NeuroImage: Clinical 25 (2020) 102182

Contents lists available at ScienceDirect

NeuroImage: Clinical
journal homepage: [Link]/locate/ynicl

Hippocampal plasticity underpins long-term cognitive gains from resistance T

exercise in MCI
⁎
Kathryn M. Broadhousea,b, , Maria Fiatarone Singhc,d, Chao Suob,e, Nicola Gatesb,f,g, Wei Wenf,h,
Henry Brodatyg,i, Nidhi Jainj, Guy C. Wilsonj, Jacinda Meiklejohnj, Nalin Singhj,
Bernhard T. Baunek, Michael Bakerc,l, Nasim Foroughim, Yi Wangm,n, Nicole Kochang,
Kevin Ashtono, Matt Brownp,r, Zhixiu Lip, Yorgi Mavrosc, Perminder S. Sachdevg,
⁎⁎
Michael [Link] ,g,q
a
Nola Thompson Centre for Advanced Imaging, Sunshine Coast Mind and Neuroscience Thompson Institute, University of the Sunshine Coast, QLD, Australia
b
Regenerative Neuroscience Group, Brain and Mind Centre and Sydney Medical School, The University of Sydney, Sydney, NSW, Australia
c
Physical Activity, Lifestyle, Ageing and Wellbeing Faculty Research Group, Faculty of Health Sciences and Sydney Medical School, The University of Sydney, Lidcombe,
NSW, Australia
d
Hebrew SeniorLife and Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University, Boston, MA, USA
e
School of Psychological Sciences and Monash Biomedical Imaging, Monash University, Melbourne, VIC, Australia
f
School of Psychiatry, University of New South Wales, Sydney, NSW, Australia
g
Centre for Healthy Brain Ageing, School of Psychiatry, University of New South Wales, Sydney, NSW, Australia
h
Neuropsychiatric Institute, Prince of Wales Hospital, Sydney, NSW, Australia
i
Dementia Collaborative Research Centre, University of New South Wales, Sydney, NSW, Australia
j
Physical Activity, Lifestyle, Ageing and Wellbeing Faculty Research Group, Faculty of Health Sciences, The University of Sydney, Lidcombe, NSW, Australia
k
Department of Psychiatry, School of Medicine, University of Adelaide, Adelaide, SA, Australia
l
School of Exercise Science, Australian Catholic University, Strathfield, NSW, Australia
m
Clinical and Rehabilitation Research Group, Faculty of Health Sciences, The University of Sydney, Lidcombe, NSW, Australia
n
Department of Medicine and the Diabetes Center, University of California, San Francisco, San Francisco, CA, USA
o
Biomedical Sciences, Faculty of Health Sciences and Medicine, Bond University, QLD, Australia
p
Institute of Health and Biomedical Innovation, Queensland University of Technology, QLD, Australia
q
School of Medical Sciences, Sydney Medical School, University of Sydney, Sydney, NSW, Australia
r
King’s College London National Institutes of Health Biomedical Research Centre, UK

A R T I C LE I N FO A B S T R A C T

Keywords: Dementia affects 47 million individuals worldwide, and assuming the status quo is projected to rise to 150
Resistance exercise million by 2050. Prevention of age-related cognitive impairment in older persons with lifestyle interventions
Mild cognitive impairment continues to garner evidence but whether this can combat underlying neurodegeneration is unknown. The Study
Plasticity of Mental Activity and Resistance Training (SMART) trial has previously reported within-training findings; the
Hippocampus
aim of this study was to investigate the long-term neurostructural and cognitive impact of resistance exercise in
Subfields
Randomised controlled trial
Mild Cognitive Impairment (MCI). For the first time we show that hippocampal subareas particularly susceptible
to volume loss in Alzheimer's disease (AD) are protected by resistance exercise for up to one year after training.
One hundred MCI participants were randomised to one of four training groups: (1) Combined high intensity
progressive resistance and computerised cognitive training (PRT+CCT), (2) PRT+Sham CCT, (3) CCT+Sham
PRT, (4) Sham physical+sham cognitive training (SHAM+SHAM). Physical, neuropsychological and MRI as-
sessments were carried out at baseline, 6 months (directly after training) and 18 months from baseline (12
months after intervention cessation). Here we report neuro-structural and functional changes over the 18-month
trial period and the association with global cognitive and executive function measures.
PRT but not CCT or PRT+CCT led to global long-term cognitive improvements above SHAM intervention at
18-month follow-up. Furthermore, hippocampal subfields susceptible to atrophy in AD were protected by PRT
revealing an elimination of long-term atrophy in the left subiculum, and attenuation of atrophy in left CA1 and
dentate gyrus when compared to SHAM+SHAM (p = 0.023, p = 0.020 and p = 0.027). These neuroprotective

⁎
Corresponding author at: Nola Thompson Centre for Advanced Imaging, Sunshine Coast Mind and Neuroscience Thompson Institute, University of the Sunshine
Coast, QLD, Australia.
⁎⁎
Corresponding author at: Centre for Healthy Brain Ageing, School of Psychiatry, University of New South Wales, Sydney, NSW, Australia.
E-mail addresses: kbroadhouse@[Link] (K.M. Broadhouse), [Link]@[Link] (M. [Link]).

[Link]
Received 2 August 2019; Received in revised form 13 January 2020; Accepted 13 January 2020
Available online 14 January 2020
2213-1582/ © 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
([Link]
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

effects mediated a significant portion of long-term cognitive benefits. By contrast, within-training posterior
cingulate plasticity decayed after training cessation and was unrelated to long term cognitive benefits. Neither
general physical activity levels nor fitness change over the 18-month period mediated hippocampal trajectory,
demonstrating that enduring hippocampal subfield plasticity is not a simple reflection of post-training changes in
fitness or physical activity participation. Notably, resting-state fMRI analysis revealed that both the hippo-
campus and posterior cingulate participate in a functional network that continued to be upregulated following
intervention cessation.
Multiple structural mechanisms may contribute to the long-term global cognitive benefit of resistance ex-
ercise, developing along different time courses but functionally linked. For the first time we show that 6 months
of high intensity resistance exercise is capable of not only promoting better cognition in those with MCI, but also
protecting AD-vulnerable hippocampal subfields from degeneration for at least 12 months post-intervention.
These findings emphasise the therapeutic potential of resistance exercise; however, future work will need to
establish just how long-lived these outcomes are and whether they are sufficient to delay dementia.

1. Introduction to replicate a positive trophic effect and rather shown attenuated

atrophy over the duration of the aerobic intervention (i.e., within-
Dementia affects 47 million individuals worldwide and assuming training effect) (Smith et al., 2010). Whether any physical exercise can
the status quo is projected to rise to 150 million by 2050 modify degenerative trajectories in MCI at either the whole-hippo-
(Alzheimer's Disease International, 2016). However, there are good campal level or within AD-vulnerable subfields is unknown (Firth et al.,
reasons to predict otherwise, as 30–40% of dementia is estimated to be 2018).
attributable to modifiable risk factors such as physical and cognitive Previously, in the Study of Mental Activity and Resistance Training
inactivity (Livingston et al., 2017; Norton et al., 2014). Consequently, (SMART) Trial (Gates et al., 2011), we found that 6 months of high
there is now intense international interest in large-scale clinical trials intensity progressive resistance training (PRT) in MCI was efficacious
targeted for the purpose of slowing cognitive-functional decline and for the trial's primary outcome, general cognition (Fiatarone
potentially delaying dementia onset (Hill et al., 2017; Ngandu et al., Singh et al., 2014; Mavros et al., 2017), and furthermore, this was
2015; Richard et al., 2009; Scarmeas, 2017; Vellas et al., 2014; therapeutically-mediated by structural plasticity in the posterior cin-
Yassine and Schneider, 2017). In this regard, delay is equivalent to gulate (PC) (Suo et al., 2016). We examined PRT because it is feasible in
population-level prevention because of death from other causes. Mod- frail and sarcopenic older individuals, is particularly effective for gen-
elling suggests that an intervention that delays onset by just one year, eral health outcomes such as cardiovascular disease, mobility impair-
implemented in 2010, could avert 1.3 million dementia cases in the US ment/falls, osteoporosis, and depression (Cooney et al., 2014;
by 2050 (Zissimopoulos et al., 2014). Irvine and Taylor, 2009; Peterson and Gordon, 2011), and continues to
Key to such prevention strategies is when to begin, how long to be under-investigated in relation to cognition (Liu-Ambrose and
intervene, and to understand if it is possible to modify the course of Donaldson, 2009). Within-training effects on the PC are promising be-
underlying neurodegeneration. Cognitive symptoms emerge many cause it is another target of degeneration and dysfunction in early AD
years before dementia diagnosis, and there is generally a transition (Choo et al., 2010), changes that coincide with hippocampal and en-
through the preclinical state, Mild Cognitive Impairment (MCI). MCI torhinal cortical degeneration (Huang et al., 2002).
increases the risk of progression to dementia by five to ten-fold Given that research on neuroplasticity has so far been limited to
(Petersen et al., 1999), with post mortem examinations indicating sub- within-training effects, investigating what happens to the brain after
stantial neuronal loss in the hippocampus (Scheff et al., 2006). In fact, training cessation is of particular importance. The SMART Trial cohort
MRI studies have shown that hippocampal atrophy can precede cog- is unique in this regard because we continue to see positive cognitive
nitive impairment by several years (Jack et al., 2000) and correlates effects up to 12 months post-training (Fiatarone Singh et al., 2014).
with the degree of cognitive dysfunction upon MCI diagnosis Understanding the mechanistic basis for this was therefore our primary
(Jack et al., 2000; Morra et al., 2009). Both imaging and histological aim. Here, we report for the first time the long-term impact of PRT on
studies suggest that hippocampal atrophy in MCI is not homogeneous, hippocampal subfields and PC structural and functional trajectories.
with the most severe volume loss being in CA1 and subiculum subfields, Finally, we model aerobic capacity, muscle strength and exercise par-
a pattern associated with progression from MCI to Alzheimer's dementia ticipation after the cessation of formal training, to determine any
(AD) (Apostolova et al., 2006, 2010; Costafreda et al., 2011; physical activity (PA) behaviour or physical fitness mediation of neu-
Rossler et al., 2002; West et al., 1994). Yet no intervention, behavioural roplasticity.
or pharmacologic, has been found to influence the trajectory of AD-
vulnerable subfield degeneration.
2. Materials and methods
Aerobic exercise has been suggested as a possible promoter of hip-
pocampal plasticity, but trials in healthy older adults have produced
The SMART trial was approved by the Human Research Ethics
mixed results. For example, ten Brinke et al., 2015 found that aerobic
Committee of the University of Sydney. Australian and New Zealand
training significantly increased hippocampal volume in older women
Clinical Trials Registry Protocol No: X08-0064. Registry No:
with probable MCI (ten Brinke et al., 2015). Furthermore,
ACTRN12608000489392.
Erickson et al., 2011 found that one year of moderate-intensity walking
not only reversed whole hippocampal atrophy in cognitively healthy
older adults but triggered volumetric expansion; conversely, memory 2.1. SMART trial design
and cognitive outcomes were not differentially impacted compared to a
control group engaged in progressive multi-modal exercise sessions Community-dwelling, non-depressed, non-demented participants
which included moderate intensity resistance training, yoga, balance aged ≥55, diagnosed with MCI (Petersen original criteria)
and stretching exercises (Erickson et al., 2011), reflecting conflicting (Petersen et al., 1999), with a Mini-mental State Examination score of
clinical trial outcomes more generally (Baker et al., 2010; Gates et al., 24–28, and a Clinical Dementia Rating Score ≤0.5, were recruited.
2013; Wang et al., 2014; Young et al., 2015). Other studies have failed Participants were randomised into one of four training groups in which
they completed 90-min supervised, centre-based training, two or three

2
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

times a week, for 26 weeks. (1) Combined high intensity progressive com/USA/[Link]), a multidomain, computer-based software
resistance and computerized cognitive training (PRT+CCT), (2) package developed for neurorehabilitation was used for CCT. Sham CCT
PRT+Sham computerized cognitive training, (3) CCT+Sham was also computer based: participants watched video clips of general
stretching and toning training and finally, (4) Sham physical+sham interest documentary topics, followed by a set of simple questions re-
cognitive training (SHAM+SHAM). Physical, metabolic, and a battery garding the presented material (Gates et al., 2011). CCT and sham
of neuropsychological tests as well as MRI assessments were carried out cognitive training were matched for duration, setting and sensorimotor
at baseline, 6 months and 18 months from baseline (Fig. 1). In addition, stimulation.
genotyping for the apolipoprotein E polymorphism ε4 allele (APOE-e4)
was carried out at baseline. Full details describing the SMART trial have 2.3. Cognitive assessments
been previously reported (Gates et al., 2011).
Cognitive status was evaluated via the Alzheimer's Disease
2.2. Training interventions Assessment Scale–cognitive subscale (ADASCog) and Executive domain,
assessed using the Matrices and Similarities subtests of the Wechsler
Physical exercise training. Pneumatic resistance machines (Keiser Adult Intelligence Scale, Third Edition (WAIS-III) and verbal fluency
Sports Health Equipment, Ltd, Fresno, CA, USA) were used for training (Controlled Oral Words Association Test (COWAT), Animal Naming).
at high intensity (80% of peak capacity, continuously progressed), three The MCI diagnosis was retrospectively sub-categorised to amnestic and
sets of eight repetitions of each of five–six exercises/session for most non-amnestic (aMCI, naMCI) to characterise the cohort at baseline.
major muscle groups (chest press, leg press, seated row, standing hip Specifically, aMCI was defined by performance falling ≥1 SD below
abduction, knee extension). PRT was supervised by experienced re- normative values or equivalent on two or more of three memory do-
search assistants (exercise physiologists and physiotherapists) in a main tests: ADASCog word list, BVRT-revised or Logical Memory.
physician-supervised clinic at the University of Sydney Faculty of
Health Sciences campus in a ratio of one trainer to four–five partici- 2.4. Exercise mediation measures
pants. Sham physical exercise included stretching and seated ca-
listhenics, designed not to notably increase heart rate or aerobic ca- To determine whether long-term benefits of resistance exercise are a
pacity or improve balance or strength. direct consequence of training dose, or rather, mediated by habitual
Cognitive training. All cognitive training was conducted at the uni- physical activity behaviour and/or physical fitness adaptations three
versity campus under supervision. COGPACK program ([Link]. different variables (Fiatarone Singh et al., 2014) were tested: peak

Fig. 1. SMART Trial design and statistical models. (A) Randomization was into one of four training groups – combined progressive resistance training and com-
puterised cognitive training (PRT+CCT), PRT and cognitive sham (PRT+SHAM), CCT and sham exercise (CCT+SHAM), sham exercise and cognitive sham, (SHAM
+SHAM). MRI, cognitive and fitness assessments were carried out at baseline, directly after 6 months of training (6M) and at 18 months (18M), following a one-year
usual care (UC) period with weekly telephone contact by research staff to administer health/adverse event checks in all participants. In addition, cognitive, physical,
social, recreational, volunteer and religious activity participation was recorded daily for the entire 18 months in a logbook by each participant. (B) Raw change in the
trial's primary outcome (ADASCog error scores) over the complete 18-month period, as well as neuropsychologically-defined Executive domain scores were compared
between SHAM+SHAM and the three training groups. Box-and-Whisker plots show the median (horizontal line), interquartile range (box) and upper and lower
quartiles (whiskers) of change in outcome scores for each training group. Dotted line indicates no change in outcome over the 18-month period. LME analysis found
significantly (*Time x Group p < 0.05 superimposed on the Box-and-Whisker plots) improved cognition in the PRT+SHAM group compared to SHAM+SHAM on
both outcomes (model including covariates age, sex and education, as well as group factor and time as continuous variable). (C) Basic Linear Mixed Effects (LME)
model including baseline covariates (age, sex, education), time (as continuous variable), group (SHAM+SHAM comparator) and Time x Group interaction.
Unadjusted locally-weighted mean trajectories were plotted over a sliding temporal window (i.e., lowess plot) for hippocampal volume as percentage of baseline and
functional connectivity z-scores for all four groups over the 18-month trial period. Temporal relationship determined the subsequent modelled time interaction
function in the LME.

3
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

oxygen consumption measurements (VO2peak) via treadmill stress test; to investigate the effect of cognitive and physical training (Suo et al.,
z-score whole-body strength via one-repetition maximum testing on 2016). However, we found no therapeutic benefit from the combined
Keiser machines; and self-reported, daily written physical activity (PA) training group (Fiatarone Singh 2014) compared to PRT alone, and
logs. PA logs were defined as activities over and above activities of daily postulated that the combined intervention could have caused possible
living (excluding intervention) (Mavros et al., 2017). Individual weekly mental and/or physical stress rather than the hypothesized augmented
averages were calculated. Baseline assessments was carried out over 2 benefits (Suo et al., 2016). As this secondary analysis aims to under-
weeks prior to intervention commencement. The 6- and 18-month PA stand the long-term impact of cognitive and physical training on neu-
scores were calculated as an average of the weekly average scores be- roplasticity, with prior knowledge of the ineffectiveness of the com-
tween baseline-6 month and 7–18-month assessments and then square bined training group, we investigated the 2 original RES (PRT+CCT
root transformed to achieve normal distribution. Baseline, 6-month and and PRT+SHAM vs CCT+SHAM and SHAM+SHAM) and COG
18-month scores for all three variables were entered into the statistical (PRT+CCT and CCT+SHAM vs PRT+SHAM and SHAM+SHAM)
linear mixed effects model for each individual so that change in beha- factors as well as the 3 individual intervention arms vs SHAM+SHAM
viour/fitness was modelled over the trial period. Full details of fitness control contrasts. Statistical contrasts are shown in Supplementary
testing are provided in Mavros et al. (2017) and described in Supple- Fig. 1 and false discovery rate (FDR) was used to correct for multiple
mentary materials and model details are given in Supplementary comparisons.
Fig. 2E. Subfield validation: Currently, the “gold standard” hippocampal
subfield segmentation approach requires a T2-weighted, high spatial
2.5. MRI acquisition resolution acquisition (Iglesias et al., 2015; Yushkevich et al., 2015).
We have previously cross-validated TI and T2 subfield segmentations in
MRI data were acquired on a 3-Tesla Philips Achieva Scanner an independent clinical MCI cohort (Broadhouse et al., 2019), defining
(Amsterdam, Netherlands) with 8-channel receive head coil, full details valid T1-weighted subfield segmentations as those with a Pearson's
are reported in Suo et al. (2016). Structural measures were derived correlation of >0.9 with T2-weighted segmentations. Presubiculum,
from 3 D-T1TFE sequences (1 mm isotropic resolution, TR/TE/ Subiculum, CA1, CA4 and dentate gyrus met our criteria and were
FA = 5.39 ms/2.43 ms/8°). Functional connectivity (FC) was analysed consequently included in this analysis.
from resting-state fMRI (rsfMRI), T2* echo-planar BOLD sequences Longitudinal modelling: Longitudinal changes in hippocampal volume
acquired with eyes closed (FOV = 250 × 250 mm, matrix (as a percentage of baseline volume), PC thickness and functional
size = 64 × 64, 29 slices, slice thickness = 4.5 mm, TR/TE = 2000/ connectivity between these two regions were investigated using a freely
30 ms, 200 volumes). available univariate linear mixed effects (LME) Matlab tool within the
FreeSufer framework (Bernal-Rusiel et al., 2013). Full details regarding
2.6. MRI post-processing model designs and analyses are given in the Supplementary materials
and Supplementary Fig. 2. Left and right hemispheres were analysed
3 D-T1TFE datasets were processed with the longitudinal analysis separately, and subsequent subfield analysis was carried out in the left
stream in FreeSurfer (v6.0) to extract reliable volume and thickness hippocampus only. Time was modelled as a continuous variable, with
estimates (Reuter et al., 2012; Iglesias et al., 2015). All datasets were Time x Group as the main contrast of interest. The marginal means
visually assessed both for data quality prior to FreeSurfer processing ( ± SEM) of idealised/collapsed time point plots (baseline, 6M, 18M)
and post-processing segmentation quality. Manual edits were then shown in subsequent figures were generated for visualisation purposes
carried out if necessary following the FreeSurfer guidelines. Hippo- and to provide a measure of effect size. Full details regarding effect size
campal subfield segmentation was then carried out on the longitudinal calculations are given in Supplementary materials. When modelling
processed output data using the longitudinal hippocampal subfield functional connectivity changes, lowess plots revealed that an ex-
segmentation tool in FreeSurfer (Iglesias et al., 2016). All subsequent ponential (not linear) time component was appropriate (Fig. 1C). Age,
reported whole-hippocampal measures are derived from the whole years of education and sex were used as covariates in all models. Fi-
hippocampal volumes generated by the final hippocampal-specific nally, the three previously defined PA and fitness variables (baseline, 6-
subfield segmentation pipeline and not the output from the preceding month and 18-month VO2peak, whole-body strength and PA values)
longitudinal whole-brain pipeline. rsfMRI datasets were again visually were added into the hippocampal model as covariates to investigate
inspected for data quality and then processed with the SPM8-based mediation effect.
Data Processing Assistant for Resting-State fMRI Advanced (DPARSFA) Correlation analysis: The relationship between left hippocampal
as part of the DPABI V2.1 toolbox ([Link]/) based on pub- subfield atrophy rate and cognitive outcome scores was investigated by
lished protocols (Chao-Gan and Yu-Feng, 2010). This involved: ex- carrying out a backward wise multiple linear regression analysis with
cluding the first ten volumes, motion correction, co-registration, nor- age at baseline scan, years of education, sex and significant subfields
malization to standard MNI space (DARTEL stream), resampling, volumes as independent variables. Partial correlation analysis was
smoothing and removing global signal trends, bypass filtering of carried out to investigate the relationship between change in left hip-
0.01–0.08 Hz and finally regressing out nuisance signals related to pocampal and PC functional connectivity and change in ADASCog be-
white matter, whole-brain and cerebrospinal fluid signal. Co-registra- tween baseline and 18-month follow up correcting for covariates. All
tion outputs between structural and functional scans were visually in- multiple linear regression and correlation analyses was carried out in
spected to ensure accurate registration and between-slice motion cor- SPSS (IBM, Armonk, NY, USA. Release 24).
rection. Poor quality, motion artefacted data was excluded from further
analysis. Individual 116 × 166 FC matrices were then generated using 3. Results
the Anatomical Automatic Labelling (AAL) template. The Fisher z-score
correlation between PC and hippocampal regions for each individual at Eighty-four (58 female) of the 100 participants had a baseline MRI
each time point were then extracted. All MRI pre-processing and ana- assessment (mean age = 69.5 (SD = 6.6)). Participant dropout or poor-
lysis was carried out by a single observer. quality 3 D-T1TFE datasets led to 79 (53 female) participants at 6-
month and 74 (50 female) participants at 18-month follow-up (Fig. 1).
2.7. Statistical analysis In addition, due to poor quality and motion artefact, seven baseline and
four 18-month rsfMRI datasets were excluded from functional con-
Contrast: The SMART trial was originally designed as a fully-fac- nectivity analysis. All subsequent analysis was carried out on this MRI
torial trial and previous statistical analyses utilized this flexible design subset (there was no significant difference in demographics (age or sex)

4
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Fig. 2. Resistance exercise preserves and protects Alzheimer-vulnerable hippocampal subfields for up to one year after training. (A) Left but not right PC cortical
thickness was enhanced by any resistance exercise training as opposed to double-sham; however, this advantage was lost one year after training cessation. (**) refers
to significant within-training effect of PRT+CCT vs SHAM+SHAM (F = 11.1, p = 0.005 FDR corrected, DF = 39.2) and PRT+SHAM vs SHAM+SHAM (F = 15.2,
p = 0.005 FDR corrected, DF = 36.4), using an identical LME model restricted to baseline and 6 months. No Time x Group effects were observed when testing the
complete 18-month model. (B) Left hippocampal volume trajectory is protected from atrophy by any training compared to double-sham (left; Time x Group F = 7.27,
p = 0.008, DF = 157.9), the effect specific to groups that included resistance training (PRT+CCT: middle, F = 5.10, p = 0.041 FDR corrected, DF = 84.0; and
PRT+SHAM: F = 7.22, p = 0.027 FDR corrected, DF = 79.6). (C) Resistance exercise alone protects Alzheimer-vulnerable hippocampal subfields, including: left
CA1 (F = 8.81, p = 0.020 FDR corrected, DF = 81.0); Subiculum (F = 7.14, p = 0.023 FDR corrected, DF = 80.2); and dentate gyrus (F = 5.53, p = 0.035 FDR
corrected, DF = 79.1). For each of these subfields, 6 months of resistance exercise protected participants from 2% to 3% volumetric loss over the complete 18-month
follow-up period. Means ( ± SEM) adjusted for baseline age, years of education, sex, and baseline volume normalized by ICV. (*) refers to significant Time x Group
interaction at p < 0.05. NB: Idealised time points (baseline, 6M, 18M) are displayed for visualisation purposes only; statistics are solely from LME models where real-
time is treated as a continuous variable.

5
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

between this subset and the whole cohort). Fisher exact tests identified left hippocampus over the 18-month period compared to SHAM
no significant differences in amnestic vs. non-amnestic MCI subtype +SHAM (Fig. 2B; FDR corrected p = 0.042 and 0.027). PRT+CCT did
diagnosis at baseline (aMCI = 11% of total cohort, p = 0.27) or in not show this effect; concordant with our previous findings on cognition
number of APOE-e4 carriers (APOE-e4 carriers = 29% of total cohort, and fitness adaptations (Fiatarone Singh et al., 2014; Mavros et al.,
p = 0.63) between groups. 2017), the neuroplastic benefits of PRT were attenuated when com-
bined with CTT. Full results for whole hippocampal LME analysis are
3.1. Resistance exercise leads to long-term cognitive benefits given in Table 1A. Focusing on the left hippocampus, subfield LME
analysis revealed an elimination of long-term atrophy in the left sub-
Previously, we found that resistance exercise led to better global iculum, and attenuation of atrophy in left CA1 and dentate gyrus for the
cognitive and executive function at 18 months compared to no re- PRT+SHAM group compared to SHAM+SHAM (Fig 2C; FDR corrected
sistance training in factorial analysis (RES contrast) and individual p = 0.023, Cohen's d = 0.99; p = 0.020 Cohen's d = 1.02; and
group investigation revealed that only PRT+SHAM was significantly p = 0.027, Cohen's d = 0.82, respectively). Full results for subfield
different from SHAM+SHAM (Fiatarone Singh et al., 2014). For con- hippocampal LME analyses are given in Tables 1B and C.
tinuity we investigated these effects at the individual group level within
the MRI subset and again found that PRT+SHAM selectively benefited
global ADASCog performance, as well as executive function, compared 3.4. Within-training preservation of PC volume predicts long-term
to SHAM+SHAM group (Time x Group p = 0.028; Cohen's d = −0.36; hippocampal preservation
p = 0.046, Cohen's d = 0.32) (Fig. 1B).
Next, the relationship between within-training PC and long-term
hippocampal structural plasticity was investigated. Whole group cor-
3.2. Within-training exercise-induced PC plasticity is not sustained long-
relation analysis established a weak yet significant negative relation-
term
ship between the within-training structural PC plasticity and the long-
term hippocampal subfield neuroprotection described above.
To investigate whether the within-training resistance exercise-in-
Specifically, individuals who underwent the most within-training in-
duced PC structural plasticity previously reported in Suo et al. (2016)
crement in PC cortical thickness had less subiculum and dentate gyrus
was sustained long-term we first replicated the within-training results
atrophy one year post-training (Pearson's correlation = −0.24,
using our updated LME within the FreeSurfer framework. As antici-
p = 0.04 and Pearson's correlation = −0.26, p = 0.03, respectively).
pated, the FreeSurfer-based LME mirrored the previous findings
(Suo et al., 2016) and indicated that resistance exercise increases cor-
tical thickness during training, although the FreeSurfer hemispheric
3.5. PRT increases functional connectivity between PC and hippocampus
differentiation of cortical regions revealed that this was localised to the
left PC. However, building on this, a second analysis modelling long-
To further investigate this link between the spatiotemporally dis-
term thickness trajectory over the 18-month period revealed that this
tinct PC and hippocampal plasticity we examined the change in func-
PC structural benefit was not evident 12 months after training cessation
tional connectivity between left PC and hippocampus. A significant RES
(Fig. 2A). Full results are given in Supplementary Table 1. Resistance
x Time interaction over the 18-month period was found, such that long-
exercise therefore seems to stimulate an important cortico-structural
term functional connectivity between PC and hippocampus was sig-
mechanism during training, which is not sustained long-term.
nificantly strengthened in PRT+CCT and PRT+SHAM compared to
CCT+SHAM and SHAM+SHAM (p = 0.018) (Fig. 3). However, sig-
3.3. Resistance exercise slows post-training CA1, subiculum and dentate nificant p-values did not survive FRD correction for the individual
atrophy group analysis. Full results for the functional connectivity LME analysis
are given in Table 2. Partial correlation analysis across the whole group
Next, LME analysis was used to investigate the long-term impact of revealed a non-significant relationship between change in left hippo-
the interventions, over the 18-month trial period, on whole hippo- campal and PC functional connectivity and change in ADASCog be-
campal volume. We found significant effects for left, but not right tween baseline and 18-month follow (p = 0.097).
hippocampal atrophy rates. Specifically, both the PRT+CCT and
PRT+SHAM groups exhibited significantly slower atrophy rates in the

Table. 1
(A)Whole hippocampal and (B) Subfield LME analysis. Significant p-values are shown in bold, individual intervention analysis shows FDR corrected p-values.
A Left Right

Long-term effect Fvalue p value DF F value p value DF

RES 3.60 0.060 159.2 0.19 0.663 159.9
COG 1.53 0.219 158.8 0.71 0.401 159.5
PRT+CCT vs SHAM+SHAM 5.10 0.041 83.8 0.77 0.650 84.0
PRT+SHAM vs SHAM+SHAM 7.22 0.027 79.6 0.62 0.650 79.9
CCT+SHAM vs SHAM+SHAM 2.70 0.104 84.7 0.18 0.650 85.0
B PRT+CCT vs SHAM+SHAM PRT+SHAM vs SHAM+SHAM
F value p value DF F value p value DF
Subiculum 0.74 0.392 83.6 7.14 0.023 80.2
CA1 2.85 0.119 84.6 8.81 0.020 81.0
Dentate gyrus 6.05 0.075 83.5 5.53 0.035 79.1
CA4 4.67 0.083 83.7 3.39 0.086 79.6
Presubiculum 2.87 0.119 83.7 0.59 0.442 80.9

RES = PRT+CCT and PRT+SHAM vs CCT+SHAM and SHAM+SHAM.

COG = PRT+CCT and CCT+SHAM vs PRT+SHAM and SHAM+SHAM.
PRT = progressive resistance training.
CCT = computerised cognitive training.

6
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Fig. 3. Resistance exercise leads to long term increase in functional connectivity between left hippocampus and posterior cingulate. (A) An individual functional
connectome from one SMART participant at baseline showing the correlation between all 116 AAL-atlas brain regions (left) in the form of a correlation heat map
(right). Individual connectomes were generated at each time point, and correlation values between the left hippocampus and PC regions extracted to investigate the
change in functional connectivity using our LME model as outlined in Fig 1E. (B) Raw change in the functional connectivity between left hippocampus and PC over
the complete 18-month period is depicted at the individual level. Box-and-Whisker plots show the median (horizontal line), interquartile range (box) and upper and
lower quartiles (whiskers) of change in functional connectivity between left PCC and hippocampus for each training group. LME analysis found significantly
increased connectivity for those two groups that underwent resistance exercise training (PRT+CCT and PRT+SHAM) compared to those that did not (CCT+SHAM
and SHAM+SHAM). *Time x Group p < 0.05; model included covariates age, sex and education, as well as group factor and time as continuous variable.

Table. 2 exercise protects and preserves subiculum volume (Fig 2), is ther-
LME analyses of PC-hippocampal functional connectivity z-scores. Significant p- apeutically effective for cognition in the long term, and the former
values are shown in bold, individual intervention analysis shows FDR corrected predicts the latter, it follows that the addition of subiculum atrophy
p-values. abolished the therapeutic effects of resistance exercise on final
Within-training Long-term ADASCog, indicating that resistance exercise-dependent plasticity of
the subiculum contributes to mediation of long-term therapeutic ben-
F value p value DF F value p value DF efits on cognition.
RES 0.07 0.789 77.3 5.72 0.018 148.0
COG 0.38 0.538 76.5 0.02 0.887 147.1
PRT+CCT vs 0.02 0.903 39.8 1.82 0.180 77.9 3.7. Long-term neuroprotection is not mediated by post-training physical
SHAM+SHAM activity or fitness adaptations
PRT+SHAM vs 0.00 0.985 38.2 1.68 0.199 70.4
SHAM+SHAM
CCT+SHAM vs 0.81 0.374 39.6 0.22 0.637 77.8
Finally, the challenging question as to whether the long-term brain
SHAM+SHAM benefits of resistance exercise are a direct and extended consequence of
a time-limited training dose, or rather, mediated by habitual PA en-
RES = PRT+CCT and PRT+SHAM vs CCT+SHAM and SHAM+SHAM. gagement outside of the intervention or fitness adaptations was in-
COG = PRT+CCT and CCT+SHAM vs PRT+SHAM and SHAM+SHAM. vestigated. There are many indirect pathways by which an exercise
PRT = progressive resistance training. intervention could continue to impact the brain and cognition.
CCT = computerised cognitive training.
However, in this context, changes in either physical fitness or post-
training “exercise behaviour”, are most relevant, and are con-
3.6. Neuroprotection of AD-vulnerable areas mediate long-term cognition
ceptualized in Fig. 5A. LME analysis revealed a significant increase in
whole body strength in the PRT+CCT group compared to SHAM
Next, backward multiple linear regression was used to investigate
+SHAM over the 18-month period (Fig. 5B). This is potentially im-
the association between left hippocampal subfield atrophy trajectory
portant therapeutically, since we have previously shown that increases
over the 18-month period and previously observed, long-term cognitive
in muscle strength during training mediate clinical improvements in
gains (Fiatarone Singh et al., 2014) (Fig. 1B). For ADASCog, sex
general cognition (Mavros et al., 2017). However, when entering the
(β = −0.32, p = 0.002), subiculum atrophy rate (β = 0.28,
three PA and fitness variables (VO2peak, whole-body strength and PA)
p = 0.011), education (β = −0.24, p = 0.022), and baseline age
as covariates into our subfield LME model, CA1, subiculum and dentate
(β = 0.22, p = 0.04) were significant independent predictors (overall
gyrus atrophy trajectories were still significantly attenuated in
model fit adjusted-R2 = 0.28; Fig. 4A), where higher ADASCog scores
PRT+SHAM compared to SHAM+SHAM (Fig. 5C). Suggesting that
indicate poorer global cognitive function. Interestingly, for the trial's
long-term protective effects of PRT on atrophic subfield trajectories are
primary outcome (ADASCog), rate of subicular atrophy accounted for
not mediated by exercise behavioural change outside the intervention.
more unique variance than age or education (subiculum: adjusted
Rather, our data indicate that this discrete 6-month PRT intervention
R2 = 9%; age: R2 = 6% and education: R2 = 7%). For executive
extends a direct and lasting neuroprotective effect on cognitively-re-
function, education (β = 0.53, p < 0.0001), baseline age (β = −0.30,
levant and AD-vulnerable hippocampal subfields. Full LME results with
p = 0.001) and CA1 atrophy (β = −0.25, p = 0.007) were significant
the addition of the three exercise and fitness variables are given in
and independent predictors (overall model fit aR2 = 0.45, Fig. 4A). To
Supplementary Table 2.
further investigate this association between subiculum atrophy on our
primary outcome measure ADASCog, we re-ran the cognitive LME
4. Discussion
model in Fig 1B with 18-month change in subiculum volume as an
additional covariate and found no significant PRT+SHAM x Time in-
Cognitive benefit directly following lifestyle intervention is now
teraction when compared to SHAM+SHAM (Fig 4B). Since resistance
well established in older persons (Mavros et al., 2017; Ngandu et al.,

7
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Fig. 4. Therapeutic benefits of resistance exercise on cognition one year after training are mediated by preservation of AD-vulnerable hippocampal subfields. (A)
Multiple regression analysis tested the unique variance accounted in the primary outcome (ADASCog) and Executive function by 18-month subiculum or CA1 atrophy
(as percent change from baseline; age, sex and education were also in the model). Partial plots here show that subiculum atrophy was an independent predictor of
ADAGCog13, whilst CA1 atrophy was an independent predictor of Executive function. The beta values for each of these (β = 0.28 and −0.25, respectively) were of
similar magnitude to clinically meaningful predictors age (β = 0.22 and −0.30) and education (β = −0.24 and 0.53). (B) Since resistance exercise protects and
preserves subiculum and CA1 volume (Fig 2C), is therapeutically effective on cognition in the long term (A), and the former predict the latter (B), we re-ran the
cognitive LME model in (A) with 18-month change in subiculum volume (Δsubic; expressed as percentage change from baseline) as an additional covariate. This
abolished the therapeutic effect of resistance exercise on ADASCog. Accordingly, resistance exercise-dependent plasticity of the subiculum mediates long-term
therapeutic benefits on global cognition.

2015; Rebok et al., 2014; Richard et al., 2009), however the underlying areas highly sensitive to AD degeneration, only becomes evident well
mechanisms are poorly understood and how these evolve long-term, after training has finished. Interestingly, whilst such structural benefits
essentially unknown. For the first time with any intervention, we show can be distinguished in time, we reveal a possible underlying ex-
that resistance exercise can not only lead to cognitive benefits but also planation in the form of upregulated hippocampal-PC functional con-
protects AD-vulnerable hippocampal subfields long-term. Moreover, nectivity, a change unique to resistance exercise (RES contrast) that
this form of neuroprotection helps to explain such lasting cognitive persisted across the entire 18-month period.
benefits, spatiotemporally distinct from PC plasticity that mediates The PC has strong reciprocal connections with several cortical and
cognitive improvement only during exercise. These two temporally subcortical regions, most pertinently with the hippocampus (Leech and
distinct targets of structural protection are linked by a long-term up- Sharp, 2014). In MCI and AD, functional and diffusion tensor imaging
regulation of functional inter-connectivity following resistance ex- studies show a decrease in connectivity between the PC and hippo-
ercise. Finally, we demonstrate that enduring hippocampal subfield campus (Zhou et al., 2008), part of a well characterized reduction in
plasticity is not a simple reflection of post-training changes in fitness or functional connectivity in the encompassing default mode network
physical activity participation. Rather, resistance exercise can effec- (DMN) (Greicius et al., 2004; Zhou et al., 2008). Previous studies sug-
tively support ongoing general cognition in at-risk elders by directly gest that disrupted connectivity between these two regions may account
protecting hippocampal subfields most at risk for degeneration in AD. for PC hypometabolism commonly detected in early AD (Greicius et al.,
It is widely assumed that the same mechanisms elicited during ex- 2004). The DMN may also be plastic to exercise, as witnessed in a re-
ercise will either persist or diminish after training cessation, which in cent trial in healthy older adults that found that stretching led to an
turn determines long-term therapeutic efficacy. Our data contradict this increase DMN functional connectivity after 6 months, while aerobic
and suggest that distinct mechanisms emerge during- and post-training, showed similar results at 12 months (Voss et al., 2010). It is thereby
in effect a cascade of therapeutic mechanisms that combine to support possible that resistance exercise triggers increased functional con-
cognition (Fig. 6). Our past work indicates that within-training cogni- nectivity between these two key hubs of the DMN, regions that then
tive gains are coupled with short-to-medium term structural plasticity undergo structural plasticity along different time courses. More gen-
in the PC (Suo et al., 2016). Here we show this decays after the offset of erally, recent studies have indicated that resting state functional con-
training. However, plasticity in left CA1, subiculum and dentate gyrus, nectivity maybe be a promising biomarker for AD as functional brain

8
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Fig. 5. Physical fitness and physical activity behaviour after training does not mediate long-term protection of hippocampal subfields. (A) A conceptual schema
showing how the long-term effects of resistance exercise on the brain could arise directly following training in the context of reverting to business-as-usual (BAU) or
could be mediated by changes in physical fitness or physical activity behaviour unrelated to the interventions. We modelled exercise-related behaviour and physical
fitness in three different ways, using long term changes in physiologic VO2peak and whole-body muscle strength measurements, as well as daily physical activity logs
over 18 months. (B) Change in fitness and exercise behaviour is charted over time (mean ( ± SEM)) using a LME model, revealing a significant Time x Group (*)
effect of PRT+CCT on whole body strength during training and over the complete 18-month period compared to SHAM+SHAM (F = 15.22, p = 0.001 FDR
corrected, DF = 38.52 and F = 5.47, p = 0.018 FDR corrected, DF = 76.03, respectively). We therefore confirm that 6 months of resistance exercise training
continues to modify whole body muscle strength for up to a year after the cessation of formal supervised study-related resistance training. (C) However, after
inclusion of these fitness measures and non-study physical activities in our LME model of hippocampal subfield change, the neuroprotective effect of resistance
exercise was not diminished. This is further visualised by partial plots of subiculum atrophy against the trial's primary cognitive outcome (ADASCog error scores),
showing no moderation of relationship before (black) or after (red) inclusion of whole-body muscle strength, aerobic capacity (VO2 peak) and total number of
physical activity sessions outside of the intervention in the model. Long-term protective effects of resistance exercise on AD-vulnerable hippocampal subfields are
hence direct, delayed and protracted, independent of PA behaviour after formal training ends.

changes are thought to precede structural changes. Decreased func- and within the hippocampus. Longitudinal human studies consistently
tional connectivity has been observed in healthy aging, mild cognitive show asymmetrical volume loss in MCI and AD, with higher atrophy
impairment, a prodromal stage of AD, and AD. However, as the exact rates in the left hippocampus (Shi et al., 2009). More specifically, this
role of aberrant functional connectivity within the cascade of patho- asymmetry is reserved to the anterior part of the hippocampus and
physiological processes of AD remains poorly understood, it is unclear asymmetry increases with disease progression (Maruszak and
how the observed upregulated hippocampal-PC functional connectivity Thuret, 2014). Since CA1 and subiculum subfields lie within the ante-
will impact or delay the onset of dementia. Further research is therefore rior hippocampus, our findings of left-lateralised effects after resistance
required to determine the temporal sequence and inter-dependency of exercise gain added relevance. Furthermore, atrophy progresses non-
functional and structural brain changes triggered by PRT. linearly, with periods of acceleration predicting transition from MCI to
dementia (Apostolova et al., 2010; Maruszak and Thuret, 2014;
Rossler et al., 2002). An intervention that effectively protects against
4.1. Neuroprotection of AD-vulnerable hippocampal subfields
volume loss in AD-vulnerable hippocampal subfields could potentially
delay transition to dementia.
We found strong evidence for left asymmetric plasticity in the cortex

9
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Fig. 6. Model in which different mechanisms are responsible for therapeutic cognitive effects during and after training. Global cognition presumably
declines before observations begin at baseline because of the MCI status of participants (dashed line). A discrete dose of resistance exercise (bottom trace) produces a
significant improvement to cognition, coupled with contemporaneous structural plasticity in the posterior cingulate (PC) that is time-limited and wanes after training
offset. Resistance exercise also leads to a time-delayed structural response within select hippocampal subvolumes, evident up to one year after training offset. This
delayed hippocampal subfield plasticity helps explains ongoing maintenance of cognition. In this conceptual schema several parameters are unknown, including: the
shape, symmetry, peak and width of α and β, how they interact, or how long therapeutic cognitive effects last.

Accordingly, the positive outcome seen here using resistance ex- agents or exercise for diabetes or cardiovascular disease.
ercise suggests this non-pharmacological intervention may not only be
clinically effective, but capable of targeting a key aspect of AD patho- 4.2. CA1 and subiculum function
genesis. Our results are also of interest because the cohort was popu-
lation-based and trial entry non-specific for subtype of MCI. In fact, our CA1 and subiculum volumetric plasticity is also interesting given
sample was generally non-amnestic MCI (89%) and only about one- their putative role in higher-order cognitive function. Classically, they
third held the AD-risk gene APOE4. Yet hippocampal AD pathology and participate in the trisynaptic circuit, a mnemonic loop where afferents
neurodegeneration are as common in amnestic MCI as in non-amnestic from CA3 synapse onto CA1 and information is passed onto the sub-
MCI (Dugger et al., 2015), primarily because pathology in MCI and AD iculum before delivery to cortical and contralateral targets. In rodents,
is highly heterogenous (Schneider et al., 2009). Furthermore, functional CA1 is strongly implicated in episodic memory (Kyle et al., 2015;
imaging studies have revealed that the hippocampus is implicated in far Reitz et al., 2009) and spatial navigation (O'Keefe, 1991). Similarly, the
more than just memory consolidation and retrieval, forming an import subiculum has been heavily implicated in mnemonic consolidation
hub in several networks involved in higher order executive functioning, processes (O'Mara et al., 2000). It may therefore appear counter-
impulse control, cognitive flexibility and decision making (Mars et al., intuitive that selective plasticity in these areas by PRT seemed to pri-
2012; Vincent et al., 2008). It has been suggested that the functional marily benefit executive function. However, this is less surprising
disruptions within these neurocircuitries lead to maladaptive responses considering the presence of dense monosynaptic CA1 to medial-orbital
during cognitive and emotional processing, which may underpin cog- prefrontal connections in primate studies (Ongur and Price, 2000),
nitive decline in this cohort (Liu et al., 2014). For these reasons, AD- suggesting a more complex role for the hippocampus (O'Mara et al.,
related mechanisms are as relevant to non-amnestic MCI as amnestic 2000), not just in memory but also in executive processes and affect
MCI, and for community-orientated preventative intervention such as regulation (Godsil et al., 2013; Ongur and Price, 2000). Resistance
resistance exercise there is no compelling justification to make the exercise may potentially alter the connections between CA1 and pre-
distinction. frontal structures, providing for more nuanced and effective integration
At the same time, caution against over interpretation is required. of mnemonic and affective information. Because a precise role for these
Annualized rates of hippocampal atrophy in those with stable versus subfields in humans is still emerging, resistance exercise may prove
declinate MCI are 2.6% and 3.7% (Jack et al., 2000) respectively – a useful in disambiguating hippocampal subfield function.
~1% difference per year. Mapping cognitive decline showed that on The primary outstanding question remains why and how any ex-
average non-converter and converter MCI subjects declined by 0.75 and ercise regime should benefit cognition. The relative impact of periph-
2.93 ADASCog points over a 12-month period (Evans et al., 2010). We eral- versus centrally-induced molecular changes remains poorly un-
observed that 6 months of resistance exercise produced a 12-month- derstood, with some evidence that exercise-induced myokines may
advantage to general cognition of 1–2 ADASCog points indicating that cross the blood-brain barrier and possibly mediate some forms of brain
the training-dependent effects observed in SMART are comparable to plasticity (Bostrom et al., 2012; Kobilo et al., 2011; Wrann et al., 2013).
~0.7–2years of naturalistic decline but in the opposite direction. Fur- Whilst some mechanisms may be shared with aerobic exercise, there is
thermore, we see a preservation of CA1 and subiculum volume by ~3%. abundant evidence that anabolic training produces its own metabolic
If such linear trends are extrapolated, our intervention could preserve signature. Resistance exercise, for example, is particularly effective at
key hippocampal substructures by ~3 years compared to usual care. counteracting insulin resistance, sarcopenia and stimulating osteogenic
This would be a significant improvement to the individual, but MCI is a mechanisms (Cooney et al., 2014; Irvine and Taylor, 2009;
long-term condition of indeterminate prognosis, and so a circumscribed Peterson and Gordon, 2011). Notably, in the EXCEL RCT, the only head-
dose of resistance exercise would be unlikely to provide indefinite to-head comparison of resistance and aerobic training in older women
protection against cognitive decline or neurodegeneration, as would be with probable MCI published to date, 6 months of resistance training
the case if it were used for sarcopenia or osteoporosis, for example. This significantly improved selective attention/conflict resolution, and as-
therefore poses an implementation challenge as prescription of PRT sociative memory compared to the sham-exercise control, along with
may need to be lifelong, similar to the prescription of pharmacologic functional changes in three regions of cortex the right lingual and

10
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

occipital-fusiform gyri, and the right frontal pole during the encoding delayed and protracted plasticity observed after training in AD-vul-
and recall of associations (ten Brinke et al., 2015). Additionally, there nerable subfields. The two phenomena are correlated, and these regions
was a significant positive correlation between change in hemodynamic participate within a functional network stimulated by resistance ex-
activity in the right lingual gyrus and change in behavioural associative ercise, but their causal relationship is unclear. Further work is needed
memory performance in the resistance training group. None of these to investigate the long-term structural and functional, resistance ex-
changes were seen in the aerobic exercise group (Nagamatsu et al., ercise-induced plasticity in MCI and whether these plastic effects are
2012). However, compared with the control group, aerobic training diverse between subtypes.
significantly increased left, right, and total hippocampal volumes, but
the increased left hippocampal volume was independently associated
4.4. Conclusion
with reduced verbal memory and learning performance (ten Brinke
et al., 2015). Consequently, whether resistance exercise has unique
Given the great challenge of dementia to modern society it is pro-
plastic effects on the older brain is still unclear and warrants further
mising to show for the first time that 6 months of high intensity re-
investigation.
sistance exercise is capable of not only promoting cognition in those
with MCI, but also protecting AD-vulnerable hippocampal subfields
4.3. Limitations
from degeneration for at least 12 months post-intervention. Future
work will need to establish just how long-lived these outcomes are and
Although the long-term, resistance exercise induced, structural and
whether they are sufficient enough to delay cognitive decline. Despite
functional plasticity reported here is encouraging in the MCI popula-
this, given the strength of our findings we recommend that resistance
tion, caution is needed when interpreting such findings. Firstly, our
exercise be considered an integral part of lifestyle-based prevention
results derive from a single community-based cohort in Sydney,
programs in older persons.
Australia, and our main analyses centred around the SHAM+SHAM
and PRT+SHAM groups, a total sample of 43 participants for MRI and
49 for cognition. For replication purposes, assuming 80% power, and Funding
the observed Cohen's d effect sizes for cognition outcomes (ADASCog
and executive domain scores), group sample sizes of 150 and 190 are This study was funded by the National Health and Medical Research
recommended, respectively. By contrast, subfield analysis revealed Council (NHMRC) of Australia Dementia Research Grant, Project Grant
large effect sizes translating to recommended sample sizes of 21–31 ID No. 512672 from 2008 to 2011. Additional funding for a research
participants in each group depending on the subfield of interest. We assistant position was sourced from the NHMRC Program Grant ID No.
also recognise that power calculations are complex for longitudinal 568969.
designs due to missing data and autocorrelation between measures over
time. Consequently, we calculated the realized power (Bernal-
CRediT authorship contribution statement
Rusiel et al., 2013a, 2013b) associated with the CA1, subiculum and
dentate gyrus linear mixed-effects model which provided an estimate of
Kathryn M. Broadhouse: Formal analysis, Investigation,
power of 0.83, 0.78 and 0.63 respectively (see Supplementary materials
Methodology, Writing - original draft, Writing - review & editing. Maria
for details). For some outcomes our study was therefore underpowered.
Fiatarone Singh: Supervision, Project administration,
Furthermore, the relatively small numbers and trial design did not
Conceptualization, Funding acquisition, Data curation, Methodology,
allow dichotomization of the groups into the MCI subtypes, amnestic
Writing - review & editing. Chao Suo: Data curation, Methodology,
and non-amnestic MCI. It is thought that amnestic MCI increases the
Writing - review & editing. Nicola Gates: Formal analysis, Data cura-
risk of conversion to AD further however, the wider generalizability of
tion, Methodology, Writing - review & editing. Wei Wen: Formal
our findings will need to be tested in large and more diverse MCI po-
analysis, Methodology, Writing - review & editing. Henry Brodaty:
pulations along with determining links to other AD biomarkers.
Conceptualization, Funding acquisition, Methodology, Writing - review
Secondly, only 3 D-T1TFE datasets were acquired for segmentation
& editing. Nidhi Jain: Formal analysis, Data curation, Methodology,
analysis, although datasets were visually inspected for data quality, the
Writing - review & editing. Guy C. Wilson: Data curation,
lower resolution and lower within structure contrast may lead to in-
Methodology, Writing - review & editing. Jacinda Meiklejohn: Data
accuracies in subfield segmentations. However, it is important to note
curation, Methodology, Writing - review & editing. Nalin Singh: Data
that although a dedicated high resolution T2-weighted sequence is
curation, Methodology, Writing - review & editing. Bernhard T.
considered by many as the “gold standard” for subfield segmentation,
Baune: Data curation, Methodology, Writing - review & editing.
as with all MR sequences there are tradeoffs. The improved grey/white
Michael Baker: Data curation, Methodology, Writing - review &
matter contrast is achieved with a spin echo sequence – a gradient
editing. Nasim Foroughi: Data curation, Methodology, Writing - re-
readout that is inherently longer and resultant contrast that provides
view & editing. Yi Wang: Data curation, Methodology, Writing - review
less signal or “entropy” than the T1-weighted counterpart.
& editing. Nicole Kochan: Data curation, Methodology, Writing - re-
Furthermore, to achieve reasonable scan times the T2 sequence is not
view & editing. Kevin Ashton: Data curation, Methodology, Writing -
3 D and as a result has anisotropic resolution (typically
review & editing. Matt Brown: Data curation, Methodology, Writing -
0.4 × 0.4 × 2 mm). These factors need to be considered when carrying
review & editing. Zhixiu Li: Data curation, Methodology, Writing -
out accurate subfield segmentation. In a cohort where hippocampal
review & editing. Yorgi Mavros: Funding acquisition, Formal analysis,
atrophy occurs along the anterior-posterior axis and speed to reduce
Data curation, Methodology, Writing - review & editing. Perminder S.
motion artefacts is priority, the faster, 3 D T1-weighted, isotropic re-
Sachdev: Supervision, Conceptualization, Funding acquisition, Data
solution sequence may often be preferable and as robust.
curation, Methodology, Writing - review & editing. Michael
Finally, we found strong evidence for long-term left asymmetric
[Link]: Supervision, Conceptualization, Funding acquisition,
plasticity in the hippocampus. Longitudinal human studies consistently
Data curation, Methodology, Writing - review & editing.
show asymmetrical volume loss in MCI and AD, with higher atrophy
rates in the left hippocampus (Shi et al., 2009). Therefore, our findings
of left-lateralised plastic effects after resistance exercise gain clinical Data for reference
relevance but are essentially unexplained. Furthermore, we do not
understand the dependency, if any, between the previously reported The data that support the findings of this study are available from
within-training PC cortical mechanisms (Suo et al., 2016) and the the corresponding author, upon reasonable request.

11
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

Declaration of Competing Interest Gates, N., Fiatarone Singh, M.A., Sachdev, P.S., Valenzuela, M., 2013. The effect of ex-
ercise training on cognitive function in older adults with mild cognitive impairment:
a meta-analysis of randomized controlled trials. Am. J. Geriatr. Psychiatry 21 (11),
The authors have no conflicts of interest to declare. 1086–1097. [Link]
Gates, N.J., Valenzuela, M., Sachdev, P.S., Singh, N.A., Baune, B.T., Brodaty, H., ...
Acknowledgements Fiatarone Singh, M.A., 2011. Study of Mental Activity and Regular Training (SMART)
in at risk individuals: a randomised double blind, sham controlled, longitudinal trial.
BMC Geriatr. 11 (19). [Link]
We would like to acknowledge the generous support and contribu- Godsil, B.P., Kiss, J.P., Spedding, M., Jay, T.M., 2013. The hippocampal-prefrontal
tion of all our SMART Trial participants. MV was supported by a pathway: the weak link in psychiatric disorders. Eur. Neuropsychopharmacol. 23
(10), 1165–1181. [Link]
National Health and Medical Research Council of Australia career de- Greicius, M.D., Srivastava, G., Reiss, A.L., Menon, V., 2004. Default-mode network ac-
velopment fellowship APP1112813 and The University of Sydney. tivity distinguishes Alzheimer's disease from healthy aging: evidence from functional
MRI. Proc. Natl. Acad. Sci. USA 101 (13), 4637–4642. [Link]
0308627101.
Supplementary materials
Hill, N.T., Mowszowski, L., Naismith, S.L., Chadwick, V.L., Valenzuela, M., Lampit, A.,
2017. Computerized cognitive training in older adults with mild cognitive impair-
Supplementary material associated with this article can be found, in ment or dementia: a systematic review and meta-analysis. Am. J. Psychiatry 174 (4),
the online version, at doi:10.1016/[Link].2020.102182. 329–340. [Link]
Huang, C., Wahlund, L.O., Svensson, L., Winblad, B., Julin, P., 2002. Cingulate cortex
hypoperfusion predicts Alzheimer's disease in mild cognitive impairment. BMC
References Neurol. 2, 9.
Iglesias, J.E., Augustinack, J.C., Nguyen, K., Player, C.M., Player, A., Wright, M., ... 2015.
A computational atlas of the hippocampal formation using ex vivo, ultra-high re-
Alzheimer's Disease International, 2016. World Alzheimer Report 2016. Retrieved from solution MRI: application to adaptive segmentation of in vivo MRI. Neuroimage 115,
London. 117–137. [Link]
Apostolova, L.G., Dutton, R.A., Dinov, I.D., Hayashi, K.M., Toga, A.W., Cummings, J.L., Iglesias, J.E., Van Leemput, K., Augustinack, J., Insausti, R., Fischl, B., Reuter, M., 2016.
Thompson, P.M., 2006. Conversion of mild cognitive impairment to Alzheimer dis- Bayesian longitudinal segmentation of hippocampal substructures in brain MRI using
ease predicted by hippocampal atrophy maps. Arch. Neurol. 63 (5), 693–699. subject-specific atlases. Neuroimage 141, 542–555. [Link]
[Link] neuroimage.2016.07.020.
Apostolova, L.G., Mosconi, L., Thompson, P.M., Green, A.E., Hwang, K.S., Ramirez, A., ... Irvine, C., Taylor, N.F., 2009. Progressive resistance exercise improves glycaemic control
de Leon, M.J., 2010. Subregional hippocampal atrophy predicts alzheimer's dementia in people with type 2 diabetes mellitus: a systematic review. Aust. J. Physiother. 55
in the cognitively normal. Neurobiol. Aging 31 (7), 1077–1088. [Link] (4), 237–246.
1016/[Link].2008.08.008. Jack Jr., C.R., Petersen, R.C., Xu, Y., O'Brien, P.C., Smith, G.E., Ivnik, R.J., ... Kokmen, E.,
Baker, L.D., Frank, L.L., Foster-Schubert, K., Green, P.S., Wilkinson, C.W., McTiernan, A., 2000. Rates of hippocampal atrophy correlate with change in clinical status in aging
... Craft, S., 2010. Effects of aerobic exercise on mild cognitive impairment: a con- and AD. Neurology 55 (4), 484–489.
trolled trial. Arch. Neurol. 67 (1), 71–79. [Link] Kobilo, T., Yuan, C., van Praag, H., 2011. Endurance factors improve hippocampal neu-
307. rogenesis and spatial memory in mice. Learn. Mem. 18 (2), 103–107. [Link]
Bernal-Rusiel, J.L., Greve, D.N., Reuter, M., Fischl, B., Sabuncu, M.R., 2013a. Statistical 10.1101/lm.2001611.
analysis of longitudinal neuroimage data with Linear Mixed Effects models. Kyle, C.T., Stokes, J.D., Lieberman, J.S., Hassan, A.S., Ekstrom, A.D., 2015. Successful
Neuroimage 66, 249–260. [Link] retrieval of competing spatial environments in humans involves hippocampal pattern
Bernal-Rusiel, J.L., Reuter, M., Greve, D.N., Fischl, B., Sabuncu, M.R., 2013b. separation mechanisms. Elife 4. [Link]
Spatiotemporal linear mixed effects modeling for the mass-univariate analysis of Leech, R., Sharp, D.J., 2014. The role of the posterior cingulate cortex in cognition and
longitudinal neuroimage data. Neuroimage 81, 358–370. [Link] disease. Brain 137 (Pt 1), 12–32. [Link]
neuroimage.2013.05.049. Liu, Y., Yu, C., Zhang, X., Liu, J., Duan, Y., Alexander-Bloch, A.F., ... Bullmore, E., 2014.
Bostrom, P., Wu, J., Jedrychowski, M.P., Korde, A., Ye, L., Lo, J.C., ... Spiegelman, B.M., Impaired long distance functional connectivity and weighted network architecture in
2012. A PGC1-alpha-dependent myokine that drives brown-fat-like development of Alzheimer's disease. Cereb. Cortex 24 (6), 1422–1435. [Link]
white fat and thermogenesis. Nature 481 (7382), 463–468. [Link] cercor/bhs410.
nature10777. Liu-Ambrose, T., Donaldson, M.G., 2009. Exercise and cognition in older adults: is there a
Broadhouse, K.M., Mowszowski, L., Duffy, S., Leung, I., Cross, N., Valenzuela, M.J., role for resistance training programmes. Br. J. Sports Med. 43 (1), 25–27. [Link]
Naismith, S.L., 2019. Memory performance correlates of hippocampal subfield vo- org/10.1136/bjsm.2008.055616.
lume in mild cognitive impairment subtype. Front. Behav. Neurosci. 13, 259. https:// Livingston, G., Sommerlad, A., Orgeta, V., Costafreda, S.G., Huntley, J., Ames, D., ...
[Link]/10.3389/fnbeh.2019.00259. Mukadam, N., 2017. Dementia prevention, intervention, and care. Lancet. https://
Chao-Gan, Y., Yu-Feng, Z., 2010. DPARSF: a MATLAB toolbox for “Pipeline” data analysis [Link]/10.1016/S0140-6736(17)31363-6.
of resting-state fMRI. Front. Syst. Neurosci. 4 (13). [Link] Mars, R.B., Neubert, F.X., Noonan, M.P., Sallet, J., Toni, I., Rushworth, M.F., 2012. On the
2010.00013. relationship between the “default mode network” and the “social brain. Front. Hum.
Choo, I.H., Lee, D.Y., Oh, J.S., Lee, J.S., Lee, D.S., Song, I.C., ... Woo, J.I., 2010. Posterior Neurosci. 6, 189. [Link]
cingulate cortex atrophy and regional cingulum disruption in mild cognitive im- Maruszak, A., Thuret, S., 2014. Why looking at the whole hippocampus is not enough-a
pairment and Alzheimer's disease. Neurobiol. Aging 31 (5), 772–779. [Link] critical role for anteroposterior axis, subfield and activation analyses to enhance
org/10.1016/[Link].2008.06.015. predictive value of hippocampal changes for Alzheimer's disease diagnosis. Front.
Cooney, G., Dwan, K., Mead, G., 2014. Exercise for depression. JAMA 311 (23), Cell. Neurosci. 8, 95. [Link]
2432–2433. [Link] Mavros, Y., Gates, N., Wilson, G.C., Jain, N., Meiklejohn, J., Brodaty, H., ... Fiatarone
Costafreda, S.G., Dinov, I.D., Tu, Z., Shi, Y., Liu, C.Y., Kloszewska, I., ... Simmons, A., Singh, M.A, 2017. Mediation of cognitive function improvements by strength gains
2011. Automated hippocampal shape analysis predicts the onset of dementia in mild after resistance training in older adults with mild cognitive impairment: outcomes of
cognitive impairment. Neuroimage 56 (1), 212–219. [Link] the study of mental and resistance training. J. Am. Geriatr. Soc. 65 (3), 550–559.
neuroimage.2011.01.050. [Link]
Dugger, B.N., Davis, K., Malek-Ahmadi, M., Hentz, J.G., Sandhu, S., Beach, T.G., ... Morra, J.H., Tu, Z., Apostolova, L.G., Green, A.E., Avedissian, C., Madsen, S.K., ... 2009.
Sabbagh, M.N., 2015. Neuropathological comparisons of amnestic and nonamnestic Automated 3 D mapping of hippocampal atrophy and its clinical correlates in 400
mild cognitive impairment. BMC Neurol. 15 (146). [Link] subjects with Alzheimer's disease, mild cognitive impairment, and elderly controls.
015-0403-4. Hum. Brain Mapp. 30 (9), 2766–2788. [Link]
Erickson, K.I., Voss, M.W., Prakash, R.S., Basak, C., Szabo, A., Chaddock, L., ... Kramer, Nagamatsu, L.S., Handy, T.C., Hsu, C.L., Voss, M., Liu-Ambrose, T., 2012. Resistance
A.F., 2011. Exercise training increases size of hippocampus and improves memory. training promotes cognitive and functional brain plasticity in seniors with probable
Proc. Natl. Acad. Sci. USA 108 (7), 3017–3022. [Link] mild cognitive impairment. Arch. Intern. Med. 172 (8), 666–668. [Link]
1015950108. 1001/archinternmed.2012.379.
Evans, M.C., Barnes, J., Nielsen, C., Kim, L.G., Clegg, S.L., Blair, M., ... 2010. Volume Ngandu, T., Lehtisalo, J., Solomon, A., Levalahti, E., Ahtiluoto, S., Antikainen, R., ...
changes in Alzheimer's disease and mild cognitive impairment: cognitive associa- Kivipelto, M., 2015. A 2 year multidomain intervention of diet, exercise, cognitive
tions. Eur. Radiol. 20 (3), 674–682. [Link] training, and vascular risk monitoring versus control to prevent cognitive decline in
Singh, Fiatarone, A., M., Gates, N., Saigal, N., Wilson, G.C., Meiklejohn, J., Brodaty, H., ... at-risk elderly people (FINGER): a randomised controlled trial. Lancet 385 (9984),
Valenzuela, M., 2014. The Study of Mental and Resistance Training (SMART) study- 2255–2263. [Link]
resistance training and/or cognitive training in mild cognitive impairment: a ran- Norton, S., Matthews, F.E., Barnes, D.E., Yaffe, K., Brayne, C., 2014. Potential for primary
domized, double-blind, double-sham controlled trial. J. Am. Med. Dir. Assoc. 15 (12), prevention of Alzheimer's disease: an analysis of population-based data. Lancet
873–880. [Link] Neurol. 13 (8), 788–794. [Link]
Firth, J., Stubbs, B., Vancampfort, D., Schuch, F., Lagopoulos, J., Rosenbaum, S., Ward, O'Keefe, J., 1991. An allocentric spatial model for the hippocampal cognitive map.
P.B., 2018. Effect of aerobic exercise on hippocampal volume in humans: a systematic Hippocampus 1 (3), 230–235. [Link]
review and meta-analysis. Neuroimage 166, 230–238. [Link] O'Mara, S.M., Commins, S., Anderson, M., 2000. Synaptic plasticity in the hippocampal
neuroimage.2017.11.007. area CA1-subiculum projection: implications for theories of memory. Hippocampus

12
K.M. Broadhouse, et al. NeuroImage: Clinical 25 (2020) 102182

10 (4), 447–456 10.1002/1098-1063(2000)10:4<447::AID-HIPO11>[Link];2-2. 1038/mp.2016.57.

Ongur, D., Price, J.L., 2000. The organization of networks within the orbital and medial ten Brinke, L.F., Bolandzadeh, N., Nagamatsu, L.S., Hsu, C.L., Davis, J.C., Miran-Khan, K.,
prefrontal cortex of rats, monkeys and humans. Cereb. Cortex 10 (3), 206–219. Liu-Ambrose, T., 2015. Aerobic exercise increases hippocampal volume in older
Petersen, R.C., Smith, G.E., Waring, S.C., Ivnik, R.J., Tangalos, E.G., Kokmen, E., 1999. women with probable mild cognitive impairment: a 6-month randomised controlled
Mild cognitive impairment: clinical characterization and outcome. Arch. Neurol. 56 trial. Br. J. Sports Med. 49 (4), 248–254. [Link]
(3), 303–308. 093184.
Peterson, M.D., Gordon, P.M., 2011. Resistance exercise for the aging adult: clinical Vellas, B., Carrie, I., Gillette-Guyonnet, S., Touchon, J., Dantoine, T., Dartigues, J.F., ...
implications and prescription guidelines. Am. J. Med. 124 (3), 194–198. [Link] Andrieu, S., 2014. Mapt study: a multidomain approach for preventing Alzheimer's
org/10.1016/[Link].2010.08.020. disease: design and baseline data. J. Prev. Alzheimers Dis. 1 (1), 13–22.
Rebok, G.W., Ball, K., Guey, L.T., Jones, R.N., Kim, H.Y., King, J.W., ... Group, A.S., 2014. Vincent, J.L., Kahn, I., Snyder, A.Z., Raichle, M.E., Buckner, R.L., 2008. Evidence for a
Ten-year effects of the advanced cognitive training for independent and vital elderly frontoparietal control system revealed by intrinsic functional connectivity. J.
cognitive training trial on cognition and everyday functioning in older adults. J. Am. Neurophysiol. 100 (6), 3328–3342. [Link]
Geriatr. Soc. 62 (1), 16–24. [Link] Voss, M.W., Prakash, R.S., Erickson, K.I., Basak, C., Chaddock, L., Kim, J.S., ... Kramer,
Reitz, C., Brickman, A.M., Brown, T.R., Manly, J., DeCarli, C., Small, S.A., Mayeux, R., A.F., 2010. Plasticity of brain networks in a randomized intervention trial of exercise
2009. Linking hippocampal structure and function to memory performance in an training in older adults. Front. Aging Neurosci. 2. [Link]
aging population. Arch. Neurol. 66 (11), 1385–1392. [Link] 2010.00032.
archneurol.2009.214. Wang, C., Yu, J.T., Wang, H.F., Tan, C.C., Meng, X.F., Tan, L., 2014. Non-pharmacological
Reuter, M., Schmansky, N.J., Rosas, H.D., Fischl, B., 2012. Within-subject template esti- interventions for patients with mild cognitive impairment: a meta-analysis of ran-
mation for unbiased longitudinal image analysis. Neuroimage 61 (4), 1402–1418. domized controlled trials of cognition-based and exercise interventions. J. Alzheimers
[Link] Dis. 42 (2), 663–678. [Link]
Richard, E., Van den Heuvel, E., Moll van Charante, E.P., Achthoven, L., Vermeulen, M., West, M.J., Coleman, P.D., Flood, D.G., Troncoso, J.C., 1994. Differences in the pattern of
Bindels, P.J., Van Gool, W.A., 2009. Prevention of dementia by intensive vascular hippocampal neuronal loss in normal ageing and Alzheimer's disease. Lancet 344
care (PreDIVA): a cluster-randomized trial in progress. Alzheimer Dis. Assoc. Disord. (8925), 769–772.
23 (3), 198–204. [Link] Wrann, C.D., White, J.P., Salogiannnis, J., Laznik-Bogoslavski, D., Wu, J., Ma, D., ...
Rossler, M., Zarski, R., Bohl, J., Ohm, T.G., 2002. Stage-dependent and sector-specific Spiegelman, B.M., 2013. Exercise induces hippocampal BDNF through a PGC-1alpha/
neuronal loss in hippocampus during Alzheimer's disease. Acta Neuropathol. 103 (4), FNDC5 pathway. Cell Metab. 18 (5), 649–659. [Link]
363–369. [Link] 09.008.
Scarmeas, N., 2017. Dementia: multimodal dementia prevention – does trial design mask Yassine, H.N., Schneider, L.S., 2017. Lessons from the multidomain alzheimer preventive
efficacy. Nat. Rev. Neurol. 13 (6), 322–323. [Link] trial. Lancet Neurol. 16 (8), 585–586. [Link]
2017.73. 30227-2.
Scheff, S.W., Price, D.A., Schmitt, F.A., Mufson, E.J., 2006. Hippocampal synaptic loss in Young, J., Angevaren, M., Rusted, J., Tabet, N., 2015. Aerobic exercise to improve cog-
early Alzheimer's disease and mild cognitive impairment. Neurobiol. Aging 27 (10), nitive function in older people without known cognitive impairment. Cochrane
1372–1384. [Link] Database Syst. Rev. 4[Link]
Schneider, J.A., Arvanitakis, Z., Leurgans, S.E., Bennett, D.A., 2009. The neuropathology CD005381.
of probable Alzheimer disease and mild cognitive impairment. Ann. Neurol. 66 (2), Yushkevich, P.A., Amaral, R.S., Augustinack, J.C., Bender, A.R., Bernstein, J.D., Boccardi,
200–208. [Link] M., ... Hippocampal Subfields, G, 2015. Quantitative comparison of 21 protocols for
Shi, F., Liu, B., Zhou, Y., Yu, C., Jiang, T., 2009. Hippocampal volume and asymmetry in labeling hippocampal subfields and parahippocampal subregions in in vivo MRI:
mild cognitive impairment and Alzheimer's disease: meta-analyses of MRI studies. towards a harmonized segmentation protocol. Neuroimage 111, 526–541. https://
Hippocampus 19 (11), 1055–1064. [Link] [Link]/10.1016/[Link].2015.01.004.
Smith, P.J., Blumenthal, J.A., Hoffman, B.M., Cooper, H., Strauman, T.A., Welsh-Bohmer, JrZhou, Y., Dougherty, J.H., Hubner, K.F., Bai, B., Cannon, R.L., Hutson, R.K., 2008.
K., ... Sherwood, A., 2010. Aerobic exercise and neurocognitive performance: a meta- Abnormal connectivity in the posterior cingulate and hippocampus in early
analytic review of randomized controlled trials. Psychosom. Med. 72 (3), 239–252. Alzheimer's disease and mild cognitive impairment. Alzheimers Dement. 4 (4),
[Link] 265–270. [Link]
Suo, C., Singh, M.F., Gates, N., Wen, W., Sachdev, P., Brodaty, H., ... Valenzuela, M.J., Zissimopoulos, J., Crimmins, E., Clair, St, 2014. The value of delaying Alzheimer's disease
2016. Therapeutically relevant structural and functional mechanisms triggered by onset. Forum Health Econ. Policy 18 (1), 25–39. [Link]
physical and cognitive exercise. Mol. Psychiatry 21 (11), 1645. [Link] 0013.

13