Normalization of aberrant resting state functional connectivity in
ﬁbromyalgia patients following a three month physical exercise therapy
, M. Löfgren
, I. Bileviciute-Ljungar
, P. Fransson
Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden
Department of Rheumatology and Inﬂammation Research, Institute of Medicine, Sahlgrenska Academy, Gothenburg University, Gothenburg, Sweden
Department of Clinical Sciences, Danderyd Hospital, Karolinska Institutet, Stockholm, Sweden
Received 1 June 2015
Received in revised form 29 July 2015
Accepted 6 August 2015
Available online 18 August 2015
Resting state fMRI
Physical exercise is one of the most efﬁcient interventions to mitigate chronic pain symptoms in ﬁbromyalgia
(FM). However, little is known about the neurophysiological mechanisms mediating these effects. In this study
we investigated resting-state connectivity using functional magne tic resonance imaging (fMRI) before and
after a 15 week standardized exercise program supervised by physical therapists. Our aim was to gain an under-
standing of how physical exercise inﬂuences previously shown aberrant patterns of intrinsic brain activity in FM.
Fourteen FM patients and eleven healthy controls successfully completed the physical exercise treatment. We
investigated post- versus pre-treatment changes of brain connectivity, as well as changes in clinical symptoms
in the patient group. FM patients reported improvements in symptom severity. Although several brain regions
showed a treatment-related change in connectivity, only the connectivity between the right anterior insula
and the left primary sensorimotor area was signiﬁcantly more affected by the physical exercise among the ﬁbro-
myalgia patients compared to healthy controls. Our results suggest that previously observed aberrant intrinsic
brain connectivity patterns in FM are partly normalized by the physical exercise therapy. However, none of the
observed normalizations in intrinsic brain connectivity were signiﬁcantly correlated with symptom changes. Fur-
ther studies conducted in larger cohorts are warranted to investigate the precise relationship between improve-
ments in ﬁbromyalgia symptoms and changes in intrinsic brain activity.
© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
Fibromyalgia (FM) is a condi tion characterized by widespread
chronic pain and is often accompanied by cognitive dysfunction and fa-
tigue. The pathogenesis is still largely unknown, but there is evidence of
both peripheral and central pathophysiology (for a review, see Clauw,
2014). The diagnosis is currently based on self-reported pain, and as of
yet, no laboratory test can directly test for FM. Intriguingly however,
several brain imaging studies of FM patients indicate an altered func-
tional brain structure which is related to aberrant pain evoked brain
activation , particularly in the anterior cingulate cortex and thalamus
(Jensen et al., 2013). Resting state brain connectivity constitutes
a promising complement to the traditional task-based fMRI studies
employing subtraction designs that disregard the brain activity reﬂecting
sustained pain states. Resting state fMRI studies have reported altered
intrinsic brain activity in FM patients including: decreased connectivity
between insula and prefrontal areas (Ichesco et al., 2014) and the
periaqueductal grey (PAG) (Pujol et al., 2014), increased connectivity
between insula and medial regions of the default mode network (DMN)
(Napad ow et al., 2010), and decreased connectivity between somatosen-
sory regions and visual and auditory cortices (Pujol et al., 2014). Using a
comprehensive set of analytical approaches to characterize intrinsic
brain activity in FM, we recently showed a decreased connectivity
between pain-related and sensorimotor brain areas during rest (Flodin
et al., 2014).
Physical exercise is a potent treatment for FM, on par with the efﬁ-
ciency of cognitive behavioural therapy, education in coping strategies
and phar macological in terve ntions (Clauw, 2014). In th is study, we
aimed to investigate if physical exercise could normalize the previously
described aberrant patterns of intrinsic brain connectivity in FM and if
this was related to symptom improvement. To our knowledge, this is
the ﬁrst study to investigate the longitudinal effects of physical exercise
NeuroImage: Clinical 9 (2015) 134–139
Abbreviations: FM, ﬁbromyalgia; PAG, periaqueductal grey; FIQ, Fibromyalgia Impact
Questionnaire; SF36BP, bodily pain subscale of the Short Form Health Survey.
* Corresponding author at: Department of Clinical Neuroscience, Karolinska Institutet,
Nobels väg 9, Stockholm SE-171 77, Sweden. Tel.: +46 709 243172.
E-mail address: parﬂodin@gmail.com (P. Flodin).
2213-1582/© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ynicl
on intrinsic brain activity in FM. Our hypothesis was that the physical
training program would inﬂuence the previously detected deviant con-
nectivity patterns in six pairs of brain ROI-to-ROI (Region-of-Interest)
connectivity in FM patients (Flodin et al., 2014). We further aimed to in-
vestigate the extent to which longitudinal changes in connectivity cor-
related with changes in symptoms.
Subject recruitment and inclusion criteria were identical to our base-
line study (Flodin et al., 2014). All FM patients satisﬁed the American
College of Rheumatology (ACR) 1990 disease criteria for FM. Of the 16
female subjects included in the baseline study, two subjects failed to
comply with the longitudinal study protocol. The mean age of the re-
maining fourteen subjects was 48.4 (range 25–64) years (all female).
The mean symptom gravity (assessed with the Fibromyalgia Impact
Questionnaire (FIQ), see Bennett, 2005 for reference) was 60.8 (SD =
11.8), and the mean FM duration was 7.3 years ( SD = 4.0). Among
the 22 female subjects originally inc luded a s healthy controls in the
baseline study, nine subjects were only scanned once, fMRI data from
one subject was discarded due to excessi ve head-motion, and fMRI
data from one sub ject was rejected due to technical failures in the
image acquisition process. Thus, fMRI res ting-state data from eleven
healthy controls were included (mean age 41.8 years, range 20–63).
2.2. Training intervention
A 15-week exercise program with two sessions each week was carried
out under supervis ion from a physio therapi st (PT). Before the part icipants
started the intervention, they had individual meetings with a PT who test-
ed their one repetition maximum (1 RM) and tolerance before deciding
the initial load of each exercise. At the same time the participants received
individual instructions for each exercise. Each session lasted for about 1 h
and included 10 min of warming up by ergometer cycling, isometric exer-
cises for the deep muscles in the back and stomach, and concentric and
non-concentric exercises for the legs, back, stomach, arms and hands.
The program ended with stretching exercises.
The participants3 individual 1 RM for the different exercises were
tested before starting and at three time points during the program. For
the legs and arms the initial loads were set at approximately 40% of
one 1 RM and the participants were instructed to repeat each exercise
15–20 times in one to three sets within symptom tolerance. In between
each set, the participants rested for at least 45 s. After 5 weeks the load
was raised up to about 50% of 1 RM with 2 sets of 12–15 repetitions,
after 8 weeks up to about 60% of 1 RM with 2 sets of 10–12 repetitions
and after 12 weeks up to about 70% of 1 RM with 2 sets of 8–10 repeti-
tions. Leg exercises for explosive strength were also included at weeks
ﬁve and eight.
The participants reported pain and other adverse effects of the exer-
cise program to the PT during every session. In the case that the level of
pain increased without returning back to normal within a few days, the
participants were instructed to lower the load, but continue to do the
exercise. The same instructions were given if the participants had a
bad day or an increase of symptoms. If any special exercise was causing
problems the participants were instructed to refrain from doing it. The
PT followed up on program compliance; if participants did not partici-
pate in a session they were instructed to give notice with the reason
of absence, on which the physiotherapist made a follow-up phone call.
The overall compliance rate was high, with an average of 29.12 (SD =
3.2) training sessions taken for the 25 participants (for FM: M = 29.4,
SD =1.6;HC:M = 28.7, SD = 4.5). There was no signiﬁcant group dif-
ference with regard to compliance (t(23) = 0.52, p
3. Behavioural measures
We used two behavioural measures to estimate pain and ﬁbromyalgia
symptomatology. We used the bodily pain subscale of the Short Form
Health Survey (SF36BP) for pain assessment, since this assesses pain dur-
ing a 4 week period and thus reﬂects long-term changes, disregarding
temporary ﬂuctuations in pain intensity (Contopoulos-Ionannidis et al.,
2009). In addition, we used the Fibromyalgia Impact Questionnaire
(FIQ), which is a general questionnaire regarding the impact of ﬁbromy-
algia on everyday life (Bennett, 2005).
2.4. MRI data acquisition
Anatomical and functional MR imaging were performed on a 3 T
General Electric 750 MR scanner. For each subject we performed one
resting state scan consisting of 200 volumes, using an echo-planar imag-
ing sequence with TR/TE = 2500/30 ms, ﬂip = 90°, 49 slices, 96 × 96
matrix size, FOV = 288×288 mm, slice thickness = 3 mm using inter-
leaved slice acquisition. In the resting state condition, subjects were
instructed to lie still and rest, and keep their eyes closed and try not to
fall asleep. Prior to the resting state fMRI data acquisition, subjects
underwent two fMRI sessions of a task-evoked pain fMRI paradigm
(approx. 7 min each), and two sessions of an fMRI adopted version of
the Stroop task (approx. 7 min each). Data from the task-evoked fMRI
sessions will be reported elsewhere.
2.5. Resting state fMRI data analysis
In this study we investigated longitudinal changes (post- versus pre-
physical exercise treatment) of functional connectivity of six seed re-
gions located in pain regions as dema rked by a metaanalysis on 314
pain studies carried out in the framework of neurosynth (Yarkoni
et al., 2011), that we previously found to be affected by FM in our base-
line study (FM: n = 16, HC: n = 22, see Flodin et al., 2014). Speciﬁcally,
the spherical Region-of-Interest (ROI, with a radius of 4 mm) were lo-
cated in the insula, sensorimotor cortex and thalamus. A detailed list
of the anatomical location of the ﬁve seed ROIs and the corresponding
six target clusters is given in Table 1. Supplementary Fig. S1 illustrates
the whole brain connectivity maps pertaining to eac h of the seed
regions shown at baseline for both the FM and the HC cohort.
Image preprocessing, seed-based correlation analysis (SCA) and
group-level analyses were carried out in Matlab (Mathworks Inc., Na-
tick, MA, USA). Prior to SCA, imaging data were preprocessed using
the Matlab toolbox SPM12 (Wellcome Trust Centre of Neuroimaging,
University College London, UK). Image preprocessing included slice
time correction to the middle slice, realignment to the mean im age
using the 4th degree of B-spline interpolation, co-registration of func-
tional and structural images, tissue segmentation of structural images,
normalization of structural and functional scans to the MNI template
using the deformation ﬁeld obtained from the segmentation (4th de-
gree B-spline function, resampling to 2 mm isotropic voxels). Finally,
functional volumes were spatially smoothed using an 8 mm FWHM
Gaussian kernel. Subject level SCA analyses were carried out using
the Conn toolbox (
http ://www.nitrc.org/p rojects/conn)(Whitﬁel d-
Gabrieli & Nieto-Castanon, 2012). Functional volumes were band pass
ﬁltered at 0.008–0.09 Hz (default values). Subject speciﬁcnuisancere-
gressors included 6 movement and their time derivatives, and 5 regres-
sors pertaining to white matter and CSF signals respe ctively, using a
component based noise correction (CompCor) approach (Behzadi
et al., 2007). Additionally, images that were regarded as movement out-
liers were regressed out. Outliers were detected using the ART toolbox
http://nitrc.org/projects/ar tifact_detect/) and deﬁned as volumes
with frame wise displacement (FD) larger than 0.5 mm or signal inten-
sity changes greater than 3 standard deviations (default thresholds). For
both pre- and post-treatment fMRI data, estimates for the strength of
135P. Flodin et al. / NeuroImage: Clinical 9 (2015) 134–139
resting-state functional connectivity were obtained from each subject
and for each of the six ROI-to-ROI connectivity pairs.
The computed estimate of the strength of connectivity (β values) be-
tween the seed ROIs and target clusters was subsequently used at a sec-
ond level group model of differences in intrinsic connectivity (Δβ
) (after controlling for age and inter-individual differences in mean
frame-wise displacement, see Flodin et al., 2014). Additionally, we test-
ed for the relationship between changes in functional connectivity and
changes in behaviour (post-minus pre-physical exercise treatment),
using the Pearson correlation coefﬁcient. Since we were interested in
connectivity change s that were speciﬁc to the FM condition (rather
than general training effects independent of group), we compared the
post- versus pre-change in intrinsic brain connectivity in the FM pa-
tients versus the HC cohort (Δβ speciﬁcity = |Δβ
|). Due to
relatively unequal group sizes (14 FM and 11 HC), we conducted per-
mutation tests that are less sensitive to unbalanced designs compared
to parametric t-tests. Thus, for each of the six seed-target connectivity
pairs, we estimated the likelihood of the observed group difference,
given the null hypothesis that there would be no difference in longitudi-
nal change in connectivity, by randomly shufﬂin g the group membershi p
of the subjects (1000 permutations, p-value threshold b0.05, Bonferroni
Pre-treatment, the mean SF36BP score for the FM cohort was
37.00 ± 9.70, (highest subjective health, i.e. no pain = 100) whereas
the post-treatment score was 37.07 ± 11.40 (t(14) = −0.026, p =
0.98). Furthe r, in the FM cohort, the pre-treatment mean FIQ was
60.8 ± 11.8 (maximally severe FM impact = 100), and for post-
training 53.3 ± 29.5 (t(13) = 2.282, p = 0.040, effect size r = 0.53).
The FIQ and SF36BP were negatively correlated both pre- (r = −0.78,
p = 0.00080) and post-treatment (r = −0.79, p = 0.00089) (Fig. 1).
3.2. Intrinsic brain connectivity
Out of the six seed connectivity pairs tested, four showed a signiﬁ-
cant normalization among the FM patients (see Table 1). The observed
changes included an increased connectivity between the right anterior
insula and left primary sensory motor areas, the right supramarginal
gyrus and primary sensorimotor areas, and the right supramarginal
gyrus and left inferior prefrontal cortex. The aberrantly high baseline
connectivity in FM between supramarginal gyrus and cerebellum was
normalized. Noteworthy, also healthy controls showed signiﬁcant lon-
gitudinal connectivity changes. Hence, in order to investigate the extent
to which the change in intrinsic connectivity was selective to the FM
group, we performed permutation tests by which the absolute differ-
ence in magnitude of the connectivity change between the two groups
was compared. In the permutation test, only the change in connectivity
between the insula and primary sensorimotor regions showed a selec-
tive normalization in the FM group (see Table 1 an d Fig. 2). None of
the treatment-related changes in intrinsic connectivity correla ted
signiﬁcantly with the FIQ or SF36BP scores (Table 1).
4. Discussion and conclusion
In this study we have shown that the previously reported abnormal
patterns of resting state connectivity in FM patients are partly normal-
ized after a 3 month schedule of regular physical exercise. Although
several intrinsic brain connectivity patterns underwent longitudinal
change, only the c onnectivity between the right insula and primary
sensorimotor cortex displayed a selectively greater change for FM com-
pared to HC. It should be noted that all the seed regions used in the cur-
rent study belongs to a set of brain areas that commonly are activated in
(seed) ↔ (target)
| t (p)
baseline t (p)
Correlation of Δβ
Insula (40, 24, 8) ↔ S1/M1 (−30, −22, 68) 3.95
Supr. gyr. (60, −36, 28) ↔ S1/M1 (10, −24, 80) 3.12
Supr. gyr. (50, −26, 28) ↔ inferior PFC (−32, 30, − 8) 3.34
Mid cing (0, −36, 28) ↔ occipital (42, −72, −12) 1.94
Thalamus (−10, −16, 8) ↔ premotor (0, 6, 56) 1.44
Supr. gyr. (60, −36, 28) ↔ cerebellum (−42, −74, −34) −3.60
mmary of changes in intrinsic brain connectivity related to physical exercise training and correlations between intrinsic connectivity and the only pain-related parameter that displayed
signiﬁcant longitudinal changes (FIQ) among FM patients.
Out of the six pairs of brain regions tested, four displayed a signiﬁcant longitudinal renormalization in the ﬁbromyalgia cohort. The degree of connectivity was signiﬁcantly changed for the
healthy control between the supramarginal gyrus and cerebellum. Of note, only the change of connectivity between the insula and S1/M1 was relatively signiﬁcantly larger for FM than
that for HC (no change in connectivity was signiﬁcantly larger for the healthy compared to ﬁbromyalgia). The level of statistical signiﬁcance (p = 0.05) was corrected for multiple com-
parisons with regard to the six tested seed regions using Bonferroni correction (p b 0.0083). Statistical signiﬁcant changes are marked with “*”. Abbreviations: Supr. gyr. = supramarginal
gyrus; S1/M1 = primary sensorimotor areas; ΔB = changes in connectivity; t = t-value; p = p-value; r = Pearson correlation coefﬁcient. All coordinates (x, y, z) are in MNI space.
Fig. 1. Average FIQ and SF36BP ratings in 14 FM patients before (solid bars) and following
(striped bars) the exercise intervention. The reduction in FIQ ratings indicates reduced FM
symptoms. No cha nge was observed in pain ratings (SF36BP). The asterisk sign (*)
signiﬁes a signiﬁcant difference at p b 0.01 between post- versus pre-treatment condi-
tions. Error bars denote standard deviations. FIQ = Fibromyalgia Impact Questionnaire,
SF36BP = short form bodily pain subscale.
136 P. Flodin et al. / NeuroImage: Clinical 9 (2015) 134–139
pain studies, as delineated by an automated meta-analysis (described in
Flodin et al., 2014). The insula plays a pivotal role in pain processin g
(Duerden et al., 2013) and activity in its anterior part has been associat-
ed with interoceptive accuracy and subjective feeling (Critchley et al.,
2004). Interestingly, a re cent study in FM patients reported a hypo-
connectivity between the bilateral anterior insula and PAG (peri-
aqueductal grey nucleus), that correlated with symptom severity (Pujol
et al., 2014). Speculatively, the shift towards increased insula connectivity
following physical exercise therapy observed here could reﬂect an in-
creased interaction between brain regions responsible for interoceptive
valiance and bodily representations. The restoration of functional con-
nectivity between the right SMG and primary sensorimotor areas, as
well as between the right SMG and inferior prefrontal cortex, consisted
in reduction of anticorrelations for FM patients.
Whereas functional connectivit y is commonly interpre ted as (at least
partly) reﬂecting a history of co-activation of brain regions interacting in
order to accomplish a speciﬁc (cognitive or housekeeping) function at
a given time, the functional signiﬁcance of anticorrelation is less well
understood. It has previously been shown that anticorrelations can be
artiﬁcially induced during preprocessing of the data, such as by global
mean regression of the fMRI signal intensity time-series (Murphy et al.,
2009). However, using electrophysiological techniques, a neurobiological
origin of anticorrelations has been conﬁrmed (Keller et al., 2013). Further-
more, preprocessing strategies that omit global signal regression, in-
cluding the CompCor strategy that is employed here, commonly render
anticorrel atio ns (Chai et al., 2012). The general functional signiﬁcance of
anticorrelations, especially outside the well studied relationship between
the default mode network and task positive networks (Fox et al., 2005;
Fransson, 2005), remains tentative. One interpretation of the normaliza-
tion (i.e. decreases in anticorrelation) observed here, is that pain- and so-
matosensory regions interact in a more coherent fashion, implying a
reduced mismatch between sensorimotor signals and activity in brain
regions involved for pain processing. Our ﬁnding complement results
from previous FM intervention studies that primarily reported decreased
insular connectivity following FM interventions. For instance, Schmidt-
Wilcke et al. (2014) found that improvements in clinical pain correlated
with decreased functional connectivity between the anterior insula and
ACC following a 6 week long milnacipran treatment. Likewise, the same
research group observed a decreased connectivity between insula
and the default mode network following an acupuncture treatment
(Napadow et al., 2012). These results were interpreted as a restoration
of FM-associated hyperconnectivity between the brain regions involved
in pain perception (anterior insula) and self-referential thought (DMN).
Thus, an emerging picture is that improvements in symptom severity in
FM could be associated either with reduced interaction between regions
sub-serving pain processing and self-related cognition, or by a decreased
asynchr ony betw een the pain and senso rimo tor regio ns. Sinc e the current
study investigated the effects of physical exercise, it is plausible that the
mechanism mediating these symptom improvements to a larger degree
involves the sensorimotor brain areas.
Few studies have directly investigated longitudinal changes of in-
trinsic brain conn ectivity following physical exercise interventions.
Becerra et al. (2014) investigated the effects of a three week physical
and psychological training program in a group of children and adoles-
cents with paediatric complex regional pain syndrome. They reported
treatment effects in several networks (e.g. DMN, salience networks
and cer ebellum). However, signiﬁcance levels were not adjusted for
testing of multiple networks, and their relatively small cohort (12 pa-
tients and an equal number of controls), prevents any strong conclu-
sions on the effect of the intervention. There are severa l studies on
effects of motoric skill training, such as sequential motor training para-
digms with durations of a couple of weeks. A general ﬁndi
these studies is that an initially increased task evoked responses in
both the primary sensory and primary motor cortices (Floyer-Lea &
Matthews, 2005; Hlustík et al., 2004). Interestingly, also resting state re-
gional cerebral blood ﬂow (rCBF, measured with positron emission to-
mography) in the primary motor cortex is increased following motor
training (Xiong et al., 2009). Focusing on resting state connectivity,
Krafft et al. (2014) studied the effects of an 8 month exercise interven-
tion in a cohort of overweight children. They report increased network
synchronicity within the motor network, and decreased synchronicity
within th e DMN and cogn itive cont rol n etworks in the intervention
group relative to controls. Another study reports on the acute effect of
a 20 minute aerobic exercise intervention on restin g state activity
(Rajab et al., 2014). Their primary ﬁnding was an increased co-
activation within the primary and sensorimotor areas.
Although the current study is the ﬁrst to investigate the resting state
changes follow ing an exercise intervention among ﬁbromyalgia pa-
tients, there are several studies coherent with our ﬁnding that physical
exercise primarily inﬂuences functional connectivity within cardinal
The beneﬁcial health effects of physical exercise are well acknowl-
edged. Following the intervention, FM patients rated lower on the FIQ
scale, reﬂecting increased healthiness and a decrease of symptom sever-
ity, physical disability and overall impact of ﬁbromyalgia. However, the
body pain levels (SF36BP) did not change. These ﬁndings indicate that
the therapeutic effects of the 3 month exercise intervention rendered
improvements of symptom s that only indirectly pertained to pain.
Thus, the observed normalizations of functional connectivity involving
sensory areas could possibly reﬂect subclinical therapeutic processes,
likely induced by increased levels of motor activity.
However, since neither SF36BP nor FIQ could be linearly correlated
with change in brain activity in the current study, the behavioural im-
plications of the obse rved normalization in intrinsic connectivity
warrants further investigation. A possible explanation for the absent
brain–behavioural coupling could be that treatment-induced chang-
es in brain connectivity follow an inverted u-shape temporal proﬁle
as proposed by the expan sion–partial n orma liz ation hypoth esis of
Fig. 2. Physical exercise induced normalization of resting state connectivity between the
right insula and the left sensorimotor region in the FM cohort. (A) Intrinsic connectivity
between a spherical seed region (radius = 4 mm) located in the right anterior insula
and a cluster extending 1490 voxels i n the left sensorimotor cortex. (B) Post- versus
pre- treatment insular-sensorimotor connectivity (arbitrary units) for ﬁbromyalgia
(blue) and controls (red). Error bars denote standard deviations.
137P. Flodin et al. / NeuroImage: Clinical 9 (2015) 134–139
neuroplastici ty (Brehmer et al., 2014). However, plotting change in
behavioural scores against change in connectivity did not indicate
any such relationship. Another possible scenario is t hat the behav-
ioural parameters monitored in the present investigation were
subserved by different neuronal processes other than the ones in-
We note several limitations in the current study. First, the resting
state scans were acquired approximately 30 mi n after two task fMRI
sessions were admin istered in the MR scanner (Stroop and evoked
pain, data will be reported elsewhere). Thus, we cannot rule out the
possibility that putative spill-over effects from task-based fMRI para-
digms occur, which cou ld have had a differential inﬂuence on the
group brain activity patterns. Second, since our study did not include a
placebo or no-treatment FM control group, the observed longitudinal
changes in brain connectivity might in part be driven by spontaneous
improvements in pain severity and accompanying changes in brain
connectivity that is not attributable to the physical intervention per se
(i.e. “regression towards the mean”, Stigler, 1997). Third, changes in in-
trinsic brain connectivity might be unrelated to the measured improve-
ments in pain (at least in relation to the pain measures used here).
Theoretically, the observed connectivity changes could be accompanied
by changes in other commonly reported FM symptoms such as mood
changes, insomnia, fatigue or impairments in cognitive function, al-
though the fact that the seed regions were placed in pain-related
brain areas supports the interpretation that there is a link between the
physical intervention and pain relevant brain processes. Finally, we sug-
gest that our ﬁndings should be treated to some extent as preliminary in
nature, given the relatively small number of participants that completed
the post-therapy imaging session. Future studies using larger study
samples would likely enable a more detailed investigation of the rela-
tionship between changes at the neurophysiological versus the behav-
ioural level. Addition ally, it might be desirable for fut ure studies to
record brain connectivity and behaviour at multiple time-points and
thereby achieve a temporally more ﬁne-gr ained characterization of
the longitudinal neurophysiological and behavioural changes related
to exercise therapy in FM patients.
In conclusion, we have shown that a therapeutic physical exercise-
training program increases the perceived degree of healthiness in patients
diagnosed with ﬁbromyalgia. Interestingly, we observed a restoration of
functional connectivity between several pain and sensorimotor regions.
In this context, it is important to note that longitudinal change of intrinsic
connectivity between the sensorimotor system and the insula was
observed exclusively in the FM cohort. This ﬁnding suggests that the
insula connectivity may play a key role for the observed physical
therapy-induced effects on FM symptomatology.
Supplementary data related to this article can be found online at
Conﬂicts of interest
This study was supported b y grants f rom the Swedish Rheuma-
tism Association, Stockholm County Council - 2010, Swedish Foun-
dation for Strategic Research (2012-0179), Swedish Research
Council K2013-5 2X-22199-01-3, and Karolinska Institut e Founda-
tion (2012FoBi34779). The funders had no role in the study design,
data collection and analysis, decision to publish, or preparation of
The authors would like to thank all the participants in this study.
Becerra, L., Sava, S., Simons, L.E., Drosos, A.M., Sethna, N., Berde, C., Lebel, A.A., Borsook, D.,
2014. Intrinsi c brain networks normalize with treatment i n pediatric complex
regional pain syndrome. Neuroimage Clin. 6, 347–369. http://dx.doi.org/10.1016/j.
Behzadi, Y., Restom, K., Liau, J., Liu, T.T., 2007. A component based noise correction meth-
od (CompCor) for BOLD and perfusion based fMRI. Neuroi mage 37 ( 1), 90–101.
Bennett, R., 2005. The Fibromyalgia Impact Questionnaire (FIQ): a review of its develop-
ment, current version, operating characteristics and uses. Clin. Exp. Rheumatol. 23 (5
Suppl 39), S154–S16216273800.
Brehmer, Y., Kalpouzos, G., Wenger, E., Lövdén, M., 2014. Plasticity of brain and cognition
in older adults. Psychol. Res. 78 (6), 790–802. http://dx.doi.org/10.1007/s00426-014-
Chai, X.J., Castañón, A.N., Ongür, D., Whitﬁeld-Gabrieli, S., 2012. Anticorrelations in resting
state networks without global signal regression. Neuroimage 59 (2), 1420–1428.
Clauw, D.J., 2014. Fibromyalgia: a clinical review. JAMA 311 (15), 1547–1555. http://dx.
Contopoulos-Ioannidis, D.G., Karvouni, A., Kouri, I., Ioannidis, J.P.A., 2009. Reporting and
interpretation of SF-36 outcomes in randomised trials: systematic review. BMJ 338,
Critchley, H.D., Wiens, S., Ro tshte in, P., Ohman, A., Dolan, R.J., 2004. Neural systems
supporting interoceptive awareness. Nat. Neurosci. 7 (2), 189–195. http://dx.doi.
Duerden, E.G., Albanese, M.-C., 2013. Localization of pain-related brain activation: a meta-
analysis of neuroimaging data. Hum. Brain Mapp. 34 (1), 109–149. http://dx.doi.org/
Flodin, P., Martinsen, S., Löfgren, M., Bileviciute-Ljungar, I., Kosek, E., Fransson, P., 2014. Fi-
bromyalgia is associated with decreased connectivity between pain- and sensorimo-
tor brain areas. Brain Connect. 4 (8), 587–594. http://dx.doi.org/10.1089/brain.2014.
Floyer-Lea, A., Matthews, P.M., 2005. Distinguishable brain activation networks for short-
and long-term motor skill learning. J. Neurophysiol. 94 (1), 512–518.
Fox, M.D., Snyder, A.Z., Vincent, J.L., Corbetta, M., Van Essen, D.C., Raichle, M.E., 2005. The
human brain is intrinsically organized into dynamic, anticorrelated functional net-
works. Proc. Natl. Acad. Sci. U. S. A. 102 (27), 9673–9678. http://dx.doi.org/10.1073/
Fransson, P., 2005. Spontaneous low-frequency BOLD signal ﬂuctuations: an
fMRI investigation of the resting-state default mode of brain function hy-
pothesis. Hum. Brain Mapp. 26 (1), 15–29. http://dx.doi.org/10.1002/hbm.
Hlustík, P., Solodkin, A., Noll, D.C., Small, S.L., 2004. Cortical plasticity during three-week
motor skill learning. J. Clin. Neurophysi ol. 21 (3), 180–191. http://dx.doi.org/10.
Ichesco, E., Schmidt-Wilcke, T., Bhavsar, R., Clauw, D.J., Peltier, S.J., Kim, J., Napadow, V.,
Hampson, J.P., Kairys, A.E., Williams, D.A., et al., 2014. Altered resting state connectiv-
ity of the insular cortex in individuals with ﬁbr
omyalgia. J. Pain 15 (8), 815–826.e1.
Fransson, P., Marcus, H., Williams, S.C.R., et al., 2013. Overlapping structural
and functional brain changes in patients with long-term exposure to ﬁbromyal-
gia pain. Arthritis Rheum. 65 (12), 3293–3303. http://dx.doi.org/10.1002/art.
Keller, C.J., Bickel, S., Honey, C.J., Groppe, D.M., Entz, L., Craddock, R.C., Lado, F.A.,
Kelly, C., Milham, M., Mehta, A.D., 2013. Neurophysiological investigation of
spontaneous correlated and ant icorrelated ﬂuctuations of the BOLD signal.
J. Neurosci. 33 (15), 6333–6342. http://dx.doi.org/10.1523/JNEUROSC I.4837- 12.
Krafft, C.E., Pierce, J.E., Schwarz, N.F., Chi, L., Weinberger, A.L., Schaeffer, D.J., Rodrigue, A.L.,
Camchong, J., Allison, J.D., Yanasak, N.E., et al., 2014. An eight month randomized con-
trolled exercise intervention alters resting state synchrony in overweight children.
Neuroscience 256, 445 –455. http://dx.doi.org/10.1016/j.neuroscience.2 013.09.
Murphy, K., Birn, R.M., Handwerker, D.A., Jones, T.B., Bandettini, P.A., 2009. The impact of
global signal regression on resting state correlations: are anti-correlated networks in-
troduced? Neuroimage 44 (3), 893–905. http://dx.doi.org/10.1016/j.neuroimage.
Napadow, V., Kim, J., Clauw, D.J., Harris, R.E., 2012. Decreased intrinsic brain connectivity
is associated with reduced clinical pain in ﬁbromyalgia. Arthritis Rheum. 64 (7),
Napadow, V., LaCount, L., Park, K., As-Sanie, S., Clauw, D.J., Harris, R.E., 2010. Intrinsic brain
connectivity in ﬁbr omyalgia is associated with chronic pain intensity. Arthritis
Rheum. 62 (8), 2545–2555. http://dx.doi.org/10.1002/art.2749720506181.
Pujol, J., Macià, D., Garcia-Fontanals, A., Blanco-Hinojo, L., López-Solà, M., Garcia-Blanco, S.,
Poca-Dias, V., Harrison, B.J., Contreras-Rodríguez, O., Monfort, J., et al., 2014. The
contribution of sensory system functional connectivity reduction to clinical pain in ﬁ-
bromyalgia. Pain 155 (8), 1492–1503. http://dx.doi.org/10.1016/j.pain.2014.04.
Rajab, A.S., Crane, D.E., Middleton, L.E., Robert son, A.D., Hampson, M., MacIntosh, B.J.,
2014. A single session of exercise increases connectivity in sensorimotor-related
brain networks: a resting-state fMRI study in young healthy adults. Front. Hum.
Neurosci. 8, 625. http://dx.doi.org/10.3389/fnhum.2014.0062525177284.
Schmidt-Wilcke, T., Ichesco, E., Hampson, J.P., Kairys, A., Peltier, S., Harte, S., Clauw, D.J.,
Harris, R.E., 2014. Resting state connectivity correlates with drug and pl acebo
138 P. Flodin et al. / NeuroImage: Clinical 9 (2015) 134–139
response in ﬁbromyalgia patients. Neuroimage Clin. 6, 252–261. http://dx.doi.org/10.
Stigler, S.M., 1997. Regression toward the mean, historically considered. Stat. Methods
Med. Res. 6 (2), 103–114. http://dx.doi.org/10.1191/0962280976763614319261910.
Whitﬁeld-Gabrieli, S., Nieto-Castanon, A., 2012. Conn: a functional connectivity toolbox
for correlate d and anticorrelated brain networks. Brain Connect. 2 (3), 125–141.
Xiong, J., Ma, L., Wang, B., Narayana, S., Duff, E.P., Egan, G.F., Fox, P.T., 2009. Long-term
motor training induced changes in regional cerebral blood ﬂow in both task and rest-
ing states. Neuroimage. 45 (1), 75–82.
Yarkoni, T., Poldrack, R.A., Nichols, T.E., Van Essen, D.C., Wager, T.D., 2011. Large-scale
automated synthesis of human functional neuroimaging data. Nat. Methods 8 (8),
139P. Flodin et al. / NeuroImage: Clinical 9 (2015) 134–139