Beyond the chronic pain stage: default mode network perturbation depends on years lived with back pain

Default mode network perturbation reflects an adaptation to years lived with back pain and related cognitive processes rather than simply reflecting a pain classification status.

Supplemental Digital Content is Available in the Text.

Abstract

Research has indicated that the default mode network (DMN) is perturbated in patients with chronic pain when compared with healthy controls, and this perturbation is correlated with the duration of pain during the chronic pain stage. It remains unclear whether DMN adaptations manifest during the subacute pain stage and progress over time because of the duration of pain experience, rather than being a specific correlate of the chronic pain stage. Furthermore, information regarding whether these adaptations are related to cognitive processes of adaptation is lacking. To this end, we examined the DMN in 31 patients with chronic back pain (CBP), 77 patients with subacute back pain (SBP), as well as 39 healthy pain-free controls (HC) applying a graph-theoretic network approach on functional resting-state magnetic resonance imaging. Beyond the comparison between groups, we used a linear analysis considering the years lived with pain (YLP) across all patients with back pain and additionally performed a mediation analysis of the role of cognitive pain coping. In line with previous studies, we found significant DMN perturbation in CBP compared with HC. However, this did not apply to the comparison of CBP with SBP. Instead, we observed a positive correlation between DMN perturbation and YLP. This was significantly mediated by coping attitudes towards pain. Default mode network perturbation may thus reflect neural adaptation processes to pain experience rather than a single correlate of the chronic pain stage and be modulated by cognitive adaption. This points to potentially underinvestigated significant adaptation processes that could enable more fine-grained patient stratification.

1. Introduction

In chronic pain, neurobiological changes in emotional-cognitive brain circuits and associated psychological processes are crucial.³⁴ In this context, and among several alterations in the brain, reduced activation, connectivity, and regulation ability of the default-mode network (DMN) have been shown in patients with chronic pain compared with healthy controls.^{5,6,8,10,15,16,43,46,60} Although this DMN perturbation may be seen as neural adaptation because of the specific stage of chronic pain, it has also been interpreted to reflect a process of continuous exposure to pain within the chronic pain stage^5,8 and might thus change also during the chronic pain stage. In line with this, significant positive correlations between the years lived with back pain (YLP) and DMN alterations were shown in patients with chronic back pain (CBP).^3,8 This implies that this adaptation could be more of a process representing a linear, exposure-based correlation with painful experiences than a direct and static correlation with the onset of a chronic pain condition.

In this respect, it may even go beyond the chronic pain stage, also reflecting symptoms of pain at the subacute stage. Previous (neuroimaging) studies classified subacute pain as at least 7 weeks of pain in the current episode but also included patients with previous episodes in case these did not exceed 3 months⁴⁰ or have not occurred for one year.^7,34,42,64 Patients with subacute back pain (SBP) who never fulfilled chronicity criteria but still indicate several YLP might show DMN perturbation closer to the one observed for CBP than to SBP with less YLP, but this has not been investigated. Although longitudinal investigations on the transition period into chronic pain provided some first evidence,^3,9,41,43 studies that have integrated chronic pain samples as well are still lacking.

Therefore, this study aimed to investigate whether the perturbation of the DMN reflects a gradual neural adaptation process, observed beyond the chronic pain stage. This would show in the degree of DMN perturbation varying among patients with back pain (both SBP and CBP) dependent on their YLP instead of differing between SBP and CBP. The importance of such questions has been emphasized in recent discussions on duration-dependent stages of pain, arguing that different mechanisms may lay behind and important information may be lost when only investigating pain mechanisms along healthy-chronic and subacute-chronic approaches as these neglect individual differentiation.^23,40

This may also relate to cognitive adaptations to the continuous exposure of pain, especially pain coping. Pain coping has been found to be positively correlated with decreased pain, accompanied by activation of DMN's key hub, the medial prefrontal cortex (MPFC),^18,22 and might thereby additionally interact with the relationship of YLP and DMN perturbation as a mediator.

To investigate these issues, we analysed the DMN at resting state in SBP and CBP as well as in a sample of healthy pain-free controls (HC). We performed (1) group comparisons between SBP, CBP, and HC regarding DMN perturbation, and then (2) moved into a linear DMN-related analysis pooling SBP and CBP according to YLP, followed by (3) a mediation analysis on the role of cognitive pain coping.

2. Methods

2.1. Study design

2.1.1. Subjects

Participants were recruited through the outpatient pain clinic of the Institute of Cognitive and Clinical Neuroscience, general practitioners, physiotherapy practices, as well as reports in local newspapers and through the institute's web site. A version of the data set has already been used in previous research.⁴² To meet inclusion criteria, participants had to be at least 18 years old. Healthy pain-free controls were included only if they were free of pain. For the SBP group, we included patients with a current back pain episode of 7 to 12 weeks of back pain. Patients with a current back pain episode and additional back pain episodes in their history were included as well if the episodes never exceeded a period of 12 weeks. For the CBP group, we only included patients with a current back pain episode of more than 100 days. The sample consisted of 31 CBP (f = 16), 77 SBP (f = 53), and 39 HC (f = 17) who had complete resting-state measurements.

Patients were assessed for general eligibility through self-report using a screening intake form, which covered comorbid health and psychological conditions, MRI safety, concomitant medication dosages and indications, current and previous illicit drug/alcohol use, and pain levels. All participants passed the MRI safety screening requirements at each scanning visit. Informed consent was obtained from all participants on their first visit. Participants were compensated with €10/hour. All procedures were approved by the Ethics Committee of the Medical Faculty Mannheim of Heidelberg University and complied with the Declaration of Helsinki in its most recent form.

2.1.2. Questionnaires and diagnostic tools

The patients participated in a pain interview, where important characteristics of their back pain and the impact of the pain on their lives were assessed. Within this interview, the subjects were asked since when the back pain had been present. Years lived with back pain were derived by calculating the difference to the date of the experiment. If not available (2 SBP), item 7a of the German Pain Questionnaire,⁵² asking for how long the pain existed, was used. Possible answers are less than 1 month, 1 month to half a year, half a year to 1 year, 1 to 2 years, 2 to 5 years, and more than 5 years. The lower boundary of each item was implemented as YLP. Pain coping was assessed from “Catastrophizing” and “Coping” of the Pain-Related Self Statements Scale (PRSS) that describe situation-specific cognitive coping strategies as well as the subscales “Resourcefulness” and “Helplessness” of the Pain-Related Control Scale (PRCS) that describe coping attitudes towards pain.²⁸ The number of pain days in the last year and pain severity were assessed using the West Haven-Yale Multidimensional Pain Inventory (MPI; German version).²⁷ Anxiety and depression were assessed using the Hospital Anxiety and Depression Scale (HADS³⁶), which consists of 14 items (7 items are computed as total score for depression, and 7 items are computed as total score for anxiety), with rating scores ranging from 0 to 3. Pain intensity on the day of the MRI measurement was queried using a numeric rating scale from 0 to 10, where 0 represents no pain at all and 10 the worst pain possible. Current and past medication were assessed as self-report before the experiment. Furthermore, all participants underwent the Structured Clinical Interview for DSM-IV (SCID-I; German version)⁷⁰ to determine axis I mental disorders.

2.1.3. Characteristics of participants

Groupwise sample characteristics and results of descriptive group comparisons are shown in Table 1 and Supplementary Table S1, http://links.lww.com/PAIN/C95. Diagnoses are reported in Supplementary Table S2, http://links.lww.com/PAIN/C95, medication in Supplementary Table S3, http://links.lww.com/PAIN/C95, and frequency of pain in locations other than the back in Table S4, http://links.lww.com/PAIN/C95.

Table 1.

Sample characteristics.

	All	CBP	SBP	HC	SBP + CBP	NA
YLP	4.561 (6.712)	7.857 (6.973)	5.544 (7.16)	0 (0)	6.208 (7.152)	0/0/0/0/0
PI	2.014 (2.041)	2.903 (1.777)	2.675 (1.976)	0.0 (0)	2.741 (1.916)	0/0/0/0/0
NPD	88.49 (101.78)	250.645 (89.038)	68.013 (40.06)	0 (0)	120.435 (101.254)	0/0/0/0/0
CPS	4.926 (4.004)	7.697 (2.605)	6.305 (3.258)	0 (0)	6.705 (3.137)	0/0/0/0/0
CA	9.421 (7.388)	11.438 (8.828)	10.388 (7.347)	5.909 (4.725)	10.689 (7.773)	44/13/18/13/31
CO	28.603 (7.422)	28.223 (7.054)	28.131 (7.428)	29.836 (7.737)	28.157 (7.29)	44/13/18/13/31
HE	12.787 (5.633)	15.487 (6.489)	12.481 (4.765)	11.246 (5.887)	13.344 (5.459)	44/13/18/13/31
RE	14.663 (6.411)	17.123 (6.604)	14.58 (5.909)	12.873 (6.738)	15.31 (6.194)	44/13/18/13/31
Anx	5.997 (3.946)	7.162 (4.156)	6.99 (3.861)	3.11 (2.17)	7.039 (3.929)	17/5/9/3/14
Dep	4.022 (3.529)	5.281 (3.919)	4.554 (3.542)	1.971 (2.109)	4.763 (3.65)	17/5/9/3/14
Age	37.034 (14.526)	42.0 (16.129)	36.078 (13.798)	34.974 (14.08)	37.778 (14.677)	0/0/0/0/0
Gender	86/61	16/15	53/24	17/22	69/39	0/0/0/0/0

Mean and SD. For gender absolute frequency (f/m).

Anx, Anxiety (HADS); CA, Catastrophizing; CBP, patients with chronic back pain; CO, Coping; Dep, Depression (HADS); HC, healthy pain-free controls; HE, helplessness; NA, not available; NPD, number of pain days last year; PI, pain intensity at day of MRI measurement; PS, pain severity (MPI); RE, resourcefulness; SBP, patients with subacute pain; YLP, years since first back pain.

2.1.4. MRI acquisition

For resting-state assessment, participants were asked to lay still with their eyes closed, remain awake, and try not to think about anything specific for 7:20 minutes. MR data were acquired using a 3-T Siemens TrioTim MRI scanner. One run of T1-weighted SP\MP\OSP GR\IR (GR\IR) single-echo structural MRI data was collected (256 slices; TR = 2300 ms; echo time, TE = 2.98 ms; FA = 9°; FOV = 192 × 240 mm; matrix size = 192 × 240; voxel size = 1 × 1 × 1 mm). For resting-state imaging, one run of 210 rest segmented k-space echo planar (EP) single-echo fMRI data was collected (36 slices in interleaved ascending order; TR = 2100 ms; TE = 23 ms; FA = 90°; FOV = 220 × 220 mm; matrix size = 96 × 96; voxel size = 2.29 × 2.29 × 3 mm; in-plane acceleration factor = 2). DICOMs were converted to NIfTI-1 format using dcm2niix (v1.0.20220720). This section was (in part) generated automatically using pybids (0.15.6).⁷¹

2.2. Data processing

2.2.1. fMRI processing and denoising

Results included in this article come from preprocessing performed using fMRIPrep 21.0.4,^23,24 which is based on Nipype 1.6.1.^30,31

2.2.2. Anatomical data preprocessing

A total of 1 T1-weighted (T1w) images were found within the input BIDS data set. The T1-weighted (T1w) image was corrected for intensity non-uniformity (INU) with N4BiasFieldCorrection,⁶³ distributed with ANTs 2.3.3,⁴ (RRID:SCR_004757), and used as T1w-reference throughout the workflow. The T1w-reference was then skull-stripped with a Nipype implementation of the antsBrainExtraction.sh workflow (from ANTs), using OASIS30ANTs as target template. Brain tissue segmentation of cerebrospinal fluid (CSF), white matter (WM), and gray matter (GM) was performed on the brain-extracted T1w using fast (FSL 6.0.5.1:57b01774, RRID:SCR_002823).⁷² Brain surfaces were reconstructed using recon-all (FreeSurfer 6.0.1, RRID:SCR_001847),²¹ and the brain mask estimated previously was refined with a custom variation of the method to reconcile ANTs-derived and FreeSurfer-derived segmentations of the cortical gray matter of Mindboggle (RRID:SCR_002438).³⁸ Volume-based spatial normalization to 2 standard spaces (MNI152NLin2009cAsym, MNI152NLin6Asym) was performed through nonlinear registration with antsRegistration (ANTs 2.3.3), using brain-extracted versions of both T1w reference and the T1w template. The following templates were selected for spatial normalization: ICBM 152 Nonlinear Asymmetrical template version 2009c²⁹ [RRID:SCR_008796; TemplateFlow ID: MNI152NLin2009cAsym] and FSL's MNI ICBM 152 nonlinear 6th Generation Asymmetric Average Brain Stereotaxic Registration Model²⁵ [RRID:SCR_002823; TemplateFlow ID: MNI152NLin6Asym].

2.2.3. Functional data preprocessing

For the BOLD run found per subject, the following preprocessing was performed. First, a reference volume and its skull-stripped version were generated using a custom methodology of fMRIPrep. Head-motion parameters with respect to the BOLD reference (transformation matrices and 6 corresponding rotation and translation parameters) are estimated before any spatiotemporal filtering using mcflirt (FSL 6.0.5.1:57b01774).³⁷ BOLD runs were slice-time corrected to 1.01 seconds (0.5 of slice acquisition range 0 s-2.03 s) using 3dTshift from AFNI,¹⁹ (RRID:SCR_005927). The BOLD time series (including slice-timing correction when applied) were resampled onto their original, native space by applying the transforms to correct for head motion. These resampled BOLD time series will be referred to as preprocessed BOLD in original space or just preprocessed BOLD. The BOLD reference was then coregistered to the T1w reference using bbregister (FreeSurfer), which implements boundary-based registration.³² Coregistration was configured with 6 degrees of freedom. Several confounding time series were calculated based on the preprocessed BOLD: framewise displacement (FD), DVARS, and 3 regionwise global signals. Framewise displacement was computed using 2 formulations following Power (absolute sum of relative motions)⁵³ and Jenkinson (relative root mean square displacement between affines).³⁷ Framewise displacement and DVARS are calculated for each functional run, both using their implementations in Nipype (following the definitions by Ref. 53). The 3 global signals are extracted within the CSF, the WM, and the whole-brain masks. In addition, a set of physiological regressors was extracted to allow for component-based noise correction (CompCor).⁹ Principal components are estimated after high-pass filtering the preprocessed BOLD time series (using a discrete cosine filter with 128-second cutoff) for the 2 CompCor variants: temporal (tCompCor) and anatomical (aCompCor). tCompCor components are then calculated from the top 2% variable voxels within the brain mask. For aCompCor, 3 probabilistic masks (CSF, WM, and combined CSF + WM) are generated in anatomical space. The implementation differs from that of Behzadi et al. in that instead of eroding the masks by 2 pixels on BOLD space, and the aCompCor masks are subtracted a mask of pixels that likely contain a volume fraction of GM. This mask is obtained by dilating a GM mask extracted from the FreeSurfer's aseg segmentation, and it ensures components are not extracted from voxels containing a minimal fraction of GM. Finally, these masks are resampled into BOLD space and binarized by thresholding at 0.99 (as in the original implementation). Components are also calculated separately within the WM and CSF masks. For each CompCor decomposition, the k components with the largest singular values are retained, such that the retained components' time series are sufficient to explain 50% of variance across the nuisance mask (CSF, WM, combined, or temporal). The remaining components are dropped from consideration. The head-motion estimates calculated in the correction step were also placed within the corresponding confounds file. The confound time series derived from head motion estimates and global signals were expanded with the inclusion of temporal derivatives and quadratic terms for each.⁵⁶ Frames that exceeded a threshold of 0.5 mm FD or 1.5 standardised DVARS were annotated as motion outliers. Automatic removal of motion artifacts using independent component analysis (ICA-AROMA)⁵⁴ was performed on the preprocessed BOLD on MNI space time series after removal of non–steady-state volumes and spatial smoothing with an isotropic, Gaussian kernel of 6-mm FWHM (full-width half-maximum). Corresponding “nonaggressively” denoised runs were produced after such smoothing. In addition, the “aggressive” noise regressors were collected and placed in the corresponding confounds file. All resamplings can be performed with a single interpolation step by composing all the pertinent transformations (ie, head-motion transform matrices, susceptibility distortion correction when available, and coregistrations to anatomical and output spaces). Gridded (volumetric) resamplings were performed using antsApplyTransforms (ANTs), configured with Lanczos interpolation to minimize the smoothing effects of other kernels.⁴⁰ Nongridded (surface) resamplings were performed using mri_vol2surf (FreeSurfer).

Many internal operations of fMRIPrep use Nilearn 0.8.1,¹ (RRID:SCR_001362), mostly within the functional processing workflow.

The above boilerplate text was automatically generated by fMRIPrep.

2.2.4. fMRI denoising

Preprocessed images were denoised using the ICAAROMA2PhysGS strategy previously described.⁵⁰ Therefore, we used the nonaggressive ICA-ARMOMA-derived output images from fMRIPrep. The first 5 images of the functional run were discarded to account for saturation effects. As confounder, we imported the mean physiological signals from white matter and cerebrospinal fluid as well as mean global signal using nilearn's load_confounds function and then denoised and high-pass filtered (0.08 Hz) using nilearn's NiftiMasker function. Within this step, the voxelwise temporal signal was standardized.

2.2.5. Quantification of network perturbation

Default mode network perturbation was defined as inversion of the graph-theoretical functional level of the DMN and derived by examining multiple common and robust graph-theoretical measurements of within network and within whole-brain network function. We examined the mean node degree (MND), mean shortest path (MSP), closeness centrality (CC), general efficiency (GE), and small worldness (SW) as common graph-theoretic metrics, that span from simple low-order to complex high-order indication of network function.¹⁷

2.2.6. Connectome extraction

First, whole-brain images were parcellated with the Dictionaries of Multiple Dimensions (DiFuMo) atlas in the 512-node version²⁰ with nilearn's NiftiMapsMasker function after applying a Gaussian kernel smoothing with a full-width at half-maximum of 5 mm. Individual connectomes were computed as the Pearson correlation of the time series between pairs of nodes. Subsequently, each connectome was binarized by selecting the top 10% strongest connections. We used 2 strategies to compute graph-theoretical features for the DMN, both within the network and within the whole brain.

2.2.7. Within-network features

For within-network features, we first extracted the binarized connectome corresponding to the DMN nodes based on the Yeo-7-network parcellation.⁶² We then combined this binarized DMN connectome with the minimum spanning tree calculated from the nonbinarized DMN connectome. This approach allowed us to obtain a fully connected, binarized connectome specifically for the DMN in each participant. From this DMN connectome, we computed the GE, MSP, and SW (number of iterations = 100) using the implementations provided by the networkx library.³³

2.2.8. Within-whole-brain-network features

For network features within the whole brain, we combined the minimum spanning tree derived from the whole-brain connectome with the binarized whole-brain connectome. We then calculated closeness centrality and node degree for each node in the connectome and subsequently averaged the measures across all nodes of the DMN. This procedure allowed us to obtain the MND and the CC specifically for the DMN. Increased MSP and SW represent DMN perturbation, whereas an increased MND, CC, and GE represent DMN network function and hence should show inverted effects.

2.2.9. Nodewise and edgewise features

For the topological analysis pertaining to network changes on nodewise and edgewise level, we assessed node degree and standardized connectivity. Node degree was computed similar to the calculations performed for MND, but without averaging over multiple nodes. For edgewise analysis, standardized connectivity was extracted as the whole-brain wide z-standardized Pearson correlation coefficient between pairs of DMN nodes.

2.3. Statistical analysis

2.3.1. Group comparisons between chronic back pain, subacute back pain, and healthy pain-free controls

To examine whether the DMN is perturbated in the chronic pain stage compared with the subacute pain stage, we first tested for group difference between the SBP and CBP in the graph-theoretic network function of the DMN at resting state. Given the previous results of DMN perturbation in the chronic pain stage compared with the healthy pain-free state, we additionally tested for group difference of CBP, SBP, as well as the pooled patients with back pain (SBP + CBP) compared with HC. Therefore, we conducted pairwise t tests between SBP, CBP, and HC as well as between the pooled CBP + SBP and HC, separately for all graph-theoretic metrics.

2.3.2. Linear analysis pooling chronic back pain and subacute back pain according to years lived with back pain

Based on our assumption that DMN perturbation reflects a gradual neural adaptation process in response to the duration of pain experience, we examined the linear relationship of DMN perturbation and YLP in the pooled SBP + CBP. For this analysis, we conducted OLS regressions with YLP as a predictor. Separate analyses for each graph-theoretical parameter of DMN perturbation were performed. For spatial analysis, the same analysis was conducted for each node on the node degree and for each edge on the Pearson correlation coefficient. Gender and age were included in the OLS as additional predictors to control for potential confounding effects.

2.3.3. Mediation analysis on the role of cognitive coping

To examine the potential mediation of YLP–DMN perturbation associations through cognitive adaptation, we extended our analysis beyond the OLS approach using structural equation modelling (SEM). In doing so, we treated DMN perturbation as a latent variable indicated by the observed graph-theoretic metrics, while adding gender and age as additional variables for control. In addition, we used 2 more latent variables, coping attitudes towards pain and situation-specific cognitive coping strategies. These were implemented on the basis of previously reported structure²⁸ with helplessness and resourcefulness as measured variables for coping attitudes towards pain and coping and catastrophizing as measured variables for situation-specific cognitive coping strategies. In the SEM, we implemented coping attitudes towards pain and situation-specific cognitive coping strategies to be predicted by YLP and to predict DMN perturbation. Wishart loglikelihood was used for optimisation.

2.3.4. Imputation of missing values

Missing values were imputed using multivariate imputation by chained equations (mice) with sklearn's IterativeImputer function⁵¹ based on the mice implementation in R.¹³

2.3.5. Alpha and false-discovery-rate correction

All statistical tests were performed 2-tailed. To control for false-discovery rates (FDR) for multiple tests targeting the same independent variable, we used the Benjamini/Hochberg procedure implemented by the statsmodels library's multipletests function.⁵⁷ In the sample description, this applied to graph-metric wise tests for group differences and graph-metric wise regression analysis. Adjusted P-values were reported alongside unadjusted P-values. For all tests, we used a significance level of α≤ .05, except for the connectivity and edgewise analysis on clusters. As in this secondary analysis, we were interested in potential spatial aggravations of associated nodes and edges, and we limited for the 10% of nodes and 1% of edges with the most significant associations with YLP.

2.3.6. Availability of data and materials

All data needed to evaluate the conclusions are present in the article and/or the Supplementary Materials, http://links.lww.com/PAIN/C95. Ethical restrictions aimed at protecting participant confidentiality prohibit us from publicly releasing anonymized study data. For access to the study data and materials, interested readers should reach out to the corresponding author through a formal collaboration agreement. This agreement stipulates that data will be shared with researchers who agree to collaborate with the authors solely for the purpose of verifying the claims made in the article. After approval of this formal collaboration agreement by the local Ethics Committee, the data and materials will be provided to requestors.

The jupyter notebook written for this analysis is available online through doi.org/10.5281/zenodo.8340363.

3. Results

3.1. Chronic back pain shows significant default mode network perturbation compared with healthy pain-free controls, but not compared with subacute back pain

Although consistent with previous reports, there were significant differences between CBP and HC in all graph-theoretic metrics; in line with our assumptions, there was no significant difference between CBP and SBP in any graph-theoretic metric as shown in Figure 1. This indicates that DMN perturbation is reflecting the chronic pain stage, only when compared with a healthy pain-free stage, but not when compared with SBP. We performed additional tests between SBP compared with HC as well as between the pooled SBP + CBP and HC. The pooled SBP + CBP compared with HC showed significant differences in CC, GE, and MSP, whereas SBP did not express significant perturbation of the DMN compared with HC after multiple comparison correction (Fig. 1). For exact test results, see Supplementary Table S5, http://links.lww.com/PAIN/C95, and for groupwise means and SDs, see Supplementary Table S6, http://links.lww.com/PAIN/C95. Given the variations in gender frequency between CBP and SBP, we conducted further analyses by comparing CBP to a subset of SBP matched by gender and age. These analyses yielded consistent results across any graph metrics (Supplementary Table S7 and S8, http://links.lww.com/PAIN/C95).

Figure 1.

CBP show significant DMN perturbation compared with HC, but not compared with SBP. CBP show significant alteration of the DMN in graph-theoretic metrics compared with HC, but in none compared with SBP. Pooled SBP + CBP show alteration of the DMN in some graph-theoretic metrics, whereas SBP show no perturbation compared with HC. Boxplot (median, first and third quartile, median ± 1.5 times interquartile range) and scatterplots. Performed tests are indicated by brackets. For detailed results, see Supplementary Table S1, http://links.lww.com/PAIN/C95. HC: green, SBP: orange, CBP: blue, SBP + CBP: pink. Significance is marked with *P_unadj ≤ 0.05, **P_unadj ≤ 0.01, (ns): not significant after multiple comparison correction (P_adj ≤ 0.05). CBP, patients with chronic back pain; CC, closeness centrality; GE, global efficiency; HC, healthy pain-free controls; MND, mean node degree; MSP, mean shortest path; SBP, patients with subacute pain; SW, small worldness.

3.2. Default mode network perturbation varies among patients with back pain according to years lived with back pain

The lack of significant difference in DMN perturbation for the chronic pain stage compared with the subacute pain stage emphasized our question, whether there are individual characteristics among patients with back pain beyond the chronic pain stage that may rather explain the perturbation of the DMN. Coherent with our hypothesis YLP significantly predicted DMN perturbation, except for small-worldness, which still showed an effect in the same direction (Table 2 and Fig. 2A). We next examined whether dedicated regions and circuits of the DMN were particularly incorporated into the association between DMN perturbation and YLP. The nodes that were most strongly associated with YLP were mainly located in the MPFC, whereas edges within the MPFC and between the MPFC and the posterior cingulate cortex showed the strongest associations with YLP (Fig. 2B).

Table 2.

Results of the regression predicting default mode network perturbation with years since first back pain in the pooled patients with subacute pain + patients with chronic pain.

	GE	MSP	SW	MND	CC
YLP	−0.205* (0.086)	0.204* (0.085)	0.152 (0.097)	−0.253* (0.092)	−0.201* (0.091)
Age	−0.246* (0.086)	0.275** (0.085)	0.083 (0.097)	−0.146 (0.092)	−0.179**** (0.091)
Gender	0.697*** (0.176)	−0.680*** (0.174)	−0.472* (0.197)	0.511** (0.187)	0.602** (0.185)
R2	0.290	0.303	0.105	0.195	0.210
R2adj	0.269	0.283	0.079	0.172	0.187
P(F)adj	<0.001	<0.001	0.009	<0.001	<0.001

Standard coefficients and SEs.

****P_adj < 0.1; *P_adj < 0.05; **P_adj < 0.01; ***P_adj < 0.001.

CC, closeness centrality; GE, global efficiency; MND, mean node degree; MSP, mean shortest path; SW, small worldness; YLP, years lived with back pain.

Figure 2.

DMN perturbation varies among patients with back pain according to YLP. (A) Combined regression and scatterplots for each graph-theoretic DMN feature. Colours indicate group. SBP: orange and CBP: green. 95%CI of the prediction is indicated by a shaded area and dotted lines. For detailed results, see Table 2. Distribution of YLP (top), as well as the graph-theoretic DMN metrics (right), was plotted using kernel density estimate separately for SBP and CBP. (B) Regression results with the same model, but edgewise with standardized Pearson correlation coefficients and nodewise with node degrees. Nodewise only significant nodes (P_adj ≤ 0.1) were coloured. Size and, for significant nodes, also colour are proportional to β coefficient. Edgewise only significant (P_adj ≤ 0.1) results are plotted and were coloured, which is proportional to β coefficient. CC, closeness centrality; GE, global efficiency; MND, mean node degree; MSP, mean shortest path; SW, small worldness; YLP, years lived with back pain.

3.3. Coping attitudes towards pain mediate the association of default mode network perturbation and years lived with back pain

Last, we examined a potential mediating effect of cognitive pain coping. In our mediation, SEM model YLP was significantly associated with coping attitudes towards pain but not DMN perturbation, whereas coping attitudes towards pain significantly predicted DMN perturbation (Fig. 3). This indicated that coping attitudes towards pain mediate the association of DMN and YLP. The goodness-of-fit parameters and the exact pathway results are stated in Supplementary Table S9 and S10, http://links.lww.com/PAIN/C95. For comparison, see the results of a model without the mediating pathways (Supplementary Table S9 and S11, Fig. S1, http://links.lww.com/PAIN/C95).

Figure 3.

Coping attitudes towards pain mediate the association of DMN perturbation and YLP. Standardized estimates are stated on the respective paths. For the paths of interest, wideness is proportional to strength of the association. Negative associations are coloured blue, positive red. ***P < 0.001, **P < 0.01, *P < 0.05. AP, coping attitudes towards pain; CA, catastrophizing; CC, closeness centrality; CO, coping; DMN, default mode network; GE, global efficiency; HE, helplessness; MND, mean node degree; MSP, mean shortest path; RE, resourcefulness; SC, situation-specific coping strategies; SW, small worldness; YLP, years lived with back pain.

4. Discussion

In this study, we showed that perturbation of the DMN reflects the linear correlate of the duration of pain experience indicated by YLP rather than a single correlate of the chronic pain stage. Moreover, we demonstrated that this relationship between YLP and DMN perturbation is modulated by cognitive adaptation concerning coping attitudes towards pain. These results align with previous assumptions that DMN perturbation reflects a neural adaptation to continuous pain^5,8 and that it is associated with pain coping,^8,35,39,47 but show that this extends beyond the chronic pain stage dependent on the individual pain experience. This improves the understanding of the course of pain chronicity by enabling fine-grained patient stratification.^26,41

Many previous studies have shown DMN alterations in chronic pain in comparison with healthy controls.^{2,5,6,8,10,15,16,43,60,61} This approach was used to provide evidence on processes of neural adaptations in the DMN because of chronic pain^5,8 but may have masked significant differences within the course of chronic pain. Notably, although consistent with previous studies, we found the DMN to be perturbated in CBP compared with HC and found no significant DMN perturbation compared with SBP. This points towards individual effects that underlie the DMN perturbation and require a fine-grained analysis beyond the chronic pain stage. Initial evidence that DMN perturbation may be driven by pain experience, came from Baliki et al.,⁸ who showed that DMN perturbation within CBP depends on the duration of the pain experience. Extending this beyond the chronic pain stage, we tested in a linear fashion whether this neural adaptation in the DMN maps the entire course of pain chronicity from the subacute stage onward. Indeed, DMN perturbation and YLP were positively correlated in the pooled CBP + SBP. This implies that DMN perturbation reflects the results of a gradual neural adaptation to pain experience manifesting from subacute stages onward rather than adaptations because of the stage of chronic pain.^5,8 This supports previous assumptions that some mechanisms in chronic pain overlap with those in the transition from acute to chronic pain⁴⁵ and that individual factors, here YLP, must be given greater consideration.⁴¹

Our finding on the significant relationship with coping attitudes towards pain, another important individual factor, also aligns with this. Pain-associated cognitive adaptation processes, specifically regarding coping, have previously been associated with the DMN.^8,35,39,47 The way an individual cognitively deals with the experience of pain can change the way pain affects the individual.^55,66,67,69 Such effect of pain coping may, moreover, influence the neural adaptation to pain experiences. Coherently, we showed a significant mediation of coping attitudes towards pain between YLP and DMN perturbation in the pooled SBP + CBP. This implies a differential role in how pain experiences affect the DMN, which may explain previous inconsistencies in studies on cognitive pain coping and the DMN. Between chronic pain samples, YLP distributions vary,⁸ thereby potentially introducing a bias in pain coping that hides correlations with DMN perturbation. Accordingly, our additional analyses in the CBP sample revealed no significant contribution of pain coping to the association between DMN perturbation and YLP. The directional effects of mediation still remained consistent, and our findings demonstrated a significant dependence of coping attitudes towards pain on YLP (Supplementary Table S9 and S12, Fig. S2, http://links.lww.com/PAIN/C95).

Throughout our analyses, we incorporated network parameters from simple to highly complex metrics to explore how neural adaptation affects the DMN across levels of network function complexity. Notably, we observed no significant disparities in the association of YLP with the different graph-theoretical network functions. However, MND exhibited the highest association with YLP, whereas SW showed the least. This suggests that beyond the network's internal architecture, as reflected in SW, its rank in brainwide processing is primarily affected by perturbations. Nonetheless, further investigations are necessary to understand how the intertwined effects on complex network function of the DMN evolve.

Finally, we investigated whether specific regions of the DMN were particularly related to the neural adaptation. We found that the MPFC expressed the most significant linear dependencies with YLP. This links to recent animal research at the microscopic level of Stegemann et al.⁵⁹ who have shown that the MPFC is involved in the storage of long-term pain-related fear memories in distinct cell populations. These showed hyperconnectivity to fear-related structures outside the DMN⁵⁹ and may potentially amplify and expand when exposed to continuous pain. This may at the macroscopic level be reflected in the reduced interconnectivity of the MPFC, that we showed. Notably, this assumed mechanism could explain the association of DMN perturbation and YLP beyond the chronic pain stage. Pain experience is a key criterion of both subacute and chronic pain, but its frequency is not yet considered within chronic pain above a specific cutoff related to the current pain episode (persisted for >3 months⁴⁸) and not over multiple potentially separate pain episodes. In addition, even subacute pain definitions include the frequency of pain experience in between specific cutoffs related to the current pain episode (eg, persistent for >7 weeks <3 months) but again not including repeated pain experience over multiple potentially separate pain episodes. Giving more weight to pain experiences as indicated by YLP could be beneficial for enhancing and specifying the neural correlates of pain chronicity.

In this regard, it is also important to note that based on previous studies on resting-state networks in chronic pain,^2,5,8,65 we only investigated the DMN, leaving open potential involvement of other brain networks that might drive or compensate the effect on the DMN. To explore such potential covarying effects of other brain networks, we additionally tested whether other networks show similar experience-dependencies perturbations as the DMN in the pooled patient sample. None of the networks showed any significant association with YLP (Supplementary Table S13, http://links.lww.com/PAIN/C95). Nonetheless, we also examined potential correlations between the degree of DMN perturbation and the degree of perturbation of other networks in the same sample (Supplementary Table S14, http://links.lww.com/PAIN/C95). Positive correlations were primarily found with the salience ventral attention and control network, which could be expected since they are functionally tightly connected.⁴⁴ Notable negative correlations were found with the somatomotor and visual network, regarding the MND. Taken together, these findings further emphasize the role of neuronal adaptation specifically within the DMN and provide significant additional insights that can inform future studies on brain networks and (chronic) pain.

This study has several limitations. The DMN is influenced by the intensity of pain during measurement^2,8 and potentially also by the extent of pain exposure occurring close in time. In our study, we focused on long-term pain experience and therefore used YLP as an indicator. Nonetheless, in our supplementary analysis, we used pain intensity on the day of MRI measurement, the pain days last year, and pain severity but did not reveal significant associations with DMN perturbation (Supplementary Tables S15, S16, and S17, http://links.lww.com/PAIN/C95). Still, long-term pain experience, particularly in patients with many YLP, may be influenced by the frequency of pain episodes. Although we inquired since when the pain persisted, we cannot determine in which frequency the pain occurred over the individual course of the pain. However, this specific aspect is commonly not assessed in questionnaires used in the research area but may be promising to implement in future studies. It is also important to note that the reports of YLP were not objectively verified. Although common practice in research and clinical interviews,^{14,49,52,58,68} objective verification may enhance accuracy in future studies. In this respect, the impact of other or additional pain types may also need further investigation. In our study, we only included patients based on their subjectively identified most important pain being back pain. Other pain sites reported in this sample are commonly associated with back pain and thus probably do not represent an additional type of pain (Supplementary Table S4, http://links.lww.com/PAIN/C95). However, patients also frequently report concurrent, alternating, or accumulating pain types, which may have additional impact on the DMN, also in relation to YLP.⁸Therefore, conducting further investigation into patient samples comprising individuals with mixed and comorbid types of pain could yield valuable insights in future studies. In light of these limitations, it is important to note that a significant proportion of the variance related to DMN perturbation remains unexplained by our analysis. Although this may be perceived as a limitation, particularly concerning the clinical significance, it could be anticipated given that the DMN serves as the central network for processing all self-referential experiences, whether pain-related or not. Adverse experiences indirectly, directly, or unrelated to pain are common in samples with back pain^11,12 and may also contribute to DMN perturbation. Moreover, although we demonstrated a mediating effect of pain coping on the association of YLP and DMN perturbation, we assessed coping simultaneously with DMN perturbation. Individual coping strategies used during YLP may provide a better explanation for DMN perturbation and should be explored in future studies.

5. Conclusions

In conclusion, we have shown that perturbation of the DMN mirrors neural adaptation manifesting from subacute pain stages and progressing because of the duration of pain experience as indicated by YLP, rather than simply reflecting chronic pain as a single stage. Furthermore, we have demonstrated that the association of DMN perturbation and YLP is influenced by cognitive adaptation that also follows the duration of pain experience. Based on our findings, the examination of potentially underinvestigated significant adaptation processes to pain experience can improve our understanding of the course of pain chronicity by allowing for a more-fine grained patient stratification. To this end, future research should determine the longitudinal effects of DMN perturbation for the individual patient regarding their YLP. This could considerably contribute to advance personalised treatment approaches.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C95.