Skip to main content
NCBI home page
Alzheimer's & Dementia : Translational Research & Clinical Interventions logoLink to Alzheimer's & Dementia : Translational Research & Clinical Interventions
. 2025 Dec 22;11(4):e70187. doi: 10.1002/trc2.70187

Association between 24‐h time‐use composition and brain age: The IGNITE study

Audrey M Collins 1,✉, Maddison L Mellow 2, Lu Wan 1, Ashleigh E Smith 2, Lauren E Oberlin 1,3, Kelsey R Sewell 1,4, Neha P Gothe 5,6,7, Jason Fanning 8, Jairo H Migueles 9, Dorothea Dumuid 2, Aaron Miatke 2, John M Jakicic 10, Chaeryon Kang 11,12, George Grove 13, Haiqing Huang 1, Bradley P Sutton 14,15, Anna L Marsland 13, M Ilyas Kamboh 16, Arthur F Kramer 14, Charles H Hillman 6,7,17, Eric D Vidoni 18, Jeffrey M Burns 18, Edward McAuley 14,19, Kirk I Erickson 1,20
PMCID: PMC12722602  PMID: 41445670

Abstract

INTRODUCTION

The relationships between 24‐h time‐use composition (i.e., sleep, sedentary behavior, light physical activity, and moderate‐to‐vigorous physical activity [MVPA]) and brain morphology in older adulthood remain poorly understood. We examined associations between 24‐h time‐use composition and brain age using compositional data analysis, predicting that 24‐h time use would be associated with brain age and that a greater amount of time engaged in MVPA would drive associations with younger brain age.

METHODS

Baseline data from the Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE; n = 648) were analyzed. Brain age was estimated using T1‐weighted magnetic resonance imaging data. Time‐use composition was derived from wrist‐worn triaxial accelerometers. Regression models examined associations between 24‐h time‐use composition (expressed as isometric log ratios) and brain age, adjusting for age, sex, apolipoprotein E4 (APOE4) carriage, education, body mass index, image quality, and site. Compositional isotemporal substitution evaluated how hypothetical reallocations of time between behaviors related to brain age.

RESULTS

The final sample included 573 adults (69.8±3.7 years, 407 females). It was found that 24‐h time‐use composition was associated with brain age (F = 2.72, p = 0.004). Post hoc modeling indicated that time spent in MVPA primarily drove these associations, such that less MVPA was associated with greater brain age, irrespective of whether time was taken from sleep, sedentary behavior, or light physical activity.

DISCUSSION

These results suggest that 24‐h time use, especially time spent in MVPA, relates to structural brain age in late adulthood. Maintaining or increasing MVPA may help preserve younger brain age, irrespective of which behaviors this time was reallocated from. Future research should examine whether systematically shifting 24‐h time use toward MVPA alters brain aging trajectories.

Clinical Trial Registration Number and Name of Trial Registry: ClinicalTrial.gov: NCT02875301

Highlights

  • Time use relates to brain age in older adults.

  • More time spent in MVPA may contribute to younger brain age.

  • Associations between time use and brain age are independent of demographic variation or genetic risk for AD.

Keywords: brain age, compositional data analysis, physical activity, sedentary behavior, sleep

1. BACKGROUND

The brain undergoes age‐related changes across the lifespan, including progressive atrophy of gray and white matter volume (particularly in the hippocampus, temporal, and frontal regions), along with changes in white matter microstructure and functional connectivity. 1 , 2 , 3 These changes are clinically important, with the rate of change predictive of cognitive impairment and increased dementia risk. 4 , 5 Individual variability in the magnitude and rate of age‐related brain atrophy may be influenced by chronological age, sex, education, apolipoprotein E4 (APOE4) carriage, body mass index (BMI), and other biological and lifestyle markers. 3 , 6 , 7 , 8

Across the 24‐h day, time is mostly spent in sleep, sedentary, and physical activity behaviors, each of which has been individually related to brain morphology. Short sleep duration (<7 h) and poor sleep quality have been related to accelerated brain atrophy in late life. 9 , 10 The impact of sedentary behavior (SB) on brain health remains inconclusive and mixed, with systematic reviews linking greater SB to poorer brain health, including structural white matter health. 11 , 12 These relationships may be dependent upon the methodology for measuring SB, the classification of SB, and brain health outcome. In contrast, accumulating active time may mitigate age‐related brain changes. Longitudinal observational studies indicate that more physically active older adults have more favorable brain structure, supported by interventional work demonstrating that exercise can increase cortical and subcortical volumes. 13 , 14 , 15 , 16 Public health guidelines conclude that moderate‐to‐vigorous physical activity (MVPA) affects brain structure and other brain health outcomes in late adulthood. 17

Although 24‐h time‐use behaviors may collectively influence markers of brain health, they are rarely analyzed together in the same statistical model due to their perfect multicollinearity. Compositional data analysis (CoDA) overcomes this by expressing the composition of time spent in each behavior as isometric log‐ratio coordinates. 18 The isometric log ratios can therefore be used to represent the 24‐h “time‐use composition” in standard statistical models. 18 Using this approach, two studies have reported inconsistent findings. Mellow and colleagues found that time use was not associated with total or regional gray matter volume in cognitively unimpaired older adults, while Balbim and colleagues reported a relationship between 24‐h time‐use composition and gray matter volume in older adults with mild cognitive impairment. 19 , 20 Further work is needed to clarify these relationships.

BrainAGE, or brain‐predicted age difference (brain‐PAD), is a personalized biomarker to detect accelerated structural brain aging based on machine‐learning algorithms that predict an individual's age from anatomical brain images. The predicted brain age is then compared with the individual's chronological age to produce a brain‐PAD score. 21 , 22 Higher brain‐PAD score has been associated with, and is predictive of, an increased risk of cognitive decline, Alzheimer's disease, mortality, and a decline in physical function. 22 , 23 , 24 , 25 Brain‐PAD may provide more precise and nuanced estimates of subsequent dementia risk and progression from mild cognitive impairment to Alzheimer's disease than chronological age, neuropsychological testing, hippocampal volume, or select cerebrospinal fluid biomarkers. 24 Its utility for early detection of structural brain aging is highlighted by studies of individuals with and without mild cognitive impairment or Alzheimer's disease. 22 , 24

RESEARCH IN CONTEXT

  1. Systematic review: The extant literature was searched systematically using traditional database searches. Previous work reported relationships between 24‐h time use and gray matter volume but did not examine associations with brain age.

  2. Interpretation: Brain age was associated with 24‐h time‐use composition. These relationships were driven by time spent in MVPA.

  3. Future directions: Longitudinal and intervention studies are needed to examine whether systematically shifting 24‐h time‐use allocation toward more MVPA would improve brain age.

We previously reported on time‐use behaviors, fitness, and brain health outcomes in this sample. 26 , 27 , 28 , 29 Here, we extend that work to examine whether 24‐h time‐use composition is associated with brain‐PAD. We hypothesized that time use would be associated with brain‐PAD and that greater time spent in MVPA (relative to time spent in sleep, SB, and light physical activity [LPA]) would be associated with a younger brain‐PAD. We also explored whether this association would be moderated by chronological age, sex, education, and APOE4 carriage.

2. METHODS

2.1. Participant recruitment and screening

The current analysis used baseline data from the Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE) study, which collected data from 648 adults 65 to 80 years of age (69.9 ± 3.7 years; 71% female) across three sites (University of Pittsburgh, University of Kansas Medical Center, Northeastern University) (ClinicalTrial.gov: NCT02875301). Baseline data were collected from September 2017 through January 2022. Details of the protocol, recruitment, and exclusion and inclusion criteria have been previously described (and elaborated on in Appendix 1). 26 , 30 In short, participants were excluded if they had a current diagnosis of an Axis I or II disorder (e.g., major depression, substance use disorder, schizophrenia), current diagnosis or history of a neurological condition (e.g., multiple sclerosis, Parkinson's disease, dementia, stroke), current or recent cardiovascular events (e.g., congestive heart failure, myocardial infarction, angioplasty), current type I diabetes, current insulin‐dependent or uncontrolled type II diabetes, magnetic resonance imaging (MRI) contraindications or inability to complete a MRI scan, and self‐reported activity levels that exceeded 20 min of structured MVPA per day for at least 3 days per week over the last 6 months. Prior to data collection, participants provided written informed consent approved by the Institutional Review Board at each site. 30

2.2. Study measures

2.2.1. Image acquisition and brain‐PAD calculation

Participants underwent a brain MRI scan. Preprocessing pipelines (including the generation of a quantitative image quality rating [IQR] metric) are available in Appendix 2. Brain age estimation and calculation of brain‐PAD was performed using the brainageR analysis pipeline (github.com/james‐cole/brainageR). 31 The T1‐weighted images were preprocessed at the voxel level using the SPM12 software, 32 including segmentation and normalization. Raw T1‐weighted images were segmented into gray matter, white matter, and cerebrospinal fluid, followed by non‐linear spatial normalization to a template using SPM12's Diffeomorphic Anatomical Registration using Exponentiated Lie Algebra (DARTEL) toolbox. Images were resampled using a voxel size of 1.5 mm and smoothed with a Gaussian spatial smoothing kernel of 4 mm at full‐width at half‐maximum (FWHM). Visual quality control of the probabilistic tissue maps was performed to ensure accuracy of the segmentation.

BrainageR was previously trained to predict age from normalized brain volumetric maps of 3377 healthy individuals (aged 18 to 92 years) using a Gaussian process regression model and has shown high accuracy and test–retest reliability with a mean absolute error of <5 years. 31 , 33 This pretrained model was applied to the preprocessed imaging data to estimate brain age for each participant. Brain‐PAD was further calculated by subtracting chronological age (in whole years) from the predicted brain age for each participant, such that higher brain‐PAD values indicated accelerated brain aging.

2.2.2. Device‐measured time‐use composition

Triaxial accelerometers (ActiGraph GT9X Link; Ametris (formerly ActiGraph), Pensacola, FL, USA) were initialized at a sampling frequency of 70 Hz, with Idle Sleep Mode disabled, and worn on the non‐dominant wrist for 7 continuous days. Participants were informed that the devices could be worn during bathing and swimming activities. Raw data (.gt3x files) were processed using Version 2.9–0 of GGIR. 34 Processing (see Appendix 3 for the GGIR configuration file) and quality control procedures (see Appendix 4 for additional details) were described previously. 29 In short, sleep analysis employed the vanHees 2015 and Heuristic algorithm looking at Distribution of Change in Z‐Angle (HDCZA) algorithms to separate sleep period time from awake time in each day. 35 , 36 Then epochs were aggregated to 60 s for physical activity analysis. Waking behaviors were classified by acceleration‐based cut points to identify time spent in SB (<35 milligravity [mg]), LPA (35 to 99 mg) and MVPA (≥100 mg). Days were defined from midnight to midnight, where the plain average of minutes per valid day (sensor worn for at least 16 h and two‐thirds of waking hours) spent in each time‐use behavior (sleep period time window, SB, LPA, and MVPA) were used when creating the time‐use compositions.

2.2.3. Covariates and moderators

Chronological age (years), sex (male/female), formal education (years), study site (University of Pittsburgh, University of Kansas Medical Center, Northeastern University), APOE4 allele carriage (carrier/noncarrier), BMI (kg/m2), and IQR were accounted for as covariates.

Chronological age was included to account for a potential age bias. 37 IQR of the T1 image, reflecting noise and motion artifacts, was included as a covariate due to its potential impact on brain‐PAD analysis and interpretation. 38 Study site was included to account for potential site‐related differences. TaqMan assays from blood samples were used for APOE genotyping, resulting in six genotypes. 26 In the current analysis, homozygous and heterozygous carriers of an APOE4 allele were identified as “carriers.” Height and weight were measured using calibrated scales and stadiometers for the calculation of BMI.

Chronological age, sex, education, and APOE4 carriage were tested as moderator variables as they are well‐recognized predictors of atrophy and dementia risk. 3 , 7 , 8 , 37

2.3. Data analyses

Analyses were performed in R (version 4.4.1). 39 Components of the time‐use composition were checked for missingness. 19 Two participants had zero minutes of MVPA, assumed to be censored zeros, that is, with a longer or more sensitive measurement, some MVPA would eventually be recorded. These zero values were replaced by small values imputed using expectation‐maximization algorithms via the lrEM function in zCompositions. 40 The outcome (brain‐PAD), predictor (24‐h time‐use composition), covariates (chronological age, sex, education, site, APOE4 carriage, BMI, and IQR), and residuals were checked for normality and extreme skewness.

We first examined associations between time use and brain‐PAD using traditional (non‐compositional) approaches that examined each time‐use component independently. Pairwise correlations explored relationships between individual time‐use behaviors (minutes/day in sleep, SB, LPA, and MVPA), brain‐PAD, covariates, and moderators. Pearson or point biserial correlations were performed for pairwise correlations outside of the time‐use composition. For correlations between the time‐use variables, symmetric balanced isometric log‐ratio coordinates were applied. 41 Four linear regression models examined associations between independent time‐use behaviors and brain‐PAD while adjusting for covariates specified earlier. Due to skewness, MVPA was log‐transformed for this analysis. Both traditional and compositional analytical methods were employed to understand how relationships may be attenuated after accounting for other time‐use behaviors.

Compositional analyses were performed using the R compositions package (version 2.0–8). 42 One 24‐h time‐use composition was created per participant to represent the average proportion of the day spent in each time‐use component (sleep, SB, LPA, and MVPA). Time‐use components were proportionally rescaled to a fixed sum of 1440 min. 42 The four‐part 24‐h time‐use compositions were transformed into three isometric log‐ratio coordinates following CoDA guidelines and the structure previously described by Mellow and colleagues. 18 , 43

Linear regression models investigated the associations between time‐use composition and brain‐PAD, adjusting for chronological age, sex, formal education, study site, APOE4 carriage, BMI, and IQR. Multivariate type III F‐tests were performed using the car package to determine overall variable significance, where alpha = 0.05. 44 To account for the possibility that relationships between time‐use composition and brain‐PAD were non‐linear (e.g., inverted U‐shaped associations with sleep), we fit additional models where time‐use composition was expressed using quadratic (squared) terms. 19 An ANOVA was performed to evaluate whether expressing time‐use composition using squared terms improved the model fit. Moderation of the association between time‐use composition and brain‐PAD was tested via interaction terms with time‐use composition and chronological age, sex, education, or APOE4 carriage. A sensitivity analysis was performed to confirm that the relationships observed were not due to minimal or excessive accelerometer wear time (by removing individuals with less than 4 total days, including less than 3 weekdays or less than 1 weekend day, or more than 10 total days).

If the 24‐h time‐use composition was significantly associated with brain‐PAD, compositional isotemporal substitution methods were used to explore how hypothetical reallocations of time were associated with brain‐PAD. 18 This was performed using the codaredistlm package, where model‐generated predictive response curves demonstrated proportional reallocations while using the sample mean as the reference composition. 45 If modeled interaction terms with 24‐h time‐use composition were significant, compositional isotemporal substitution methods were conducted separately by level of the corresponding variables.

3. RESULTS

3.1. Sample characteristics

Baseline data from 573 participants with complete brain‐PAD, actigraphy, and covariate data were included with the descriptive characteristics displayed in Table 1. There were no significant differences in demographic characteristics between this subsample and the total sample enrolled in IGNITE (n = 648) (Appendix 5), suggesting that missingness was likely at random (see Appendix 4 for flow diagram). The 24‐h time‐use footprint indicates that, on average, participants spent approximately 7.5 h in sleep, 12.0 h in SB, 4.0 h in LPA, and 0.5 h in MVPA. These are visualized in Figure 1 and proportionally represent 31% of the 24‐h day in sleep, 50% in SB, and 19% in PA (combining LPA and MVPA).

TABLE 1.

Sample characteristics.

Characteristic Level

Mean (SD)

or n (%)

Range

(min, max)
No. participants 573
Site Pittsburgh 199 (35.0%)
Kansas City 198 (35.0%)
Boston 176 (30.0%)
Chronological age (years) 69.8 (3.7) 65.0, 80.0
Sex (female) 407 (71.0%)
Education (years) 16.3 (2.2) 10.0, 20.0
BMI (kg/m2) 29.6 (5.6) 18.0, 50.9
APOE4 carriers 157 (27.0%)
Predicted brain age (years) 66 (8) 41, 89
Brain‐PAD (years) −4 (7) −27, 14
24‐h time‐use composition (arithmetic means, (min/day) Sleep 454 (78)
SB 719 (114)
LPA 236 (84)
MVPA 31 (26)
24‐h time‐use composition (geometric means under closure, (min/day) Sleep 460
SB 730
LPA 228
MVPA 22
Open in a new tab

Note: 24‐h time‐use composition means are presented as both arithmetic means and geometric means (both under closure to 1440 min).

Abbreviations: BMI = body mass index; kg = kilograms; m = meters; APOE = apolipoprotein E; SB = sedentary behavior; LPA = light physical activity; MVPA = moderate‐to‐vigorous physical activity; Brain‐PAD = brain‐predicted age difference (i.e., difference between chronological age and predicted brain age).

FIGURE 1.

FIGURE 1

Open in a new tab

Ternary plots of time‐use compositions. Each gray dot represents an individual participant's 24‐h time‐use composition (i.e., the geometric mean after closure to 1440 min). The center black dot represents the mean 24‐h time‐use composition of the entire sample. PA = sum of LPA and MVPA. Values along the axes represent the proportion of the day in each behavior. Black ellipses represent 75%, 95%, and 99% confidence intervals, respectively. On average, participants spent 31% of the 24‐h day in sleep, 50% in SB, and 19% in PA.

3.2. Associations between individual time‐use components and brain‐PAD

A correlation matrix displaying the correlations between time‐use components in a non‐compositional framework, brain‐PAD, and covariates can be found in Appendix 6. When considering each time‐use behavior independently in linear regression models, there were no associations between sleep, SB, or LPA and brain‐PAD after adjusting for covariates (Appendix 7). In contrast, log‐transformed MVPA was significantly associated with brain‐PAD (p = 0.01).

3.3. Association between time‐use composition and brain‐PAD

There was a significant relationship between 24‐h time‐use composition and brain‐PAD using linear regression, after covarying for chronological age, sex, education, site, APOE4 carriage, BMI, and IQR (F = 3.64, p = 0.01; see Appendices 8 and 9 for comparisons by site and sex, respectively). ANOVA results were statistically significant when comparing the linear and polynomial models (F = 2.23, p = 0.04), confirming the preferred expression of the 24‐h time‐use composition using squared terms (second‐order polynomial terms). Therefore, the final model expressed the 24‐h time‐use composition using polynomial terms, where a significant relationship was found with brain‐PAD after controlling for covariates (F = 2.72, p = 0.004; Table 2). When performing a sensitivity analysis by removing individuals with less than 4 (including less than 3 weekdays or less than 1 weekend day) or more than 10 valid days of accelerometer data, brain‐PAD remained associated with 24‐h time‐use composition (F = 2.98, p = 0.002).

TABLE 2.

Association between time‐use composition and brain‐PAD.

Brain‐PAD
Variable F P
Site 11.31 <0.001
Sex 32.64 <0.001
APOE4 carriage 1.24 0.27
Chronological age 0.66 0.42
Education 0.10 0.76
BMI 0.07 0.79
IQR 2.52 0.11
Time‐use composition 2.72 0.004
Open in a new tab

Note: Multivariate F test outputs (F = F statistic; where bold p values indicate statistical significance at an alpha level of 0.05) for regression models investigating the association between 24‐h time‐use composition and brain‐PAD. 24‐h time‐use composition was expressed using squared (second‐order polynomial) terms. See Appendices 8 and 9 for statistical output of comparisons by site and sex, respectively.

Abbreviations: APOE4 carriage = apolipoprotein E4 allele carriage (carrier or non‐carrier); BMI = body mass index; IQR = image quality rating.

The final model was a post hoc compositional isotemporal substitution analysis. Proportional reallocation plots can be found in Figure 2. Given that the reallocation curves for sleep, SB, and LPA are generally flat with confidence intervals crossing zero on the x‐axis, it can be concluded that these associations were not significant and that time spent in MVPA drove the relationship between 24‐h time‐use composition and brain‐PAD. Therefore, hypothetically increasing or decreasing time spent performing MVPA (while proportionally taking time from sleep, SB, and LPA) accounted for model‐predicted differences in brain‐PAD. One‐for‐one reallocation plots can be found in Appendix 10, which represent direct reallocations of time between individual behaviors (e.g., hypothetically increasing time in one behavior while directly drawing time from another behavior).

FIGURE 2.

FIGURE 2

Open in a new tab

Proportional reallocation plots. Note that Figure 2 displays the associations between model‐predicted differences in brain‐PAD (y‐axis) and reallocations of time spent performing time‐use behaviors (listed in header). Reallocations range from −20 min to +20 min (x‐axis) away from the geometric mean of each time‐use behavior presented in the header (e.g., mean MVPA=22 min, thus −20 min and +20 min reallocations equate to engaging in 2 min or 42 min of MVPA per day, respectively). Reallocations are based on a linear regression model with brain‐PAD as the outcome, and the following predictors (based on an identified “average case” participant): site = Northeastern University; chronological age = 69.8 years; sex = female; education = 16.3 years; BMI = 29.6 kg/m2; APOE4 carriage = carrier; IQR = 0.79. Zero on the y‐axis represents the conditional mean brain‐PAD of −3.45 years (i.e., the mean brain‐PAD associated with the average case previously specified). Shading of reallocation curves represent 95% confidence intervals. For simplicity, we have modeled the pattern using the linear term.

Figure 2 shows the hypothetical impact of reallocating time toward or away from the sample‐mean MVPA of 22 min per day (at zero on the x‐axis; fourth panel) on brain‐PAD. Reallocations involving MVPA displayed an asymmetric and non‐linear pattern, where the predicted detrimental association of reallocating time away from MVPA was larger in absolute terms than the beneficial association observed when the equivalent duration of time was hypothetically reallocated toward MVPA (Figure 2). For example, when hypothetically reallocating 20 min/day away from MVPA (i.e., −20 on the x‐axis), brain‐PAD was estimated to increase by approximately 2.5 years (95% CI: 0.85, 4.21). However, when 20 min/day was reallocated towards MVPA (i.e., +20 on the x‐axis), brain‐PAD was estimated to decrease by approximately 0.65 years (95% confidence interval [CI]: −1.09, −0.22). Nonetheless, once surpassing the mean of ∼22 min of MVPA per day (indicated by the zero on the x‐axis), hypothetically spending more time in MVPA (at the equal expense of remaining behaviors) corresponded with a better brain‐PAD. Of note, the largest estimated difference in predicted brain‐PAD was observed when comparing the −20 and −15 min reallocations. Given the geometric mean MVPA (at 0 on the x‐axis) was 22 min/per day, this can be approximated as the cross‐sectional difference between engaging in 2 min of MVPA per day and 7 min of MVPA per day, respectively.

3.4. Moderation analyses

Chronological age, sex, education, and APOE4 carriage did not moderate the association between 24‐h time use composition and brain‐PAD (Appendix 11).

4. DISCUSSION

This study examined cross‐sectional associations between 24‐h time‐use composition and brain‐PAD in cognitively unimpaired older adults. As hypothesized, 24‐h time‐use composition was associated with brain‐PAD after covariate adjustment. Post hoc isotemporal substitution analysis indicated that theoretically decreasing time in MVPA while proportionally or directly reallocating time toward sleep, SB, or LPA was associated with less favorable brain‐PAD. In contrast, hypothetically increasing time in MVPA was protective but with smaller effects. The greatest hypothetical (cross‐sectional) difference was observed for reallocations of −20 and −15 min of MVPA, which approximates the difference between engaging in 2 min of MVPA to 7 min of MVPA per day. This hypothetical difference predicts approximately a 1.4‐year difference in brain‐PAD relative to the average case in this sample. These findings highlight the importance of MVPA, suggesting that maintaining or increasing MVPA, particularly for those engaging in minimal MVPA, may be favorable for brain morphology in late adulthood.

These results align with evidence from non‐CoDA studies demonstrating the benefits of MVPA on brain structure in late adulthood. 15 , 16 To our knowledge, only two studies have applied CoDA and isotemporal substitution analysis to examine 24‐h time‐use composition and brain volume in older adults. Mellow and colleagues found no direct associations with volumetric outcomes, although spending more time in MVPA was associated with better long‐term memory in those with smaller frontal volumes. 19 Balbim and colleagues found that reallocating 30 min between SB and MVPA in older adults with mild cognitive impairment was associated with gray matter volume differences in the right inferior temporal gyrus (TE2a region). 20 Discrepancies across studies may reflect differences in populations (e.g., cognitively unimpaired, inactive older adults) and brain measures (e.g., brain‐PAD vs gray matter volume). 19 , 20 Notably, brain‐PAD integrates whole‐brain voxel‐wise patterns of structural atrophy rather than relying on region‐of‐interest volumetric or thickness measures.

Mechanistic pathways linking MVPA to brain health may include increases in cardiorespiratory fitness, associated with preserved region‐specific brain integrity in this sample and others. 15 , 27 , 28 MVPA‐induced cellular and molecular changes, such as modulation of neurotrophic factors and insulin‐like growth factor, also likely partially explain these associations. 46 However, additional research is needed to better understand mechanistic pathways linking time use to brain structure. While none of the tested factors moderated the association between 24‐h time‐use composition and brain‐PAD in the current analysis, future longitudinal work should examine other factors (e.g., sleep quality, diet, or social engagement) to better understand heterogeneous patterns of aging.

We found no significant associations between brain‐PAD and most 24‐h time‐use variables using traditional correlation and multiple regression, except for log‐transformed MVPA. With a CoDA approach, we observed a non‐linear relationship between MVPA and brain‐PAD, whereby hypothetically increasing time in MVPA from almost none (e.g., 2 min/day, at the −20 min reallocation on Figure 2) to a little each day (e.g., 7 min/day, at the −15 min reallocation) was associated with substantial reductions in brain‐PAD, with benefits persisting but diminishing in magnitude at higher levels. 18 This is consistent with the relative nature of time reallocations where a 5‐min/day increase in MVPA would represent a much larger proportional change for someone with low baseline MVPA. These findings align with evidence that health benefits are greatest when sedentary individuals increase activity from none to some, consistent with public health recommendations that any physical activity is better than none. 47 , 48 Taken together, these results suggest that future studies could benefit from examining 24‐h time‐use behaviors using both CoDA and traditional analytical approaches across a wider range of MVPA to test for threshold benefits and ceiling effects. Given that our sample was recruited to be low active, we cannot generalize findings to more active individuals.

The mean brain‐PAD of the sample was −4 ± 7 years. The negative mean brain‐PAD value suggests that, on average, participants had a younger predicted brain age than their chronological age, indicative of a “younger appearing” brain relative to the reference population used to train the algorithm. This pattern may reflect that our sample was, on average, healthier than the normative sample used in model development. This difference has implications for generalizability, as a healthier and more selective sample may not fully represent the broader aging population. Consequently, the associations observed in our study may underestimate age‐ or disease‐related brain changes that would be more pronounced in more heterogeneous or clinically diverse cohorts.

As a cross‐sectional study, the compositional isotemporal substitution analysis reflects theoretical between‐person reallocations of time, limiting insight into within‐person changes or causal inference. Previous work suggested that risk of a future dementia diagnosis increases by 3% with every additional year of brain‐PAD. 49 Therefore, longitudinal and intervention studies are needed to better understand how changes in 24‐h time use or MVPA affect brain‐PAD and to understand the clinical relevance of the predicted changes in brain‐PAD. Although brain‐PAD provides a personalized and global index of structural brain integrity, it lacks regional specificity, which may limit targeted interventions. In addition, our analysis focused only on duration parameters of lifestyle behaviors but did not account for other dimensions that may relate to brain structure (e.g., sleep quality or context of time‐use behaviors). 9 , 50 Lastly, GGIR quantified SB without postural context, capturing inactivity rather than postural definitions of SB. Future studies should also evaluate cut points for movement intensity for this age range (65 to 80 years) and for wrist‐worn devices such as the ActiGraph GT9X Link.

In conclusion, these findings hold clinical significance as brain‐PAD is a biological marker of brain health based on structural neuroimaging data that is predictive of clinical outcomes in late adulthood. 22 , 23 , 24 , 25 Using rigorous objective measurement techniques (including neuroimaging and accelerometry) in a sizable and well‐characterized sample of older adults, our findings link a marker of estimated brain age to time spent in lifestyle behaviors across the 24‐h day for the first time. These relationships were driven by time spent in MVPA, underscoring the need for public health campaigns to promote and preserve MVPA for structural brain health in aging.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest relevant to this article. KIE consults for MedRhythms, Inc., and Neo Auvra, Inc. JMJ is on the Scientific Advisory Board for Wondr Health, Inc. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Prior to data collection, participants provided written informed consent approved by the Institutional Review Board at each site (Northeastern University, University of Kansas Medical Center, University of Pittsburgh).

Supporting information

Supporting Information

TRC2-11-e70187-s001.docx (201.8KB, docx)

Supporting Information

TRC2-11-e70187-s002.pdf (1.7MB, pdf)

ACKNOWLEDGMENTS

The authors would like to thank all of the participants, staff, faculty, and students that contributed to the IGNITE study. Study data were managed using REDCap electronic data capture tools through National Institutes of Health (NIH) support of the Clinical and Translational Sciences Institute (CTSI) at the University of Pittsburgh (UL1‐TR‐001857). This study was funded by the NIH (R01AG053952) awarded to KIE, JMB, CHH, AFK, and EM. This manuscript was also supported by NIH grant R35 AG072307 to KIE. AES is funded by a Dementia Australia Research Foundation Henry Brodaty Mid‐Career Research Fellowship. AM is supported by an Australian Government Research Training Program scholarship. DD was supported by an Australian Research Council (ARC) Discovery Early Career Researcher Award (DECRA) DE230101174. LEO is supported by National Institute of Mental Health grant K23MH129882. JHM is supported by the Spanish Ministry of Science, Innovation and Universities under Beatriz Galindo's 2022 fellowship program (ref: BG22/00075).

Collins AM, Mellow ML, Wan L, et al. Association between 24‐h time‐use composition and brain age: The IGNITE study. Alzheimer's Dement. 2025;11:e70187. 10.1002/trc2.70187

REFERENCES

  • 1. Giorgio A, Santelli L, Tomassini V, et al. Age‐related changes in grey and white matter structure throughout adulthood. Neuroimage. 2010;51(3):943‐51. doi: 10.1016/j.neuroimage.2010.03.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Farras‐Permanyer L, Mancho‐Fora N, Montalà‐Flaquer M, et al. Age‐related changes in resting‐state functional connectivity in older adults. Neural Regen Res. 2019;14(9):1544‐1555. doi: 10.4103/1673-5374.255976 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Crivello F, Tzourio‐Mazoyer N, Tzourio C, Mazoyer B. Longitudinal assessment of global and regional rate of grey matter atrophy in 1,172 healthy older adults: modulation by sex and age. PLoS One. 2014;9(12):e114478. doi: 10.1371/journal.pone.0114478 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Jack CR, Jr. , Shiung MM, Weigand SD, et al. Brain atrophy rates predict subsequent clinical conversion in normal elderly and amnestic MCI. Neurology. 2005;65(8):1227‐31. doi: 10.1212/01.wnl.0000180958.22678.91 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Defrancesco M, Egger K, Marksteiner J, et al. Changes in white matter integrity before conversion from mild cognitive impairment to Alzheimer's disease. PLoS One. 2014;9(8):e106062. doi: 10.1371/journal.pone.0106062 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Wrigglesworth J, Ryan J, Ward PGD, et al. Health‐related heterogeneity in brain aging and associations with longitudinal change in cognitive function. Front Aging Neurosci. 2022;14:1063721. doi: 10.3389/fnagi.2022.1063721 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Boller B, Mellah S, Ducharme‐Laliberté G, Belleville S. Relationships between years of education, regional grey matter volumes, and working memory‐related brain activity in healthy older adults. Brain Imaging Behav. 2017;11(2):304‐317. doi: 10.1007/s11682-016-9621-7 [DOI] [PubMed] [Google Scholar]
  • 8. Honea RA, Vidoni E, Harsha A, Burns JM. Impact of APOE on the healthy aging brain: a voxel‐based MRI and DTI study. J Alzheimers Dis. 2009;18(3):553‐64. doi: 10.3233/jad-2009-1163 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Sexton CE, Storsve AB, Walhovd KB, Johansen‐Berg H, Fjell AM. Poor sleep quality is associated with increased cortical atrophy in community‐dwelling adults. Neurology. 2014;83(11):967‐73. doi: 10.1212/wnl.0000000000000774 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Spira AP, Gonzalez CE, Venkatraman VK, et al. Sleep Duration and Subsequent Cortical Thinning in Cognitively Normal Older Adults. Sleep. 2016;39(5):1121‐8. doi: 10.5665/sleep.5768 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Maasakkers CM, Weijs RWJ, Dekkers C, et al. Sedentary behaviour and brain health in middle‐aged and older adults: A systematic review. Neurosci Biobehav Rev. 2022;140:104802. doi: 10.1016/j.neubiorev.2022.104802 [DOI] [PubMed] [Google Scholar]
  • 12. Gogniat MA, Won J, Cruz C, et al. Sedentary behavior, cognition, and brain health in older adults: a systematic review. Front Aging Neurosci. 2025;17:1622049. doi: 10.3389/fnagi.2025.1622049 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Erickson KI, Raji CA, Lopez OL, et al. Physical activity predicts gray matter volume in late adulthood: the Cardiovascular Health Study. Neurology. 2010;75(16):1415‐22. doi: 10.1212/WNL.0b013e3181f88359 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Gow AJ, Bastin ME, Muñoz Maniega S, et al. Neuroprotective lifestyles and the aging brain: activity, atrophy, and white matter integrity. Neurology. 2012;79(17):1802‐8. doi: 10.1212/WNL.0b013e3182703fd2 [DOI] [PubMed] [Google Scholar]
  • 15. Erickson KI, Voss MW, Prakash RS, et al. Exercise training increases size of hippocampus and improves memory. Proc Natl Acad Sci U S A. 2011;108(7):3017‐22. doi: 10.1073/pnas.1015950108 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Colcombe SJ, Erickson KI, Scalf PE, et al. Aerobic exercise training increases brain volume in aging humans. J Gerontol A Biol Sci Med Sci. 2006;61(11):1166‐70. doi: 10.1093/gerona/61.11.1166 [DOI] [PubMed] [Google Scholar]
  • 17. Erickson KI, Hillman C, Stillman CM, et al. Physical activity, cognition, and brain outcomes: a review of the 2018 physical activity guidelines. Med Sci Sports Exerc. 2019;51(6):1242‐1251. doi: 10.1249/mss.0000000000001936 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Dumuid D, Stanford TE, Martin‐Fernández J‐A, et al. Compositional data analysis for physical activity, sedentary time and sleep research. Stat Methods Med Res. 2018;27(12):3726‐3738. doi: 10.1177/0962280217710835 [DOI] [PubMed] [Google Scholar]
  • 19. Mellow ML, Dumuid D, Olds T, et al. Cross‐sectional associations between 24‐hour time‐use composition, grey matter volume and cognitive function in healthy older adults. Int J Behav Nutr Phys Act. 2024;21(1):11. doi: 10.1186/s12966-023-01557-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Balbim GM, Boa Sorte Silva NC, Falck RS, Kramer AF, Voss MW, Liu‐Ambrose T. 24‐hour activity cycle behaviors and gray matter volume in mild cognitive impairment. Alzheimers Dement. 2025;21(7):e70496. doi: 10.1002/alz.70496 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Franke K, Ziegler G, Klöppel S, Gaser C. Estimating the age of healthy subjects from T1‐weighted MRI scans using kernel methods: exploring the influence of various parameters. Neuroimage. 2010;50(3):883‐92. doi: 10.1016/j.neuroimage.2010.01.005 [DOI] [PubMed] [Google Scholar]
  • 22. Franke K, Gaser C. Longitudinal changes in individual BrainAGE in healthy aging, mild cognitive impairment, and Alzheimer's disease. GeroPsych: The Journal of Gerontopsychology and Geriatric Psychiatry. 2012;25(4):235‐245. doi: 10.1024/1662-9647/a000074 [DOI] [Google Scholar]
  • 23. Liem F, Varoquaux G, Kynast J, et al. Predicting brain‐age from multimodal imaging data captures cognitive impairment. Neuroimage. 2017;148:179‐188. doi: 10.1016/j.neuroimage.2016.11.005 [DOI] [PubMed] [Google Scholar]
  • 24. Gaser C, Franke K, Klöppel S, Koutsouleris N, Sauer H. BrainAGE in Mild Cognitive Impaired Patients: Predicting the Conversion to Alzheimer's Disease. PLoS One. 2013;8(6):e67346. doi: 10.1371/journal.pone.0067346 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Cole JH, Ritchie SJ, Bastin ME, et al. Brain age predicts mortality. Mol Psychiatry. 2018;23(5):1385‐1392. doi: 10.1038/mp.2017.62 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Oberlin LE, Wan L, Kang C, et al. Cardiorespiratory fitness is associated with cognitive function in late adulthood: baseline findings from the IGNITE study. Br J Sports Med. 2024:bjsports‐2024‐108257. doi: 10.1136/bjsports-2024-108257 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Ripperger HS, Reed RG, Kang C, et al. Cardiorespiratory fitness, hippocampal subfield morphology, and episodic memory in older adults. Front Aging Neurosci. 2024;16:1466328. doi: 10.3389/fnagi.2024.1466328 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Tinney EM, Warren AEL, O'Brien A, et al. White matter microstructure mediates the association between cardiorespiratory fitness and cognitive performance in older adults. Alzheimers Dement (N Y). 2025;11(3):e70125. doi: 10.1002/trc2.70125 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Collins AM, Mellow ML, Smith AE, et al. 24‐Hour time use and cognitive performance in late adulthood: results from the Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE) study. Age and Ageing. 2025;54(4):afaf072. doi: 10.1093/ageing/afaf072 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Erickson KI, Grove GA, Burns JM, et al. Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE): Protocol. Contemp Clin Trials. 2019;85:105832. doi: 10.1016/j.cct.2019.105832 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Cole JH, Poudel RPK, Tsagkrasoulis D, et al. Predicting brain age with deep learning from raw imaging data results in a reliable and heritable biomarker. NeuroImage. 2017;163:115‐124. doi: 10.1016/j.neuroimage.2017.07.059 [DOI] [PubMed] [Google Scholar]
  • 32. Gaser C, Dahnke R, Thompson PM, Kurth F, Luders E, Alzheimer's Disease Neuroimaging I. CAT—A Computational Anatomy Toolbox for the Analysis of Structural MRI Data. bioRxiv. 2023:2022.06.11.495736. doi: 10.1101/2022.06.11.495736 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Dörfel RP, Arenas‐Gomez JM, Fisher PM, et al. Prediction of brain age using structural magnetic resonance imaging: A comparison of accuracy and test‐retest reliability of publicly available software packages. Hum Brain Mapp. 2023;44(17):6139‐6148. doi: 10.1002/hbm.26502 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Migueles JH, Rowlands AV, Huber F, Sabia S, van Hees VT. GGIR: A Research Community–Driven Open Source R Package for Generating Physical Activity and Sleep Outcomes From Multi‐Day Raw Accelerometer Data. J Meas Phys Behav. 2019;2(3):188‐196. doi: 10.1123/jmpb.2018-0063 [DOI] [Google Scholar]
  • 35. van Hees VT, Sabia S, Anderson KN, et al. A Novel, Open Access Method to Assess Sleep Duration Using a Wrist‐Worn Accelerometer. PLOS ONE. 2015;10(11):e0142533. doi: 10.1371/journal.pone.0142533 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. van Hees VT, Sabia S, Jones SE, et al. Estimating sleep parameters using an accelerometer without sleep diary. Scientific Reports. 2018;8(1):12975. doi: 10.1038/s41598-018-31266-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. de Lange A‐MG, Cole JH. Commentary: Correction procedures in brain‐age prediction. NeuroImage: Clinical. 2020;26:102229. doi: 10.1016/j.nicl.2020.102229 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Hanson JL, Adkins DJ, Bacas E, Zhou P. Examining the reliability of brain age algorithms under varying degrees of participant motion. Brain Inform. 2024;11(1):9. doi: 10.1186/s40708-024-00223-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. R: A language and environment for statistical computing. R Foundation for Statistical Computing; 2021. [Google Scholar]
  • 40. Palarea‐Albaladejo J, Martín‐Fernández JA. zCompositions — R package for multivariate imputation of left‐censored data under a compositional approach. Chemometr Intell Lab Syst. 2015;143:85‐96. doi: 10.1016/j.chemolab.2015.02.019 [DOI] [Google Scholar]
  • 41. Kynčlová P, Hron K, Filzmoser P. Correlation Between Compositional Parts Based on Symmetric Balances. Mathematical Geosciences. 2017;49(6):777‐796. doi: 10.1007/s11004-016-9669-3 [DOI] [Google Scholar]
  • 42. van den Boogaart KG, Tolosana‐Delgado R. “compositions”: A unified R package to analyze compositional data. Computers & Geosciences. 2008;34(4):320‐338. doi: 10.1016/j.cageo.2006.11.017 [DOI] [Google Scholar]
  • 43. Mellow ML, Dumuid D, Wade AT, et al. Twenty‐four‐hour time‐use composition and cognitive function in older adults: Cross‐sectional findings of the ACTIVate study. Front Hum Neurosci. 2022;16:1051793. doi: 10.3389/fnhum.2022.1051793 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44. Fox J, Weisberg S. An R Companion to Applied Regression. Third ed. Sage; 2019. [Google Scholar]
  • 45. Compositional Data Linear Models with Composition Redistribution. Version 0.1.0. 2022. https://cran.r‐project.org/web/packages/codaredistlm/codaredistlm.pdf [Google Scholar]
  • 46. Stillman CM, Cohen J, Lehman ME, Erickson KI. Mediators of Physical Activity on Neurocognitive Function: A Review at Multiple Levels of Analysis. Front Hum Neurosci. 2016;10:626. doi: 10.3389/fnhum.2016.00626 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Piercy KL, Troiano RP, Ballard RM, et al. The Physical Activity Guidelines for Americans. Jama. 2018;320(19):2020‐2028. doi: 10.1001/jama.2018.14854 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Ross R, Chaput JP, Giangregorio LM, et al. Canadian 24‐Hour Movement Guidelines for Adults aged 18‐64 years and Adults aged 65 years or older: an integration of physical activity, sedentary behaviour, and sleep. Appl Physiol Nutr Metab. 2020;45(Suppl. 2):S57‐S102. doi: 10.1139/apnm-2020-0467 [DOI] [PubMed] [Google Scholar]
  • 49. Biondo F, Jewell A, Pritchard M, et al. Brain‐age is associated with progression to dementia in memory clinic patients. Neuroimage Clin. 2022;36:103175. doi: 10.1016/j.nicl.2022.103175 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Mellow ML, Dumuid D, Wade A, et al. Should We Work Smarter or Harder for Our Health? A Comparison of Intensity and Domain‐Based Time‐Use Compositions and Their Associations With Cognitive and Cardiometabolic Health. The Journals of Gerontology: Series A. 2024;79(11):glae233. doi: 10.1093/gerona/glae233 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

TRC2-11-e70187-s001.docx (201.8KB, docx)

Supporting Information

TRC2-11-e70187-s002.pdf (1.7MB, pdf)

Articles from Alzheimer's & Dementia : Translational Research & Clinical Interventions are provided here courtesy of Wiley

RESOURCES