International Journal of Clinical and Health Psychology 22 (2022) 100317

International Journal
of Clinical and Health Psychology
www.elsevier.es/ijchp

ORIGINAL ARTICLE

Abnormalities in the default mode network in late-life

depression: A study of resting-state fMRI
Joan Gua
rdia-Olmosa,e,f,*, Carles Soriano-Masa,b,c,**, Lara Tormo-Rodríguezg,
Cristina Can  a,f, Ine
~ ete-Masse  s del Cerrob, Mikel Urretavizcayab,c,d, Jose  nb,c,d,
 M. Mencho
Virgina Soriab,c,d, Maribel Pero  -Cebolleroa,e,f

a
Departament de Psicologia Social & Psicologia Quantitativa, Faculty of Psychology, University of Barcelona, Spain
b
Bellvitge Biomedical Research Institute-IDIBELL, Department of Psychiatry, Bellvitge University Hospital, Barcelona, Spain
c
Network Center for Biomedical Research on Mental Health (CIBERSAM), Carlos III Health Institute (ISCIII), Madrid, Spain
d
Department of Clinical Sciences, Bellvitge Campus, University of Barcelona, Spain
e
Institut of Neuroscience, Universitat de Barcelona
f
UB Institute of Complex Systems
g
Faculty of Psychology, University of Barcelona, Spain

Received 4 February 2022; accepted 10 May 2022

Available online xxx

KEY WORDS Abstract

Default mode Background/Objective: Neuroimaging studies have reported abnormalities in the examination of
network; functional connectivity in late-life depression (LLD) in the default mode network (DMN). The
Late-life major present study aims to study resting-state functional connectivity within the DMN in people diag-
depressive disorder; nosed with late-life major depressive disorder (MDD) compared to healthy controls (HCs). More-
Functional over, we would like to differentiate these same connectivity patterns between participants with
connectivity; high vs. low anxiety levels. Method: The sample comprised 56 participants between the ages of
fMRI; 60 and 75; 27 of them were patients with a diagnosis of MDD. Patients were further divided into
Resting State; two samples according to anxiety level: the four people with the highest anxiety level and the five
Experiment with the lowest anxiety level. Clinical aspects were measured using psychological questionnaires.
Each participant underwent functional magnetic resonance imaging (fMRI) acquisition in different
regions of interest (ROIs) of the DMN. Results: There was a greater correlation between pairs of
ROIs in the control group than in patients with LLD, being this effect preferentially observed in
patients with higher anxiety levels. Conclusions: There are differences in functional connectivity
within the DMN depending on the level of psychopathology. This can be reflected in these correla-
tions and in the number of clusters and how the brain lateralizes (clustering).
© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

* Corresponding author: Departament de Psicologia Social & Psicologia Quantitativa, Faculty of Psychology, University of Barcelona, Spain. **
Co-corresponding author: Department of Psychiatry, Bellvitge University Hospital, Bellvitge Biomedical Research Institute-IDIBELL, Spain.

rdia-Olmos), csoriano@idibell.cat (C. Soriano-Mas).
E-mail addresses: jguardia@ub.edu (J. Gua

https://doi.org/10.1016/j.ijchp.2022.100317
1697-2600/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

rdia-Olmos, C. Soriano-Mas, L. Tormo-Rodríguez et al.
J. Gua

Introduction (Jacob et al., 2020). Therefore, comorbid anxiety symptoms

may LLD may indeed play an essential role in DMN disruption
According to recent estimates, major depressive disorder in LLD.
(MDD) may be observed in more than 7% of the population of In this study, we aimed to assess abnormal DMN activity in
elderly individuals (i.e., late-life depression, or LDD) this population of individuals compared to a group of healthy
(Wen et al., 2022). This disorder, therefore, should be con- controls comparable in age and sex distribution. Moreover,
sidered as one of the major health issues of this population, we also aimed to assess the moderating effect of anxiety on
not only because of the direct effect of MDD on individuals’ our findings. Anxiety symptoms have also been related to
quality of life and functionality (Zhang, Chen & Ma, 2018) DMN alterations (Andreescu, Sheu, Tudorascu, Walker &
but also because it is well known that subjects with LLD Aizenstein, 2014; Laird et al., 2019; Zugman et al., 2022),
have a twofold increased risk of progression to neurodegen- and since this is a symptom typically observed in LLD
erative disorders such as Alzheimer’s disease (Habes et al., (Beekman et al., 2000), they may add or interact with MDD
2021). In this context, developing accurate diagnostic meas- symptoms in accounting for DMN alterations. Finally, in this
ures and effective prevention and treatment strategies spe- study, we used a clustering approach to assess DMN altera-
cifically targeting this population is of paramount tions. Clustering analysis in fMRI allows data stratification in
importance. For this purpose, in addition to an accurate a hierarchical structure, building a dendrogram of all the
clinical and neurocognitive characterization of patients with members (Wang et al., 2021; Zhou, Zemanova , Zamora, Hil-
LLD, it is also important to obtain more information about getag & Kurths, 2006).
the neurobiological correlates of this condition, in order to
maximize the effectiveness of clinical strategies using a bio-
logically informed perspective. Methods
Functional magnetic resonance imaging (fMRI) provides a
noninvasive means to explore brain function. Moreover, it Participants and measurements
allows not only the assessment of task-related activations
but also interregional functional connectivity (i.e., synchro- The study sample consisted of 56 participants. Patients with
nous patterns of neural activity fluctuations) at rest. This LLD (n = 27, 20 women, mean age M = 68.2 and SD = 4.01)
not only facilitates comparison across studies due to the were consecutively recruited at the Department of Psychia-
lack of differences related to varying task performance but try of Bellvitge University Hospital. A control group of 29
also allows for the description of brain activity at the net- subjects (19 women, mean age M = 67.7 and SD = 4.23) was
work level, that is, in terms of coordinated patterns of activ- recruited from the same sociodemographic environment
ity across distant brain regions underpinning cognitive and through advertisements and word-of-mouth. Inclusion crite-
emotional functioning (Geng et al., 2019; Mancho- ria for patients included a primary diagnosis of MDD and
Fora et al., 2020). aged between 60 and 75 years. MDD diagnoses were estab-
The most studied resting-state network is the default lished by two experienced psychiatrists according to DSM-IV-
mode network (DMN), which encompasses anterior and pos- TR criteria (which do not substantially differ from DSM-5 cri-
terior regions of the medial wall as well as inferior parietal teria and are aligned with the diagnostic criteria of the
areas (Damoiseaux et al., 2008). However, activity in other interview used to identify comorbid symptoms). Disorder
brain regions has also been correlated with DMN activity severity was estimated with the Hamilton Depression Rating
(Van den Heuvel & Pol, 2010; Wei et al., 2019). Activity in Scale (HDRS) (Hamilton, 1960) and the Geriatric Depression
the DMN is allegedly related to inward attentional processes, Scale (GDS) (Sheikh & Yesavage, 1986; Yesavage et al.,
and the pattern of correlations across its different compo- 1982), which was not used for diagnostic purposes. Higher
nents is therefore preferentially observed during resting- scores in these scales denote higher severity of depression
state (i.e., no-task) acquisitions (Harrison et al., 2008). symptoms. State and trait anxiety were measured through
Moreover, abnormal functional connectivity patterns across the State-Trait Anxiety Inventory (STAI, Spielberger,1983).
the different components of the DMN have been described in Similar to above, higher scores in these scales indicate more
different disorders of mental health, including MDD severe anxiety symptoms. To identify the current or past
(Wise et al., 2017) and neurodegenerative disorders presence of other than depression symptoms, all partici-
(Xue et al., 2019). pants were also interviewed using the Mini-International
Different studies have provided evidence of disturbances Neuropsychiatric Interview (MINI) (Sheehan et al., 1998),
in DMN activity in LLD. Gandelman et al. (2019) observed dif- which provided a fast but accurate assessment of the major
ferences in intrinsic functional connectivity, while psychiatric diagnoses. Finally, the Vocabulary subtest of the
Manning, Wang and Steffens (2019) presented an interesting Wechsler Adult Intelligence Scale, Third Edition (WAIS-III)
review regarding DMN alterations and those in other impor- (Wechsler, 1999), was administered to all participants to
tant networks, such as the salience network, in LLD, which estimate the premorbid intelligence quotient (IQ; higher
may contribute to shared behavioral syndromes. In this scores, higher premorbid IQ values). Importantly, medica-
sense, they highlighted the importance of anxiety symptoms tion was not changed in patients and was kept at stable
in LLD, which may affect 50% of LLD patients doses for at least one month before MRI acquisition.
(Beekman et al., 2000), in relation to the DMN dysfunction Exclusion criteria included: 1) ages <60 or >75 years
observed in this population. Indeed, some basic alterations (we set this superior age limit to minimize effects of
in psychological functioning that are observed both in altered neurovascular coupling), 2) past or current diag-
depression and anxiety samples, such as rumination (Smith nosis of other major psychiatric disorders including sub-
& Alloy, 2009), may partially account for DMN alterations stance abuse or dependence (except nicotine), 3)

2
 International Journal of Clinical and Health Psychology 22 (2022) 100317

intellectual disability/neurodevelopmental disorders, 4) (MNI) stereotactic space, and images were resliced to a 2-
neurological disorders, 5) Hachinski Ischemic Score >5, mm isotropic resolution. Finally, images were smoothed
6) presence of dementia according to the DSM-IV-TR cri- with an 8-mm full-width at half-maximum (FWHM) isotropic
teria and/or a CDR score>1, 7) severe medical condi- Gaussian kernel.
tions, 8) electroconvulsive therapy in the previous year, After preprocessing, data were denoised from residual
9) conditions preventing neuropsychological assessment movement and physiological noise. Denoising steps included
or MRI procedures (e.g., blindness, deafness, claustro- temporal despiking, regressing out confounding factors (i.
phobia, pacemakers, or cochlear implants), and 10) gross e., effect of BOLD signal small ramping effects at the begin-
abnormalities in the MRI scan. Moreover, although 59 par- ning of each scan session and the six rigid body realignment
ticipants were recruited initially, three participants (two parameters, as well as their first-order derivatives), control-
patients and one control) were excluded from the study ling for total gray matter (GM) signal, the ART scrubbing pro-
sample because of excessive movement (half of the voxel tocol, linear detrending, and bandpass filtering
size as criteria) (the two patients) or outlier values in (0.008 0.09 Hz). Physiological noise was removed with the
the psychometric assessment (the control subject). anatomical component-based noise correction method
The study was approved by The Clinical Research Ethics (aCompCor) (Behzadi, Restom, Liau & Liu, 2007). Impor-
Committee (CEIC) of Bellvitge University Hospital (reference tantly, after implementing these different steps, none of
PR156/15, 17th February 2016) and performed following the the subjects was removed from the analysis because,
ethical standards laid down in the 1964 Declaration of Hel- according to current guidelines (Van Dijk et al., 2010), all
sinki and its later amendments (revised in 2013). All partici- individual functional series included at least 95% of the origi-
pants gave written informed consent to participate in the nal volumes after scrubbing (volume censoring) and spike
study. regression.

Imaging data acquisition and preprocessing

Statistical analyses
Each participant underwent an 8-minute resting-state func-
tional MRI (fMRI) scan in a 3T Philips Ingenia scan (Philips Sociodemographic and clinical data were analyzed with IBM
Health care, Best, The Netherlands) using a 32-channel head SPSS v 24.0.0.0 for Mac (SPSS Inc., Chicago, IL) and with
coil. The functional sequence consisted of 240 echo-planar libraries and own programming in R. Shapiro-Wills tests
image volumes (excluding the four initial dummy volumes) were used to ascertain the normality distribution of these
comprising 40 interleaved slices acquired in the oblique variables. Between-group differences in quantitative varia-
axial direction perpendicular to the floor of the fourth ven- bles were assessed with Student's t-test or the nonparamet-
tricle (repetition time = 2000 ms; echo time= 25 ms; flip ric U test of the Mann-Wittney test, when appropriate. In
angle = 90°; 3 mm isotropic voxels; field of view = 24 cm, contrast, between-group differences in qualitative variables
80 £ 80 pixel matrix). For anatomical reference and imaging were explored with the x2 test.
preprocessing purposes, we also acquired for each partici- Regarding analyses of imaging data, according to our
pant a whole-brain T1-weighted anatomical three-dimen- study hypotheses, we focused on the DMN, which, following
sional inversion-recovery prepared spoiled gradient echo previous research (Huang et al., 2015), was split into three
sequence (233 axial slices; repetition time = 10.46 ms; echo different components: the anterior DMN (DMNa), the ven-
time = 4.79 ms; flip angle = 8°; 0.75 mm isotropic voxels; tral DMN (DMNv), and the posterior DMN (DMNp). Specifi-
field of view = 24 cm; pixel matrix =320 £ 318; total cally, within these components, we defined six, twelve,
duration = 5 min, 04 s). and six regions of interest (ROIs) using cortical parcellations
Functional time series were initially despiked using the from the automated anatomical atlas (AAL) (Tzourio-
BrainWavelet toolbox v2.028 (Patek et al., 2014). Next, Mazoyer et al., 2002) anatomically corresponding to such
using MATLAB version 9.3 (R2017b) (The MathWorks Inc, components. All contrasts derived from the image data
Natick, Massachusetts) and the MATLAB-based CONN-fMRI were corrected for significance using Family-Wise Error
Functional Connectivity toolbox version 17.f29, imple- Rates (FWER) according to Flandin & Friston, (2019) for the
mented in SPM12 (Wellcome Department of Imaging reduction of nominal type I errors. More details are pro-
Neuroscience, London, UK; www.fil.ion.ucl.ac.uk/spm), vided in Table 1.
functional images were aligned to the first volume of the The MATrix LABoratory program (MATLAB) was used to
time series using a six-parameter rigid body spatial trans- analyze functional connectivity. Specifically, for each
formation and least-squares minimization in combination subject, we extracted the time-series BOLD signal fluctu-
with an unwarping algorithm aimed at correcting motion ations from the above-described ROIs with the different
and motion-related distortions. Slice-timing correction was components of the DMN. Next, we computed a region-by-
then applied. ART-based automatic volume outlier detec- region correlation matrix using Pearson correlation coeffi-
tion (www.nitrc.org/projects/artifact_detect/) was also cients for each pair of ROIs. Autocorrelations (rxy for
run for later scrubbing. Likewise, both functional and struc- x = y) and anticorrelations (rxy < 0) were eliminated
tural images were subjected to simultaneous gray matter, from this correlation matrix. Moreover, correlations were
white matter, and cerebrospinal fluid segmentation, and a transformed to partial correlations by eliminating the
bias correction was performed to remove smoothly varying effects of years of schooling. For this, the means and
intensity differences across images. Such image segments standard deviations of the observed distribution of years
were subsequently spatially normalized through nonlinear of schooling were estimated, and distributions adjusted
transformations to the Montreal Neurological Institute to these values were simulated to obtain a plausible

3

rdia-Olmos, C. Soriano-Mas, L. Tormo-Rodríguez et al.
J. Gua

Table 1 Regions of Interest and Their Names.

DMNa DMNv DMNp

#ROI Name of Region #ROI Name of Region #ROI Name of Region

1 Insula Left 5 Cingulum Post Left 13 Parietal Superior Left
2 Insula Right 6 Cingulum Post Right 14 Parietal Superior Right
3 Cingulum Ant Left 7 Hippocampus Left 15 Parietal Inferior Left
4 Cingulum Ant Left 8 Hippocampus Right 16 Parietal Inferior Right
23 Lob Temp Med Left 9 Circ ParaHippo Left 21 Temporal Med Left
24 Lob Temp Med R. 10 Circ ParaHipp Right 22 Temporal Med Right
11 Gyrus Fusiform Left
12 Gyrus Fusiform Right
17 Angular Gyrus Left
18 Angular Gyrus Right
19 Precuneus Left
20 Precuneus Right

estimate of the effect of years of schooling. This proce- individuals with very low anxiety levels from those with very
dure is based on the proposal of Ponsoda et al. (2017). high anxiety levels (low anxiety group [mean§SD]: 65.25§
The initial expression is the following: 5.0; high anxiety group [mean§SD]: 68.80§3.30. The low

 anxiety group was made up of the five individuals with the
rxy rxz ¢ ryz
rxy=z ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : lowest state anxiety scores, while the high anxiety group
1 rxz2 ¢ 1 ryz2 included the four individuals with the highest state anxiety
ratings. Obviously, the sample sizes do not allow for inferen-
Moreover, at the group level, we computed a correlogram
tial estimates, but the extreme groups have been kept small
by averaging the subject-level correlation matrices. We also
in order to maximize the differences between them, and
conducted a clustering analysis that allowed classifying, for
these results should be interpreted for a more descriptive
each group, the different ROIs based on a graph theory
than inferential purpose. To improve the robustness of the
approach, effectively quantifying the degree of functional
tests, all variability estimates have been carried out using
connectivity of each ROI (vertex) with its neighbors (Watts &
bootstrap estimates according to Turner, Paul, Miller and
Strogatz, 1998). Cluster analysis is often used to study func-
Barbey (2018) and are of special cliical interest.
tional connectivity (Shakil et al., 2014). We used a hierarchi-
cal clustering analysis to construct two models based on the
connectivity distance, using the Euclidean distance between
each pair of vectors. This distance exists between two points Results
in a Euclidean space (full vector with internal product). To
optimize the visualization of these results, dendrograms Analysis of sociodemographic and clinical data
(graphs of ROI groupings) were computed for each group.
Finally, due to the objectives of the study, the group of Table 2 displays the descriptive statistics of the study sam-
depressed patients was split into two different groups ple. Groups did not differ in gender [x2 = 0.487; df= 1;
according to their level of state anxiety prior to MRI. We p=.487] or age [t = 0.42; df= 54; p= .68]. Conversely, sig-
estimated two cutoff points from both the mean and the nificant differences were observed in the different clini-
standard deviation of the MDD group to discriminate cal variables (i.e., HDRS, GDSY, MMSE, vocabulary, STAI-S,

Table 2 Descriptive characteristics of the participants.

Control LLD p value
x (SD) x (SD)
Age 67,7 (4,23) 68,2 (4) .68
Years of Education 13,17 (4,33) 6,93 (3,69) < 0.0001
GDS 0,9 (1,24) 5,85 (4,44) < 0.001
MMSE 28,93 (1,4) 26,7 (2,28) < 0.001
VocWAIS 45,45 (8,53) 29,11 (7,62) <0.001
IQ 111,62 (9,81) 101,3 (8,15) <0.0001
HDRS 0,86 (1,19) 11,59 (7,27) <0.0001
STAI S 7,79 (5,91) 24,48 (13,78) <0.0001
STAI T 11,31 (6,32) 29,81 (13,46) <0.0001
Note: x: Mean; SD: Standard Deviation.

4
 International Journal of Clinical and Health Psychology 22 (2022) 100317

Table 3 Clustering between groups.

Note: L (Left), R(Right). If the cell is green, then there is a good lateralization. If the cell is red, there is a bad lateralization when
clustering.

5

rdia-Olmos, C. Soriano-Mas, L. Tormo-Rodríguez et al.
J. Gua

Fig. 1 Cluster analysis between groups. A: Clustering of the control group; B: Clustering of the LLD group; C: Clustering of the low
anxiety group; D: Clustering of the high anxiety group.

and STAI-T). We also observed significant differences were extracted, with a small distance range (0.7 1.6), in
between groups in years of schooling [t = 5.8; df = 54; p patients with LLD, only 5 clusters with a greater distance
<0.001], although, as described above, the effect of range (0.8 1.8) were extracted. Additionally, as shown in
this variable was removed from data before statistical Table 3, control group clusters were, in general, more later-
analyses. alized than clusters observed in patients with LLD.

Between-group differences in correlation values Low anxiety group vs. high anxiety group

We observed that the average of pairwise correlations in the We repeated the above analyses contrasting the subgroups
control group was 0.212 (SD = 0.22), while in the MDD group, of subjects with low and high anxiety derived from the LLD
this value was 0.181 (SD = 0.24). Although these values were group. In Table 3 and Fig. 1, the results of the grouping by
not significantly different between the study groups anxiety are also shown. The average pairwise correlation in
[t = 1.558; df = 550; p = 0.120], when estimating the differ- low anxiety subjects was 0.218 (SD = 0.26) and 0.179
ence between the group-level correlation means (i.e., (SD = 0.23) in high anxiety subjects. This difference was not
0.212 0.181 = 0.031), this value showed a 95% confidence significant [t = 1.914; df = 540.682; p = .056]. Nevertheless,
interval ranging between 0.01 and 0.05. Although this is a similar to what we observed when comparing MDD patients
subtle between-group difference, typical of the type of sig- to healthy controls, the 95% confidence interval of the dif-
nal used in the study, this range of correlation difference ference between these mean values (0.218 0.179 = 0.039)
values indicates that average correlation values in healthy ranged between 0.01 and 0.07, indicating that this differ-
controls are consistently higher than average pairwise corre- ence in correlation values may reach 7 correlation units in
lations in MDD patients. the population of origin of the samples, with low anxiety
subjects showing larger pairwise correlation values across
Cluster analyses the different ROIs.
In the cluster analysis, we observed that regions clus-
Table 3 displays the results of the clusters between groups. tered differently in low- and high-anxiety subjects, as shown
Moreover, Fig. 1 displays the graphical representation. The in Fig. 1. While anatomical ROIs were grouped into 6 clusters
clusters of ROIs within the DMN differed between the study in low anxiety subjects, with a distance range between 0.6
groups. Thus, while 6 clusters in the healthy control group and 1.8, in high anxiety subjects, we only observed 5

6
 International Journal of Clinical and Health Psychology 22 (2022) 100317

clusters with a distance range between 0.8 and 1.8. Finally, account for the differences between studies assessing DMN
low-anxiety subjects showed greater lateralization than connectivity in MDD samples.
high-anxiety individuals. As mentioned before, alterations in the DMN have been
found in several mental illnesses, such as neurodegenerative
disorders or dementias. In this sense, the findings of our
study are aligned with others dealing with disorders such as
Discussion mild cognitive impairment, Alzheimer’s disease, or autism.
In all cases, a disruption in the DMN is found, always in the
The aim of this paper was to assess abnormal DMN activity in sense of decreased connectivity (Farras-Permanyer et al.,
an LLD group compared to an age- and sex-matched group of 2019). Therefore, this network seems to be of great impor-
healthy controls. Moreover, a hierarchical clustering analysis tance across mental health disorders. Although the DMN
was performed, first comparing LLD patients with healthy function was initially associated with spontaneous neural
controls and thereafter dividing the late-life depression activity during resting periods (Raichle et al., 2001), recent
group by anxiety symptoms. The results of our study show studies suggest that DMN activation is important for differ-
the value of the methods used to discriminate patients with ent the cognitive processes involved in abstract tasks,
LLD vs. age- and sex-comparable healthy controls. Specifi- including reading comprehension or generating mental con-
cally, patients with MDD displayed lower interregional corre- tent using information from memory (Zhang et al., 2022).
lations, as well as a decreased specialization of the different Our findings suggest these domains may be preferentially
areas of the DMN, as reflected by a fewer number of clus- altered in disorders showing DMN alterations at the neural
ters, which were also less lateralized. Finally, high anxiety level.
levels contribute to such alterations since the pattern Our study is not without limitations. First, we assessed a
observed in patients with low anxiety levels resembles that relatively small sample of patients with LLD, although the
of healthy controls, while patients with high anxiety levels number of subjects recruited for this research was similar to
display the pattern of alterations characterizing the whole previous studies. Second, and probably related to the first
MDD group (i.e., decreased interregional correlations and point, some of our analyses did not reach statistically signifi-
decreased regional specialization). cant between-group differences. Nevertheless, these same
The combination of lower interregional pairwise correla- analyses were significant when using alternative significance
tions with an equally decreased number of clusters within testing approaches, such as the estimation of confidence
the DMN suggests that this network is less efficiently orga- intervals. Thirdly, we did not assess the correlations of DMN
nized in patients with LLD than in healthy controls. This find- regions with other brain networks. Although such compari-
ing concurs with previous findings indicating that patients sons are beyond the scope of this manuscript, future
with MDD show both increased static functional connectivity research studies may want to assess such potential internet-
and decreased variability within the DMN (Demirtaş et al., work connectivity alterations, including subcortical net-
2016). We observed that activity within the DMN is organized works, to further characterize network-level disruptions in
in larger bilateral clusters, although, at the same time, the LLD at the whole-brain level. Likewise, we did not exclude
lack of major fluctuations in brain activity may reduce the individuals with vascular and metabolic conditions (i.e.,
strength of interregional correlations in pairwise analyses. hypertension or diabetes) which show a very high prevalence
This weaker intramodular functional connectivity has also within the age range of the individuals assessed here.
been recently reported in a review of graph-theory Although this is true both for the control and the MDD
approaches to network-level organization in MDD (Yun groups, it is also true that some of this health conditions
&Kim, 2021). Nevertheless, other studies found opposite may show higher prevalence in individuals with MDD (Alexo-
findings (Eyre et al., 2016); therefore, although it seems poulos, 2019), and, therefore, a potential effect on our find-
clear that connectivity within the DMN is altered in MDD, ings cannot be ruled out. Finally, the inclusion of a group of
further research is warranted to elucidate the clinical varia- individuals with anxiety, but no mood, disorders would have
bles that may be significantly modulating resting-state activ- allowed to clarify the effects of anxiety on our findings.
ity and interregional connectivity within the DMN in MDD Such comparison is warranted for future research.
samples. In sum, this is, to our knowledge, the first study of late-
One such variable may be anxiety levels, which we life depression using a clustering approach, and the results
observed significantly modified the pattern of DMN alter- appear to be promising. In this sense, our results show that
ation in patients with MDD, since patients with low anxiety patients with LLD show a less efficiently organized DMN,
levels did not differ from controls in the DMN connectivity with a decrease in pairwise interregional correlations and
and clustering assessments. Anxiety itself is known to more extended intranetwork clusters. Moreover, anxiety
decrease connectivity within the DMN (Northhoff, 2020; seems to significantly contribute to such alterations. These
Tumati, Paulus & Northoff, 2021) and between the DMN and results can help to further characterize the neurobiological
other resting-state networks (Xu et al., 2019), while comor- correlates of MDD in the elderly and should allow the specific
bid anxiety symptoms are indeed associated with a worse comparison with samples of younger patients with MDD and
outcome of depression (Maarsingh, Heymans, Verhaak, Pen- older patients with cognitive impairments. Such compari-
ninx & Comijs, 2018) and specific neurobiological alterations sons should eventually permit the identification of the brain
(Laird et al., 2019) in the elderly. Therefore, it seems that functional alterations linking depression with neurodegener-
varying anxiety levels in MDD samples may significantly ative disorders, thus allowing the development of treatment
affect functional connectivity within the DMN and disorder strategies specifically targeting the brain networks where
severity. Likewise, these anxiety levels may also partially such alterations have been detected.

7

rdia-Olmos, C. Soriano-Mas, L. Tormo-Rodríguez et al.
J. Gua

Acknowledgments Habes, M., Pomponio, R., Shou, H., Doshi, J., Mamourian, E.,
Erus, G., , & iSTAGING consortium, the Preclinical AD con-
The authors would like to thank the study participants sortium, the ADNI, and the CARDIA studies. (2021). The
and the staff from Bellvitge University Hospital and Insti- Brain Chart of Aging: Machine-learning analytics reveals
tut de Diagno
stic per la Imatge (IDI) who contributed to links between brain aging, white matter disease, amyloid
recruiting the study sample. This study was supported by burden, and cognition in the iSTAGING consortium of 10,216
harmonized MR scans. Alzheimer's & Dementia, 17(1), 89–
the Agency for Management of University and Research
102.
Grants of the Catalan Government (2017SGR1247), the Hamilton, M. (1960). A rating scale for depression. Journal of Neu-
Carlos III Health Institute (Grants PIE14/00034, PI19/ rology, Neurosurgery, and Psychiatry, 23, 56–62. https://doi.
01040 and INT21/00055), the European Regional Develop- org/10.1136/jnnp.23.1.56.
ment Fund (ERDF) "A way to build Europe", and CIBER- Harrison, B. J., Pujol, J., Lo pez-Sola

, M., Herna ndez-Ribas, R.,
SAM. We also thank CERCA Programme/Generalitat de Deus, J., Ortiz, H., et al. (2008). Consistency and functional spe-
Catalunya for institutional support and Ministerio de cialization in the default mode brain network. In Proceeding of
Ciencia, Innovacion y Universidades, Agencia Estatal de National Academy of Sciences (pp. 9781 9786). https://doi.
n. Grant Number: PGC2018 095829-B-I00.
Investigacio org/10.1073/pnas.0711791105.
Huang, C. C., Hsieh, W. J., Lee, P. L., Peng, L. N., Liu, L. K.,
Lee, W. J., et al. (2015). Age-Related Changes in Resting-State
Networks of A Large Sample Size of Healthy Elderly. CNS Neuro-
science and Therapeutics, 21(10), 817–825. https://doi.org/
References 10.1111/cns.12396.
Jacob, Y., Morris, L. S., Huang, K. H., Schneider, M., Rutter, S.,
Alexopoulos, G. S. (2019). Mechanisms and treatment of late-life Verma, G., et al. (2020). Neural correlates of rumination in
depression. Translational Psychiatry, 9(1), 188. major depressive disorder: A brain network analysis. Neuro-
Andreescu, C., Sheu, L. K., Tudorascu, D., Walker, S., & image Clinical, 25, 102142. https://doi.org/10.1016/j.nicl.
Aizenstein, H. (2014). The ages of anxiety—Differences across 2019.102142.
the lifespan in the default mode network functional connectivity Laird, K. T., Siddarth, P., Krause-Sorio, B., Kilpatrick, L.,
in generalized anxiety disorder. International Journal of Geriat- Milillo, M., Aguilar, Y., et al. (2019). Anxiety symptoms are
ric Psychiatry, 29(7), 704–712. associated with smaller insular and orbitofrontal cortex vol-
Beekman, A. T., De Beurs, E., Van Balkom, A. J., Deeg, D. J., umes in late-life depression. Journal of Affective Disorders,
Van Dyck, R., & Van Tilburg, W. (2000). Anxiety and depression in 256, 282–287.
later life: Co-occurrence and communality of risk factors. Ameri- Maarsingh, O. R., Heymans, M. W., Verhaak, P. F.,
can Journal of Psychiatry, 157(1), 89–95. Penninx, B. W. J. H., & Comijs, H. C. (2018). Development and
Behzadi, Y., Restom, K., Liau, J., & Liu, T. T. (2007). A component- external validation of a prediction rule for an unfavorable course
based noise correction method (CompCor) for BOLD and perfu- of late-life depression: A multicenter cohort study. Journal of
sion-based fMRI. NeuroImage, 37(1), 90–101. https://doi.org/ Affective Disorders, 235, 105–113.
10.1016/j.neuroimage.2007.04.042. Mancho-Fora, N., Montala
-Flaquer, M., Farra
s-Permanyer, L.,
Damoiseaux, J. S., Beckmann, C. F., Sanz Arigita, E. S., Barkhof, F., Zarabozo-Hurtado, D., Gallardo-Moreno, G. B.,
Scheltens, P., Stam, C. J., et al. (2008). Reduced resting-state Gudayol-Farre , E., et al. (2020). Network change point detection
brain activity in the “default network” in normal aging. Cerebral in resting-state functional connectivity dynamics of mild cogni-
Cortex, 18(8), 1856–1864. tive impairment patients. International Journal of Clinical and
Demirtaş, M., Tornador, C., Falco n, C., Lo pez-Sola
, M., Health Psychology, 20(3), 200–212. https://doi.org/10.1016/j.
Herna ndez-Ribas, R., Pujol, J., et al. (2016). Dynamic functional ijchp.2020.07.005.
connectivity reveals altered variability in functional connectivity Manning, K., Wang, L., & Steffens, D. (2019). Recent advances in the
among patients with major depressive disorder. Human Brain use of imaging in psychiatry: Functional magnetic resonance
Mapping, 37(8), 2918–2930. imaging of large-scale brain networks in late-life depression.
Eyre, H. A., Yang, H., Leaver, A. M., Van Dyk, K., Siddarth, P., F1000Research, 8. https://doi.org/10.12688/f1000research.
Cyr, N. S., et al. (2016). Altered resting-state functional connec- 17399.1.
tivity in late-life depression: A cross-sectional study. Journal of Northoff, G. (2020). Anxiety Disorders and the Brain’s Resting State
Affective Disorders, 189, 126–133. https://doi.org/10.1016/j. Networks: From Altered Spatiotemporal Synchronization to Psy-
jad.2015.09.011. chopathological Symptoms. In: Y. K. Kim (eds) Anxiety disorders.
Farras-Permanyer, L., Mancho-Fora, N., Montala
-Flaquer, M., advances in experimental medicine and biology, vol 1191.
Bartres-Faz, D., Vaque
-Alca
zar, L., Pero
-Cebollero, M., et al. (2019). Springer, Singapore. https://doi.org/10.1007/978-981-32-9705-
Age-related changes in resting-state functional connectivity in 0_5
older adults. Neural Regeneration Research, 14(9), 1544. Patek, A. X., Kundu, P., Rubinov, M., Jones, P. S., Vertes, P. E.,
Flandin, G., & Friston, K. J. (2019). Analysis of family-wise error Ersche, K. D., et al. (2014). A wavelet method for modeling and
rates in statistical parametric mapping using random field the- despiking motion artifacts from resting-state fMRI time series.
ory. Human Brain Mapping, 40(7), 2052–2054. https://doi.org/ NeuroImage, 95, 287–304. https://doi.org/10.1016/j.neuro-
10.1002/hbm.23839. image.2014.03.012.
Gandelman, J. A., Albert, K., Boyd, B. D., Park, J. W., Riddle, M., Ponsoda, V., Martínez, K., Pineda-Pardo, J. A., Abad, F. J., Olea, J.,
Woodward, N. D., et al. (2019). Intrinsic functional network con- Roma n, F. J., et al. (2017). Structural brain connectivity and cog-
nectivity is associated with clinical symptoms and cognition in nitive ability differences: A multivariate distance matrix regres-
late-life depression. Biological Psychiatry: Cognitive Neurosci- sion analysis. Human Brain Mapping, 38(2), 803–816. https://
ence and Neuroimaging, 4(2), 160–170. doi.org/10.1002/hbm.23419.
Geng, J., Yan, R., Shi, J., Chen, Y., Mo, Z., Shao, J., et al. (2019). Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J.,
Altered regional homogeneity in patients with somatic depres- Gusnard, D. A., & Shulman, G. L. (2001). A default mode of brain
sion: A resting-state fMRI study. Journal of Affective Disorders, function. In Proceeding of the National Academy of Sciences USA
246, 498–505. (pp. 676 682). https://doi.org/10.1073/pnas.98.2.676.

8
 International Journal of Clinical and Health Psychology 22 (2022) 100317

Shakil, S., Magnuson, M. E., Keilholz, S. D., & Lee, C. (2014). Cluster Wei, Y., de Lange, S. C., Scholtens, L. H., Watanabe, K.,
- based analysis for characterizing dynamic functional connectiv- Ardesch, D. J., Jansen, P. R., et al. (2019). Genetic mapping and
ity. 36th Annual International Conference of the IEEE Engineer- evolutionary analysis of human-expanded cognitive networks.
ing in Medicince and Biology Society (pp. 982 985). Nature Communications, 10(1), 1–11.
Sheehan, D. V., Lecrubier, Y., Sheehan, K. H., Amorim, P., Wen, J., Fu, C. H. Y., Tosun, D., Veturi, Y., Yang, Z.,
Janavs, J., Weiller, E., et al. (1998). The Mini-International Neu- Abdulkadir, A., et al., , & iSTAGING consortium, ADNI, BIOCARD,
ropsychiatric Interview (MINI): The development and validation and BLSA. (2022). Characterizing Heterogeneity in Neuroimag-
of a structured diagnostic psychiatric interview for DSM-IV and ing, Cognition, Clinical Symptoms, and Genetics Among Patients
ICD-10. Journal of Clinical Psychiatry, 59(20), 22–33. With Late-Life Depression. JAMA psychiatry e220020. https://
Sheikh, J. I., & Yesavage, J. A. (1986). Geriatric Depression Scale doi.org/10.1001/jamapsychiatry.2022.0020.
(GDS): Recent evidence and development of a shorter version. Wise, T., Marwood, L., Perkins, A. M., Herane-Vives, A., Joules, R.,
Clinical Gerontologist: The Journal of Aging and Mental Health, Lythgoe, D. J., et al. (2017). Instability of default mode network
5(1 2), 165–173. https://doi.org/10.1300/J018v05n01_09. connectivity in major depression: A two-sample confirmation
Smith, J. M., & Alloy, L. B. (2009). A roadmap to rumination: A study. Translational Psychiatry, 7(4). https://doi.org/10.1038/
review of the definition, assessment, and conceptualization of tp.2017.40 e1105-e1105.
this multifaceted construct. Clinical Psychology Review, 29, 116– Xu, J., Van Dam, N. T., Feng, C., Luo, Y., Ai, H., Gu, R., et al. (2019).
128. Anxious brain networks: A coordinate-based activation likelihood
Spielberger, C. D. (1983). State-Trait Anxiety Inventory for Adults estimation meta-analysis of resting-state functional connectivity
(STAI-AD) [Database record]. APA PsycTests. https://doi.org/ studies in anxiety. Neuroscience & Biobehavioral Reviews, 96,
10.1037/t06496-000. 21–30.
Tumati, S., Paulus, M. P., & Northoff, G. (2021). Out-of-step: Brain- Xue, C., Yuan, B., Yue, Y., Xu, J., Wang, S., Wu, M., et al. (2019).
heart desynchronization in anxiety disorders. Molecular Psychia- Distinct disruptive patterns of default mode subnetwork connec-
try, 26, 1726–1737. https://doi.org/10.1038/s41380-021-01029-w. tivity across the spectrum of preclinical Alzheimer’s disease.
Turner, B. O., Paul, E. J., Miller, M. B., & Barbey, A. K. (2018). Small Frontiers in Aging Neuroscience, 11, 307–321. https://doi.org/
sample sizes reduce the replicability of task-based fMRI studies. 10.3389/fnagi.2019.00307.
Communications Biology, 1(1), 1–10. https://doi.org/10.1038/ Yesavage, J. A., Brink, T. L., Rose, T. L., Lum, O., Huang, V.,
s42003-018-0073-z. Adey, M., et al. (1982). Development and validation of a geriatric
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., depression screening scale: A preliminary report. Journal of Psy-
Etard, O., Delcroix, N., et al. (2002). Automated anatomical chiatric Research, 17(1), 37–49.
labeling of activations in SPM using a macroscopic anatomical Yun, J. Y., & Kim, Y. K. (2021). Graph theory approach for the struc-
parcellation of the MNI MRI single-subject brain. NeuroImage, 15 tural-functional brain connectome of depression. Progress in
(1), 273–289. https://doi.org/10.1006/nimg.2001.0978. Neuro-Psychopharmacology and Biological Psychiatry, 111,
Van Den Heuvel, M. P., & Pol, H. E. H. (2010). Exploring the brain e110401.
network: A review on resting-state fMRI functional connectivity. Zhang, M., Bernhardt, B. C., Wang, X., Varga, D.,
European Neuropsychopharmacology, 20(8), 519–534. Krieger-Redwood, K., Royer, J., et al. (2022). Perceptual cou-
Van Dijk, K. R. A., Hedden, T., Venkataraman, A., Evans, K. C., pling and decoupling of the default mode network during mind-
Lazar, S. W., & Buckner, R. L. (2010). Intrinsic functional connec- wandering and reading. eLife, 11, e74011. https://doi.org/
tivity as a tool for human connectomics: Theory, properties, and 10.7554/eLife.74011.
optimization. Journal of Neurophysiology, 103(1), 297–321. Zhang, Y., Chen, Y., & Ma, L. (2018). Depression and cardiovascular
https://doi.org/10.1152/jn.00783.2009. disease in elderly: Current understanding. Journal of Clinical
Wang, R., Liu, M., Cheng, X., Wu, Y., Hildebrandt, A., & Neuroscience, 47, 1–5.
Zhou, C. (2021). Segregation, integration, and balance of large- Zhou, C., Zemanova , L., Zamora, G., Hilgetag, C. C., &
scale resting brain networks conFig. different cognitive abilities. Kurths, J. (2006). Hierarchical organization unveiled by func-
Proceedings of the National Academy of Sciences, 118(23). tional connectivity in complex brain networks. Physical Review
Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of “small- Letters, 97,(23) e238103.
world” networks. Nature, 393(6684), 440–442. https://doi.org/ Zugman, A., Harrewijn, A., Cardinale, E. M., Zwiebel, H.,
10.1038/30918. Freitag, G. F., Werwath, K. E., et al. (2022). Mega-analysis meth-
Wechsler, D. (1999). WISC-III: Wechsler intelligence scale for chil- ods in ENIGMA: The experience of the generalized anxiety disor-
dren (finnish version).Psykologien Kustannus. der working group. Human Brain Mapping, 43(1), 255–277.