HHS Public Access

Author manuscript
Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.
Author Manuscript

Published in final edited form as:

Neurobiol Aging. 2022 January ; 109: 195–203. doi:10.1016/j.neurobiolaging.2021.10.007.

Brain-predicted age difference is associated with cognitive

processing in later-life
Jo Wriggleswortha, Nurathifah Yaacoba, Phillip Wardb,c,d, Robyn L. Woodsa, John McNeila,
Elsdon Storeya, Gary Eganb,d, Anne Murrayf,g, Raj C. Shahh, Sharna D. Jamadarb,c,d, Ruth
Trevaksa, Stephanie Warda,i,j, Ian H. Hardingb,e,†, Joanne Ryana,†,* on behalf of the ASPREE
investigator group
Author Manuscript

aSchool of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria,
Australia
bMonash Biomedical Imaging, Monash University, Melbourne, Victoria, Australia
cTurner Institute for Brain and Mental Health, Monash University, Melbourne, Victoria, Australia
dAustralian Research Council Centre of Excellence for Integrative Brain Function, Clayton,
Victoria, Australia
eDepartment of Neuroscience, Central Clinical School, Monash University, Melbourne, Victoria,
Australia
fBermanCenter for Outcomes & Clinical Research, Hennepin Healthcare Research Institute,
Minneapolis, MN, USA
Author Manuscript

gDepartment of Medicine, Division of Geriatrics, Hennepin Healthcare, University of Minnesota,

Minneapolis, MN, USA
hDepartment of Family Medicine and the Rush Alzheimer’s Disease Centre, Rush University
Medical Centre, Chicago, IL, USA
iCentrefor Healthy Brain Ageing (CHeBA), University of New South Wales, Sydney, New South
Wales, Australia

*
Corresponding author at: J. Ryan, School of Public Health and Preventive Medicine, Monash University, Level 5, The Alfred Centre
99 Commercial, Melbourne, Victoria 3004, Australia. Tel.: (03) 9903 0200; fax: (03) 9903 0979. joanne.ryan@monash.edu (J. Ryan).
Author Manuscript

†Joint last authors

Verification
No part of this manuscript has been published previously and it has not been submitted for publication elsewhere. All authors approve
the publication of this manuscript.
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.neurobiolaging.2021.10.007.
CRediT authorship contribution statement
Jo Wrigglesworth: Conceptualization, Methodology, Formal analysis, Writing – original draft, Visualization. Nurathifah Yaacob:
Writing – review & editing. Phillip Ward: Supervision, Conceptualization, Resources, Writing – review & editing. Robyn L. Woods:
Supervision, Conceptualization, Resources, Writing – review & editing. John McNeil: Resources, Writing – review & editing. Elsdon
Storey: Resources. Gary Egan: Resources, Writing – review & editing. Anne Murray: Resources, Writing – review & editing. Raj
C. Shah: Resources, Writing – review & editing. Sharna D. Jamadar: Writing – review & editing. Ruth Trevaks: Writing – review
& editing. Stephanie Ward: Resources, Writing – review & editing. Ian H. Harding: Supervision, Conceptualization, Methodology,
Writing – review & editing. Joanne Ryan: Supervision, Conceptualization, Writing – review & editing.
 Wrigglesworth et al. Page 2

jDepartment of Geriatric Medicine, Prince of Wales Hospital, Randwick, New South Wales,
Author Manuscript

Australia

Abstract
Brain age is a neuroimaging-based biomarker of aging. This study examined whether the
difference between brain age and chronological age (brain-PAD) is associated with cognitive
function at baseline and longitudinally. Participants were relatively healthy, predominantly white
community-dwelling older adults (n = 531, aged ≥70 years), with high educational attainment
(61% ≥12 years) and socioeconomic status (59% ≥75th percentile). Brain age was estimated from
T1-weighted magnetic resonance images using an algorithm by Cole et al., 2018.
After controlling for age, gender, education, depression and body mass index, brain-PAD
was negatively associated with psychomotor speed (Symbol Digit Modalities Test) at baseline
(Bonferroni p < 0.006), but was not associated with baseline verbal fluency (Controlled Oral Word
Author Manuscript

Association Test), delayed recall (Hopkins Learning Test Revised), or general cognitive status
(Mini-Mental State Examination). Baseline brain-PAD was not associated with 3-year change in
cognition (Bonferroni p > 0.006).

These findings indicate that even in relatively healthy older people, accelerated brain aging is
associated with worse psychomotor speed, but future longitudinal research into changes in brain-
PAD is needed.

Keywords
Estimated brain age; Brain aging; Predicted age difference; Neuroimaging; Magnetic resonance
imaging; Cognitive function
Author Manuscript

1. Introduction
Aging is an inevitable biological process characterized by an accumulation of cell and tissue
damage over the lifespan (Lopez-Otin et al., 2013). The inability to repair this damage
subsequently impairs normal physiological functions (Lopez-Otin et al., 2013). The brain is
particularly sensitive to the effects of aging, which manifest as changes in brain structure
and associated cognitive functions (Oschwald et al., 2019). Neuroimaging has made it
possible to view these changes in vivo. The most commonly reported age-related brain
changes include cortical atrophy (i.e., a loss of brain tissue volume) (Grajauskas et al., 2019)
and the decline in white matter integrity and connectivity (Coelho et al., 2021; Damoiseaux,
2017).
Author Manuscript

Age-related changes in the brain have a wide range of neurocognitive consequences. This
includes a decline in domain-specific cognitive functions, such as memory, processing
speed, and verbal fluency (Armstrong et al., 2020; Ritchie et al., 2015). Cognitive decline is
a key symptom of dementia, which is a major cause of disability and loss of independence
among older people worldwide (Organization, 2019; Tucker-Drob, 2019). Biological brain
aging and associated changes in cognitive functioning, however, are highly heterogeneous
among the population. While some people will retain brain tissue volume and high cognitive

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 3

proficiency throughout life (so called “super agers”), another subgroup will experience
Author Manuscript

earlier and more severe cortical atrophy and cognitive decline (so-called “accelerated
agers”). As such, chronological age alone is a poor marker of cognitive function and risk of
decline (Tucker-Drob, 2019).

In order to provide a more individually-relevant measure of overall brain health, increasing

attention has been paid to the development of neuroimaging-based biomarkers that reflect
a person’s “brain age,” as distinct from their chronological age. An older brain age,
relative to chronological age is considered a sign of accelerated brain aging (i.e., greater
age-related changes to the brain, such as cortical atrophy); while a lower brain age suggests
decelerated brain aging (i.e., fewer age-related changes, and thus, preserved brain tissue).
A number of different models have been developed and validated for estimation of brain
age from structural magnetic resonance imaging (MRI) data (Dinsdale et al., 2021; Franke
et al., 2010; Kaufmann et al., 2019; Liem et al., 2017), and these have shown promise as
Author Manuscript

biomarkers of aging (Oschwald et al., 2019). One commonly used approach is that of Cole
and colleagues (Cole et al., 2018) “brain predicted age difference” (brain-PAD) model. This
algorithm has been trained and validated using large population samples encompassing ages
18–92 years. Its source code is freely available, and growing use of this algorithm across
a wide number of studies and populations also facilitates direct between study comparison
and replication (Biondo et al., 2020; Ly et al., 2020; Soros and Bantel, 2020). It has been
associated with other markers of aging (i.e., grip strength, lung function, walking speed and
allostatic load), subsequent dementia diagnosis, and mortality risk, and thus has potential as
a biomarker of age-related cognitive decline (Biondo et al., 2020; Cole et al., 2018).

A growing number of cross-sectional studies have investigated the relationship between

brain age and cognitive function in unimpaired community dwelling adults, covering a
Author Manuscript

range of domains including psychomotor speed, verbal fluency, delayed recall, and general
cognitive status. Most consistently reported is the association between accelerated brain
aging and worse psychomotor speed (Boyle et al., 2021; Elliott et al., 2019; Jonsson et al.,
2019; Smith et al., 2019), while evidence for all other domains is relatively sparse, and
inconsistent. Such discrepancies may relate to varying participant characteristics (i.e., wide
age ranges that cover developmental and degenerative processes, or cohorts are limited to
young and/or middle aged adults), and methods of measuring brain age.

Longitudinal investigations of the association between brain age and cognitive function
have so far been limited to clinical populations. This body of literature consistently reports
associations between accelerated brain aging and poor cognitive status in people with mild
cognitive impairment and Alzheimer’s Disease (refer to Wrigglesworth et al., 2021 for a
Author Manuscript

systematic review of these findings) (Franke and Gaser, 2012; Gaser et al., 2013; Lowe et
al., 2016). Though 1 prospective study has investigated brain age in middle aged, community
dwelling adults (~45 years), declines in cognitive function reflects changes occurring from
child- to middle age adulthood (Elliott et al., 2019). Thus, to our knowledge, no study has
investigated this association prospectively in community dwelling older adults.

Using the Cole and colleagues brain-PAD model (Cole et al., 2018), we determined whether
(1) brain-PAD is cross-sectionally associated with 4 domains of cognitive function, selected

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 4

for their propensity to decline in later life, and (2) if brain-PAD is predictive of changes in
Author Manuscript

cognitive function over a 3 year follow-up. In the context of cognitive function, brain-PAD
compares as a marker of local tissue volume (i.e., gray, white matter and cerebrospinal fluid)
from across the whole brain, while other models are specific to 1 or more brain tissues types
(i.e., gray matter density), or modalities. Our study is a novel examination of the 3-year
change in 4 domains of cognitive function in cognitively unimpaired older participants.
We hypothesized an association between accelerated brain aging and both poorer cognitive
function and cognitive decline.

2. Materials and methods

2.1. Study design
ASPREE-NEURO is a neuroimaging sub-study of the ASPirin in Reducing Events in the
Elderly (ASPREE) clinical trial (Ward et al., 2017). The NEURO sub-study recruited 572
Author Manuscript

cognitively healthy individuals from ASPREE, aged 70 years and over. Full eligibility
criteria have been published elsewhere (Group, 2013), but in brief, exclusion criteria
included established cardiovascular disease; diagnosis of dementia or a Modified Mini-
Mental State Examination score <78 (Teng and Chui, 1987); severe disability or inability to
perform 6 basic activities of daily living (Woods et al., 2020); a condition with a high current
or recurrent risk of bleeding; anemia; and current use of aspirin or any other antiplatelet
or anti-coagulated medications (McNeil et al., 2017). Of the 572 NEURO participants, 531
(95%) had usable MRI and cognitive assessment data at baseline. Fourteen participants
did not complete a baseline MRI, and 27 were excluded due to poor image quality.
Two hundred and fifty-nine participants had complete neuropsychological assessments at
baseline, 1- and 3-year visits, and comprise our longitudinal subsample. The other 272
participants did not have complete longitudinal assessments. Reasons include death (n =
Author Manuscript

7), no longer attending in-person visits (i.e., assessments conducted via phone or through
medical review [n = 11]), or the trial phase of ASPREE had ended before the third-year
cognitive assessment (n = 254). ASPREE is registered with the International Standard
Randomized Controlled Trial Number Register (ISRCTN83772183) and Clinicaltrials.gov
(NCT01038583). NEURO is independently registered with the Australian New Zealand
Clinical Trial Registry (AC-TRN12613001313729). The study was approved by the Monash
University Human Research and Ethics Committee, and all participants provided written
informed consent.

2.3. Brain age prediction

Brain age was estimated using a trained, publicly sourced processing pipeline and prediction
model developed by Cole and colleagues (BrainageR, version 2.1 [https://github.com/james-
Author Manuscript

cole/brainageR]) (Cole et al., 2018). Full details of the data acquisition, image quality
assessment, and brain age protocol are provided in the Supplementary Methods. In brief,
MRI data was acquired using a 3 Tesla Siemens Skyra scanner (Siemens, Erlangen,
Germany) with a 32-channel head coil, located at the Monash Biomedical Imaging facility,
Melbourne, Australia. Raw 3D T1-weighted images were preprocessed using the Statistical
Parametric Mapping toolbox (University College London, London, UK) for MATLAB
(Cole et al., 2015). This includes segmentation into gray matter (GM), white matter (WM)

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 5

and cerebrospinal fluid, and normalization to Montreal Neurological Institute space using
Author Manuscript

nonlinear registration (Ashburner, 2007). Images were modulated and resampled to a voxel
size of 1.5 mm. Finally, spatial smoothing was applied using a kernel of 4 mm at FWHM.

Features of brain volume were combined, and divided into 435 principle components
identified for the training cohort (n = 3377 healthy adults, 18–92 years old). Brain age
was derived from these components, using a Gaussian process regression algorithm. Our
primary outcome of interest is the brain-predicted age difference (brain-PAD), defined as
the difference between an individual’s estimated brain age and their chronological age. A
positive brain-PAD defines a predictive brain age that is older than one’s chronological age,
while a negative value suggests brain age is younger than expected.

Estimated brain age had a significant positive correlation of ρ = 0.46 with chronological age,
indicating moderate accuracy (p < 0.0001; Supplementary Figure 1A). Estimated brain age
Author Manuscript

was evaluated for age bias (i.e., overestimation in younger, while underestimated in older
participants) using the correlation between brain-PAD and chronological age. Brain-PAD
was not associated with chronological age (ρ = −0.01, p = 0.83; Supplementary Figure 1B).
These results were consistent among the longitudinal subsample (brain age: ρ = 0.45, p <
0.0001; brain-PAD: ρ = 0.01, p = 0.87).

2.4. Cognitive function

As a part of the ASPREE clinical trial, participants underwent cognitive assessments at
baseline, year 1 and year 3 of follow-up. Cognitive tests include the Controlled Oral Word
Association Test (COWAT) to measure verbal fluency (Ross, 2003), the Hopkins Verbal
Learning Test - Revised Delayed Recall task for episodic memory (Benedict et al., 1998),
the Symbol Digit Modalities Test (SDMT) for psychomotor speed (Smith, 1982), and the
Author Manuscript

Modified Mini-Mental State (3MS), an overall screening of general cognitive status (Teng
and Chui, 1987). A higher test score represents a better cognitive performance.

2.5. Study covariates

Extensive lifestyle and health information was available for all participants, allowing for
an analysis of their contributing or confounding influence on the relationship between
brain age and cognition. These include self-reported demographics (e.g., education, living
arrangement), lifestyle factors (e.g., smoking and alcohol), and physical measures (e.g.,
blood pressure, body mass index [BMI]) that were collected at baseline, and at each follow-
up of the ASPREE study. The diagnosis of diabetes, hypertension, dyslipidemia, and chronic
kidney disease were derived from data collected at baseline, and have been described in
detail elsewhere (McNeil et al., 2017).
Author Manuscript

2.6. Statistical analysis

Two key analyses were undertaken in line with our 2 key aims: (1) baseline brain-PAD
versus baseline cognitive measures in all participants; and (2) baseline brain-PAD versus
longitudinal change in cognitive measures in a subgroup of participants with cognitive data
available over the 3-year follow-up (see Results).

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 6

Univariable linear regression models were used to assess the association between baseline
Author Manuscript

brain-PAD and cognitive function (Aim 1), and baseline brain-PAD relative to changes
in cognitive function (Aim 2), with brain-PAD treated as the exposure, and cognitive
function the outcome of interest. The 3-year change in cognitive function was defined
as the difference between baseline and year 3 cognitive assessment (Year 3 – Baseline).
The coefficient of determination, and normality of residuals were used to evaluate the fit
of each model. Multivariable linear regressions were subsequently undertaken to account
for the influence of confounding covariates. Relevant covariates were identified using a
univariable linear regression model, and a liberal criterion (p ≤ 0.20). In this sample of
participants who were without major cognitive impairment and cardiovascular disease, we
adjusted for chronological age (Model 1), in addition to gender, education, depression and
BMI (Model 2). Our final model was specific to the analysis of change in cognitive function,
and includes all covariates, along with baseline cognitive function (Model 3). With the
Author Manuscript

exception of gender, chronological age, and education, covariates not associated with both
cognition and brain-PAD were not included as this does not fit the criteria of confounding.
These include socioeconomic percentile (SES), smoking, alcohol consumption, diabetes,
hypertension, and chronic kidney disease diagnoses (refer to Supplementary Table 1). The
multivariable models considered consecutive adjustment for covariates, and interactions to
determine whether individual covariates were having a strong influence on the results (refer
to Supplementary Table 2). For all models, we account for multiple testing (i.e., 4 measures
of cognitive function at 2 time-points) by applying a Bonferroni correction of p ≤ 0.006.

We conducted a number of sensitivity analyses to assess the robustness of our main findings.
This included: (1) analyses focused on individuals aged 70–75 years only, to ensure that the
findings were not driven by the oldest participants in the cohort; and (2) investigating the 2
years (Year 3 – Year 1) change in cognitive function, to help account for potential practice
Author Manuscript

effects with the cognitive tests administered in 2 consecutive years (baseline and year 1).
Results were compared to the 1-year change in cognitive function (Baseline – Year 1) - a
shorter re-test interval considered more susceptible to the effects of practice (Tombaugh,
2005). Analyses were performed using Stata software, release 16 (StataCorp).

3. Results
3.1. Participant characteristics
At baseline, the total cohort ranged in age from 70 to 88 years, but the majority were
aged 70–74 years (Table 1). Most participants resided in more advantaged areas (higher
SES percentile), completed tertiary education (i.e., 12 years or more), were over-weight and
hypertensive.
Author Manuscript

Participants from the longitudinal subsample (n = 259) had completed a 1 and 3-year
cognitive assessment a median duration 358 and 1072 days after baseline, respectively
(Table 1). These participants were similar to the remaining study sample that had
not attended the follow-up cognitive assessments, with the exception that there were
significantly more men (57% vs. 46%) and fewer women (43% vs. 54%; Pearson χ2[1]
= 6.22, p = 0.013).

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 7

3.2. Estimated brain age and brain-predicted age difference (brain-PAD)

Author Manuscript

T1-weighted MRI scans were acquired a median of 14 (IQR = 11–20) days after cognitive
assessment. The median brain-PAD was −1.5 (IQR = −5.5 to 2.8) and ranged between −19.5
and 14.6 years. Relative to the remaining total cohort, the longitudinal subsample had a
similar estimated brain age (mean difference = 0.50 [95% CI = −0.68, 1.69] years, p = 0.51),
and brain-PAD (mean difference = 0.30 [95% CI = −0.72, 1.33] years, p = 0.56).

3.3. Baseline and change in cognitive function

Baseline and longitudinal measures of cognitive performance are presented in Table 2. The
distribution of baseline scores in 3MS, and HVLT-R were positively skewed, and thus the
majority were high scoring on these tests. For COWAT and SDMT, baseline scores were
normally distributed.
Author Manuscript

In the longitudinal subsample there was evidence of practice effects, with the average scores
for COWAT increasing slightly from baseline to year 3 (Table 2). Scores on the other 3 tests
remained stable, with on average (across all participants) no significant difference between
the 2 time points. With the exception of SDMT (n = 257, mean difference = 0.30 [95% CI =
−0.43, 1.02], p = 0.43), the shorter follow-up of 1 year resulted in a significant improvement
in the average score of multiple domains, including 3MS (n = 259, mean difference = 1.04
[95% CI = 0.56, 1.53], p < 0.0001), COWAT (n = 258, mean difference = 1.26 [95% CI =
0.81, 1.72], p < 0.0001) and HVLT-R (n = 256, mean difference = 0.31 [95% CI = 0.05,
0.57], p = 0.02). In contrast, when removing baseline scores from measures of change (i.e.,
2-year change), a slight significant decrease in 3MS was observed over the year 1 and year
3 follow-up (mean difference = −0.58 [95% CI = −1.10, −0.05], p = 0.03), while change
in all other domains were not significant (COWAT: mean difference = 0.05 [95% CI =
Author Manuscript

−0.45, 0.56], p = 0.83; SDMT: mean difference = −0.67 [95% CI = −1.39, 0.06], p = 0.07;
HVLT-R: mean difference = −0.02 [95% CI = −0.28, 0.24], p = 0.88).

3.4. Brain aging and baseline cognitive function

Univariable analyses of baseline brain-PAD and cognitive function show a positive brain-
PAD (i.e., older brain age relative to chronological age) was significantly associated with
lower scores on the SDMT and HVLT-R (Fig. 1A and B; Table 3). When adjusting
for multiple comparisons, only the association between brain-PAD and SDMT remained
significant (p ≤ 0.006).

In multivariable analyses including chronological age (Model 1), gender, BMI, education
and depression (Model 2), brain-PAD remained significantly associated with SDMT,
and HVLT-R (Table 3). Only the association between brain-PAD and SDMT remained
Author Manuscript

significant following adjustment for multiple comparisons (p ≤ 0.006). Interactions between

brain-PAD and any of the chosen covariates were not significantly associated with cognitive
function (p > 0.05) (Supplementary Table 2). No significant associations were observed with
3MS or COWAT (Table 3).

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 8

3.5. Brain aging and cognitive change

Author Manuscript

Univariable analyses were first performed using data from the longitudinal cohort. From
this analysis, a greater positive baseline brain-PAD was significantly associated with a
decline in HVLT-R over 3 years of follow-up (Fig. 2). The association between brain-PAD
and the 3-year change in HVLT-R was no longer significant after correction for multiple
comparisons (Table 3). Exclusion of 2 visual outliers with a very large decline in cognition
(decrease of 9 and 11 words recalled; refer to Fig. 2) had minimal influence on the results
with HVLT-R (n = 255, r = −0.16, b = −0.06 [95% CI = −0.11, −0.02], p = 0.01) and were
thus retained in subsequent analyses.

Associations with the decline in HVLT-R remained significant when adjusting for
chronological age (Model 1), or all covariates (Model 2), including baseline cognitive
function (Model 3; Table 3). All models passed standard fit and regression assumptions
Author Manuscript

(Supplementary Figure 2A–F). When adjusted for baseline cognitive function, positive
values of brain-PAD were significantly associated with a greater decline in 3MS (Table
3). After adjusting for multiple comparisons, the association between brain-PAD and the
3-year change in HVLT-R and 3MS did not reach significance (p > 0.006).

Interactions between positive brain-PAD and depression were also significantly associated
with the 3-year decline in SDMT. Both interactions, however, no longer remained following
correction for multiple comparisons (education: n = 257, b = −0.13 [95% CI = 0.01, 0.24], p
= 0.04; depression (n = 258, b = 0.37[0.05, 0.70], p = 0.02).

3.6. Sensitivity analyses

Sensitivity analyses were performed to further explore the relationship between baseline
brain-PAD and cognitive function by limiting the analysis to only participants aged below
Author Manuscript

76 years. Univariable findings were consistent to our corrected main results, and remained
when adjusting for covariates (Supplementary Table 3). Removing these participants from
the longitudinal subsample (n = 71), we observed no association between brain-PAD and
any measure of change in cognitive function (Supplementary Table 3 and 4).

Further analyses were performed on a 1 year (baseline to year 1) and 2 years (year 1 to
year 3) change in cognitive function. Univariable analyses show no association between
brain-PAD and the 1 year change in cognitive function (Supplementary Table 5). Similar
to our main findings, a more positive brain-PAD was significantly associated with a greater
decline in HVLT-R over the 2-year follow-up, which remained consistent when adjusting
for chronological age, and all covariates (i.e., chronological age, gender, BMI, education
and depression; Supplementary Table 3, Models 1 and 2). Models passed standard fit and
Author Manuscript

regression assumptions (Supplementary Figure 3A–D).

4. Discussion
This study aimed to determine whether interindividual differences in brain-PAD are
associated with baseline cognitive function, and changes in cognitive function over time.
In a sample of 531 older community dwelling adults aged 70 years and above, accelerated
brain aging (i.e., older predicted brain age relative to chronological age) was associated with

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 9

slower baseline psychomotor speed and worse delayed recall. After correcting for multiple
Author Manuscript

comparisons, only the association with psychomotor speed remained.

In a subsample of 259 participants followed longitudinally, accelerated brain aging at

baseline was predictive of a modest decline in delayed recall and general cognitive status
over 3 years, although these findings didn’t reach Bonferroni-corrected significance levels.
Given the likely practice effects observed (i.e., changes in cognitive performance that may
be attributable to a prior test experience rather than normal age-related change) (Salthouse,
2019), in particular in relation to COWAT, we attempted to account for this by performing a
sensitivity analysis that examined the change in cognitive function from year 1 to year 3. In
this respect, we ignored baseline test scores, which was the first-time participants had likely
undertaken such a test, and examined change when participants were equally familiar with
the cognitive assessment. Through this analysis, we confirm our findings regarding delayed
recall, but observed no association with general cognitive status. However, when excluding
Author Manuscript

the few participants aged 76 years and above, we no longer identified an association with
the change in cognition, and thus findings may be influenced by these few older participants.
No significant associations were found with baseline scores of general cognitive status and
verbal fluency, nor the decline in verbal fluency.

4.1. Cross-sectional associations between brain aging and cognitive function

Although an increasing number of studies have investigated the relationship between brain-
PAD and health conditions, including dementia (Wrigglesworth et al., 2021), few have
determined associations with cognitive function in unimpaired older populations. In the
published literature, psychomotor speed has been most extensively examined, with the
majority of studies reporting associations between accelerated brain aging and impaired or
worse cognitive function in this domain (Boyle et al., 2021; Elliott et al., 2019; Jonsson et
Author Manuscript

al., 2019; Smith et al., 2019). This aligns with the findings of our study, and is consistent
with the slower processing speed identified in older populations (Ebaid et al., 2017). Only 1
study found no association between accelerated brain aging and psychomotor function, but
they defined psychomotor function using computational modelling of responses to a letter-
based task rather than directly from neuropsychological tests (Bundesen, 1990; Richard et
al., 2018).

The cognitive domain of memory has received less attention, with only 3 studies examining
this association (Boyle et al., 2021; Elliott et al., 2019; Richard et al., 2018). Of these
studies, 2 found no association between brain aging and memory (Boyle et al., 2021;
Richard et al., 2018), while the third observed a significant association. Although we
initially identified a significant association with memory, these findings had not survived
Author Manuscript

multiple correction. These distinctions may relate to differences in the methodologies,

measures of brain age and cognitive function, as well as the cohort characteristics of the
study samples. For example, compared to our target group of late-life participants (aged 70–
88 years old), Elliot and colleagues identified an association in adults aged approximately
45 years (aged 43–47 years old) (Elliott et al., 2019). The evidence regarding phonological
verbal fluency is also inconsistent, and may also be explained by these distinctions. For
example, 2 cross-sectional studies have investigated phonological fluency – 1 in community

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 10

dwelling cohorts aged ~47–94 years and the second in patients with a traumatic brain
Author Manuscript

injury (Boyle et al., 2021; Cole et al., 2015). Interestingly, both studies reported associations
with brain age estimated from GM, rather than overall GM and WM as undertaken herein,
and reported in other studies of verbal fluency (Armstrong et al., 2020; Richard et al.,
2018). Thus, discrepancies may also relate to tissue-specific sensitivity of brain aging to
impairment in specific cognitive domains, but more research is needed.

We also failed to find any evidence of a significant relationship between brain-PAD and
general cognitive status (i.e., as measured using the 3MS). In the context of brain aging,
the evidence on general cognitive status is extensive, but primarily related to cognitively
impaired or Alzheimer’s disease populations (Beheshti et al., 2018; Franke et al., 2010).
In line with our current results, 2 previously mentioned studies that used screening tools
to assess general cognitive status (Mini Mental State Examination, Montreal Cognitive
Assessment), similar to the 3MS assessment we used, did not find an association or reported
Author Manuscript

mixed results across different cohorts (Boyle et al., 2021; Richard et al., 2018). Thus,
participant characteristics may play an important role when investigating brain-PAD and
general cognitive status, and warrants further investigation.

4.2. Longitudinal associations between brain aging and the change in cognitive function
over time
Our study also identified an association between accelerated brain-PAD at baseline and a
3-year change in delayed recall. This, however, was only a small effect, and no longer
significant when adjusting for multiple comparisons. Further, we identified no association
with change when excluding participants aged 76 years and above, and thus initial findings
may be biased by these few outliers. We also did not find an association with longitudinal
declines in psychomotor speed, verbal fluency, and general cognitive status. An association
Author Manuscript

with general cognitive status was observed when adjusting for covariates including baseline
cognitive function, but was no longer significant after applying a multiple correction. Given
the high educational attainment and negative brain-PAD of most participants, it is possible
our findings represent contributing factors like cognitive reserve and brain reserve which
have enabled some individuals to tolerate more age-related change, and thus, prolong the
onset of cognitive decline (Stern et al., 2020). Findings, however, corroborate evidence from
neuroimaging studies which also report no association between baseline brain volume and
change in cognitive function in middle aged, and older adults (Armstrong et al., 2020;
Charlton et al., 2010; Ritchie et al., 2015). As there are currently no longitudinal studies
investigating brain age and cognitive decline, further evidence is required.

4.3. Strengths and limitations

Author Manuscript

A number of strengths of this study are notable. The prospective investigation of cognitive
function from a large community-based cohort of older cognitively unimpaired participants
enabled inferences to be targeted to the stage of life associated with cognitive decline and
heterogeneous aging trajectories. We also estimated brain age using a publicly sourced
model that has been previously validated using an independent test cohort (Cole et al.,
2018), providing for transparency and replicability of our results, while also undertaking
a robust, quantitative evaluation of image quality prior to data analysis. Further, estimates

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 11

are derived from local tissue volume (i.e., GM and WM volume) from across the whole
Author Manuscript

brain, which change with aging (Grajauskas et al., 2019). Finally, we accounted for possible
learning effects on our neuropsychological tests by performing a sensitivity analysis that
assessed associations with a 1- and 2-year change in cognitive function.

Several limitations must also be considered, and are as following. Although we targeted a
community-based sample, most participants who enrolled in the study were predominantly
white, highly educated (completed tertiary education), and were in the higher SES
percentile. Further, the proportion of participants who completed at least secondary school
education is a bit higher than the overall ASPREE populations, and estimates from the
broader Australian population (Tunny, 2006), which is around 50%. Though we account for
the effects of education, this, along with the inclusion of cognitively unimpaired individuals,
may have influenced cognitive outcomes and trajectories, given the established relationship
between wealth or SES and cognitive reserve. Considering the HVLT-R delayed recall is
Author Manuscript

prone to ceiling effects, and that eligible ASPREE participants required a 3MS score of 78
or above, these factors reduced the variability in cognitive scores in the population and thus
may have reduced the power to detect significant associations (Ryan et al., 2019).

Though longitudinal, this study was limited to a 3-year follow-up period. A longer follow-up
period would allow for greater variability in cognitive trajectories, including the onset of
clinically-relevant cognitive decline in some people, to emerge (Wu et al., 2021). Further,
findings were limited to a select few cognitive domains (i.e., delayed recall, psychomotor
speed, and verbal fluency and general cognitive status), that were evaluated using single
tests. Longitudinal analyses were also limited to a smaller sub-sample of participants with
cognitive data available at each follow-up time point. In future, the release of longitudinal
cognitive follow-up data for the NEURO participants, who had not yet attended a year 3
Author Manuscript

visit at the time the intervention phase of the ASPREE study was truncated in mid-2017,
will allow this to be addressed. Though as few as 3 repeated measures may not be enough
to produce reliable estimates, alternative approaches like linear mixed models have an innate
flexibility to handle missing values, and thus may be more appropriate for subsequent
studies (Brauer and Curtin, 2018).

The fit of the brain age prediction model was relatively accurate in our population; however,
the strength of the association is weaker than in the training sample and the independent
validation cohort used to develop the model. This may relate to the relatively narrow age
range included in our sample, in comparison to data from the full adult lifespan included
in the training cohort. The less accurate model fit may have introduced a measure of noise
into our estimates, alongside true interindividual variability, but does not invalidate our
Author Manuscript

inferences (Cole et al., 2017).

Despite a statistically significant association with baseline psychomotor speed and delayed
recall, and the 3-year change in memory and general cognitive status, the overall effect
of brain-PAD was relatively small, and only psychomotor speed remains significant when
applying a conservative Bonferroni corrected p value of 0.006 (4 cognitive tests at 2 time
points) to reduce the risk of type 1 error due to multiple comparisons. Thus, there is the
possibility that our findings reflect a false positive. Though the direction of the association

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 12

between brain-PAD and cognitive function are as predicted, our interpretation of a positive
Author Manuscript

brain-PAD – considered to reflect an older, “less healthy” looking brain – being associated
with a worse cognitive performance may be over simplistic. Further, we infer “brain aging”
from our cross-sectional measure of brain-PAD. While brain-PAD is often interpreted as a
proxy marker of accelerated or decelerated biological aging (Franke and Gaser, 2019), it
does not reflect a measure of longitudinal change in brain anatomy. Finally, further research
will be necessary to determine how different sub regions of the brain may be contributing to
the brain age estimation and cognition relationships, and whether inclusion of other imaging
modalities (e.g., white matter integrity) may improve the fit and sensitivity of the model
(Scheinost et al., 2019).

5. Conclusions
This study investigated whether brain aging is associated with cognitive function and
Author Manuscript

cognitive decline, in community dwelling older adults. Using estimates of local tissue
volume from across the whole brain, we provide additional evidence of an association
between accelerated brain aging and impaired psychomotor speed. Evidence of a weak
effect of accelerated brain aging on delayed recall memory, the decline in memory and
general cognitive status were also identified, but these didn’t remain significant after
adjustment for multiple comparisons. Given few studies have investigated brain age and
cognitive decline in later life, future research will benefit from further prospective analyses
of associations between brain age, cognitive function, and the risk of clinically-relevant
cognitive impairment in initially healthy, community dwelling older adults.

Supplementary Material
Author Manuscript

Refer to Web version on PubMed Central for supplementary material.

Acknowledgements
The authors would like to thank participants of the Australian ASPREE and NEURO studies for volunteering
their time, and general practitioners for supporting participants throughout the trial. Authors also acknowledge the
hard work of all ASPREE field staff, and those at the Monash Biomedical Imaging facility who attained magnetic
resonance images. Supported by grants (U01AG029824 and U19AG062682) from the National Institute on Aging
and the National Cancer Institute at the National Institutes of Health, by grants (334047; 1086188 and 1127060)
from the National Health and Medical Research Council of Australia, and by Monash University and the Victorian
Cancer Agency. Funders did not direct the conduction of this study, nor the decision to publish these findings.

Disclosure statement
Jo Wrigglesworth is the recipient of a Research Training Program stipend, awarded by Monash University and the
Australian government. Ian H. Harding is supported by NHMRC Fellowship [APP1106533]. Joanne Ryan is funded
by a NHMRC Dementia Research Leader Fellowship [APP1135727]. Sharna Jamadar is supported by an Australian
Author Manuscript

National Health and Medical Research Council Fellowship [APP1174164] and the Australian Research Council
Centre of Excellence for Integrative Brain Function [CE140100007]. The remaining authors report no conflicts of
interest.

References
Armstrong NM, An Y, Shin JJ, Williams OA, Doshi J, Erus G, Davatzikos C, Ferrucci L, Beason-Held
LL, Resnick SM, 2020. Associations between cognitive and brain volume changes in cognitively
normal older adults. Neuroimage 223, 117289. [PubMed: 32835822]

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 13

Ashburner J, 2007. A fast diffeomorphic image registration algorithm. Neuroimage 38 (1), 95–113.
[PubMed: 17761438]
Author Manuscript

Beheshti I, Maikusa N, Matsuda H, 2018. The association between “Brain-Age Score” (BAS) and
traditional neuropsychological screening tools in Alzheimer’s disease. Brain Behav. 8 (8), e01020.
[PubMed: 29931756]
Benedict RH, Schretlen D, Groninger L, Brandt J, 1998. Hopkins verbal learning test–revised:
normative data and analysis of inter-form and test-retest reliability. Clin. Neuropsychol 12 (1),
43–55.
Biondo F, Jewell A, Pritchard M, Mueller C, Steves CJ, Cole JH, 2020. Brain-age predicts subsequent
dementia in memory clinic patients: Neuroimaging/Optimal neuroimaging measures for early
detection. Alzheimer’s & Dementia 16 (S5), e037378. doi: 10.1002/alz.037378, In press.
Boyle R, Jollans L, Rueda-Delgado LM, Rizzo R, Yener GG, McMorrow JP, Knight SP, Carey
D, Robertson IH, Emek-Savas DD, Stern Y, Kenny RA, Whelan R, 2021. Brain-predicted age
difference score is related to specific cognitive functions: a multi-site replication analysis. Brain
Imaging Behav. 15 (1), 327–345. [PubMed: 32141032]
Brauer M, Curtin JJ, 2018. Linear mixed-effects models and the analysis of non-independent data:
Author Manuscript

a unified framework to analyze categorical and continuous independent variables that vary within-
subjects and/or within-items. Psychol. Methods 23 (3), 389–411. [PubMed: 29172609]
Bundesen C, 1990. A theory of visual attention. Psychol. Rev 97 (4), 523–547. [PubMed: 2247540]
Charlton RA, Schiavone F, Barrick TR, Morris RG, Markus HS, 2010. Diffusion tensor imaging
detects age related white matter change over a 2 year follow-up which is associated with working
memory decline. J. Neurol. Neurosurg. Psychiatry 81 (1), 13–19. [PubMed: 19710050]
Coelho A, Fernandes HM, Magalhaes R, Moreira PS, Marques P, Soares JM, Amorim L, Portugal-
Nunes C, Castanho T, Santos NC, Sousa N, 2021. Signatures of white-matter microstructure
degradation during aging and its association with cognitive status. Sci. Rep 11 (1), 4517. [PubMed:
33633204]
Cole JH, Leech R, Sharp DJ Alzheimer’s Disease Neuroimaging, I., 2015. Prediction of brain age
suggests accelerated atrophy after traumatic brain injury. Ann. Neurol 77 (4), 571–581. [PubMed:
25623048]
Cole JH, Ritchie SJ, Bastin ME, Valdes Hernandez MC, Munoz Maniega S, Royle N, Corley J,
Pattie A, Harris SE, Zhang Q, Wray NR, Redmond P, Marioni RE, Starr JM, Cox SR, Wardlaw
Author Manuscript

JM, Sharp DJ, Deary IJ, 2018. Brain age predicts mortality. Mol. Psychiatry 23 (5), 1385–1392.
[PubMed: 28439103]
Cole JH, Underwood J, Caan MW, De Francesco D, van Zoest RA, Leech R, Wit FW, Portegies
P, Geurtsen GJ, Schmand BA, Schim van der Loeff MF, Franceschi C, Sabin CA, Majoie CB,
Winston A, Reiss P, Sharp DJ collaboration, C., 2017. Increased brain-predicted aging in treated
HIV disease. Neurology 88 (14), 1349–1357. [PubMed: 28258081]
Council, M., 2009. Australian Guidelines to Reduce Health Risks from Drinking Alcohol. NHMRC,
Canberra.
Damoiseaux JS, 2017. Effects of aging on functional and structural brain connectivity. Neuroimage
160, 32–40. [PubMed: 28159687]
Dinsdale NK, Bluemke E, Smith SM, Arya Z, Vidaurre D, Jenkinson M, Namburete AIL, 2021.
Learning patterns of the ageing brain in MRI using deep convolutional networks. Neuroimage 224,
117401. [PubMed: 32979523]
Ebaid D, Crewther SG, MacCalman K, Brown A, Crewther DP, 2017. Cognitive processing speed
Author Manuscript

across the lifespan: beyond the influence of motor speed. Front. Aging Neurosci 9, 62. [PubMed:
28381999]
Elliott ML, Belsky DW, Knodt AR, Ireland D, Melzer TR, Poulton R, Ramrakha S, Caspi A, Moffitt
TE, Hariri AR, 2019. Brain-age in midlife is associated with accelerated biological aging and
cognitive decline in a longitudinal birth cohort. Mol. Psychiatry 26, 3829–3838 In press. [PubMed:
31822815]
Franke K, Gaser C, 2012. Longitudinal changes in individual BrainAGE in healthy aging, mild
cognitive impairment, and Alzheimer’s disease. GeroPsych 25 (4), 235.

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 14

Franke K, Gaser C, 2019. Ten years of BrainAGE as a neuroimaging biomarker of brain aging: what
insights have we gained? Front. Neurol 10, 789. [PubMed: 31474922]
Author Manuscript

Franke K, Ziegler G, Kloppel S, Gaser C Alzheimer’s Disease Neuroimaging, I., 2010. Estimating
the age of healthy subjects from T1-weighted MRI scans using kernel methods: exploring the
influence of various parameters. Neuroimage 50 (3), 883–892. [PubMed: 20070949]
Gaser C, Franke K, Kloppel S, Koutsouleris N, Sauer H Alzheimer’s Disease Neuroimaging, I., 2013.
BrainAGE in mild cognitive impaired patients: predicting the conversion to Alzheimer’s disease.
PLoS One 8 (6), e67346. [PubMed: 23826273]
Grajauskas LA, Siu W, Medvedev G, Guo H, D’Arcy RCN, Song X, 2019. MRI-based evaluation of
structural degeneration in the ageing brain: pathophysiology and assessment. Ageing Res. Rev 49,
67–82. [PubMed: 30472216]
Group, A.I., 2013. Study design of ASPirin in Reducing Events in the Elderly (ASPREE): a
randomized, controlled trial. Contemp. Clin. Trials 36 (2), 555–564. [PubMed: 24113028]
Jonsson BA, Bjornsdottir G, Thorgeirsson TE, Ellingsen LM, Walters GB, Gudbjartsson DF,
Stefansson H, Stefansson K, Ulfarsson MO, 2019. Brain age prediction using deep learning
uncovers associated sequence variants. Nat. Commun 10 (1), 5409. [PubMed: 31776335]
Author Manuscript

Kaufmann T, van der Meer D, Doan NT, Schwarz E, Lund MJ, Agartz I, Alnaes D, Barch DM,
Baur-Streubel R, Bertolino A, Bettella F, Beyer MK, Boen E, Borgwardt S, Brandt CL, Buitelaar
J, Celius EG, Cervenka S, Conzelmann A, Cordova-Palomera A, Dale AM, de Quervain DJF,
Di Carlo P, Djurovic S, Dorum ES, Eisenacher S, Elvsashagen T, Espeseth T, Fatouros-Bergman
H, Flyckt L, Franke B, Frei O, Haatveit B, Haberg AK, Harbo HF, Hartman CA, Heslenfeld
D, Hoekstra PJ, Hogestol EA, Jernigan TL, Jonassen R, Jonsson EG, Karolinska Schizophrenia
P, Kirsch P, Kloszewska I, Kolskar KK, Landro NI, Le Hellard S, Lesch KP, Lovestone S,
Lundervold A, Lundervold AJ, Maglanoc LA, Malt UF, Mecocci P, Melle I, Meyer-Lindenberg
A, Moberget T, Norbom LB, Nordvik JE, Nyberg L, Oosterlaan J, Papalino M, Papassotiropoulos
A, Pauli P, Pergola G, Persson K, Richard G, Rokicki J, Sanders AM, Selbaek G, Shadrin AA,
Smeland OB, Soininen H, Sowa P, Steen VM, Tsolaki M, Ulrichsen KM, Vellas B, Wang L,
Westman E, Ziegler GC, Zink M, Andreassen OA, Westlye LT, 2019. Common brain disorders
are associated with heritable patterns of apparent aging of the brain. Nat. Neurosci 22 (10), 1617–
1623. [PubMed: 31551603]
Liem F, Varoquaux G, Kynast J, Beyer F, Kharabian Masouleh S, Huntenburg JM, Lampe L, Rahim
Author Manuscript

M, Abraham A, Craddock RC, Riedel-Heller S, Luck T, Loeffler M, Schroeter ML, Witte AV,
Villringer A, Margulies DS, 2017. Predicting brain-age from multimodal imaging data captures
cognitive impairment. Neuroimage 148, 179–188. [PubMed: 27890805]
Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G, 2013. The hallmarks of aging. Cell
153 (6), 1194–1217. [PubMed: 23746838]
Lowe LC, Gaser C, Franke K Alzheimer’s Disease Neuroimaging, I., 2016. The effect of the APOE
genotype on individual BrainAGE in normal aging, mild cognitive impairment, and Alzheimer’s
disease. PLoS One 11 (7), e0157514. [PubMed: 27410431]
Ly M, Yu GZ, Karim HT, Muppidi NR, Mizuno A, Klunk WE, Aizenstein HJ Alzheimer’s Disease
Neuroimaging, I., 2020. Improving brain age prediction models: incorporation of amyloid status in
Alzheimer’s disease. Neurobiol. Aging 87, 44–48. [PubMed: 31843257]
McNeil JJ, Woods RL, Nelson MR, Murray AM, Reid CM, Kirpach B, Storey E, Shah RC, Wolfe RS,
Tonkin AM, Newman AB, Williamson JD, Lockery JE, Margolis KL, Ernst ME, Abhayaratna WP,
Stocks N, Fitzgerald SM, Trevaks RE, Orchard SG, Beilin LJ, Donnan GA, Gibbs P, Johnston CI,
Grimm RH Group, A.I., 2017. Baseline characteristics of participants in the ASPREE (ASPirin
Author Manuscript

in Reducing Events in the Elderly) study. J. Gerontol. A Biol. Sci. Med. Sci 72 (11), 1586–1593.
[PubMed: 28329340]
Organization, W.H., 2019. Risk reduction of cognitive decline and dementia: WHO guidelines.
Oschwald J, Guye S, Liem F, Rast P, Willis S, Rocke C, Jancke L, Martin M, Merillat S, 2019. Brain
structure and cognitive ability in healthy aging: a review on longitudinal correlated change. Rev.
Neurosci 31 (1), 1–57. [PubMed: 31194693]
Radloff LS, 1977. The CES-D scale: A self-report depression scale for research in the general
population. Appl. Psychol. Meas 1 (3), 385–401.

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 15

Richard G, Kolskar K, Sanders AM, Kaufmann T, Petersen A, Doan NT, Monereo Sanchez J, Alnaes
D, Ulrichsen KM, Dorum ES, Andreassen OA, Nordvik JE, Westlye LT, 2018. Assessing distinct
Author Manuscript

patterns of cognitive aging using tissue-specific brain age prediction based on diffusion tensor
imaging and brain morphometry. PeerJ 6, e5908. [PubMed: 30533290]
Ritchie SJ, Dickie DA, Cox SR, Valdes Hernandez Mdel C, Corley J, Royle NA, Pattie A, Aribisala
BS, Redmond P, Munoz Maniega S, Taylor AM, Sibbett R, Gow AJ, Starr JM, Bastin ME,
Wardlaw JM, Deary IJ, 2015. Brain volumetric changes and cognitive ageing during the eighth
decade of life. Hum. Brain Mapp 36 (12), 4910–4925. [PubMed: 26769551]
Ross TP, 2003. The reliability of cluster and switch scores for the controlled oral word association test.
Arch. Clin. Neuropsychol 18 (2), 153–164. [PubMed: 14591467]
Ryan J, Woods RL, Britt C, Murray AM, Shah RC, Reid CM, Kirpach B, Wolfe RS, Nelson
MR, Lockery JE, Orchard SG, Trevaks RE, McNeil JJ, Storey E Group, A.I., 2019. Normative
performance of healthy older individuals on the Modified Mini-Mental State (3MS) examination
according to ethno-racial group, gender, age, and education level. Clin. Neuropsychol 33 (4),
779–797. [PubMed: 29976121]
Ryan J, Woods RL, Britt CJ, Murray AM, Shah RC, Reid CM, Wolfe R, Nelson MR, Orchard SG,
Author Manuscript

Lockery JE, Trevaks RE, Storey E behalf of the, A.I.G.o., 2020. Normative data for the symbol
digit modalities test in older White Australians and Americans, African-Americans, and Hispanic/
Latinos. J. Alzheimers Dis. Rep 4 (1), 313–323. [PubMed: 33024939]
Salthouse TA, 2019. Trajectories of normal cognitive aging. Psychol. Aging 34 (1), 17–24. [PubMed:
30211596]
Scheinost D, Noble S, Horien C, Greene AS, Lake EM, Salehi M, Gao S, Shen X, O’Connor
D, Barron DS, Yip SW, Rosenberg MD, Constable RT, 2019. Ten simple rules for predictive
modeling of individual differences in neuroimaging. Neuroimage 193, 35–45. [PubMed:
30831310]
Smith A, 1982. Symbol digit modalities test (SDMT) manual (Revised). Western Psychological
Services, Los Angeles.
Smith SM, Vidaurre D, Alfaro-Almagro F, Nichols TE, Miller KL, 2019. Estimation of brain age delta
from brain imaging. Neuroimage 200, 528–539. [PubMed: 31201988]
Soros P, Bantel C, 2020. Chronic noncancer pain is not associated with accelerated brain aging as
assessed by structural magnetic resonance imaging in patients treated in specialized outpatient
Author Manuscript

clinics. Pain 161 (3), 641–650. [PubMed: 31764393]

Statistics, A.B.o., 2016. Socio-Economic Indexes for Areas (SEIFA), in: Statistics, A.B.o. (Ed.).
Australian Bureau of Statistics, Canberra.
Stern Y, Arenaza-Urquijo EM, Bartres-Faz D, Belleville S, Cantilon M, Chetelat G, Ewers M,
Franzmeier N, Kempermann G, Kremen WS, Okonkwo O, Scarmeas N, Soldan A, Udeh-Momoh
C, Valenzuela M, Vemuri P, Vuoksimaa E the Reserve, R., Protective Factors, P.I.A.E.D.,
Conceptual Frameworks, W., 2020. Whitepaper: defining and investigating cognitive reserve, brain
reserve, and brain maintenance. Alzheimers Dement 16 (9), 1305–1311. [PubMed: 30222945]
Teng EL, Chui HC, 1987. The Modified Mini-Mental State (3MS) examination. J. Clin. Psychiatry 48
(8), 314–318. [PubMed: 3611032]
Tombaugh TN, 2005. Test-retest reliable coefficients and 5-year change scores for the MMSE and
3MS. Arch. Clin. Neuropsychol 20 (4), 485–503. [PubMed: 15896562]
Tucker-Drob EM, 2019. Cognitive aging and dementia: a life-span perspective. Annu. Rev. Dev.
Psychol 1, 177–196. [PubMed: 34046638]
Author Manuscript

Tunny G, 2006. Educational attainment in Australia. Economic Round-up (Autumn 2006), 1–9.
Ward SA, Raniga P, Ferris NJ, Woods RL, Storey E, Bailey MJ, Brodtmann A, Yates PA, Donnan GA,
Trevaks RE, Wolfe R, Egan GF, McNeil JJon behalf of the, A.i.g., 2017. ASPREE-NEURO study
protocol: a randomized controlled trial to determine the effect of low-dose aspirin on cerebral
microbleeds, white matter hyperintensities, cognition, and stroke in the healthy elderly. Int. J.
Stroke 12 (1), 108–113. [PubMed: 27634976]
Wolfe R, Murray AM, Woods RL, Kirpach B, Gilbertson D, Shah RC, Nelson MR, Reid CM, Ernst
ME, Lockery J, Donnan GA, Williamson J, McNeil JJ, 2018. The aspirin in reducing events in the
elderly trial: statistical analysis plan. Int. J. Stroke 13 (3), 335–338. [PubMed: 29111960]

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 16

Woods RL, Espinoza S, Thao LTP, Ernst ME, Ryan J, Wolfe R, Shah RC, Ward SA, Storey E, Nelson
MR, Reid CM, Lockery JE, Orchard SG, Trevaks RE, Fitzgerald SM, Stocks NP, Williamson
Author Manuscript

JD, McNeil JJ, Murray AM, Newman AB Group, A.I., 2020. Effect of aspirin on activities of
daily living disability in community-dwelling older adults. J. Gerontol. A Biol. Sci. Med. Sci 76,
2007–2014.
Wrigglesworth J, Ward P, Harding IH, Nilaweera D, Wu Z, Woods RL, Ryan J, 2021. Factors
associated with brain ageing-a systematic review. BMC Neurol. 21 (1), 1–23. [PubMed:
33390161]
Wu Z, Woods RL, Wolfe R, Storey E, Chong TTJ, Shah RC, Orchard SG, McNeil JJ, Murray AM,
Ryan J Group, A. I., 2021. Trajectories of cognitive function in community-dwelling older adults:
a longitudinal study of population heterogeneity. Alzheimers Dement (Amst) 13 (1), e12180.
[PubMed: 33969173]
Author Manuscript
Author Manuscript
Author Manuscript

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 17
Author Manuscript

Fig. 1.
Scatterplots correlating brain-predicted age difference (brain-PAD) with baseline SDMT (A)
Author Manuscript

and HVLT-R delayed recall (B).

Author Manuscript
Author Manuscript

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 18
Author Manuscript
Author Manuscript

Fig. 2.
The association between baseline brain-predicted age difference (brain-PAD) and the 3 year
change in delayed recall task of HVLT-R.
Author Manuscript
Author Manuscript

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 19

Table 1

Baseline characteristics of participants included in the study

Author Manuscript

Total cohort Longitudinal cohort

N 531 259
Chronological age (y), median (IQR) 72.5 (71.2–75.5) 72.3 (71.2–75.2)
Age group (y), n (%)
70–74 386 (72.7) 188 (72.6)
≥75 145 (27.3) 71 (27.4)
Female gender, n (%) 257 (48.4) 111 (42.9)
a
Race, n (%)
White 517 (97.6) 250 (96.5)
Other 13 (2.5) 9 (3.5)
Education (y), n (%)
Author Manuscript

<12 204 (38.4) 102 (39.4)

≥12 327 (61.6) 157 (60.6)
SES (percentile), n (%)
≤25 34 (6.4) 13 (5.0)
25–74 184 (34.7) 99 (38.2)
≥75 313 (59.0) 147 (56.8)
Living alone (yes), n (%) 143 (26.9) 62 (23.9)
Smoking (yes), n (%)
Current 20 (3.8) 8 (3.1)
Former 207 (39.0) 94 (36.3)
Never 304 (57.3) 157 (60.6)
Alcohol consumption, n (%)
Author Manuscript

Never 68 (12.8) 34 (13.1)

Former 17 (3.2) 8 (3.1)
Low risk 200 (37.7) 85 (32.8)
High risk 246 (46.3) 132 (51.0)
a
BMl (kg/m2), n (%)
<20 5 (1.0) 2 (0.8)
20–24 139 (26.3) 63 (24.3)
25–29 239 (45.3) 126 (48.7)
≥30 145 (27.5) 68 (26.3)
Diabetes (yes), n (%) 58 (10.9) 29 (11.2)
Dyslipidemia (yes), n (%) 312 (58.8) 151 (58.3)
Depression (yes), n (%) 54 (10.2) 26 (10.0)
Author Manuscript

Hypertensive (yes), n (%) 371 (69.9) 184 (71.0)

a 91 (18.2) 43 (17.5)
Chronic Kidney Disease (yes), n (%)
Frail, n (%)
Not frail 377 (71.0) 191 (73.8)
Prefrail 150 (28.3) 67 (25.9)

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Wrigglesworth et al. Page 20

Total cohort Longitudinal cohort

N 531 259
Frail 4 (0.75) 1 (0.4)
Author Manuscript

Aspirin treatment (yes), n (%) 265 (49.9) 135 (52.1)

a
Number of participants from total cohort who are missing from body mass index (BMI, n = 3), race (n = 1), and Chronic Kidney Disease
(CKD, n = 31); participants missing from longitudinal sub-sample who are missing from CKD (n = 13). Education is defined as less than 12
years of education (i.e., did not complete secondary education in Australia), or 12 or more years of education (i.e., completed tertiary education
in Australia) (Tunny, 2006). Baseline socioeconomic status (SES) was derived using the Index of Relative Socio-Economic Advantage and
Disadvantage percentile score (SEIFA) (Statistics, 2016). Current alcohol consumption defined as low (<10 standard drinks per week/4 standard
drinks on any 1 day), or high risk (≥10 standard drinks per week/4 standard drinks on any 1 day) (Council, 2009). Diabetes is defined as a
fasting glucose level at least 126mg per deciliter (≥7 mmol per liter), or a recipient of diabetes medication or self-report. Dyslipidemia is defined
as the receipt of cholesterol lowering medications, or as serum cholesterol levels of at least 212mg per deciliter (≥5.5 mmol per liter), or as a
lipoprotein density level of more than 160 mg per deciliter (≥4.1 mmol per liter). Depression is defined as a score of ≥8 on the 10-item Centre for
Epidemiologic Studies Depression Scale (Radloff, 1977). Hypertension is defined as receipt of treatment for high blood pressure(BP) or a BP more
than 140/90 at ASPREE trial entry. Chronic Kidney Disease is defined as having an estimated glomerular filtration rate less than 60ml per min per
1.73m2, or a ratio of albumin (milligrams per liter) to creatinine (mmols per liter) in urine of 3 or more. Frailty is defined using the modified Fried
Frailty criteria, that include low body weight, low grip strength, low gait speed, exhaustion, slow walking speed and low levels of physical activity
Author Manuscript

(Wolfe et al., 2018). Participants were categorized as “prefrail” if they met 1 or 2 criteria, while “frail” individuals met 3 or more.
Author Manuscript
Author Manuscript

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

Table 2

Cognitive function at baseline and year 3

Total cohort Longitudinal cohort

Change in cognitive function

e e
Cognitive measure n Mean (SD) n Baseline Mean (SD) Follow-up Mean (SD) Mean (95% CI) p
Wrigglesworth et al.

a 531 93.8 (4.3) 258 93.6 (4.2) 94.0 (4.9) 0.5 (−0.1, 1.0) 0.10
3MS, total score
b 531 12.4 (4.5) 259 12.1 (4.2) 13.5 (5.0) 1.3 (0.8, 1.9) <0.0001
COWAT
c 528 39.1 (8.9) 258 39.5 (9.0) 39.0 (9.2) −0.5 (−1.1, 0.3) 0.21
SDMT
d 530 8.0 (2.7) 257 8.0 (2.7) 8.3 (3.2) 0.3 (−0.02, 0.6) 0.07
HVLT-R, delayed recall

Bold indicates significant change in scores of cognitive functions assessed at baseline to year 3 follow-up (p < 0.05). Key: CI, confidence interval; COWAT, Controlled Oral Word Association Test; HVLT,
Hopkins Verbal Learning Test-Revised, delayed recall; 3MS, Modified Mini-Mental State Examination, overall score; SDMT, Symbol Digit Modalities Test.
a
Total 3MS scores range from 78 to 100, with higher scores representing a better general cognitive status (Teng and Chui, 1987).
b
Values represent the total number of words starting with F, that were produced over three 60 second time trials, minus any unacceptable responses (Ross, 2003). A greater number of words suggests better
phonemic verbal fluency.
c
Represents the number of correct items from form C of the SDMT, in 90 seconds (Ryan et al., 2020). Participants completing more task are considered to have faster psychomotor speed.
d
The number of words recalled for the delayed recall task. Words were 12 nouns consisting of 4 words drawn from 3 semantic categories. A greater number of words indicates a better delayed recall.
e
Individuals could not be included as they declined to provide data for SDMT (n = 3) and HVLT-R (n = 1) at baseline; declined to provide data at the 3-year assessment for SDMT (n = 1), and HVLT-R (n =
2), or the study ceased before the year 3 assessment of 3MS (n = 1).

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

Page 21
 Author Manuscript Author Manuscript Author Manuscript Author Manuscript

Table 3

Analyses of linear associations between brain-predicted age difference and cognitive function at baseline, and the change in cognition (Year 3 – Baseline)

Cognitive measure 3MS, total score Baseline cognitive function Three-year Change
a r b (95% CI) p b r b (95% CI) p
n n
Univariable 531 −0.02 −0.01 (−0.07, 0.05) 0.72 258 −0.1 −0.08 (−0.17, 0.01) 0.09
Wrigglesworth et al.

Model 1 531 −0.01 (−0.07, 0.05) 0.7 258 −0.08 (−0.17, 0.01) 0.10
Model 2 528 −0.01 (−0.07, 0.05) 0.75 258 −0.08 (−0.17, 0.02) 0.10
Model 3 258 −0.09 (−0.17, −0.01) 0.04
COWAT
Univariable 531 −0.08 −0.06 (−0.13, 0.003) 0.06 259 −0.04 −0.03 (−0.12, 0.06) 0.51
Model 1 531 −0.06 (−0.13, 0.003) 0.06 259 −0.03 (−0.13, 0.06) 0.47
Model 2 528 −0.06 (−0.12, 0.004) 0.07 259 −0.04 (−0.13, 0.06) 0.43
Model 3 259 −0.03 (−0.12, 0.06) 0.48
SDMT
Univariable 528 −0.14 −0.21 (−0.34, −0.08) 0.001 258 0.06 0.06 (−0.06, 0.17) 0.35
Model 1 528 −0.21 (−0.34, −0.09) 0.001 258 0.06 (−0.06, 0.17) 0.34
Model 2 525 −0.22 (−0.34, −0.10) <0.0001 258 0.06 (−0.05, 0.18) 0.29
Model 3 258 0.04 (−0.08, 0.15) 0.51
HVLT-R, delayed recall
Univariable 530 −0.09 −0.04 (−0.08, −0.003) 0.04 257 −0.15 −0.07 (−0.12, −0.01) 0.02
Model 1 530 −0.04 (−0.08, −0.004) 0.03 257 −0.06 (−0.11, −0.01) 0.02
Model 2 527 −0.04 (−0.08, −0.002) 0.04 257 −0.06 (−0.11, −0.01) 0.02
Model 3 257 −0.06 (−0.11, −0.01) 0.02

Neurobiol Aging. Author manuscript; available in PMC 2022 February 11.

Model 1: adjusting for age; Model 2: adjusting for age, gender, education, body mass index (BMI), depression; Model 3: adjusting for all covariates, including baseline cognitive function.

Bold indicates significant association between PAD and cognitive function after Bonferroni correction (p < 0.006).

Key: b, unstandardized beta coefficient; CI, confidence interval; COWAT, Controlled Oral Word Association Test; HVLT, Hopkins Verbal Learning Test-Revised, delayed recall; 3MS, Modified
Mini-Mental State Examination, overall score; SDMT, Symbol Digit Modalities Test.
a
Individuals could not be included as they declined to provide baseline data for SDMT (n = 3) and HVLT-R (n = 1), or participants were missing baseline body mass index (n = 3).
b
Individuals could not be included as they declined to provide baseline and/or follow-up data for the 3-year change in SDMT (n = 1), and HVLT-R (n = 2), or the study had ceased before year 3 assessment
for 3MS (n = 1).
Page 22