominant wrist for older adults by Migueles et al. (2021), the following intensity-specific cut-points were applied to the raw data: ST = all activity below 1224 cpm (counts per minute); Light Physical Activity (LPA) = all activity between cpm greater or equal to 1224 and below 2184 cpm. MVPA was defined as all activity ≥2184 cpm. Cut-points were based on vector magnitude of the three axes.

Given previous research that has shown that increased incidental PA—not planned or structured and the result of daily activities (Strath et al., 2013)—is not only associated with better cognition among older adults (Sanchez-Lopez et al., 2018), but also may be misclassified due to activities (e.g., home chores) considered to fall under the LPA category due to different cut-points (Fanning et al., 2022), an additional variable of Lifestyle PA was created in accordance with the cut-points for older adults by Lohne-Seiler et al. (2014). The following intensity-specific cut-points using vector magnitude were applied to the raw data: Lifestyle PA (e.g., slow walking, grocery shopping) = all activity between 760 and 2019 cpm. For the accelerometer-based measures, each intensity reflects min per day spent in MVPA, LPA, Lifestyle PA, and sedentary.

2.4. The International Physical Activity Questionnaire-Short Form (IPAQ-SF)

The IPAQ-SF (Craig et al., 2003) is a widely used, self-report questionnaire designed to cross-nationally assess PA and ST for adults in the past week. The IPAQ-SF has been shown to be a valid tool for measuring PA and ST in older adults when compared against a valid accelerometer measure, with moderate levels of validity for MVPA (r = .43-.56) and good levels of validity for ST (r = .70; Cleland et al., 2018). In this study, time spent in MVPA, walking, and sitting were computed as minutes per day spent in each behavior.

2.5. EEG Task: Modified Go/No-Go

Participants were asked to complete a modified visual go/no-go task wherein three types of visual stimuli were presented that varied as a function of stimulus probability (i.e., frequent vs. infrequent) and response demands (i.e., go vs. no-go). Participants were instructed to click the left mouse button when presented with a picture of an alien (i.e., Go-Frequent; 70% of the trials; 280 trials) or an asteroid (i.e., i.e., Go-Infrequent; 15% of the trials; 60 trials). When presented with a picture of an astronaut (i.e., No-Go; 15% of the trials; 60 trials), participants were told to withhold their response.³ Each trial consisted of the stimulus presented for 200 ms, followed by a blank screen for 1000 ms and then a fixation cross presented for 300-700 ms (in 100 ms intervals). The order of the stimuli was randomized across participants. A depiction of the task is displayed in Figure 1.

Figure 1. Modified Three-Stimulus Visual Go/No-Go Task Depiction.

Note. This is an example trial during the Go/No-Go task variant that was used in the current study. Participants first saw a fixation cross that appeared focally on the computer screen for 300-700 ms. Then, one of three stimuli appeared on the screen for 200 ms, which was followed by a 1000-ms response window.

In total, the task consisted of 10 blocks of 40 trials each. After each block of 40 trials, participants received performance feedback and were either instructed to “please try to be more accurate” if their accuracy was at or below 85% for a given block of trials, or if their accuracy exceeded 95%, then feedback indicating “please try to respond faster” was presented on the screen; otherwise, a message indicated “you’re doing a great job” was displayed as feedback. Behavioral performance measures derived from the task included response accuracy (% correct) and reaction time (RT; ms). Correct Go trials were defined as trials with a detectable response up to 800 ms after stimulus onset, while correct No-Go trials were defined as trials without a detectable response. For each participant, response accuracy was computed by averaging the percentage of correct trials for each separate trial type, while RT was computed by averaging RTs on correct frequent and infrequent go trials (i.e., trials requiring a response), separately.

2.6. EEG Recording, Signal Processing, Data Reduction, and Measurement

Continuous EEG was recorded with an active 10-electrode system (LiveAMP, Brain Products, GmBH; Munich, Germany) while participants completed the modified Go/NoGo task. The electrodes were placed into a cap that was situated on the scalp in accordance with the international 10/20 system. Electrode site FCz was used as the online reference, while FPz was the ground electrode. Electrodes were also placed on the left and right mastoids, which correspond to TP9 and TP10, respectively. Electro-oculogram (EOG) was recorded from four electrodes: one placed above the left eye, one placed below the left eye, one on the outer side of the right eye and one on the outer side of the left eye. The remaining two electrodes were placed at Cz and Pz. Impedance was kept at or below 20 kΩ during recording.

EEG signal processing and data reduction was performed offline using BrainVision Analyzer 2.1 software (Brain Products GmbH, Gilching, Germany). Data were re-referenced offline to the average of the left and right mastoids, then filtered with zero-phase shift fourth-order Butterworth filters with low and high filter cutoffs set at 0.01 and 30 Hz, respectively. EEG data were then segmented in a 1200-ms window, beginning 200 ms prior to stimulus presentation. Ocular artifacts were corrected using the Gratton et al. (1983) regression-based procedure and automatic artifact rejection was implemented, rejecting any epochs with a voltage step greater than 50 μV, a voltage difference of 175 μV within a 400 ms interval, or a maximum voltage difference of less than 0.50 μV within 100-ms intervals. Stimulus-locked ERPs were then averaged for each trial type separately. The average number of segments that were retained for analyses after correcting and rejecting artifacts for each stimulus type included: 275 (range=238-280) for Go-Frequent; 59 (range=50-60) for Go-Infrequent; and 38 (range=13-58) for No-Go. Lastly, the 200-ms pre-stimulus interval was used for baseline correction.

The P300 was analyzed for each stimuli type: aliens (i.e., Go-Frequent), asteroids (i.e., Go-Infrequent), and astronauts (i.e., No-Go). Consistent with previous work from our group using the same task (e.g., Brush et al., 2022; Sheffler et al., 2022), P300 amplitude was scored as the mean voltage at the Pz electrode site between 350 ms and 600 ms following stimulus presentation. To isolate brain activity of interest, we used a subtraction-based difference score measurement approach for P300: Asteroid-Alien Difference Wave (Go-Infrequent – Go-Frequent; i.e., target P300), and Astronaut-Alien Difference Wave (No-Go – Go-Frequent; i.e., inhibitory P300).

2.7. Sensitivity Power Analyses

Due to the relative novelty of investigating the P300 across self-reported and accelerometer-based methods in older adults, sensitivity power analyses were conducted for the zero-order correlation tests in G*Power software (Version 3.1.9.7; Faul et al., 2009). Given that the sample sizes differed depending on data availability from the accelerometer-based (N = 75) and self-reported PA/ST measures, we conducted two sensitivity power analyses using power of .80 and a two-tailed alpha of .05. Based on both sensitivity power analyses, the present study had sufficient power to reliably detect small effect sizes (analyses using accelerometer-based PA/ST measures: r ≥ ∣.316∣; lower and upper critical r = ∣.227∣; analyses using self-reported PA/ST measures: r ≥ ∣.332∣; lower and upper critical r = ∣.239∣).

2.8. Data Analyses

Data analyses were performed using SPSS Statistics software (Version 25.0; IBM Corp., Armonk, N.Y., USA) and jamovi software (Version 2.2.5; the jamovi project, 2021) with a criterion of statistical significance of p < .05. For task manipulation checks, repeated measures analyses of variance (rANOVAs) were conducted to assess within-person effects of trial type on error rate, false alarm hits, RT, and P300 amplitude. In the rANOVA, the Greenhouse-Geisser epsilon (ε) correction was used if the sphericity assumption was violated, and Bonferroni-adjusted post hoc decompositions of the main effect were conducted. Then, bivariate correlations were conducted to examine associations between P300 amplitude difference scores, self-reported and accelerometer-based PA and ST, age, and sex. Convergent validity was also assessed by conducting correlations between the self-reported and accelerometer-based PA and ST measures.

Next, following any significant associations between the P300 and the self-reported and accelerometer-based PA and ST measures at the zero-order level, multilevel models (MLMs) were fit to examine whether the PA or ST measures accounted for unique or overlapping variance in the P300 difference scores (i.e., target and inhibitory control P300). These MLMs were employed rather than ordinary least squares regression due to the nested structure of the data (i.e., repeated measures nested within people). In these models, level one represented the repeated measurements of P300, while level two included person-specific variables.

To fit these models, MLMs were specified by first fitting an unconditional random-intercept model to determine the amount of variance partitioned at the between- and within-person levels, which is reported as an intraclass correlation coefficient (ICC). Then, a likelihood ratio test (LRT) was conducted to justify the use of a random intercept in subsequent models. A random intercept was retained in subsequent models if it was significant based on the LRT in the unconditional model.⁴ Conditional models were then constructed, and specific details are outlined below.

In the first conditional model (model 1), we tested whether the target and inhibitory P300 were explained uniquely by self-reported or accelerometer-based measures. Therefore, this model consisted of a random intercept and fixed effect of trial type. At level two, self-reported and accelerometer-based measures significantly associated with the P300s in the zero-order correlations were included as level two variables. Then, we conducted another model (model 2) retaining significant terms from model 1 and added age as a person-specific covariate at level two as a sensitivity analysis to determine the robustness of the significant effects. Two additional models (models 3 and 4) were then tested as exploratory analyses, as outlined in section 3.6.2.

The Mixed Model module in the GAMLj suite (Galluci, 2019) implemented in jamovi software was used to fit the MLMs. The maximum likelihood estimator was used to estimate model parameters, the Satterthwaite method was used to estimate degrees of freedom, and the Wald method was used to compute the 95% confidence intervals. Additionally, the bobyqa optimizer was employed to help with model convergence. In the models, trial type (i.e., the two P300 difference scores) was deviation coded (target = −1; inhibitory = 1), such that the intercept estimates reflected the sample mean for P300. All level two continuous variables were standardized.

3. Results

3.1. Sample Characteristics

At the initial study visit, the mean age of the full sample (N = 75) was 71.92 years (SD = 15.55) with 62.7% identifying as female. Most participants identified as White (80%), while those remaining identified as Black or African American (16%), or other (4%).

3.2. Go/No-Go Task Behavioral Performance Effects

Responses to Go-Frequent stimuli (alien; RT: M = 363.80 ms, SD = 45.76) were significantly faster than response to the Go-Infrequent stimuli (asteroid; RT: M = 403.52 ms, SD = 59.41), t(74) = −12.17, p < 0.001, Cohen’s d_z = −0.67). For Go trials, there was a 1.1% rate of missed trials (i.e., Go trials in which the participant failed to respond correctly). For No-Go trials, there was a 36% (SD = 16%; t(74) = 19.51, p < .001, Cohen’s d = 2.25) rate of false alarms (i.e., No-Go trials in which the participant failed to withhold their response).

3.3. Go/No-Go Task ERP Effects

Means and standard deviations for P300 amplitude are presented in Table 1. A rANOVA indicated a significant main effect of trial type (F[1.84, 135.79] = 65.61, p < .001, ${\eta }_p^2=.47$), such that P300 amplitude was larger on No-Go trials (M = 12.26 μV, SD = 6.79) compared to Go-Infrequent (M = 9.70 μV, SD = 6.54; p < .001) and Go-Frequent trials (M = 7.37 μV, SD = 4.97; p < .001). P300 amplitude was also larger on Go-Infrequent relative to Go-Frequent trials (p < .001). In terms of the subtraction-based difference scores, the target P300 was significantly smaller than the inhibitory P300, t(74) = −6.25, p < 0.001, Cohen’s d_z = −0.67.

Table 1.

Zero-order correlations among study variables

Variable	1	2	3	4	5	6	7	8	9	10	11
1. Target P300 (μV)	1
2. Inhibitory P300 (μV)	.57**	1
3. Accelerometer-based ST (min/day)	.07	.11	1
4. Accelerometer-based LPA (min/day)	.13	.19	.48**	1
5. Accelerometer-based Lifestyle PA (min/day)	.22	.30**	.32**	.80**	1
6. Accelerometer-based MVPA (min/day)	.24*	.30**	.19	.59**	.85**	1
7. Self-reported Sitting (min/day)	−.21	−.11	.23	.15	.17	.07	1
8. Self-reported Walking (min/day)	.21	.12	−.29*	−.05	.12	.27*	−.05	1
9. Self-reported MVPA (min/day)	.29*	.05	−.14	−.07	.14	.29*	−.17	.53**	1
10. Age (years)	−.18	−.26*	.03	.13	.05	−.18	.01	−.14	−.22	1
11. Sex	.22	.22	−.15	−.11	−.10	.05	−.41**	.10	−.01	−.01	1
N	75	75	75	75	75	75	68	68	68	75	75
M	2.32	4.89	745.85	79.07	192.16	256.55	300.65	63.66	155.07	71.92	-
SD	3.28	4.20	167.61	21.37	55.53	81.67	157.29	57.05	170.74	5.56	-

Note. Target P300 represents the Asteroid-Alien P300 difference score, while the Inhibitory P300 represents the Astronaut-Alien difference score. ST = sedentary time; LPA = light physical activity; MVPA = moderate-to-vigorous physical activity; sex was coded as 1 for males and 2 for females. **p < .01; * p < .05.

3.4. Zero-Order Correlations

Bivariate correlations between the target and inhibitory P300, accelerometer-based and self-reported measures of PA and ST, age, and sex are reported in Table 1.⁵ Older participants had a reduced inhibitory P300, r(73) = −.26, p = .023; there was no relationship between age and the target P300. There were also no associations observed between the other study variables and sex, all ps > .054, except for the significant correlation between sex and self-reported sitting, r(66) = −.42, p < .001, indicating that males reported more time spent sitting (M = 381.11 min/day; SD = 171.49) compared to females (M = 247.66 min/day; SD = 122.77).

Accelerometer-based MVPA was associated with a larger target P300, r(73) = .24, p = .035, and inhibitory P300, r(73) = .30, p = .008, while increased Lifestyle PA was also related to a larger inhibitory P300, r(73) = .30, p = .008. None of the accelerometer-based ST measures related to the P300, ps > .344. Participants who self-reported more MVPA demonstrated an increased target P300, r(66) = .29, p = .018, while there were no other associations between self-reported walking and sitting and the P300s, ps > .081. ⁶

3.5. Convergent Validity

Given that the accelerometer-based and self-report measures assess the same underlying constructs of PA and ST, we examined the validity of these measures among the 68 participants who had complete data. Overall, there were small amounts of overlap among the accelerometer-based and self-reported measures; however, there was significant overlap among three of the measures. That is, accelerometer-based MVPA shared 8.3% variance with self-reported MVPA (r[66] = .29, p = .017, R² = .083) and 7.5% variance with self-reported walking (r[66] = .27, p = .024, R² = .075). Although the relationship was nonsignificant, there was 5.4% variance shared between self-reported sitting and accelerometer-based ST (r[66] = .23, p = .055, R² = .054).

3.6. MLMs

Given the shared variance between the self-reported and accelerometer-based measures of MVPA and their significant associations with the P300, we were interested in whether either of these measures explained unique variance in the target or inhibitory P300. Therefore, we conducted a series of MLMs to investigate this question.

3.6.1. Unconditional model

For the unconditional model, the mean intercept, b = 3.70, t(68) = 9.22, p < .001, and the variance of the intercept across people was significant, χ²(1) = 11.39, p < 0.001; therefore, we retained a random intercept in all the following MLMs. The ICC was 0.39, indicating that there was 39% between-person variance and 61% within-person variance.

3.6.2. Conditional models

In model 1, self-reported and accelerometer-based MVPA were examined in relation to the target and inhibitory P300. There was a trial type main effect, ^F(1, 68) = 34.66, p < .001, indicating that the inhibitory P300 was larger than the target P300. Accelerometer-based MVPA was significantly associated with the P300, regardless of trial type, F(1, 68) = 4.33, p = .041, while there was no association between self-reported MVPA and P300, regardless of trial type, ^F(1, 68) = 0.63, p = .429 (Table 2). Findings from this model indicate that accelerometer-based MVPA uniquely explains P300 amplitude over and above self-reported MVPA; therefore, the next models only retain accelerometer-based MVPA.

Table 2.

Model 1: Multilevel model effects on the mean amplitude of the target and inhibitory P300 as a function of trial type, self-reported and accelerometer-based MVPA

Fixed Effects Parameter Estimates
Effect	Estimate	SE	95% CI		df	t	p
			LL	UL
Intercept	3.70	0.38	2.95	4.46	68	9.65	<.001
Trial Type	1.27	0.22	0.85	1.69	68	5.89	<.001
Self-reported MVPA	0.32	0.40	−0.47	1.11	68	0.80	.429
Accelerometer-based MVPA	0.84	0.40	0.05	1.63	68	2.08	.041
Random Components
Groups		SD	Variance	ICC	df	χ 2	p
ID	Intercept	2.62	6.85	0.52	1	21.39	<.001
Residual		2.52	6.34

Note. Number of Observations: 136; Number of Groups (ID): 68. MVPA = moderate-to-vigorous physical activity.

In model 2, we aimed to determine whether accelerometer-based MVPA remained a significant correlate of P300 amplitude after covarying for age. Like model 1, there was a significant trial type main effect, F(1, 68) = 34.66, p < .001. Accelerometer-based MVPA remained a significant correlate of P300 amplitude, F(1, 68) = 4.45, p = .039, even after accounting for the unique relationship between age and P300, F(1, 68) = 4.21, p = .044, indicating that older adults have reduced P300 amplitudes across trial types (Table 3).

Table 3.

Model 2: Multilevel model effects on the mean amplitude of the target and inhibitory P300 as a function of trial type, accelerometer-based MVPA, and age

Fixed Effects Parameter Estimates
Effect	Estimate	SE	95% CI		df	t	p
			LL	UL
Intercept	3.70	0.37	2.97	4.44	68	9.9	<.001
Trial Type	1.27	0.22	0.85	1.69	68	5.89	<.001
Accelerometer-based MVPA	0.8	0.38	0.06	1.55	68	2.11	.039
Age	−0.78	0.38	−1.53	−0.03	68	−2.05	.044
Random Components
Groups		SD	Variance	ICC	df	χ 2	p
ID	Intercept	2.52	6.35	0.5	1	19.63	<.001
Residual		2.52	6.34

Note. Number of Observations: 136; Number of Groups (ID): 68. MVPA = moderate-to-vigorous physical activity.

As an exploratory analysis, in model 3, we assessed whether accelerometer-based MVPA was a moderator of these age-related reductions in P300 amplitude. Therefore, we specified a model that included all main effects and interactions between trial type, accelerometer-based MVPA, and age. Although model 3 echoed the results from model 2, with main effects observed for trial type, accelerometer-based MVPA, and age, there were no significant interactions in this model (Table 4).

Table 4.

Model 3: Multilevel model effects on the mean amplitude of the target and inhibitory P300 as a function of trial type, accelerometer-based MVPA, age, and their interactions

Fixed Effects Parameter Estimates
Effect	Estimate	SE	95% CI		df	t	p
			LL	UL
Intercept	3.71	0.38	2.97	4.45	68	9.8	<.001
Trial Type	1.30	0.21	0.88	1.72	68	6.1	<.001
Accelerometer-based MVPA	0.81	0.36	0.06	1.56	68	2.1	.039
Age	−0.78	0.38	−1.53	−0.02	68	−2.02	.047
Accelerometer-based MVPA x Age	0.04	0.37	−0.68	0.77	68	0.12	.906
Trial Type x Age	−0.22	0.22	−0.64	0.21	68	−1.01	.317
Trial Type x Accelerometer-based MVPA	0.21	0.22	−0.21	0.64	68	0.98	.329
Trial Type x Accelerometer-based MVPA x Age	0.19	0.21	−0.21	0.6	68	0.93	.354
Random Components
Groups		SD	Variance	ICC	df	χ 2	p
ID	Intercept	2.55	6.5	0.52	1	21.22	<.001
Residual		2.46	6.05

Note. Number of Observations: 136; Number of Groups (ID): 68. MVPA = moderate-to-vigorous physical activity.

Lastly, we conducted a follow-up analysis to determine whether the target and inhibitory P300s were better explained by accelerometer-based MVPA or Lifestyle PA. In this analysis (model 4), after covarying for age, accelerometer-based MVPA and Lifestyle PA were both not significantly associated with P300 amplitude, indicating that they accounted for overlapping variance in P300 amplitude (Table 5).

Table 5.

Model 4: Multilevel model effects on the mean amplitude of the target and inhibitory P300 as a function of trial type, accelerometer-based MVPA and lifestyle PA, and age

Fixed Effects Parameter Estimates
Effect	Estimate	SE	95% CI		df	t	p
			LL	UL
Intercept	3.70	0.37	2.98	4.43	68	10.06	<.001
Trial Type	1.27	0.22	0.85	1.69	68	5.89	<.001
Accelerometer-based MVPA	−0.25	0.80	−1.82	1.32	68	−0.31	.758
Accelerometer-based Lifestyle PA	0.02	0.01	−0.01	0.05	68	1.49	.142
Age	−1.00	0.40	−1.79	−0.21	68	−2.48	.016
Random Components
Groups		SD	Variance	ICC	df	χ 2	p
ID	Intercept	2.46	6.06	0.49	1	18.55	<.001
Residual		2.52	6.34

Note. Number of Observations: 136; Number of Groups (ID): 68. MVPA = moderate-to-vigorous physical activity; PA = physical activity.

4. Discussion:

The goal of the current study was to examine cross-sectional associations between P300 amplitude, self-reported and accelerometer-based measures of PA and ST in a community sample of older adults. Specifically, we aimed to determine whether self-reported and accelerometer-based measures of PA and ST explained unique or overlapping variance in P300 amplitude elicited during a Go/No-Go task. Overall, results showed associations between MVPA and P300 amplitude across measurement modalities such that a greater PA was significantly associated with larger P300s across inhibitory and target detection processes. However, there were no relationships between the self-reported sitting and accelerometer-based ST measures. Moreover, using MLMs exploring whether accelerometer-based or self-report measures of MVPA better explained the inhibitory and target P300s, we found that accelerometer-based MVPA provided unique insight into both P300s over and above self-reported MVPA in older adults. Lastly, our work suggests potential methodological advantages and limitations to the current protocols and thresholds used within the field of PA measurement and its applications.

A primary goal of our study was to examine the relationships between accelerometer-based and self-report measures of PA and ST. Moreover, based on the novelty of accelerometer-based measures in work involving the P300, we sought to assess whether the two methods (i.e., self-report and accelerometer-based) were comparable or if one method enables a better view into PA or ST’s association with P300 amplitude. Although some of the self-reported and accelerometer-based PA and ST measures were correlated, their associations were fairly weak at best—only sharing approximately 5-8% variance. For measurement techniques that are believed to tap into the same constructs, the weak relationships observed herein suggest that the self-reported and accelerometer-based measures of PA are not one-in-the same variables and should not be used interchangeably. Upon further examination, we demonstrated that accelerometer-based MVPA explains unique variance in the P300 over and above self-reported MVPA. These relationships were not observed between P300 activity and accelerometer-based or self-reported ST. This observed variance hints at the unique role accelerometer-based MVPA may have on both target detection and inhibitory control processes during aging compared to ST, instead of the two opposite behavior types equally impacting cognitive functioning. It is important to note, however, that our work is cross-sectional in nature, therefore we cannot establish a cause-effect relationship. Future work should look at increasing PA in daily living as an interesting and plausible avenue of further investigating cognitive effects in older adults, who have been found to spend an estimated 60% of their day sedentary (Leung et al., 2021. In sum, these findings provide neurophysiological and accelerometer-based evidence of the positive association between physical activity and multiple facets of cognitive functioning in older adults (Kumar et al., 2022; Xu et al., 2023).

Interestingly, much like the recent work of Maasakkers et al. (2022), our work found no relationships between STs and cognitive functioning. However, it is important to note that these null associations may be due to the limitation posed by measuring ST in both our self-report instrument (i.e., IPAQ) and accelerometers. That is, we did not assess the types of activities that older adults engaged in while sedentary. Some activities that are commonly engaged in are cognitively stimulating (e.g., puzzles, reading, knitting) whereas others are less so (e.g., watching television). To better understand ST collected via accelerometer-based measures, some form of in-vivo self-report (e.g., ecological momentary assessment) may be necessary. Given that we only assessed overall time spend sedentary, future avenues of research may aim to assess the contributions of different activities performed while sedentary in relation to cognitive function during aging.

Based on previous studies demonstrating that wrist-worn accelerometers can mis-classify different light-intensity physical activities into different intensities (e.g., MVPA; Fanning et al., 2022) and the role of incidental physical activity in relation to cognition during aging (Sanchez-Lopez et al., 2018), an additional variable was created in accordance with the cut-points for older adults by Lohne-Seiler et al. (2014), which we referred to as accelerometer-based Lifestyle PA. It was noted that both accelerometer-based Lifestyle PA and MVPA demonstrated a high correlation with each other; both variables were also associated in the same direction with the P300 difference scores. Therefore, we conducted a MLM to determine whether they accounted for overlapping or unique variance in both P300s and found that both variables accounted for shared or overlapping variance. This suggests that PA’s influence on P300 and thus, cognitive functioning may be more about engaging in any type of physical activity, rather than specific intensities of activity.

As accelerometer-based measures become more prevalent in research settings, the above findings highlight an important methodological consideration regarding the use of wrist-worn accelerometers in the measurement of PA in older adults. Activities of daily living that older adults engage in might be mis-classified, which has been noted previously by Fanning et al. (2022), wherein cut-points for accelerometer-based measures may need to be more narrowly defined. For example, to our knowledge, there is only one validated set of cut-points used for accelerometers worn on the wrist among older adults, which were created from laboratory-based data and may not have successfully mimicked older adults’ daily life activities (Migueles et al., 2021). Future studies are needed to validate additional cut-points across different populations and in free-living settings to provide more accurate estimates of PA (and ST) derived from wrist-worn accelerometers, which can then have application in studies of individual differences.

Lastly, we believe it is important to draw attention to the negative relationship present between P300 amplitude and age within our study, which is a known association that has been demonstrated in previous work (e.g., Brush et al., 2021; Ford et al., 1997; Friedman et al., 1997; Polich, 1997), and could be viewed as a proxy for age-related cognitive decline. To account for this association, we included age as a covariate in sensitivity MLM analyses, and interestingly, MVPA, as assessed using an accelerometer, remained significantly associated with the P300s across trial types. Future work should look to see if cognitive processing impairments are more obvious in subgroups based on age and how physical activity is able to either predict or impact this relationship.

This study is among the first to assess relationships between cognitive functioning as measured by the P300 and both self-reported and accelerometer-based PA and ST in older adults. Overall, our work showed associations between PA, but not ST, and P300 activity, within self-report and accelerometer-based measures. These associations remained after accounting for age, which demonstrates an inverse relationship with P300 amplitude. In addition, we found that accelerometer-based measures provided unique insight across target detection and inhibitory control processes in older adults over and above the self-reported measures. Lastly, our work shows evidence demonstrating the need for standardizing and harmonizing scoring efforts of accelerometer-based measures in older adults, since findings may be influenced by the scoring method applied. Although our results point to the importance of MVPA in relation to cognitive function in older adulthood, as measured via the P300, further examination suggests that engaging in any type of physical activity (regardless of intensity) may be equally as important during this age range. Therefore, engagement in PA should continue to be emphasized as a possible means for supporting cognitive function during aging.

Figure 2. Grand-averaged stimulus-locked event-related potential (ERP) difference waveforms at the Pz electrode site (N = 75).

Note. Negative is plotted up and the time window in which P300 amplitude was scored was between 350 and 600 ms post-stimulus. The inhibitory waveform reflects the subtraction-based Astronaut-minus-Alien (i.e., No-Go-minus-Frequent-Go) difference score. The target waveform reflects the subtraction-based Asteroid-minus-Alien (i.e., Infrequent-Go-minus-Frequent-Go) differences score. FGo = Frequent-Go; Igo = Infrequent-Go.

Figure 3. Grand-averaged stimulus-locked event-related potential (ERP) difference waveforms at the Pz electrode site split by accelerometer-based MVPA levels.

Note. Negative is plotted up and the time window in which P300 amplitude was scored was between 350 and 600 ms post-stimulus. The inhibitory waveform reflects the subtraction-based Astronaut-minus-Alien (i.e., No-Go-minus-Frequent-Go) difference score. The target waveform reflects the subtraction-based Asteroid-minus-Alien (i.e., Infrequent-Go-minus-Frequent-Go) differences score. FGo = Frequent-Go; Igo = Infrequent-Go. The high MVPA (n = 37) and low MVPA (n = 38) groups were created for visual purposes only using a median-split of the accelerometer-based MVPA levels.

Figure 4. Scatterplot depicting the multilevel model effects on the mean amplitude of the target and inhibitory P300 as a function of trial type, accelerometer-based MVPA, and age.

Note. Number of Observations: 136; Number of Groups (ID): 68. MVPA = moderate-to-vigorous physical activity. This plot uses the model-implied estimates.

Highlights:

• Increased physical activity but not sedentary time was associated with increased inhibitory control and target detection.

• Relative to self-reports, accelerometer-based measures may better explain individual differences in inhibitory control

• Accelerometer-based findings may be influenced by the scoring method applied, therefore establishing more standards for scoring accelerometer-based measures in older adults is crucial.