Professional Documents
Culture Documents
NeuroImage
journal homepage: www.elsevier.com/locate/ynimg
Review
a r t i c l e
i n f o
a b s t r a c t
Introduction: This review aimed to produce hippocampal atrophy rate estimates from healthy ageing studies as
well as control samples from observational studies across the adult lifespan which can be used as benchmarks
to evaluate abnormal changes in pathological conditions.
Methods: The review followed PRISMA guidelines. PUBMED (to February 2014) was searched for longitudinal
MRI studies reporting hippocampal atrophy or volume change in cognitively healthy individuals. Titles were
screened and non-English, duplicate or irrelevant entries were excluded. Remaining record abstracts were
reviewed to identify studies for full text retrieval. Full text was retrieved and screened against inclusion/exclusion
criteria. Bibliographies and previous reviews were examined to identify additional studies. Data were
summarised using meta-analysis and age, segmentation technique and study type were tested as potential moderators using meta-regression. It was hypothesised that population studies would produce higher atrophy rates
than clinical observational studies.
Results: The systematic search identied 4410 entries and 119 studies were retrieved with 58 failing selection or
quality criteria, 30 were excluded as multiple reports and 3 studies were unsuitable for meta-analysis. The remaining 28 studies were included in the meta-analysis, n = 3422, 44.65% male, 11,735 person-years of followup, mean age was 24.50 to 83 years. Mean total hippocampal atrophy for the entire sample was 0.85% per year
(95% CI 0.63, 1.07). Age based atrophy rates were 0.38% per year (CI 0.14, 0.62) for studies with mean age
b 55 years (n = 413), 0.98% (CI 0.27, 1.70) for 55 to b 70 years (n = 426), and 1.12% (CI 0.86, 1.38) for
70 years (n = 2583). Meta-regression indicated age was associated with increased atrophy rates of 0.0263%
(CI 0.0146, 0.0379) per year and automated segmentation approaches were associated with a reduced atrophy
rate of 0.466% (CI 0.841, 0.090). Population studies were not associated with a signicant effect on atrophy. Analyses of 11 studies separately measuring left and right hippocampal atrophy (n = 1142) provided little
evidence of laterality effects. While no study separately reported atrophy by gender, a number tested for gender
effects and 2 studies reported higher atrophy in males.
Conclusions: Hippocampal atrophy rates increase with age with the largest increases occurring from midlife onwards. Manual segmentation approaches result in higher measured atrophy rates.
2015 Elsevier Inc. All rights reserved.
Article history:
Received 25 October 2014
Accepted 14 March 2015
Available online 20 March 2015
Keywords:
Hippocampus
MRI
Longitudinal
Ageing
Epidemiology
Controls
Contents
Introduction . . . . . . . . . . . .
Methods . . . . . . . . . . . . .
Search strategy . . . . . . . .
Inclusion and exclusion criteria
Data extraction . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
365
365
365
365
366
http://dx.doi.org/10.1016/j.neuroimage.2015.03.035
1053-8119/ 2015 Elsevier Inc. All rights reserved.
Statistical analysis . . . . . .
Missing data . . . . . . . . .
Meta-analysis . . . . . . . .
Reporting bias . . . . . . . .
Results . . . . . . . . . . . . . .
Meta-analysis . . . . . . . .
Reporting bias . . . . . . . .
Discussion . . . . . . . . . . . .
Laterality effects . . . . . . .
Moderators & heterogeneity . .
Reporting bias . . . . . . . .
Dropouts . . . . . . . . . .
Gender . . . . . . . . . . .
Limitations of the study . . . .
Conclusions . . . . . . . . .
Acknowledgments . . . . . . . .
Appendix A.
Supplementary data
References . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Introduction
The hippocampus plays an essential role in memory function, goal
selection, and mood regulation. Hippocampal volume changes have
been associated with neurological conditions including Alzheimer's
disease (Jack et al., 2000; West et al., 1994), Parkinson's disease
(Camicioli et al., 2003), Huntington's disease (Majid et al., 2011),
epilepsy (Liu et al., 2001), schizophrenia (Wang et al., 2008), and
depression (Arnone et al., 2012; Steffens et al., 2011). Hippocampal
volume changes also occur across the typical adult lifespan (Raz et al.,
2010). However, the magnitude of normal hippocampal age related
change is unclear and this presents a challenge when evaluating
abnormal changes in pathological conditions such as Alzheimer's
disease.
In order to accurately estimate hippocampal change in pathological
conditions it is critical that reliable and precise estimates be available
for generally healthy populations of different ages. This review
has focused on estimates from longitudinal studies in preference to
cross-sectional estimates because cross-sectional estimates can be confounded by individual subject baseline volumes. Studies where both
longitudinal and cross-sectional analyses were used indicate that
cross-sectional studies are less able to detect hippocampal volume
change effects (Du et al., 2006; Raz et al., 2005; Ridha et al., 2006).
There is now a substantial body of research investigating longitudinal
hippocampal volume change across multiple domains encompassing
the entire adult lifespan. The domain covering younger individuals focuses on neurodegenerative conditions that become apparent in adolescence or young adulthood such as schizophrenia, temporal lobe epilepsy
and mood disorders (Geuze et al., 2005). The studies in these younger
age groups tend to have small sample sizes and small effect sizes
(b0.5% annualised atrophy). A second domain of research focuses on
conditions that become apparent later in life including AD, other forms
of dementia, Parkinson's disease, Huntington's disease and other age related pathologies (Geuze et al., 2005). Hippocampal atrophy rates increase prior to the appearance of AD symptoms and continue to
increase as the disease progresses (Fox et al., 2001; Ridha et al., 2006;
Whitwell et al., 2007). Given that the incidence of dementia is increasing
as populations worldwide age (Fratiglioni et al., 1999), a growing body
of research on dementia with many large samples primarily focused
on people over 50 years of age has emerged. In a review of AD studies,
Barnes et al. (2009) estimated annualised atrophy rates of 4.66% per
year (95% CI 3.92, 5.40) for AD subjects and 1.41% per year (95% CI
0.52, 2.30) for healthy elderly controls. A third domain investigates
changes in healthy ageing in normal individuals. Available evidence suggests that hippocampal volumes change throughout adult life in a nonlinear manner (Raz et al., 2010), with hippocampal volume being
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
365
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
366
366
366
366
366
367
367
369
370
371
372
372
372
372
372
372
372
372
366
point; and (3) included at least one group of healthy participants, cognitively normal or derived from a population sample. Studies were excluded if they (a) reported exclusively on clinical treatment groups
(including placebo treatments); (b) were case studies or samples with
less than twenty participants; (c) had an MRI follow-up period of less
than twelve months; (d) the age or gender of the sample could not be
ascertained; or (e) did not provide information to allow calculation of
hippocampal atrophy. Studies meeting the inclusion and exclusion
criteria were assessed for quality using a checklist adapted from previous reviews and the Cochrane collaboration handbook (Anstey et al.,
2011; Harlein et al., 2009; JPT and Green, 2011; Stroup et al., 2000).
Mandatory quality criteria were a prospective design, a dened study
period, specication of population characteristics, standardised data
collection and the absence of signicant bias in the sample selection.
For example, samples that had been constructed to increase the proportion of participants with a family history of AD were excluded.
Data extraction
Data were extracted by two of the authors (MF and MS) and discrepancies were resolved by consensus. Where a particular sample was reported in multiple studies, the study that best t the selection criteria
and provided information in the format most suitable for metaanalysis was included and other studies were excluded. Studies that
measured hippocampal volume using manual techniques were
classied as using Manual segmentation and studies that used
automated or semi-automated segmentation techniques were classied
as using Automated segmentation. Follow-up periods were converted
to years and atrophy measures were converted to % per year with
positive atrophy representing a loss of hippocampal tissue. Variance
information was converted to standard deviations (SD). Where
atrophy was not provided, it was calculated using the formula:
Atrophy = ((volume_time1 volume_time2) / volume_time1) /
(time1 time2). Total atrophy was calculated by averaging left and
right atrophy weighted by left and right hippocampal volumes. Authors
of the selected studies were contacted via email to gain additional information or seek clarication where required. The authors contacted provided additional volume, atrophy or correlation information to enable
studies to be included in the meta-analyses.
Where studies provided separate atrophy rates for age based subsets, the age based rates were included separately. For studies where a
normal sample had been split into sub-samples by category such as
APOE variant and the atrophy information was only available at the
sub-sample level, a single weighted atrophy rate was calculated from
the sub-sample information. Where studies provided separate atrophy
rates for consecutive time points, a single mean atrophy rate was
calculated.
Meta-analysis
A random-effects model using a restricted maximum likelihood estimator (REML) was used for all meta-analyses. A random effects model
was chosen based on the assumption that included studies are heterogeneous because they sample populations with different characteristics
using a range of methodologies and therefore one cannot assume that
there is a single effect size (Borenstein et al., 2011). A random effects
meta-analysis estimates the mean of a distribution of effects rather
than estimating a unique effect (Borenstein et al., 2011). We assessed
heterogeneity across studies with the Q statistic (with p b .10 being suggestive of signicant heterogeneity) and the I2 statistic (values of 25%,
50% and 75% were indicative of low, medium, and high heterogeneity).
Separate meta-analyses were performed for total, left and right hippocampal atrophy. Sensitivity analysis was performed to assess the impact
of including SDs imputed with the multiple imputation procedure.
The impact of age on the observed atrophy rate was explored
through a meta-analysis stratied by age groups. The stratication procedure consisted of grouping studies such that it maximised the number
of age groups, containing at least 3 samples, where most of the participants from those samples would t within the age range of the group.
The distribution of samples in terms of mean age and SD was examined
to identify the number of age based groups that could be practically implemented for total, left and right hippocampal atrophy. For each age
group, the mean and SD of the included samples were used to estimate
the proportion of participants that would be fall below, within and
above the group age range. Sensitivity analysis was used to determine
group boundaries that would optimise the proportion of the sample
participants included within the groups.
Meta-regression was used to investigate the inuence of the moderators of sample type (population vs clinical), segmentation approach
(manual vs automated) and age using linear mixed-effects models
(Borenstein et al., 2011; Viechtbauer, 2010). A number of additional
non-linear meta-regression models using quadratic and cubic terms
were tested post hoc but they provided a poorer t than the linear
models.
Reporting bias
Studies that report signicant results are more likely to be published
than studies resulting in non-signicant outcomes (Song et al., 2010)
and this is known to bias the results of meta-analytic reviews. It is therefore important to formally assess publication bias and interpret results
accordingly. Reporting bias was assessed by visual inspection of funnel
plots which are scatterplots where the effect size is plotted against the
standard error of the effect size. Asymmetry of the funnel plots may
be an indication of reporting bias. The trim and ll method (Duval and
Tweedie, 2000a,b) was used to estimate the number of studies that
may be missing from the meta-analysis and to estimate adjusted effect
sizes.
Statistical analysis
Results
R version 3.1.1 (R Core Team, 2014) was used for statistical analysis.
The Amelia II R package version 1.7.2 was used for multiple imputations
(Honaker et al., 2011) and meta-analyses were performed using the
Metafor version 1.9-4 R package (Viechtbauer, 2010).
Missing data
In a few instances (n = 3) where atrophy SDs were missing, it was
possible to impute them from other published information (JPT and
Green, 2011). In other cases (total hippocampus n = 6, left/right hippocampi n = 3), missing SDs were estimated by multiple imputation using
an expectation-maximisation (EM) bootstrapping method with 5 imputations using Amelia II (Thiessen Philbrook et al., 2007).
367
automated segmentation approaches. The delineation of the hippocampus in most of the automated studies differed from the manually segmented studies by including the tail past the crus of the fornix and
excluding the mbria and alveus. A number of studies tested for gender
effects and two found increased atrophy in males; Cherbuin et al.
(2012) found greater hippocampal atrophy and Driscoll et al. (2009)
found greater age related temporal lobe atrophy.
Meta-analysis
Fig. 1. Screening and selection process for studies included in the meta-analysis.
process used for inclusion in the review. The reviewed studies are
summarised in Table 1. Additional information including imaging parameters and segmentation protocols are described in Supplementary
tables S1S4.
Quality ratings of included studies met the following criteria: denition of exposure variable and outcome, prospective design, specication
of population characteristics, description of study period, description of
the sampling procedure, standardised and described data collection,
multivariate statistics described, and reproducibility. Ten studies had
more than 100 participants across all groups. Dropout rates varied
with 16 studies reporting a dropout rate of less than 20%, 9 studies
reporting greater than 20%, 1 study did not include dropout information, 4 studies selected participants that had two MRI scans from larger
prospective datasets and 1 study invited participants from existing
studies to have a follow-up scan.
Manual segmentation was used in 19 studies. The anatomical denitions used in manually segmented studies were generally consistent in
including the dentate gyrus, hippocampus proper (CA1 through CA4),
subiculum, mbria and the alveus. All manually segmented studies included the head, the body and the tail up to the crus of the fornix except
for Kaye et al. (2005) who only measured the hippocampus body and
Whitworth et al. (2005) who included some amygdala tissue in the segmentation. Twelve studies used a range of different automated or semi-
368
Table 1
Studies included in the review.
Study
Group
Male %
Age mean
(SD)
Time
(SD)
Segmentation
NC
Population
Population
Population
Population
Population
CN
NC
NC
CN-stable
CN
NC
1434
35 to 54
Over 54
Healthy
NC
Normal
Young b50
Old 50+
Healthy
Controls
Population
3039 years
4049 years
5059 years
6069 years
7084 years
Controls
Controls
Healthy
NCI-S
Healthy
CN
NC
NC
Population
NC
20
225
249
1186
244
120
25
30
24
40
88
113
44
37
9
41
22
58
32
40
30
25
281
8
10
10
6
5
20
199
72
26
37
26
62
20
204
20
50
56.4
57
36.7
49
60.8
44
36.7
33.3
42.5
47.7
67.3
59.1
45.9
66.7
100
31.8
46.6
46.7
36
42.3
50
50
50
50
25
50
53.3
19.4
19.2
43.2
46.2
54.8
55
54.4
100
69 (7)
71.4 (6.8)
62.6 (1.4)
72.3 (3.9)
73.5 (7.9)
70.6 (6.1)
76 (7.8)
43.6 (13.1)
81.0 (3.8)
79 ()
83.0 (7.0)
35.4 (12.3)
24.5 (6.6)
44.5 (5.7)
67.9 (6.4)
67.3 (1.3)
40.1 (12.2)
74.1 (6.7)
39.4 (8.3)
63.1(7.0)
63.1 (7.0)
46.5 (10.2)
72.3 (3.8)
36.1 (2.5)
45.6 (2.9)
53.9 (3.5)
62.7 (2.3)
76.8 (5.5)
45.8 (6.8)
76 (5.1)
69.4 (6.2)
78 (6)
70.3 (5.8)
73 (7)
36.2 (14.5)
75.1 (3.7)
78.4 (5.0)a
31.5 (4.9)j
1 (01)
2.55 (0.41)
4 (0.21)
4 ()
3.4 (0.3)
6.02 (2.91)
2 (0.7)
3 (0)
1.96 (0.75)
4.3 () (2.55.2)e
2.04 (1.42)
4.94 (0.32)
3.57 (0.13)
3.53 (0.08)
3.53 (0.09)
6 ()
1 (0.1)
3.4 (1.4)
5.27 (0.3)
Manual
Manual
Manual
Automated
Automated
Automated
Automated
Manual
Manual
Manual
Manual
Manual
Manual
2.61
1.5 (0.8)
4 ()
1.58 (1.19)
1.83 (0.87)
1.91 (1.11)
2.07 (1.21)
0.98 (0.42)
1.41 (0.8)
1 ()
2 ()
5
3 ()
2.2 () 1.02.6e
2.24 (5.8)
1.88 () 0.892.9e
1.25 ()
3.7 (1.63)
Manual
Automated
Automated
Manual
Manual
Manual
Automated
Manual
Manual
Automated
Automated
Manual
Manual
Automated
Automated
Manual
Automated
Manual
Left
0.28 (0.93)
9.86 (4.17)a
2.03 (0.033)
1.204 (2.5402)b
0.51 (0.43)
0.32 (0.049)a
0.8 (1.7)
0.20c (1.65)
1.55 (1.38)
1.4 (0.7)f
2.2 (6.0)
0.8043 (.0267)a
0.11 (3.34)
0 (2.31)
0.64 (2.78)
0.16 ()
0.02 (0.72)
1.1 (1.4)
0.48 (0.83)a
1.04 (0.797)a
2.18 (2.46)
0.31 (1.25)
0.9455 (0.6841)a
0.75 (1.25)
0.5 (0.375)
1.05 (0.475)
0.875 (0.5875)
1.9375 (1.7375)
0.12c (1.83)
1.01 (1.72)
0.84 (5.9)h,a
1.4188 (0.2444)a
1.98 (3.92)b
2.34 (1.853)
0.105c (1.84)
1.0 (0.07)
0.50 (2.22)a
1.25 ()
Right
0.02 (1.25)
0.52 (1.37)
2.56 (0.034)a
1.46 (0.036)a
0.52 (0.60)
0.51 (0.88)
1 (2.6)
0.31 (2.15)
0.6 (2.5)
0.09 (2.36)
0.8028 (.0285)a
0.8054 (.0285)a
0.27 (1.94)g
0.04 (1.90)g
1.0 (0.8)
0.9 (0.8)
0.45 (1.63)
0.21 (3.00)
1.07 (7.19)h
0.63 (6.28)h
1.82 (2.04)
0.26 (2.15)
2.50 (2.03)
0.03 (2.50)
1.95 (3.12)
0.53 (3.21)
NC = normal controls; CN = cognitively normal; CN-stable = CN that does not progress to MCI or AD; NCI-S = no cognitive impairment at baseline, stable.
Multiply imputed SDs are shown in italics.
a
Personal communication with author.
b
Values pooled from subsets.
c
Calculated from left and right atrophy.
d
CN split into subsets of stable & converters only stable subset included in analysis.
e
Median plus range.
f
Source Barnes et al. (2009).
g
SD imputed using p-value.
h
Atrophy calculated from volume change.
i
Included MCSA sample only.
j
Age at follow-up.
Table 2
Meta-analysis estimates of total, left and right hippocampal atrophy rates with age based subsets.
Qp
T2
I2
1.07
0.62
1.38
1.70
1.38
b.0001
b.0001
b.0001
b.0001
b.0001
0.36
0.12
0.31
0.76
0.22
0.6
0.34
0.56
0.87
0.47
99.96
85.45
99.92
96.60
98.75
0.04
0.45
0.14
1.23
0.76
1.75
b.0001
b.0001
b.0001
0.89
0.30
1.05
0.94
0.55
1.02
99.99
84.45
99.75
0.25
0.21
0.31
1.15
0.87
1.52
b.0001
b.0001
b.0001
0.45
0.19
0.55
0.67
0.43
0.74
99.99
68.75
99.30
Description
Age
Atrophy %/yr
s.e.
95% CI
Total hippocampus
Young: b55 years
Old: 55+ years
55 to b70 years
70+ years
35
12
23
7
16
3422
413
3009
426
2583
68.59
39.53
72.58
64.24
73.95
0.85
0.38
1.12
0.98
1.12
0.11
0.12
0.13
0.37
0.13
0.63
0.14
0.86
0.27
0.86
Left hippocampus
b55 years
N55 years
11
4
7
1142
225
917
64.34
40.06
70.29
0.64
0.16
0.94
0.30
0.31
0.41
Right hippocampus
b55 years
N55 years
11
4
7
1142
225
917
64.34
40.06
70.29
0.70
0.33
0.92
0.23
0.28
0.31
k = number of samples or sub-samples included in analysis; s.e. = standard error; Qp = p-value for the signicance test of the Q statistic; T2 = heterogeneity = estimated variance of true
effects; T = estimated standard deviation of true effects; I2 = proportion of observed variance (heterogeneity) that is real.
369
Fig. 2. Random effects model of total hippocampal atrophy with age based subsets. Studies are ordered by mean age.
estimated one missing study for the left hippocampus and three missing
for the right hippocampus, inclusion of the missing studies produced
adjusted estimates for left hippocampal atrophy of 0.74% per year
(95% CI 0.15, 1.33) and right hippocampal atrophy of 0.96% per year
(95% CI 0.50, 1.42). The adjusted estimates should be treated with caution for two reasons. Firstly the estimated missing studies had very large
effect sizes and are therefore unlikely to result from reporting bias, and
secondly, the estimates produced by the trim and ll method have been
shown to be unreliable in the presence of signicant heterogeneity as
observed in the present results (Peters et al., 2007; Terrin et al., 2003).
Discussion
The aim of this review was to estimate hippocampal atrophy rates
across the adult lifespan in cognitively normal individuals and to investigate the impact of age, sample type and segmentation approach on observed atrophy rates.
Overall, the estimated rate of total hippocampal atrophy across all
studies of 0.85% per year was consistent with previous longitudinal
ndings and higher than the atrophy rates of 0.280.35% per year (Raz
et al., 2005; Raz et al., 2004b; Scahill et al., 2003) reported in cross-
Fig. 3. Random effects model of left hippocampal atrophy with age based subsets. Studies are ordered by mean age.
370
Fig. 4. Random effects model of right hippocampal atrophy with age based subsets. Studies are ordered by mean age.
sectional studies. By bringing together the results for control participants in a range of research domains we were able to show that there
is a small yet signicant rate of atrophy from young adulthood to middle
age of 0.38% per year. In contrast, individual studies that have measured
hippocampal change in younger age groups have provided equivocal
evidence of atrophy with many nding non-signicant hippocampal
volume changes, likely due to insufcient power to detect small effects.
Hippocampal atrophy rates increase with age (Du et al., 2006) and
this meta-analysis further demonstrates that the most signicant atrophy occurs from midlife onwards. The hippocampal atrophy rate increased to 0.98% per year for studies with a mean age of 55 to less
than 70 years. The estimate from this transition group was the least precise estimate due to the smaller number of studies included. The rate of
atrophy further increased to 1.12% per year in the 70 years and older age
group. While the estimates for the older age groups were lower than the
rate of 1.41% per year estimated in a previous review, the difference was
not signicant (Barnes et al., 2009). The pattern of atrophy change with
age is consistent with previous cross-sectional and longitudinal ndings
reporting a non-linear trajectory in hippocampal ageing (Fjell et al.,
2013; Raz et al., 2010; Schuff et al., 2012), but with few samples covering middle age, it was not possible to model the critical point that is believed to occur after the age of 50 years (Fjell et al., 2013). More
longitudinal research covering this critical time should be undertaken.
Laterality effects
The estimated atrophy rates for the left and right hippocampus were
consistent with total hippocampal atrophy rates. There was little evidence of laterality differences in the atrophy estimates produced by
the meta-analysis. This nding is of particular interest because a
Table 3
Mixed effects models of hippocampal atrophy rates. Model 1: age, segmentation technique and sample type; Model 2: age and segmentation technique; Model 3: age effect on studies
using manual segmentation; Model 4: age effect on studies using automated segmentation.
k
Model 1
Intercept
Age
Segmentation
Sample type
35
Model 2
Intercept
Age
Segmentation
35
Model 3
Intercept
Age
23
Model 4
Intercept
Age
12
r2
0.1529
0.0375
0.1155
0.6857
0.4558
41.84
1.318
0.0146
0.8413
0.1129
0.0379
0.0899
0.4521
42.78
0.0976
b.0001
1.5798
0.0142
0.1325
0.0425
0.4542
45.26
0.3170
0.0494
2.2862
0.0001
0.7408
0.0438
0.4827
29.98
Coef
s.e.
95% CI
0.5720
0.0256
0.5222
0.1878
0.3699
0.0061
0.2075
0.2540
1.5466
4.2053
2.5165
0.7394
0.1220
b0.0001
0.0119
0.4597
1.2970
0.0136
0.9289
0.3100
0.6026
0.0263
0.4656
0.365
0.006
0.1917
1.6508
4.414
2.429
0.0988
b.0001
0.0151
0.7237
0.0284
0.4368
0.0072
1.6567
3.9188
0.7727
0.0219
0.7722
0.0112
1.0006
1.9651
k = number of samples or sub-samples included in analysis; s.e. = standard error; = standard deviation of true effects; r2 = proportion of observed dispersion accounted for by the
model.
371
Fig. 7. Funnel plot of left hippocampal atrophy using trim and ll method. Filled circles
represent studies included in the meta-analysis. Open circles represent possible missing
studies.
under half of the heterogeneity. Unsurprisingly, age had the largest effect, accounting for most of the explained heterogeneity. Given that
many studies had wide age ranges, exceeding 25 years in some cases,
it is possible that age may have a greater moderating effect than quantied by the meta-regression. Another salient feature of this metaanalysis is that heterogeneity between studies was higher in studies
with a mean age greater than 55 years. This is probably due to chronic
disease and non-clinical brain changes which become more prevalent
in ageing. However, more targeted research is required to understand
the factors driving increased variability in older samples.
The other signicant moderator of heterogeneity identied was the
segmentation technique used to measure hippocampal volume. The
meta-regression suggested that automated segmentation results in
lower atrophy rate estimates. It is possible that automated approaches
may include some non-hippocampal tissue that has a lower rate of atrophy than hippocampal tissue (Wenger et al., 2014). Indeed, Cherbuin
et al. (2009) found that the hippocampal segmentation implemented
in Freesurfer produced signicantly larger volumes compared to
Fig. 6. Funnel plot of total hippocampal atrophy using trim and ll method. Filled circles
represent studies included in the meta-analysis. Open circles represent possible missing
studies.
Fig. 8. Funnel plot of right hippocampal atrophy using trim and ll method. Filled circles
represent studies included in the meta-analysis. Open circles represent possible missing
studies.
Fig. 5. Meta-regression of atrophy rate and mean age. The size of circles is proportional to
the weight given to the study. Larger studies and more precise studies are given more
weight in the meta-analysis. Magenta circles represent studies using manual segmentation and blue circles represent studies using automated segmentation. The magenta line
is the predicted mean atrophy rate change with age with manual segmentation
(Model 3) and the blue line represents the predicted mean atrophy rate with age for automated segmentation (Model 4).
number of studies have suggested earlier and faster atrophy in the left
hippocampus. Consequently if this effect is not present in generally
healthy individuals it may be more indicative of developing neurodegenerative pathology which has been reported to be asymmetrical
(Cherbuin et al., 2010; Thompson et al., 2003).
Moderators & heterogeneity
372
Gender
References
Hippocampal atrophy was not separately measured in males and females in any of the studies. Therefore it was not possible to perform any
gender based meta-analyses. However, a number of studies controlled
or tested for gender effects, with one study reporting greater hippocampal
atrophy rates in males while a second reported higher temporal lobe atrophy in males (Cherbuin et al., 2012; Driscoll et al., 2009). In addition, one
of the male-only studies, Whitworth et al. (2005), found annualised atrophy of 1.25% per year in young men, which is somewhat higher than the
meta-analysis estimate, but in line with the limited information from the
few studies investigating gender effects in hippocampal atrophy. Signicant negative correlations have been found between hippocampal volume and age in men, but not in women, from around 20 to 46 years of
age (Pruessner et al., 2001; Raz et al., 2004a). Eberling et al. (2003) suggested that oestrogen may protect against age related hippocampal atrophy. Given the small sample sizes of many MRI studies it is not surprising
that few studies measure hippocampal atrophy separately in each gender.
The absence of studies that investigate gender effects represent a potential gap in the current research literature and needs to be investigated in
detail.
Anstey, K., Cherbuin, N., Budge, M., Young, J., 2011. Body mass index in midlife and late life as
a risk factor for dementia: a meta-analysis of prospective studies. Obes. Rev. 12,
e426e437.
Arnone, D., McKie, S., Elliott, R., Juhasz, G., Thomas, E.J., Downey, D., Williams, S., Deakin,
J.F., Anderson, I.M., 2012. State-dependent changes in hippocampal grey matter in
depression. Mol. Psychiatry 18, 12651272.
Barnes, J., Scahill, R.I., Frost, C., Schott, J.M., Rossor, M.N., Fox, N.C., 2008. Increased hippocampal atrophy rates in AD over 6 months using serial MR imaging. Neurobiol. Aging 29,
11991203.
Barnes, J., Bartlett, J.W., van de Pol, L.A., Loy, C.T., Scahill, R.I., Frost, C., Thompson, P., Fox, N.C.,
2009. A meta-analysis of hippocampal atrophy rates in Alzheimer's disease. Neurobiol.
Aging 30, 17111723.
Borenstein, M., Hedges, L.V., Higgins, J.P., Rothstein, H.R., 2011. Introduction to Metaanalysis. John Wiley & Sons.
Callisaya, M.L., Beare, R., Phan, T.G., Blizzard, L., Thrift, A.G., Chen, J., Srikanth, V.K., 2013.
Brain structural change and gait decline: a longitudinal population-based study.
J. Am. Geriatr. Soc. 61, 10741079.
Camicioli, R., Moore, M.M., Kinney, A., Corbridge, E., Glassberg, K., Kaye, J.A., 2003.
Parkinson's disease is associated with hippocampal atrophy. Mov. Disord. 18, 784790.
Cherbuin, N., Anstey, K.J., Rglade-Meslin, C., Sachdev, P.S., 2009. In vivo hippocampal measurement and memory: a comparison of manual tracing and automated segmentation
in a large community-based sample. PLoS One 4.
Cherbuin, N., Rglade-Meslin, C., Kumar, R., Sachdev, P., Anstey, K.J., 2010. Mild cognitive disorders are associated with different patterns of brain asymmetry than normal aging: the
PATH through life study. Front. Psychiatry 1.
373
Mungas, D., Harvey, D., Reed, B.R., Jagust, W.J., DeCarli, C., Beckett, L., Mack, W.J., Kramer,
J.H., Weiner, M.W., Schuff, N., Chui, H.C., 2005. Longitudinal volumetric MRI change
and rate of cognitive decline. Neurology 65, 565571.
Patenaude, B., Smith, S.M., Kennedy, D.N., Jenkinson, M., 2011. A Bayesian model of shape
and appearance for subcortical brain segmentation. Neuroimage 56, 907922.
Peters, J.L., Sutton, A.J., Jones, D.R., Abrams, K.R., Rushton, L., 2007. Performance of
the trim and ll method in the presence of publication bias and betweenstudy heterogeneity. Stat. Med. 26, 45444562.
Pruessner, J., Collins, D., Pruessner, M., Evans, A., 2001. Age and gender predict volume decline in the anterior and posterior hippocampus in early adulthood. J. Neurosci. 21,
194200.
R. Core Team, 2014. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Raz, N., Gunning-Dixon, F., Head, D., Rodrigue, K.M., Williamson, A., Acker, J.D.,
2004a. Aging, sexual dimorphism, and hemispheric asymmetry of the cerebral
cortex: replicability of regional differences in volume. Neurobiol. Aging 25,
377396.
Raz, N., Rodrigue, K.M., Head, D., Kennedy, K.M., Acker, J.D., 2004b. Differential aging of
the medial temporal lobe: a study of a ve-year change. Neurology 62, 433438.
Raz, N., Lindenberger, U., Rodrigue, K.M., Kennedy, K.M., Head, D., Williamson, A., Dahle,
C., Gerstorf, D., Acker, J.D., 2005. Regional brain changes in aging healthy adults: general trends, individual differences and modiers. Cereb. Cortex 15, 16761689.
Raz, N., Ghisletta, P., Rodrigue, K.M., Kennedy, K.M., Lindenberger, U., 2010. Trajectories of
brain aging in middle-aged and older adults: regional and individual differences.
Neuroimage 51, 501511.
Reuter, M., Schmansky, N.J., Rosas, D.H., Fischl, B., 2012. Within-subject template estimation for unbiased longitudinal image analysis. Neuroimage 61, 14021418.
Ridha, B.H., Barnes, J., Bartlett, J.W., Godbolt, A., Pepple, T., Rossor, M.N., Fox, N.C., 2006. Tracking atrophy progression in familial Alzheimer's disease: a serial MRI study. Lancet
Neurol. 5, 828834.
Samieri, C., Maillard, P., Crivello, F., Proust-Lima, C., Peuchant, E., Helmer, C., Amieva, H.,
Allard, M., Dartigues, J.F., Cunnane, S.C., Mazoyer, B.M., Barberger-Gateau, P., 2012. Plasma long-chain omega-3 fatty acids and atrophy of the medial temporal lobe. Neurology
79, 642650.
Scahill, R.I., Frost, C., Jenkins, R., Whitwell, J.L., Rossor, M.N., Fox, N.C., 2003. A longitudinal
study of brain volume changes in normal aging using serial registered magnetic resonance imaging. Arch. Neurol. 60, 989994.
Schott, J.M., Fox, N.C., Frost, C., Scahill, R.I., Janssen, J.C., Chan, D., Jenkins, R., Rossor, M.N.,
2003. Assessing the onset of structural change in familial Alzheimer's disease. Ann.
Neurol. 53, 181188.
Schott, J.M., Bartlett, J.W., Barnes, J., Leung, K.K., Ourselin, S., Fox, N.C., 2010. Reduced sample sizes for atrophy outcomes in Alzheimer's disease trials: baseline adjustment.
Neurobiol. Aging 31, 14521462 (1462 e1451-1452).
Schuff, N., Tosun, D., Insel, P.S., Chiang, G.C., Truran, D., Aisen, P.S., Jack Jr., C.R., Weiner,
M.W., 2012. Nonlinear time course of brain volume loss in cognitively normal and
impaired elders. Neurobiol. Aging 33, 845855.
Song, F., Parekh, S., Hooper, L., Loke, Y.K., Ryder, J., Sutton, A.J., Hing, C., Kwok, C.S., Pang, C.,
Harvey, I., 2010. Dissemination and publication of research ndings: an updated review of related biases. Health Technol. Assess. 14.
Steffens, D.C., McQuoid, D.R., Payne, M.E., Potter, G.G., 2011. Change in hippocampal volume on magnetic resonance imaging and cognitive decline among older depressed
and nondepressed subjects in the neurocognitive outcomes of depression in the elderly study. Am. J. Geriatr. Psychiatry 19, 412.
Stoub, T.R., Rogalski, E.J., Leurgans, S., Bennett, D.A., deToledo-Morrell, L., 2010. Rate of entorhinal and hippocampal atrophy in incipient and mild AD: relation to memory
function. Neurobiol. Aging 31, 10891098.
Stroup, D.F., Berlin, J.A., Morton, S.C., Olkin, I., Williamson, G.D., Rennie, D., Moher, D.,
Becker, B.J., Sipe, T.A., Thacker, S.B., 2000. Meta-analysis of observational studies in
epidemiology: a proposal for reporting. Meta-analysis of Observational Studies in
Epidemiology (MOOSE) group. JAMA 283, 20082012.
Terrin, N., Schmid, C.H., Lau, J., Olkin, I., 2003. Adjusting for publication bias in the presence of heterogeneity. Stat. Med. 22, 21132126.
Thiessen Philbrook, H., Barrowman, N., Garg, A., 2007. Imputing variance estimates do not
alter the conclusions of a meta-analysis with continuous outcomes: a case study of
changes in renal function after living kidney donation. J. Clin. Epidemiol. 60, 228240.
Thompson, P.M., Hayashi, K.M., De Zubicaray, G., Janke, A.L., Rose, S.E., Semple, J., Herman,
D., Hong, M.S., Dittmer, S.S., Doddrell, D.M., 2003. Dynamics of gray matter loss in
Alzheimer's disease. J. Neurosci. 23, 9941005.
Valenzuela, M.J., Sachdev, P., Wen, W., Chen, X., Brodaty, H., 2008. Lifespan mental activity
predicts diminished rate of hippocampal atrophy. PLoS One 3, e2598.
Viechtbauer, W., 2010. Conducting meta-analyses in R with the metafor package. J. Stat.
Softw. 36, 148.
Wang, L., Swank, J.S., Glick, I.E., Gado, M.H., Miller, M.I., Morris, J.C., Csernansky, J.G., 2003.
Changes in hippocampal volume and shape across time distinguish dementia of the
Alzheimer type from healthy aging. Neuroimage 20, 667682.
Wang, L., Mamah, D., Harms, M.P., Karnik, M., Price, J.L., Gado, M.H., Thompson, P.A., Barch,
D.M., Miller, M.I., Csernansky, J.G., 2008. Progressive deformation of deep brain nuclei
and hippocampalamygdala formation in schizophrenia. Biol. Psychiatry 64,
10601068.
Wang, P.N., Liu, H.C., Lirng, J.F., Lin, K.N., Wu, Z.A., 2009. Accelerated hippocampal atrophy
rates in stable and progressive amnestic mild cognitive impairment. Psychiatry Res.
171, 221231.
Wenger, E., Mrtensson, J., Noack, H., Bodammer, N.C., Khn, S., Schaefer, S., Heinze, H.-J.,
Dzel, E., Bckman, L., Lindenberger, U., Lvdn, M., 2014. Comparing manual and automatic segmentation of hippocampal volumes: Reliability and validity issues in
younger and older brains. Hum. Brain Mapp. 35, 42364248.
374
West, M.J., Coleman, P.D., Flood, D.G., Troncoso, J.C., 1994. Differences in the pattern of
hippocampal neuronal loss in normal ageing and Alzheimer's disease. Lancet 344,
769772.
Whitwell, J.L., Przybelski, S.A., Weigand, S.D., Knopman, D.S., Boeve, B.F., Petersen, R.C.,
Jack Jr., C.R., 2007. 3D maps from multiple MRI illustrate changing atrophy patterns
as subjects progress from mild cognitive impairment to Alzheimer's disease. Brain
130, 17771786.
Whitwell, J.L., Wiste, H.J., Weigand, S.D., Rocca, W.A., Knopman, D.S., Roberts, R.O., Boeve,
B.F., Petersen, R.C., Jack Jr., C.R., 2012. Comparison of imaging biomarkers in the
Alzheimer Disease Neuroimaging Initiative and the Mayo Clinic Study of Aging.
Arch. Neurol. 69, 614622.
Whitworth, A.B., Kemmler, G., Honeder, M., Kremser, C., Felber, S., Hausmann, A., Walch,
T., Wanko, C., Weiss, E.M., Stuppaeck, C.H., Fleischhacker, W.W., 2005. Longitudinal
volumetric MRI study in rst- and multiple-episode male schizophrenia patients.
Psychiatry Res. 140, 225237.
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.