Age-related differences in functional connectivity associated with pain modulation

Growing evidence suggests that aging is associated with impaired endogenous pain modulation, and that this likely underlies the increased transition from acute to chronic pain in older individuals. Resting-state functional connectivity (rsFC) offers a valuable tool to examine the neural mechanisms behind these age-related changes in pain modulation. RsFC studies generally observe decreased within-network connectivity due to aging, but its relevance for pain modulation remains unknown. We compared rsFC within a set of brain regions involved in pain modulation between young and older adults and explored the relationship with the efficacy of distraction from pain. This revealed several age-related increases and decreases in connectivity strength. Importantly, we found a significant association between lower pain relief and decreased strength of three connections in older adults, namely between the periaqueductal gray and right insula, between the anterior cingulate cortex (ACC) and right insula, and between the ACC and left amygdala. These findings suggest that the functional integrity of the pain control system is critical for effective pain modulation, and that its function is compromised by aging.


Introduction
Given current demographic trends, there is growing clinical and empirical interest in the impact of aging on pain perception and modulation, and the underlying neural mechanisms (Cruz-Almeida & Cole, 2020;Gagliese, 2009;Lautenbacher et al., 2017).Both acute pain and chronic pain conditions are disproportionally prevalent among older adults (Lautenbacher, 2012;Fayaz et al., 2016;Domenichiello and Ramsden, 2019), and their prevalence increases with age (Dahlhamer et al., 2018;Domenichiello and Ramsden, 2019).One potential mechanism underlying the increased risk of chronic pain in older adults is an age-related impairment in endogenous pain modulation.Endogenous pain modulation is the body's innate capacity to regulate nociceptive transmission to ensure that the experienced pain is adapted to specific internal and environmental demands.This capacity is mediated through a complex set of mechanisms, involving several cortical and subcortical brain areas (i.e., prefrontal cortex (PFC), anterior cingulate cortex (ACC), insula, thalamus and amygdala) with connections via the brainstem (in particular, the periaqueductal gray (PAG) and rostral ventromedial medulla (RVM)) to the spinal cord, that can either facilitate or inhibit afferent nociceptive signals (Fields et al., 2005;Tracey & Mantyh, 2007).These pathways are referred to as the "descending pain modulation system" (Tracey & Mantyh, 2007).Affective and cognitive factors (such as emotions, memories, expectations, beliefs, attention or reappraisal) can influence pain by triggering these pathways (Wiech et al., 2008).
Many chronic pain conditions are characterized by impaired pain modulation, and it is increasingly clear that augmented pain facilitation and/or impaired pain inhibition may precede the development of chronic pain (Ossipov et al., 2014;Staud, 2012;Bushnell et al., 2013;Yarnitsky, 2015;Yarnitsky et al., 2008).Growing evidence now suggests that aging is also associated with impaired pain modulation, and that this likely underlies the increased transition from acute to chronic pain in older individuals.The majority of these studies relied on conditioned pain modulation (CPM), a paradigm commonly used to assess endogenous pain inhibition, and have demonstrated a decline in CPM efficiency with aging (see Hackett et al., 2020 for a review).
Despite accumulating evidence, the mechanisms behind the agerelated changes in endogenous pain modulation remain poorly understood.One promising approach is to examine the functional integrity of brain regions involved in pain modulation by measuring the restingstate functional connectivity (rsFC).This modality of functional magnetic resonance imaging (fMRI) data acquisition measures fluctuations in activity across separate brain regions and detects patterns of temporal intercorrelation between the functioning of these regions (Biswal et al., 1995;van den Heuvel and Hulshoff Pol., 2010).The rationale behind this approach is that the brain, rather than being inactive during rest, displays highly coherent intrinsic activity that is organized in networks and indicative of healthy brain functioning (Buckner et al., 2008;Fox & Raichle, 2007).Unlike typical fMRI studies, rsFC studies do not require participants to perform any specific task, and they are quick, reliable and replicable (Guo et al., 2012;Shehzad et al., 2009), which makes the method particularly feasible in older adults (Ferreira & Busatto, 2013).
The most consistent finding in studies investigating the influence of aging on the rsFC of the brain is that of reduced functional connectivity within large-scale resting state networks in older compared to younger adults, with age-related increases also observed but less frequently (Ferreira & Busatto, 2013;Deery et al., 2023).The within-network decreases in rsFC are most pronounced for higher-order functional networks (e.g., default mode, fronto-parietal and executive control networks), and tend to be more variable for primary sensory and motor networks, as well as for sub-cortical and attention networks (Deery et al., 2023).Many of the areas involved in pain modulation are amongst the regions commonly identified as showing age-related decreases in rsFC, in particular, the prefrontal cortex (PFC), anterior cingulate cortex (ACC), amygdala and insula.These regions have also been implicated in rsFC changes in different chronic pain conditions (Farmer et al., 2012;Pfannmöller & Lotze, 2019;Thorp et al., 2018).However, since these regions are involved in a myriad of other cognitive and affective functions, the relevance of decreased connectivity for pain modulation in healthy aging remains unknown.So far, very few studies have examined specific pain-related functional connectivity changes in healthy aging.In one previous study, we have demonstrated reduced rsFC, amongst others, between some regions that are part of the pain modulation system in healthy older adults (González Roldán et al., 2020a).However, this study did not explore how this change in rsFC translates to the capacity to modulate pain.
In the current study, our aim was to investigate the relationship between pain inhibition efficacy and rsFC within a set of regions that are involved in pain modulation, in older adults.To assess pain modulation, we focused on distraction from pain as a cognitive pain inhibition process.Successful distraction from pain depends on the pain modulation system and some recent evidence suggests that distraction is less efficient in older adults (Zhou et al., 2015;Rischer et al., 2022;González Roldán et al., 2020b).We recruited healthy young and older adults who underwent a resting-state fMRI scan and a pain distraction task involving simultaneous thermal pain stimuli and a cognitive task.Functional connectivity analysis was performed using seed regions involved in pain modulation, including prefrontal areas, the periaqueductal gray (PAG), rostral ventromedial medulla (RVM), thalamus, insula, anterior cingulate cortex (ACC), and amygdala.We expected that older adults would show decreased rsFC within this network as compared to young adults, and crucially, that this decrease would be associated with a smaller pain distraction effect.

Participants
A total of 64 participants took part: 32 young (13 male, 26.7 ± 4.2 years, range: 19-35) and 32 older adults (17 male, 68.2 ± 7.0 years, range: 60-85).A smaller subsample of the current sample has been described previously in Rischer et al. (2022).Participants were recruited through advertisement at the University of Luxembourg (targeted at regular students and senior guest students), interviews in local media and advertisements at organizations and services for senior citizens in Luxembourg.
All participants were in good health and had normal or corrected-tonormal vision.All of them were of Caucasian race, apart from 3 young adults of Asian race.One young and 2 older participants reported to take medication for hypothyroidism, 9 older adults were on antihypertensives (5 received an angiotensin receptor blocker (ARB), one received a selective beta-blocker (SBB), one a combination of ARB and SBB, one a combination of ARB and a calcium channel blocker (CCB), and one a combination of ARB, SBB, and CCB) and 5 older adults were on a very low dose of antidepressants (4 selective serotonin reuptake inhibitors (SSRIs) and one serotonin-noradrenaline reuptake inhibitor (SNRI)).Ten young and 20 older participants had drunk coffee on the day of the MRI scan, but most drank this right after getting up and well before functional imaging.None of the participants smoked on the day of the scan.Exclusion criteria were the presence of psychiatric or neurological disorders, substance or alcohol abuse, acute or chronic pain, and injuries or large tattoos on the arms where pain stimuli were applied.Furthermore, none of the participants presented with contraindications for an MRI scan (i.e., epilepsy, claustrophobia, pregnancy, metal implants, heart pacemakers or insulin pumps).They were requested not to take any pain medication or other drugs known to have an impact on sensory perception or cognition prior to the experimental session.
Participants received a compensation of €40 (in form of gift vouchers) for their time and effort and had the chance to win an additional gift voucher worth €80 in a prize draw.The study was conducted in accordance with the Declaration of Helsinki and ethical approval was obtained from the Ethics Review Panel of the University of Luxembourg and the Luxembourg national ethics committee for research (CNER).All participants gave their informed consent at the start of the study.

General procedure
The study took place at the Robert Schuman hospital in Luxembourg.In a single session, we collected structural MRI images as well as functional images during a 6-minute eyes-open rest period and during a pain distraction paradigm (task-related fMRI data are reported in Rischer et al., 2022).In the distraction paradigm, innocuous warm and painful heat stimuli were administered to the participants' left arm while they performed an easy or difficult working memory task (see Section 2.4).Before being installed in the scanner, participants first practiced the pain distraction task and the temperatures for warm and painful stimuli were calibrated to adjust for individual differences in sensitivity (see Section 2.4).During the resting state scan, participants were instructed to relax with their eyes open (looking at a fixation target).Visual stimuli were presented to participants in the scanner on an MRI-compatible LED monitor (Optostim, Cologne, Germany) placed directly behind the bore, and visible to participants via an angled mirror mounted on the head coil.The thermal stimulation was triggered, and the working memory task was presented using E-Prime 2 (Psychology Software Tools Inc, Pittsburgh, PA, United States).Sessions lasted 2 h on average.

Questionnaires
Participants completed several standard questionnaires to record demographic, medical, and psychological characteristics, cognitive status and affective state.In a first general screening form, they indicated their educational level and answered questions about their health status, pain experiences, medication use and nicotine, caffeine and drug consumption on the days before the test session.They also completed the Edinburgh Handedness Inventory (EHI; scale range: -100-100; a higher score indicates more frequent use of the right hand) (Oldfield, 1971).To assess pain-related cognitions, they were given the Fear of Pain Questionnaire (FPQ-III; scale range: 30-150; a higher score indicates more fear of pain) (McNeil and Rainwater, 1998); the Pain Catastrophizing Scale (PCS; scale range: 0-52; a higher score indicates more pain catastrophizing) (Sullivan et al., 1995); and the Pain Vigilance and Awareness Questionnaire (PVAQ; scale range: 0-80; a higher score indicates more vigilance to pain) (McCracken, 1997).
To screen for cognitive impairments, we administered the Mini Mental State Examination (MMSE) (Folstein et al., 1975).And finally, to assess affective state, participants completed the Depression Anxiety and Stress Scale (DASS-42; scale range: 0-126; a higher score indicates more emotional distress) (Lovibond and Lovibond, 1995) as well as the Positive and Negative Affect Schedule (PANAS) (Watson et al., 1988) (scale range for positive and negative affect subscales: 10-50; a higher score indicates more positive or negative affect).
Given the multilingual environment in Luxembourg, validated versions of all questionnaires were available in three languages; the French version of the FPQ-III was partly adapted from a validated short version (Albaret et al., 2004) and the remaining items were translated by a French native speaker.The French version of the PVAQ (Desrochers et al., 2009) was based on a 5-point scale and scores were later transformed to a 6-point scale, to be compatible with the other language versions.

Distraction paradigm
Participants engaged in an n-back task while they were exposed to thermal stimuli.Each trial lasted 20 s and started with the task.The thermal stimulus commenced 4 s after the start of the task, and therefore lasted 16 s.Both task and thermal stimulus ended simultaneously, followed by an interval of 4-8 s and pain ratings (see Figure 1).
In half of the trials, the task was an easy 0-back task, whereas in the other half of the trials, the task was a difficult 2-back task.Thermal stimuli were either innocuous (warm) or painful (hot), and participants received 8 trials for each combination of task difficulty with temperature (i.e.32 trials in total).The different types of trials (pain/easy, pain/ difficult, warm/easy and warm/difficult) were presented in randomized order, with the constraint that no more than two difficult or two painful trials were presented in a row.
Each trial started with the presentation of a cue word that signaled the instruction for the upcoming n-back task (0-back or 2-back).Participants were then presented with a series of letters.In the 2-back condition, they had to indicate for each letter whether it was the same as the letter presented two steps back in the sequence or not.The 0-back task required participants to indicate whether the current letter was an "X" or not.Presentation speed of the letters was individually calibrated to account for age-related differences in perceived task difficulty.
Thermal stimuli were delivered to the lower left forearm, with an MRI compatible 3 × 3 cm Peltier contact stimulator (Somedic, Sweden).They consisted of a plateau phase of 10 s and ramp-up/ramp-down phases of 3 s each.Baseline temperature was set to 34•C.Participants were asked to rate the perceived intensity and unpleasantness of stimuli on 200-point visual analog scales (VAS), with the pain threshold (or neutral (un)pleasantness) in the middle (corresponding to a VAS rating of 100), using a hand-held button box (Current Designs, USA).Temperatures for the warm and painful stimuli were predetermined in a calibration procedure as corresponding to VAS intensity ratings of 60 and 140, respectively.More details about the pain distraction paradigm, can be found in Rischer et al. (2022).

fMRI acquisition
Whole-brain functional images were acquired on a 1.5 T MRI system (Magnetom Aera, Siemens, Germany) with a 20-channel head coil.Before being placed in the scanner, participants were provided with inear hearing protection and headphones to reduce scanner noise and their head position was stabilized with foam cushions to restrict movements.The scanning protocol consisted of the acquisition of a magnetic fieldmap, four functional runs of the distraction paradigm, a 6-minute resting state functional run, several anatomical scans and a diffusion weighted (DTI) scan.Here we only analyzed the functional images acquired during the rest period and high-resolution structural scans.All structural images were assessed for age-related atrophy, vascular alterations, and other brain lesions by a neuroradiologist.

Data analysis 2.6.1. Behavioral data
For all behavioral data analyses, we used IBM SPSS Statistics for Windows, Version 28.0 (Armonk, NY: IBM Corp).Group differences in demographic and behavioral measures were assessed by means of oneway ANOVAs.VAS pain intensity and unpleasantness ratings collected during the distraction paradigm were analyzed to evaluate the magnitude of the distraction effect (i.e., the reduction in perceived pain during the difficult vs the easy task condition).They were submitted to threeway mixed ANOVAs with temperature (warm vs. pain) and task difficulty (easy vs. difficult) as within-subject factors and age group (young vs. older adults) as between-subject factor.Given a group difference in fear of pain, pain catastrophizing and positive affect, and their potential influence on pain modulation (Rischer et al., 2020;Van Damme et al., 2008; McPhee & Graven-Nielsen, 2022), we ran separate ANOVAs adding these scores as covariate.
To quantify the individual degree of distraction-related pain relief, we created a Distraction Effect score from the intensity (DE-I) and unpleasantness ratings (DE-U), by subtracting the average VAS scores of the pain trials in the difficult task condition from those in the easy task condition.A higher score indicates more efficient distraction from pain.These scores were used as covariates to assess potential relationships with functional connectivity results.

Resting state fMRI data processing
Functional connectivity analyses were performed using the CONN toolbox v22a (Nieto-Castanon & Whitfield-Gabrieli, 2021) running under Matlab R2021a (MathWorks Inc, USA).First, functional images were realigned and unwarped to correct for participant motion and susceptibility distortion (the latter by estimating the inhomogeneity  inside the scanner using a double-echo fieldmap sequence and the creation of voxel-displacement maps).We then applied slice-timing correction and performed an outlier detection (ART-based scrubbing) based on the observed global blood oxygenation level-dependent (BOLD) signal and the amount of subject-motion in the scanner.We applied a conservative outlier detection criterion, marking all images with either a framewise displacement above 0.5 mm or global BOLD signal changes above 3 SD as outliers.After that, functional and structural images were coregistered and normalized into Montreal Neurological Institute (MNI) standard space, segmented into grey matter, white matter, and CSF tissue classes, and resampled to 2 mm isotropic voxels following an indirect normalization procedure.Finally, functional data were smoothed using spatial convolution with an 8 mm FWHM Gaussian kernel.
After these preprocessing steps, images were further processed using CONN's default denoising pipeline to correct for residual non-neural variability due to a combination of physiological, outlier and residual subject-motion effects (Nieto-Castanon, 2020).The pipeline implements an anatomical component-based noise correction procedure (aComp-Cor), which uses linear regression to remove confounding effects from the BOLD signal, characterized by white matter timeseries (10 noise components), CSF timeseries (5 components), motion parameters and their first order derivates (12 factors), outlier scans (below 69 factors), session effects and their first order derivates (2 factors), and linear trends (2 factors) within the functional run.Finally, we removed temporal frequencies below 0.01 Hz from the BOLD signal.

Resting state fMRI data analysis
To examine age-related differences in functional connectivity within the pain modulation system, we performed a region of interest (ROI) to ROI analysis.Thirteen anatomical brain regions commonly activated during top-down cognitive pain modulation (e.g., Bingel et al., 2007;Tracey, 2010;Wiech et al., 2008) were included as ROIs, namely: the bilateral insula, thalamus, amygdala, dorsolateral prefrontal cortex (dlPFC) and ventrolateral prefrontal cortex (vlPFC), as well as the periaqueductal gray (PAG), rostral ventromedial medulla (RVM) and dorsal anterior cingulate cortex (ACC).Masks of the bilateral insula, thalamus and amygdala were created using the WFU PickAtlas tool v3.0.5b (Maldjian et al., 2003), based on the AAL atlas (Tzourio-Mazoyer et al., 2002).The bilateral dlPFC ROIs were created using the WFU PickAtlas by combining BA9 and BA46 (Petrides, 2005), as defined in the Talairach Daemon database atlas for Brodmann areas (Lancaster et al., 1997) and applying a 2D dilation factor of 1.Similarly, the left and right vlPFC ROIs were created using the same procedure by combining BA44, BA45, and BA47 (Petrides & Pandya, 2002).A mask of the ACC was created on the basis of the dorsal ACC region (BA24) as defined in the Human Brainnetome Atlas (Fan et al., 2016).The ACC ROI spanned both hemispheres, as functional connectivity studies have demonstrated that left and right ACC subregions showed highly similar patterns of connectivity with the rest of the brain (e.g., Margulies et al., 2007;Yu et al., 2011).The ROI for the PAG was created on the basis of a peak activation likelihood estimate reported in a review by Linnman et al. (2012), as a 7 mm radius sphere centered on MNI coordinates x=1; y=-29; z=-12.Finally, a mask of the RVM was built as a 4 mm sphere centered on MNI coordinates x=0; y=-36; z=-51 (Mills et al., 2018).
We then estimated ROI-to-ROI connectivity matrices, representing the functional connectivity strength between each pair of regions among our 13 ROIs.Functional connectivity strength was characterized as Fisher-transformed bivariate correlation coefficients from a weighted general linear model, estimated separately for each pair of ROIs, defining the association between their BOLD signal timeseries.Analyses of covariance (ANCOVAs) with sex as covariate were performed to examine group differences in connectivity measures.To assess if agerelated changes in functional connectivity were associated with cognitive pain modulation, we added the distraction effect scores (DE-I and DE-U) as second-level covariates (regressors) to the second-level general linear model in the CONN toolbox, included those ROIs that showed a significant group difference, and then performed a regression analysis.This procedure allows testing for any (positive or negative) correlations between a covariate and the first-level connectivity estimates.For those connections showing a significant correlation, we extracted the connectivity estimates from the CONN toolbox, and entered them in excel to create scatter plots for illustration purposes.ROI-level inferences were based on parametric multivariate statistics, and an FDR-corrected threshold of p < 0.05 was applied.
As several of the older participants were on antidepressants and/or antihypertensive medication, which might influence functional connectivity measures (Carnevale et al., 2023;Brakowski et al., 2017), we conducted separate group comparisons and correlation analyses while controlling for medication use.Finally, given a group difference in handedness which could potentially influence functional connectivity results (Tejavibulya et al., 2022) we repeated the group comparisons and correlation analyses while controlling for handedness.

Group characteristics
An overview of demographic and behavioral data is presented in Table 1.Young and older adults did not differ in terms of their MMSE or DASS-42 scores, indicating no age-effect on cognitive status or emotional distress.However, older adults reported less fear of pain and pain catastrophizing, more positive affect, and a stronger right-hand dominance (see Table 1).

Distraction from pain
VAS pain intensity and unpleasantness ratings from the different conditions of the distraction paradigm can be found in Figure 2.For intensity ratings, repeated measures ANOVAs with temperature (warm vs. painful) and task difficulty (easy vs difficult) as within-subject factors, and group (young vs. older) as between-subject factor revealed a significant main effect of temperature [F(1,62) = 220.49,p < 0.001, η 2 p = 0.781], and of task difficulty [F(1,62) = 10.68,p = 0.002, η 2 p = 0.147].Importantly, we also observed a significant interaction between temperature and task difficulty [F(1,62) = 33.59,p < 0.001, η 2 p = 0.351].We found no significant main effect of group, nor any interaction of temperature or difficulty with group (all p >.102).Group-wise bonferroni corrected two-tailed paired sample t-tests (critical p =.0125) revealed that both young and older participants rated painful stimuli significantly lower in intensity when these were presented during the difficult task as compared to the easy task [t(31) = − 2.66, p =.012 and t

Table 1
Overview of demographic and questionnaire data.(31) = − 6.56, p <.001, respectively], which was not the case for warm stimuli (both p >.366) (see Figure 2).Results for the unpleasantness ratings were similar, with a significant main effect of temperature [F(1,62) = 128.05,p < 0.001, η 2 p = 0.674] and interaction between temperature and task difficulty [F(1,62) = 45.88,p < 0.001, η 2 p = 0.425], but no main effect of task difficulty or of group, nor any interaction with group (all p >.075).Group-wise bonferroni corrected two-tailed paired sample t-tests (critical p =.0125) revealed significantly lower unpleasantness ratings for painful stimuli during the difficult vs. the easy task for both young [t(31) = − 2.98, p =.006] as well as older adults [t(31) = − 4.70, p < 0.001], with no differences in ratings for warm stimuli (both p >.025] (see Figure 2).Together, these results indicated a robust distraction effect for painful stimuli, with both age groups showing a similar degree of pain relief.Given that we found group differences in fear of pain, pain catastrophizing and positive affect, and that these could influence pain modulation (Rischer et al., 2020;Van Damme et al., 2008; McPhee & Graven-Nielsen, 2022), we added these scores as covariates to the repeated measures ANOVAs.This did not change the results regarding group effects (all remained non-significant, see Supplementary Materials).

Age-related differences in functional connectivity
Functional connectivity analyses revealed decreased connectivity in older compared to young adults between the right insula and PAG, between the right insula and ACC, and between the right vlPFC and left amygdala.In contrast, older adults demonstrated stronger connectivity than young adults between the left dlPFC/vlPFC and ACC, between the left/right vlPFC and left insula and between the left vlPFC and right insula (see Table 2 and Figure 3).
As around a third of the older participants were on medication that might influence functional connectivity measures (Carnevale et al., 2023;Brakowski et al., 2017), we controlled for medication intake.This changed the results slightly, with only connectivity between insula and PAG remaining significantly stronger in young than older, and only connectivity between insula and vlPFC remaining stronger in older than young participants (see Supplementary Materials).In addition, given the age-difference in handedness, which could potentially influence functional connectivity results (Tejavibulya et al., 2022), we also controlled for EHI scores.Again, this changed the results slightly, with only the insula-PAG connectivity appearing significantly stronger for young than older participants, and the connections between bilateral vlPFC and left insula in the opposite direction (see Supplementary Materials).

Association between functional connectivity and the distraction effect
We then assessed whether any age-related changes in functional connectivity were associated with the degree of pain relief related to distraction.This revealed three connections showing a significant positive association with the distraction effect intensity scores (DE-I) in older adults, namely between the PAG and right insula, between right insula and ACC and between ACC and the left amygdala (see Figure 4).Therefore, decreased connectivity between these regions in older adults is indeed associated with a smaller distraction effect.Controlling for medication use or handedness did not change these results (all three connections still showed a positive correlation, see Supplementary Materials).
We found no functional connections showing an age-related increase in strength which correlated significantly with the distraction effect in the older participants.We also found no significant correlations in young participants, nor between functional connectivity scores and the distraction effect score derived from the unpleasantness ratings (DE-U), in young or older participants.

Decreased functional connectivity in older adults
We revealed several significant reductions in resting-state functional connectivity (rsFC) within a set of regions involved in pain modulation in older adults compared to young adults.While we have previously reported age-related reductions in rsFC specifically between nodes of the pain control system (González Roldán et al., 2020a), this latter study did not demonstrate any relevance of these neural changes for the efficacy of pain modulation.Many of the ROIs included in our analyses are not specific to pain modulation and are involved in a variety of other affective and cognitive functions.Therefore, in the current study we  explicitly probed for any correlations between our connectivity changes and the behavioral distraction effect scores.This revealed several connections showing age-related decreases in strength which were associated with smaller reductions in pain in older adults.This finding confirms that aging is associated with a decline in the functional integrity of the pain control system, and that this decline indeed results in less effective pain inhibition.

Insula-PAG connectivity
The specific functional connections that were found to be altered in older adults all have a clear relevance for the modulation of pain.One pair of ROIs with reduced connectivity strength consisted of the right insula and PAG, and their connectivity strength also correlated with the distraction effect in older adults.The PAG is generally considered one of the most pivotal nodes of the pain modulation system.It is activated in many pain modulation paradigms (Linnman et al., 2012), including during distraction from pain (Tracey et al., 2002;Valet et al., 2004).The insula is part of the 'pain matrix' as well as of a more general salience network, and is involved in interoception and integration of autonomic, sensory, affective and cognitive functions (Uddin et al., 2017).Importantly, the insula also plays a key role in top-down pain inhibition (Lu et al., 2016).The PAG receives direct projections from the insula (Saleh et al., 2017) and functional connections found between the PAG and insula are commonly found during pain modulation.For example, increased PAG-insula connectivity strength was found to be associated with more efficient conditioned pain modulation (CPM) in both fibromyalgia patients and controls (Harper et al., 2018).Abnormal functional connectivity between the insula and PAG has been found in different groups of chronic pain patients (Truini et al., 2016;Xu et al., 2022).Interestingly, pre-stimulus functional connectivity between the insula and PAG was found to differ between physically identical trials that were rated as painful and trials perceived as non-painful (Ploner et al., 2010).In particular, a stronger insula-PAG connectivity predicted the occurrence of no pain trials, and suggests that insula-PAG connectivity is crucial in the ability to inhibit (impending) pain.Our results extend these findings by demonstrating the involvement of insula-PAG connectivity in the efficacy of distraction from pain.They also reveal that in older adults, the insula-PAG connection was disturbed, and suggest that this weakening might be associated with a reduced capacity to modulate susceptibility to (impending) pain.

ACC-insula connectivity
A second connection we identified with reduced strength in older adults is that between the anterior cingulate cortex (ACC) and right insula.Again, connectivity strength between these regions in older adults correlated with the distraction effect.The ACC preferentially  encodes the affective-motivational component of pain perception (Xiao & Zhang, 2018;Rainville et al., 1997), but is also crucial for pain modulation (Tracey & Mantyh, 2007;Wiech et al., 2008).Anatomical and functional connectivity studies have demonstrated that the insula and posterior ACC form the core of a system involved in the integration of interoceptive with affective information (Taylor et al., 2008).In the context of pain modulation, connectivity between the anterior insula and dorsal ACC/MCC has been suggested to encode the perceived threat value or salience of impending pain to guide perceptual decision-making (Wiech et al., 2010).Functional connectivity between ACC and insula is also commonly observed to be altered in chronic pain conditions (Ichesco et al., 2012;Malinen et al., 2010;Kim et al., 2017), and in one study it was negatively correlated with clinical pain intensity (Ichesco et al., 2012).In our study, the positive correlation between the distraction effect magnitude and ACC-insula connectivity strength in older adults highlights the involvement of this connection in the attentional modulation of pain.The age-related decrease in ACC-insula connectivity strength thus points to a possible reduced ability to accurately encode the salience or affective content of pain stimuli, resulting in maladaptive perceptual decisions.

vlPFC-amygdala connectivity
A third connection showing decreased strength in older adults was between the right vlPFC and left amygdala, which both again have very specific roles in the processing of pain.The vlPFC is a key player in the cognitive modulation of pain, in particular in encoding the controllability and perceived threat of pain (Wiech et al., 2006;Salomons et al., 2007).Self-controlled versus externally controlled pain stimuli were characterized by increased activity in the right vlPFC, and this activity was negatively correlated with perceived pain intensity (Wiech et al., 2006).The right vlPFC is also consistently activated during emotion regulation paradigms, especially during the use of reappraisal (Ochsner & Gross, 2008).This region has, therefore, been suggested to play a central role in the cognitive modulation of pain based on reappraisal of the emotional significance of a stimulus (Wiech et al., 2006;2008).Consistent with our finding, the mechanism through which the vlPFC is able to reduce pain is likely via top-down voluntary regulation of emotional processing in the amygdala (Ochsner & Gross, 2008).The amygdala is involved in affective and autonomic responses to pain and in the release of endogenous opioids (Bagley & Ingram, 2020), and has been described as a crucial subcortical node for endogenous pain inhibition (Rouwette et al., 2012).Successful pain inhibition through cognitive reappraisal has indeed been shown to result in parallel downregulation of amygdala activity (Lapate et al., 2012).Consistent with the exertion of an inhibitory influence on the amygdala by the vlPFC, strong functional connectivity between these two areas has been shown in different contexts, especially in studies employing emotional processing paradigms (Kerestes et al., 2017).In the context of pain, it has been demonstrated that participants who perceived control over pain showed increased functional connectivity between amygdala and vlPFC (Salomons et al., 2015).The age-related reduction in vlPFC-amygdala connectivity in our study thus suggests that older adults may have an impaired ability to reappraise the negative affect or threat value of pain stimuli, and thereby to down-regulate an affective and autonomic response to pain.However, this ability may not have been directly relevant for distraction from pain, given the absence of a correlation between vlPFC-amygdala connectivity strength and the distraction effect.

ACC-amygdala connectivity
Finally, we demonstrated that stronger connectivity between the dorsal ACC and amygdala was associated with a greater distraction effect in older adults.Similar to the vlPFC, the dorsal ACC is also implicated in the down-regulation of amygdala responsivity in the context of emotion processing (Ochsner & Gross, 2005;Modinos et al., 2010;Kim & Hamann, 2007), supported by the presence of strong anatomical connections between these two regions (Vogt & Pandya, 1987).Several studies have also found similar inhibitory influences of ACC on amygdala in the context of pain.Stronger ACC activation was demonstrated during the experience of pain in a negative compared to a positive emotional context, and during this condition, the ACC was functionally connected to the amygdala (Yoshino et al., 2010).The ACC has also been shown to covary with activity in bilateral amygdala during placebo analgesia (Bingel et al., 2006), suggesting the involvement of this connection in expectation-driven pain modulation.Further, in one study, longer pain exposure, during which participants reported using more coping strategies, was associated with greater activity in the ACC, together with reduced activity in the amygdala (Petrovic et al., 2004).And finally, connectivity between the ACC and amygdala has been demonstrated to be reduced in chronic pain, e.g., fibromyalgia (Jensen et al., 2012).ACC-amygdala connectivity thus appears involved in various forms of pain modulation, including affective and expectation-driven modulation, and is related to spontaneous coping strategies.Our results extend these findings by demonstrating a role for ACC-amygdala connectivity in distraction from pain.

Increased functional connectivity in older adults
We also identified several ROI pairs with increased connectivity strength in older compared to young adults, in particular, between the dorso-and ventrolateral PFC on the one hand, and the ACC and insula on the other.Observations of age-related increases in rsFC are less common than decreases, and tend to be more frequent for connectivity within lower-order (compared to higher-order) networks such as primary sensory and motor networks, or sub-cortical and attention networks (Deery et al., 2023).Increases in rsFC in older vs young adults are also regularly found in studies investigating between-network connectivity of large-scale networks (Deery et al., 2023).Since we found no correlations between age-related increases in connection strength and the distraction effect, the specific connections with increased strength in older versus young do not seem directly relevant for pain modulation.However, the ACC and insula are also part of the salience network, whereas the dlPFC and vlPFC are associated with executive control networks.Therefore, our results would be consistent with the pattern of increased between-network connectivity in aging.Greater connectivity between the salience network and executive control networks with aging has been observed before (Archer et al., 2016).
Observations of reduced within-network combined with greater between-network connectivity in older adults are consistent with the dedifferentiation theory of aging (Park et al., 2004).This theory describes the loss of functional specialization in the brain with increasing age, which results in a decrease in efficiency of neural processes and may explain the decline of cognitive abilities in aging.Differentiation/specialization can be operationalized in terms of the ratio of within-network to between network connectivity (i.e., the degree of segregation) in the brain, with dedifferentiation being characterized by a lower ratio or decreased segregation, which is indeed what is found in older adults (Koen et al., 2020).
Another possible interpretation of the age-related increases in rsFC is that aging preferentially affects long-range connections, while isolated regions in close proximity may display increased local connectivity (Tomasi and Volkow, 2012;Meunier et al., 2009).This fits with the increase we observed, especially between the vlPFC and insula, which are indeed adjacent brain regions.However, since we found no correlation between increased connection strengths and the distraction effect, we cannot draw any conclusions about the functional significance of these increases.Further research is needed to elucidate the underlying mechanisms and functional implications of increased rsFC with aging.

Influence of age on behavioural distraction effect
Despite the alterations in functioning of the pain modulation system and their association with pain distraction efficacy in older adults, we did not observe a group difference in the degree of pain relief in the pain distraction paradigm.This discrepancy between aging effects on behavioural and neural distraction efficacy aligns with a previous study on distraction from pain in aging (González Roldán et al., 2020b), and is likely due to a recruitment bias.Older participants in both studies were relatively highly educated (despite a group difference in years of education), motivated and healthy.Some of the older adults from the present sample, and all of the older adults in the other study, were enrolled as senior guest students at the university where the study was conducted.They also reported lower fear of pain and pain catastrophizing, and more positive affect than young adults in the current sample.And finally, we excluded individuals with chronic pain and psychiatric or neurological disorders.Our older participants may thus have been particularly healthily aging individuals, showing few signs of neurocognitive decline, and therefore, having relatively preserved endogenous pain modulation capacity.In accordance with this, we previously demonstrated that the degree of impairment in pain modulation is related to cognitive functioning, especially executive functions (Rischer et al., 2020;2022).However, despite a preserved behavioural pain modulation effect, in the current study we demonstrated clear age-related alterations in functional connectivity underlying pain modulation, as we did previously in terms of electrophysiology (González Roldán et al., 2020b).Therefore, rsFC and event-related potentials may be more sensitive to early signs of aging, when behavioural changes are still absent.Future studies should attempt to overcome this recruitment bias and select a more representative sample of older adults, or specifically focus on older individuals with cognitive complaints, to explore if the neurological alterations are also reflected in behavioural changes in these populations.

Implications and future directions
Distraction is a common, easily implemented, low-cost and intuitive pain coping strategy.Older adults suffering from chronic pain prefer to use distraction over conventional interventions such as pharmacological treatment, exercise or physiotherapy (Lansbury, 2000).Distraction is also increasingly employed in clinical settings, for example using virtual reality technology, or as part of cognitive behavioral therapy (Levy et al., 2010;Malloy and Milling, 2010;Vázquez-Rivera, et al., 2009).Our findings emphasize the importance for clinicians to be aware of the age-related reduction in the functional integrity of the neural network underlying distraction, and take this into account when considering the employment of pain treatment strategies mediated by these processes.An increased reliance on pharmacological interventions may be required for adequate pain treatment.
Since abnormal functioning of the pain modulation system has been associated with the development and maintenance of chronic pain (Ossipov et al., 2014;Staud, 2012;Bushnell et al., 2013;Yarnitsky, 2015;Yarnitsky et al., 2008), our results may also help to explain the high prevalence of chronic pain conditions in the aging population (Lautenbacher, 2012;Fayaz et al., 2016;Domenichiello & Ramsden, 2019).Whereas impaired functioning of the pain control system has been speculated as a potential reason for this increased risk in older adults, our study is the first to provide more direct evidence for this link.Further research is needed to explore potential interventions that may enhance pain inhibition processes in older adults.A better understanding of the neural mechanisms underlying age-related changes in pain modulation helps to guide this development of more targeted treatment strategies in older individuals.

Limitations
We collected functional MRI images on a 1.5 T scanner, and our acquisition parameters were not specifically optimized for targeting small brain structures (such as our brainstem ROIs).Even though this does not compromise our results (acquisition using these parameters is still common and robust), it should be taken into account in future research.In particular, we recommend corroboration of our findings by studies striving to adhere to the most recent scan protocol guidelines, such as those of the Human Connectome Project (e.g., Harms et al., 2018), including using higher magnetic field strength, longer scan duration and smaller voxel sizes.This would increase signal quality and may further improve the efficacy of post-acquisition noise reduction.
Despite the common finding of a decrease in connectivity strength between the ACC and insula in older vs young adults, there is a lack of further correspondence between the current study and our previous study investigating pain-related functional connectivity changes due to aging (González Roldán et al., 2020a).This could be explained by several methodological differences between the studies, amongst others, more rigorous artifact/noise reduction in the current study (including an unwarping step during preprocessing and application of a more conservative outlier detection criterion for movement-related noise).Additionally, different sample characteristics between the two studies may have contributed to subtle differences in findings.In particular, the young participants' age range went up to 26 only in the previous study, whereas we included participants up to 35 in the current study.As a consequence, the age differences found in the previous study may have been more pronounced.Furthermore, the older participants in the previous study may have been less healthy than in the current study, as evidenced by a more significant medication intake in the previous study: 77 different types of medication were taken (in a group of 37 older adults), including anti-hypoglycemics, anxiolytics, anti-hypertensives and antidepressants.A greater general age-related decline may have been associated with more pronounced age-related differences in functional connectivity.Given the general paucity of research on this topic, we strongly advocate for further studies, allowing for meta-analyses to be conducted and more solid conclusions to be drawn.
Around a third of the older (but none of the young) participants in the current study were on medication that might influence functional connectivity measures, such as antidepressants and antihypertensive drugs (Carnevale et al., 2023;Brakowski et al., 2017).In addition, handedness, which may also influence functional connectivity (Tejavibulya et al., 2022), differed significantly between groups, with older participants reporting more frequent use of the right hand.When adding medication status and handedness scores as covariates, the age-related differences in rsFC were only partially replicated, suggesting that medication intake and handedness indeed influenced our results slightly.However, at least part of this loss of significant results can be explained by the fact that both medication status and handedness strongly covaried with age.Therefore, removing any variance related to these covariates unavoidably also removes true age-related differences.This unintended removal of meaningful variance when controlling for covariates is a common problem in neuroimaging studies (Hyatt et al., 2020).In support of this interpretation, the results regarding correlations between connectivity strength and the distraction effect within age groups did not change when controlling for medication or handedness.However, future studies are necessary to either systematically investigate the effect of these factors, or preselect participants in order to control for their influence.

Conclusion
We report age-related reductions in rsFC within a set of central nervous system regions involved in pain modulation (in particular, involving the PAG, insula, ACC, vlPFC and amygdala), and demonstrate their functional relevance by showing an association with the efficacy of distraction from pain.Our data thus confirm that the integrity of this system is critical for effective pain modulation, and suggest that the decreased connectivity observed in older adults points to a disruption in the coordination of pain inhibition processes in aging, possibly involving functions such as the capacity to modulate susceptibility to (impending) pain, the ability to accurately encode the salience or affective content of pain stimuli, and the ability to reappraise the negative affect or threat value of pain stimuli.Importantly, our results also highlight the feasibility of resting-state fMRI scans in providing brain functional connectivity markers for the efficacy of pain modulation in older adults.These markers may be useful in predicting the functioning of the pain control system, which would have clear clinical benefits.Since this type of scan is non-invasive, fast and does not require any active task performance, these markers are particularly suitable for older adults or individuals with cognitive problems or chronic pain.

Fig. 1 .
Fig. 1.Trial timeline Visual representation of the trial procedure of the distraction paradigm.

M
.van der Meulen et al.

Fig. 2 .
Fig. 2. Intensity and unpleasantness ratings for thermal stimuli.Beeswarm plots of the intensity (A) and unpleasantness (B) VAS ratings for warm and painful stimuli in the easy and difficult task condition, for young and older adults.Each grey dot represents one data point.The solid red dot represents the mean and the whiskers represent the spread (in quartiles).*p <.05.**p <.01.***p <.001.

Fig. 4 .
Fig. 4. Association between connectivity strength and distraction effect size.Scatterplots illustrating the association between the distraction effect intensity scores (DE-I) and functional connectivity strength between the ACC and the left amygdala (A), between the ACC and the right insula (B) and between the PAG and the right insula (C), in young (yellow triangles) and older participants (green circles).**p <.01, *p <.05.

Table 2
Functional connectivity differences between young and older adults.