Neural, psychological, and daily life evidence for a transdiagnostic process of affective dysregulation in depression and chronic widespread pain

Supplemental Digital Content is Available in the Text.

The transdiagnostic process of affective dysregulation in depression and fibromyalgia is modulated by daily life stress. Amygdala functional magnetic resonance imaging neurofeedback suggests a top-down sensitization in fibromyalgia.

Abstract

Chronic pain and depression are leading causes of disability, frequently co-occurring and exacerbating each other. This cross-sectional study investigated putative transdiagnostic processes of affective dysregulation in fibromyalgia (FMS) and major depressive disorder (MDD) using psychometric questionnaires (Beck Depression Inventory-II, Hospital Anxiety and Depression Scale, Cognitive Emotion Regulation Questionnaire, Perceived Stress Scale, Widespread Pain Index, Somatic Symptom Disorder B Criteria Scale 12), ecological momentary assessments, and real-time functional magnetic resonance imaging amygdala neurofeedback during an emotion regulation task. We compared clinical symptoms, stress sensitivity, and emotion regulation in patients with FMS (N = 46) and MDD (N = 48) with healthy controls (N = 34). Patients with fibromyalgia syndrome and major depressive disorder reported similar psychopathological and affective dysregulation profiles, and they exhibited more psychopathology and emotion regulation deficits than healthy controls (HC). Differences between MDD and FMS were limited to pain-specific pathology in FMS (pain spread and frequency P < 0.001, intensity P < 0.05) and more rumination (P < 0.05) and self-blame (P < 0.01) in MDD. Momentary stress predicted higher subsequent pain and worse affective states across groups, with FMS and MDD exhibiting stronger stress responses (all P's < 0.05). Directly after neurofeedback training, FMS and MDD were less able to downregulate left amygdala activity than HC (P = 0.039) compared to baseline performance, and this brain marker predicted daily life psychopathology (negative affect, anxiety, and rumination, all P's < 0.05). Patients with fibromyalgia syndrome additionally exhibited unique deficits in right amygdala regulation (P = 0.004). Our findings highlight transdiagnostic affective dysregulation patterns in FMS and MDD, specific differences in emotion regulation strategies, and a potential neuronal marker of a shift towards right amygdala sensitization during affective processing in FMS.

1. Introduction

Chronic pain and depression are the 2 leading causes of years lived with disability globally^21,87 and are co-occurring in 30% to 50% of cases.^6,56,67,74 Comorbid patients suffer from heightened disability and face worse prognoses compared to patients with either condition alone.^4,6,59,73

Among all chronic pain disorders, the most striking example of this double burden may be fibromyalgia syndrome (FMS): not only are mood disturbances part of the diagnostic criteria,⁹² prevalence rates of comorbid major depressive disorder (MDD) are among the highest of all pain disorders, with lifetime prevalence of up to 86%.^1,34 This striking clinical overlap and the triad of shared genetic predispositions,^12,37,66,100 environmental adversity,³ and pathophysiological mechanisms such as difficulties in recovering after acute stress experiences^10,34,94 also suggests a shared intermediary predisposing phenotype of FMS and MDD on the cognitive-emotional level.^49,52

Dysfunctional emotion regulation (ER) has long been recognized as a hallmark and major (psycho) therapeutic target in MDD,^38,47 with patients reporting higher levels of alexithymia, rumination, catastrophizing, and other maladaptive ER strategies^46,85 as well as higher affective instability^62,78 compared to healthy controls (HC). Furthermore, functional alterations in neuronal emotion regulatory circuits are well-known in MDD.⁷⁰ In contrast, ER deficits in FMS have not yet been sufficiently researched, although it is known that emotional processes influence nociceptive stimulus processing.³⁹ Compared to HC, patients with FMS show consistently higher alexithymia, more difficulties in ER, and greater affective instability in daily life.^26,72,82 Patients with fibromyalgia syndrome show disruptions of supraspinal processes associated with positive affect and emotional modulation of pain,⁶⁹ and some aspects of ER deficits have been shown to correlate with pain or mediate the relationship with psychological distress.^25,84 An imbalance in ER has recently been proposed as an integrative model for FMS⁶⁵ but was criticized for lack of empirical evidence.¹¹

However, most primary research focuses on only one of the 2 diseases, impeding the integration of evidence across diagnoses. Our study addressed this unmet need by investigating a putative transdiagnostic process of affective dysregulation in a sample of patients with FMS and MDD together with HC, using a multimodal approach including clinical assessments, ecological momentary assessments, and a functional magnetic resonance imaging (fMRI) neurofeedback paradigm challenging ER. We aimed to capture ER deficits and associated clinical symptoms both in the laboratory and in daily life and to find potential neurobiological correlates of a shared predisposing phenotype in the brain. We expected to find substantial overlap between patients with MDD and FMS in terms of clinical symptoms and affective dysregulation processes, with pronounced differences in comparison to HC, implying that mental health interventions may also be beneficial for patients with FMS. In addition, we expected that HC would be better able than patients with FMS and MDD to learn downregulate of amygdala activity using fMRI neurofeedback.

2. Methods

2.1. Participants

Of the total N = 132 subjects participating in this study, N = 48 suffered from FMS and N = 49 from MDD; N = 35 participants served as HC. All participants underwent a diagnostic clinical interview with a trained psychologist (Mini-DIPS),⁵³ 1 week of smartphone-based daily life assessments and one 30-minute session of real-time fMRI neurofeedback. In addition, they answered a battery of psychological and pain-related self-assessment questionnaires. The study sample originated from the CRC1158 collaborative research project funded by the German Research Foundation (https://www.sfb1158.de/), some of the psychometric data presented here have been previously reported for a subset of the participants.^14,75 All participants provided written informed consent. All procedures were conducted according to the Declaration of Helsinki and approved by the Ethics Committee of the Medical Faculty of the Heidelberg University. The research protocol is registered in the German Clinical Trials Register (DRKS00010559). Recruitment channels comprised study wards and tertiary care clinics in the Rhine-Neckar area, online advertisements and the resident's registration office between January 2020 and July 2022. Diagnosis of FMS (American College of Rheumatology criteria) or MDD (Diagnostic and Statistical Manual of Mental Disorders, version 5 criteria) was confirmed by a trained physician or psychologist. To adequately represent clinical reality, we recruited based on the primary diagnosis thus refraining from excluding participants with mixed clinical presentations. The full list of inclusion and exclusion criteria is presented in SI 1, http://links.lww.com/PAIN/C379, and a CONSORT style flow diagram of study participation is shown in Figure 1.

Figure 1.

CONSORT-style flowchart of study participation per group. Inclusion and exclusion criteria are detailed in SI 1 Table 1, http://links.lww.com/PAIN/C379. Note that exclusion because of MRI incompatibility was discontinued for patients with FMS during study progression because of recruitment challenges during the ongoing COVID-19 pandemic, leading to a smaller fraction of the total FMS sample having fMRI data. fMRI, functional magnetic resonance imaging; FMS, fibromyalgia syndrome.

2.2. Psychometric assessments

Participants reported on aspects of their pain experience, psychopathology, general health, and behaviors relevant to the putative joint affective dysregulation phenotype using well-established questionnaires and structured interviews:

The West-Haven Yale Multidimensional Pain Inventory²³ and the Chronic Pain Grade Scale⁸⁶ were used to broadly assess chronic pain. The Widespread Pain Index⁹¹ and the Somatic Symptom Disorder B Criteria Scale 12⁸⁰ were used to assess specific aspects of widespread pain and functional pain syndromes. Only participants who indicated being in pain over the last 3 months (irrespective of pain intensity) reported on these pain and somatization measures.

Psychopathological measures covered perceived stress, depression, anxiety, and affect and were assessed using the Perceived Stress Scale,⁴² the Beck Depression Inventory⁹⁰ (BDI-II), the Hospital Anxiety and Depression Scale⁶⁴ (HADS), and the Positive and Negative Affect Scale,⁹ respectively. In terms of symptoms, we also assessed health anxiety using the Whiteley Index³⁰ and disorder-agnostic health and function using the World Health Organisation Disability Assessment Schedule 2.0.⁸³

Finally, all participants answered questionnaires covering aspects contributing to affective dysregulation: regarding ER behaviors, pain-focused catastrophizing and the use of 9 different cognitive coping strategies were assessed using the Pain Catastrophizing Scale⁵⁵ (PCS) and the Cognitive Emotion Regulation Questionnaire⁵⁰ (CERQ), respectively. Regarding relatively more stable personality features, alexithymia and trait anxiety were assessed using the Toronto Alexithymia Scale⁵ (TAS-20) and the State Trait Anxiety Inventory⁴⁵ (STAI-X2), respectively.

We identified group differences using ANCOVA analyses with type III sums of squares, with group as factor and age and sex as covariates. For data that were not normally distributed, we confirmed results using Kruskal–Wallis tests. For categorical data, we used χ² tests. Differences between patients with FMS and MDD were assessed post-hoc using Tukey honest significant differences (HSD), and differences between both patient groups and HC using contrast coding. P-values were corrected for multiple testing using the Benjamini–Hochberg approach for false discovery rate correction, with no clustering of questionnaires or subscales of questionnaires into families.⁷ False discovery rate correction was implemented using the P.adjust() function in R (version 4.3.1) with method = “BH.” We report the returned Bonferroni-Holmes-adjusted P-values, also known as q-values as “p = …, FDR-corrected.” Ties in adjusted P-values may occur because of the monotonicity constraint inherent to the Bonferroni-Holmes procedure, which ensures that q-values do not decrease with increasing rank and can lead to identical values when raw P-values are similar in magnitude.

We conducted additional exploratory analysis using subscales and single item scores of the BDI-II⁹⁰ and HADS⁶⁴ to provide a more granular view on the different dimensions of negative affect. We used Kruskal–Wallis and Wilcoxon rank-sum tests on the residuals of a linear model including age and sex for item level analyses to account for the changed assumptions of data being ordinal scaled without giving up on the statistical correction for age and sex effects. We calculated subscales based on published analyses of the questionnaires' factor structures.^15,40,54 In addition, we also explored core negative affective and somatic symptoms in the same sample, regrouped into patients that exhibit FMS without clinically relevant depression (pure FMS), MDD without clinically relevant chronic pain (pure MDD), and a mixed group suffering from both chronic pain and depression to some degree (mixed). These exploratory analyses were not corrected for multiple testing.

All statistical analyses were conducted in R (version 4.3.1).

2.3. Daily life ambulatory assessments

Participants used study smartphones (Nokia 4.2) equipped with the movisensXS experience sampling software (movisens GmbH, Karlsruhe, Germany) to report on their daily life pain experience, affective state, and psychological distress for 1 week. We used a flexible time- and location-dependent sampling schedule, resulting in an average of 10.92 ± 1.89 prompts per day, as previously described.^27,79 This enriched sampling method, which includes GPS-linked data sampling during eventful periods of daily life, addresses the special challenge of obtaining sufficient variance and covering the variety of daily life events.¹⁶ A schematic representation of the design is depicted in Figures 2A–C, and a detailed description of the assessment is given in SI 3, http://links.lww.com/PAIN/C379. Briefly, we assessed different pain (being in pain, pain intensity, unpleasantness, interference, and spread) and mood (valence, energetic arousal, calmness) dimensions as well as anxiety and perceived stress at each prompt.

Figure 2.

Schematic of EMA and fMRI neurofeedback method. (A) Schematic of example EMA diary item as seen by participant on study smartphone. Note that during the experiment, each item was presented after the other, here 2 items are depicted together for illustration purposes. (B) Schematic of time-based triggering of prompts and analysis of resulting time series data where variable x predicts variable y at the subsequent time point. (C) In addition, prompts were also triggered by changes in location as measured via GPS to enrich sampling of eventful periods during daily life. (D) Schematic of neurofeedback process: participants regarded aversive visual stimuli and engaged in emotion regulation when prompted to do so. Brain activation was continuously measured using fMRI, preprocessed and analyzed on a dedicated external device. An adaptively scaled feedback of amygdala activation was calculated and fed back to participants. (E) Single trial schematic. (F) Experiment design and contrast of interest: the experiment consisted 5 equally sized blocks of 10 trials each with balanced numbers of conditions. In the first and last block, no feedback was provided to assess baseline regulation capacity and transfer of learned regulation, respectively. Training with feedback consisted of 3 blocks. The contrast of interest was the difference between regulate and view trials at baseline compared to the difference between regulate and view trials at transfer (lower row). fMRI, functional magnetic resonance imaging.

We conducted multilevel analyses of daily life data in SAS (version 9.4, SAS Institute Inc., Cary, NC) to test for group differences, thereby nesting within-subject e-diary assessments (level 1) within participants (level 2) and using a categorical group variable (level 2).⁸ To assess the temporal dynamics of relationships between antecedent and subsequent daily life experiences, we used time-shifted multilevel regression analysis. These models investigate stress reactivity by using lagged (ie, backwards shifted in time) perceived stress as a predictor for psychological and pain symptoms at the subsequent time point. All multilevel models statistically controlled for age, sex, and group, as well as time of day and time of day squared. Following standard quality control procedures, we included only data from subjects who answered at least 30% of prompts to ensure representativeness and visually inspected multilevel model residuals for deviations from normality.⁷⁹ Further details on the multilevel models can be found in SI 4, http://links.lww.com/PAIN/C379. Daily life data were also analyzed at an aggregated daily life data by calculating averages across the assessment period; these data were treated as described in the psychometrics section.

2.4. Functional magnetic resonance imaging amygdala neurofeedback task

Participants completed one 30-minute session of fMRI-based real-time amygdala neurofeedback training (NFB). During this session, they viewed aversive pictures and received instructions to either downregulate negative emotions (“regulate” condition) or to simply attend to the image (“view” condition). Whole-brain fMRI data were recorded during the entire session and continuously processed on an external device, see Figure 2D for a schematic of the neurofeedback rationale. Quality control and movement parameters are reported in SI 5 and 10, http://links.lww.com/PAIN/C379.

2.4.1. Experimental setup

The fMRI-NFB training session consisted of 50 trials. Each trial comprised the presentation of one visual stimulus for 18 seconds and an interstimulus interval of 12 seconds, see Figure 2E. Stimuli were selected from the International Affective Picture Set (IAPS) standardized picture series⁴⁴ to elicit moderate to high negative valence and arousal, details on IAPS stimuli can be found in the SI. Participants received instructions whether to regulate negative emotions (“regulate” condition) or only attend the stimulus (“view” condition) during stimulus presentation.

The 50 trials form 5 equally sized blocks of 10 trials each: during the first block, we assessed emotion regulation without providing feedback, we refer to this as the baseline block. In the following 3 blocks, participants received feedback of their amygdala activity in form of a thermometer bar to the left and right of the stimulus presentation. This feedback signal was calculated as the percent signal change of the amygdala BOLD response relative to the global mean as previously described.⁹⁸ The feedback signal was shown in both “regulate” and “view” conditions and continuously updated at every repetition time. In the last of the 5 blocks, referred to as the transfer block, the feedback signal was removed to assess the stability of potential improvements in self-regulation in the absence of provided feedback.

Apart from the feedback being present or not, the 5 blocks did not differ: each block comprised 10 trials, 5 of each condition, in semi-randomized order (switching enforced after 2 similar conditions). Stimuli were presented with the Presentation software (Neurobehavioral Systems, Berkeley, CA).

Before entering the scanner, participants received standardized instructions and visual examples to familiarize themselves with the task. These included the design and aim of the task, examples of stimulus material, and specific instructions for the “regulate” and “view” conditions. The verbatim German version and a translation into English can be found in SI 5, http://links.lww.com/PAIN/C379. Participants were free to choose any strategy to regulate their emotions and control the thermometer that they might find beneficial.

2.4.2. Scanning parameters

We acquired fMRI data using a multiband echoplanar imaging (EPI) sequence developed at the Center for Magnetic Resonance Research, University of Minnesota (Release R016a) with repetition time (TR) = 1000 milliseconds, multiband acceleration factor of 6, field of view = 204 mm, voxel size 2.4 mm³ isometric and 60 slices. A subset of N = 2 participants received feedback using a TR of 800 milliseconds (all other settings identical); these data were not included in the present study for comparability reasons and because we observed previously unreported irregularities in image reconstruction timings.⁶⁸ Whole-brain T1-weighted images were collected using an MP2RAGE protocol (repetition time [TR] = 2000 ms, echo time [TE] = 3.03 ms; voxel size = 1.0 mm³ isometric; flip angle = 9°; parallel imaging using GRAPPA with acceleration factor = 2). All MRI data were collected using a 3 T Siemens Prismafit scanner with Windows 7-based VE11C console software and a 64-channel head coil (Siemens, Berlin) at the Central Institute of Mental Health, Mannheim.

2.4.3. Analysis

For online fMRI analysis and feedback calculation, EPI images were directly transferred to a separate PC running a custom MATLAB script (version R2016b) and SPM12 (Wellcome Department of Cognitive Neurology, London, United Kingdom) for online preprocessing and calculation of the feedback signal. For a detailed description of the hardware setup, see our previous publication.⁶⁸ The amygdala neurofeedback was calculated as previously described,⁹⁸ and details on the procedure are provided in SI 6, http://links.lww.com/PAIN/C379.

Standard preprocessing was performed using fMRIPrep 20.2.1,^19,20 which is based on Nipype 1.5.1.^32,33 A containing detailed description of all preprocessing steps can be found in SI 7, http://links.lww.com/PAIN/C379.

After preprocessing, the first 4 functional images were discarded to mitigate EPI equilibration effects, the remaining images were spatially smoothed using an 8-mm full-width at half maximum Gaussian kernel. First and second level analyses were implemented in SPM12.

A general linear model was built for each participant using boxcar regressors for each condition type (REG, VIEW) within each block type (Baseline, Training, Transfer). For each trial, an 18-second window of volumes of interest time-locked to the onset of stimulus presentation was defined. The 6 combinations of trial types and block types were modeled as separate regressors in the same model and convolved with the canonical hemodynamic response function. FAST pre-whitening was used to correct for autocorrelation⁶¹ and high-pass filtering at 256 Hz to remove drift. To correct for physiological noise and movement, we followed established procedures⁷⁷ to estimate and add 18 RETROICOR regressors to the model, based on a Fourier expansion of cardiac and respiratory phase and their first-order multiplicative terms. In addition, we estimated a heart-rate variability and a respiratory-volume-by-time regressor by convolving the cardiac trace and respiratory trace with their respective cardiac and respiration response functions.

The primary contrast of interest for this study was the difference between “regulate” and “view” trials at baseline compared to the difference between “regulate” and “view” trials in the transfer block: Baseline (REG > VIEW) > Transfer(REG > VIEW). Thus, higher T-values in this contrast reflect decreases in amygdala activation, indicating more successful learning of emotion regulation. See Figure 2F for an illustration.

Contrasting “regulate” and “view” trials ensures that activation unique to regulatory processes is captured and effects of stimulus material and habituation to the task are removed from the results. Contrasting baseline with transfer blocks ensures that regulation of brain activation is measured with regard to individual preexisting activation patterns and regulation capabilities and reduces potential confounds of meta-cognitive processes of the training itself.⁶³

Planned contrasts of activation in “regulate” over “view” trials (REG > VIEW) for each block type were modeled from the 6 trial x block regressors. For the target contrast Baseline (REG > VIEW) > Transfer (REG > VIEW), the planned contrasts of “regulate” over “view” at baseline and at transfer were combined at the subject level (first level), training run contrasts were only used to establish the validity of the paradigm.

We then analyzed the statistical parametric maps of this target contrast for each subject at the second level using a univariate analysis of variance (ANCOVA) to assess group differences while controlling for age and sex. Statistical significance was assessed using peak voxel intensity at P < 0.05, peak voxel-level familywise error (FWE) corrected for multiple comparisons in an a priori defined mask of the bilateral amygdalae derived from the Automated Anatomical Labeling atlas (region of interest analysis).⁷¹ We expected that HC would be better able than patients with FMS and MDD to learn to downregulate their amygdala activity in the transfer compared to the baseline block. To corroborate the real-life significance of amygdala downregulation success, individual mean contrast estimates of downregulation in the region of interest were used as second-level predictors of daily life affective well-being.

3. Results

3.1. Clinical characterization

Core clinical characteristics of the 3 participant groups are shown in Table 1 (full statistical summaries, results of omnibus tests for group differences, and post-hoc comparisons). The 3 domains psychological pathology, pain pathology, and behaviors contributing to or indicative of affective dysregulation are visualized in Figure 3. The full set of measures, all pairwise comparisons, and an overview of comorbidities can be found in SI 8 (Tables 6–11, http://links.lww.com/PAIN/C379) along with an assessment of the potential impact of differing recruitment channels across patient groups (SI 2, Tables 2–4, http://links.lww.com/PAIN/C379).

Table 1.

Core clinical and demographic characteristics by group.

Variable	FMS			MDD			HC			Omnibus test*		Tukey HSD MDD − FMS		Post-hoc FMS + MDD > HC
	N	Mean ± SD	CI	N	Mean ± SD	CI	N	Mean ± SD	CI	Test statistic	adj. P	Estimate	adj. P	Test statistic	adj. P
Sex (female: N, %)	42 of 46 (91%)			31 of 48 (65%)			21 of 34 (62%)			11.83	0.004	−1.89	0.007	0.76	0.154
Age	46	52.07 ± 8.82	49.45, 54.69	48	36.25 ± 12.56	32.60, 39.90	35	42.7 ± 11.93	38.67, 46.87	35.57	<0.001	−16.50	<0.001	0.02	0.880
Recruitment (primary channel: N, %)	14	Tertiary care pain clinic: N = 10 (71%)		41	Website of CIMH†: N = 16, (39%)		31	Website of CIMH†: N = 12, (34%)		—‡	<0.001	0.86	<0.001	0.66	<0.001
Pain intensity (CPG)§	44	61.67 ± 16.23	56.73, 66.60	23	40.72 ± 18.96	32.52, 48.92	0	—	—	—	—	−14.69	0.012	—	—
Pain spread (WPI)§	41	12.88 ± 4.46	11.47, 14.29	20	4.95 ± 3.79	3.18, 6.72	0	—	—	—	—	−7.09	<0.001	—	—
Somatization (SSD-12)§	44	25.55 ± 8.04	23.10, 27.99	23	18.89 ± 9.24	14.89, 22.89	0	—	—	—	—	−6.96	0.023	—	—
Perceived stress (PSS)	44	15.77 ± 4.37	14.44, 17.10	47	15.53 ± 4.75	14.14, 16.93	34	7.44 ± 4.80	5.77, 9.12	38.01	<0.001	−1.29	0.504	74.61	<0.001
Depression (BDI-II)	44	21.64 ± 8.89	18.93, 24.34	47	21.98 ± 12.98	18.17, 25.79	34	3.88 ± 3.80	2.56, 5.21	38.69	<0.001	1.47	0.823	77.44	<0.001
Anxiety (HADS-A)	44	9.52 ± 3.47	8.47, 10.58	47	8.55 ± 4.04	7.37, 9.74	34	3.09 ± 2.93	2.07, 4.11	31.36	<0.001	0.15	0.985	63.19	<0.001
Catastrophization (CERQ)	44	5.80 ± 2.96	4.90, 6.70	47	6.04 ± 2.82	5.21, 6.87	34	4.29 ± 1.68	3.71, 4.88	5.00	0.009	0.95	0.328	7.89	0.006
Rumination (CERQ)	44	7.77 ± 3.40	6.74, 8.81	47	9.38 ± 3.15	8.46, 10.31	34	6.09 ± 2.85	5.09, 7.08	11.09	<0.001	2.35	0.010	12.55	<0.001
Self-blame (CERQ)	44	6.25 ± 3.48	5.19, 7.31	47	8.77 ± 3.44	7.76, 9.78	34	4.32 ± 1.84	3.68, 4.96	17.90	<0.001	2.59	0.003	22.74	<0.001

Unless otherwise specified, values provided are mean and standard deviation. Bold P-values are those that remain statistically significant at alpha level <0.05 after FDR correction.

^*ANCOVA, Kruskal–Wallis or χ²/Fisher exact test, depending on measurement level and normality of data. ANCOVA with age and sex as covariates.

^†The Central Institute of Mental Health is a university medical clinic and constantly informs about current studies offering participation.

^‡Fisher exact test does not provide a test statistic, for post-hoc comparisons, Cramér V was calculated.

^§Questionnaires relating to chronic pain only assessed for FMS and subset of MDD who endorsed having chronic pain; thus, no omnibus test necessary and no post-hoc FMS + MDD > HC test possible.

BDI, Beck Depression Inventory; CERQ, Cognitive Emotion Regulation Questionnaire; CIMH, Central Institute of Mental Health; CPG, Chronic Pain Grade Scale; FMS, fibromyalgia syndrome; HADS, Hospital Anxiety and Depression Scale; HC, healthy controls; MDD, major depressive disorder; PSS, Perceived Stress Scale; SSD-12, Somatic Symptom Disorder B Criteria Scale 12; WPI, Widespread Pain Index.

Figure 3.

Psychometric characteristics of participants by group. Error bars depict standard error of the mean, significant group differences between FMS and MDD and between patients and healthy controls are denoted with asterisks, *P < 0.05, **P < 0.01, ***P < 0.001, corrected for multiple testing. (A-C) Pain symptoms based on CPG, WPI, and SSD-12. (D-F) Psychological symptoms based on BDI-II, STAIX2-T, and PSS. (G-I) Cognitive emotion regulation strategies, assessed using the CERQ. BDI, Beck Depression Inventory; CERQ, Cognitive Emotion Regulation Questionnaire; CPG, Chronic Pain Grade Scale; FMS, fibromyalgia syndrome; MDD, major depressive disorder; PSS, Perceived Stress Scale; SSD-12, Somatic Symptom Disorder B Criteria Scale 12; WPI, Widespread Pain Index.

Compared to healthy controls, patients reported significantly higher psychological symptom load in all aspects assessed, higher level of disease-agnostic impairments of activities and participation, more pain catastrophizing, health anxiety, and alexithymia (all P values < 0.001). Healthy controls also reported significantly less maladaptive emotion regulation strategies than patients, including “catastrophizing,” “self-blame,” and “rumination” (P = 0.006, P < 0.001, and P < 0.001, respectively) and significantly more adaptive strategies, such as “positive reappraisal” and “positive refocusing” (both P values < 0.001). We did not find any significant differences between patients and controls with regard to the use of “other blame,” “putting into perspective,” “refocusing on planning,” and “acceptance” as ER strategies (all P values > 0.05 in omnibus test).

The 2 patient groups did not differ significantly in their psychological symptom burden (Figs. 3A–C), including perceived stress, depressive symptoms, and clinical anxiety (all P values > 0.05, see SI 8 Table 9, http://links.lww.com/PAIN/C379). The 2 patient groups differed in pain-specific domains, with patients with FMS reporting consistently higher symptom burden in terms of pain intensity, pain spread, and somatization (P = 0.012, P < 0.001 and P = 0.023, respectively), although only those patients with MDD who endorsed having chronic pain answered these questionnaires and were included in these comparisons (Figs. 3D–F). In addition, patients with FMS reported more health anxiety than patients with MDD and a higher level of disease-agnostic impairments of activities and participation (P = 0.023 and P < 0.001, respectively, see SI 8 Table 9, http://links.lww.com/PAIN/C379). We found no evidence of group differences in terms of neuroticism, pain catastrophizing, or alexithymia (all P values > 0.05, see SI 8 Table 9, http://links.lww.com/PAIN/C379). General catastrophizing scores were similar between FMS and MDD (Figs. 3G), but patients with MDD reported more use of the maladaptive ER strategies “rumination” and “self-blame” (P = 0.010 and P = 0.003, respectively) than patients with FMS (Figs. 3H and I).

We performed additional exploratory analyses using subscales and single item scores of the BDI-II and HADS to provide a more granular view on the different dimensions of negative affect. These analyses indicated that patients with MDD suffer more from the cognitive dimensions of negative affect than do patients with FMS (SI 8.7, Tables 12–14, http://links.lww.com/PAIN/C379).

Because many patients with FMS suffer from depression at some point in their lifetime, and many depressed patients report pain as part of their phenotype, we performed additional exploratory analysis of 3 patient subgroups. In this study, N = 12 or 26.1% of the 46 patients with FMS additionally fulfilled diagnostic criteria of MDD. Our analyses suggested a higher and more diverse symptom load in the mixed pain/depression patient group than in patients with either pure FMS or pure MDD, for details see SI 8.7, Tables 12-14, http://links.lww.com/PAIN/C379.

3.2. Daily life assessments

The results of the daily-life assessments reflected the pattern of the psychometric evaluations: the 2 patient groups differed most significantly in pain-specific symptoms, but not in psychological symptoms (Fig. 4 and Table 2). Specifically, pain frequency (the proportion of prompts in which pain was reported) was significantly higher in patients with FMS than in patients with MDD (P < 0.001, FDR-corrected). All further questions about different characteristics of momentary pain referred only to those prompts that endorsed currently being in pain. Although momentary pain intensity, unpleasantness, and interference were all higher in patients with FMS than in patients with MDD, these differences were not statistically significant (all P values > 0.05, FDR-corrected). In contrast, momentary pain was significantly more widespread in patients with FMS than in patients with MDD (P = 0.014, FDR-corrected). There were no significant differences in average daily life valence, calmness, or energetic arousal between FMS and MDD and no significant differences in terms of momentary stress, state anxiety, or rumination (all P values > 0.05, FDR-corrected). Daily-life emotional instability, as indexed by the mean successive squared differences (MSSD) in valence, did not differ significantly either between the 2 patient groups (P > 0.05, FDR-corrected).

Figure 4.

Average daily life assessments by group, error bars show standard error of the mean, significant group differences are denoted with asterisks, *P < 0.05, **P < 0.01, ***P < 0.001, corrected for multiple testing. (A-C) Pain symptoms. Note that data for pain intensity and spread only refers to data points where momentary pain was present. (D-F) Psychological symptoms.

Table 2.

Average daily life measures by group.

Variable	FMS			MDD			HC			Omnibus test*		Tukey HSD MDD − FMS		Post-hoc FMS + MDD > HC
	N	Mean ± SD	CI	N	Mean ± SD	CI	N	Mean ± SD	CI	Test statistic	adj. P	Estimate	adj. P	Test statistic	adj. P
Pain frequency [%]	44	85.96 ± 22.23	[79.19, 92.72]	47	21.21 ± 26.90	[13.31, 29.11]	33	2.76 ± 6.99	[0.28, 5.23]	117.16	<0.001	−59.73	<0.001	57.25	<0.001
Pain intensity	44	47.96 ± 17.21	[42.73, 53.19]	38	34.32 ± 15.13	[29.35, 39.3]	20	23.41 ± 14.48	[14.48, 32.35]	9.61	<0.001	−8.70	0.137	15.10	<0.001
Pain spread [% of body]	44	48.41 ± 21.17	[41.98, 54.85]	38	24.75 ± 18.23	[18.76, 30.74]	19	28.44 ± 25.89	[15.96, 40.92]	4.83	0.011	−15.99	0.014	1.30	0.258
Pain interference	44	45.41 ± 18.92	[39.66, 51.16]	38	33.41 ± 15.74	[39.66, 51.16]	20	25.51 ± 21.54	[15.43, 35.59]	4.90	0.011	−7.33	0.344	7.76	0.008
Pain unpleasantness	44	47.50 ± 18.29	[41.94, 53.06]	38	34.16 ± 19.82	[27.65, 40.68]	20	25.75 ± 22.23	[15.34, 36.15]	5.16	0.009	−7.93	0.251	7.64	0.008
Valence	44	61.78 ± 14.86	[57.26, 66.3]	47	64.55 ± 16.33	[59.75, 69.34]	34	87.31 ± 11.81	[83.19, 91.43]	32.81	<0.001	4.65	0.435	63.81	<0.001
Calmness	44	56.46 ± 14.35	[52.1, 60.83]	47	58.69 ± 14.92	[54.3, 63.07]	34	81.05 ± 14.29	[76.07, 86.04]	29.90	<0.001	1.58	0.906	60.03	<0.001
Energetic arousal	44	40.26 ± 12.01	[36.61, 43.91]	47	46.22 ± 13.33	[42.31, 50.13]	34	71.60 ± 13.95	[66.73, 76.46]	57.71	<0.001	7.30	0.066	106.74	<0.001
Stress	44	48.06 ± 14.69	[43.59, 52.52]	47	42.95 ± 19.92	[37.1, 48.8]	34	16.69 ± 13.60	[11.95, 21.44]	35.93	<0.001	−6.85	0.234	68.24	<0.001
State anxiety	44	39.66 ± 14.25	[35.32, 43.99]	47	37.09 ± 15.49	[32.54, 41.64]	34	16.47 ± 13.28	[11.84, 21.11]	26.82	<0.001	−3.28	0.651	52.95	<0.001
Rumination	44	20.33 ± 15.43	[15.64, 25.02]	47	22.61 ± 19.42	[16.9, 28.31]	34	8.00 ± 9.88	[4.55, 11.44]	8.76	<0.001	2.09	0.866	17.37	<0.001
Emotional instability [MSSD valence]	44	397.28 ± 273.17	[314.22, 480.33]	47	403.42 ± 353.74	[299.55, 507.28]	34	236.88 ± 279.72	[139.28, 334.48]	3.60	0.030	−53.50	0.773	6.76	0.011

Unless otherwise specified, values provided are mean and standard deviation. Bold P-values are those that remain statistically significant at alpha level <0.05 after FDR correction.

^*ANCOVA with age and sex as covariates.

FMS, fibromyalgia syndrome; HC, healthy controls; MDD, major depressive disorder; MSSD, mean successive squared differences.

As in the psychometric assessments, both patient groups differed very clearly from healthy controls on all symptom dimensions assessed, except for pain spread. Specifically, HC reported higher average valence, calmness, and energetic arousal throughout their daily life (all P values < 0.001, FDR-corrected) as well as lower pain intensity, unpleasantness, and interference (P < 0.001, P = 0.008, and P = 0.008, respectively, all P-values FDR-corrected) for prompts at which momentary pain was endorsed. In addition, HC experienced lower levels of perceived stress, anxiety, and rumination (all P values < 0.001, FDR-corrected) and exhibited significantly less emotional instability as assessed by the MSSD in valence (P = 0.011, FDR-corrected).

We further investigated the association between momentary variations in perceived stress and subsequent variations of valence, anxiety, rumination, and pain as indicators of potential effects of stress on psychological well-being and pain experience. Across all groups, relative increases of perceived stress were followed by lower valence and heightened anxiety, rumination, and pain intensity (all P values < 0.001, FDR-corrected). In addition, we observed significant group differences in the relationship between within-subject centered perceived momentary stress and valence (P = 0.009), momentary anxiety (P = 0.044), and rumination (P = 0.031) at the following prompt, see Figure 5 for a visualization of the multilevel modelled fixed effects by group.

Figure 5.

Differential reaction to daily life stress at subsequent prompt. Visualization of multilevel models of fixed effects for each group and over entire study population (black), standard errors as shaded area. (A) Momentary stress and subsequent valence. (B) Momentary stress and subsequent anxiety. (C) Momentary stress and subsequent rumination. (D) Momentary stress and subsequent pain intensity.

We observed significant differences of the negative affective stress response between groups. Specifically, increased perceived stress was linked to greater reductions of valence in both patients with MDD (P = 0.025, FWE-corrected for pairwise comparisons) and FMS (P = 0.002, FWE-corrected for pairwise comparisons) compared to HC. Similarly, increased perceived stress was linked to greater increases in subsequent anxiety and rumination in patients with MDD (P = 0.132 and P = 0.014, respectively, FWE-corrected for pairwise comparisons) and patients with FMS (P = 0.013 and P = 0.023, respectively, FWE-corrected for pairwise comparisons) compared to HC. As before, differences between the 2 patient groups were not statistically significant (all P values > 0.05).

Although we did find significant correlations of perceived momentary stress, valence, anxiety, and rumination (P < 0.001, P < 0.001, P = 0.003, and P = 0.027, respectively, FDR-corrected) with pain intensity at the subsequent prompt, these effects were independent of group (all P values of interaction term > 0.05, see Fig. 5D). Notably, although stress predicted pain intensity at the subsequent time point, it did not predict whether or not an individual would be in pain or not (P = 0.093).

Group effects from the multilevel models are visualized in Figure 5, and detailed results presented in Table 3.

Table 3.

Type III tests of fixed effects of time-lagged multilevel models of the relationship between perceived stress and subsequent affective state and pain and of affective state and subsequent pain.

Model	F Statistic	p-value	Degrees of freedom
Stress → subsequent valence	174.59	<0.001	1, 123
Stress × group → subsequent valence	4.96	0.009	2, 120
Stress → subsequent anxiety	190.87	<0.001	1, 128
Stress × group → subsequent anxiety	3.21	0.044	2, 126
Stress → subsequent rumination	52.94	<0.001	1, 116
Stress × group → subsequent rumination	3.59	0.031	2, 115
Stress → subsequent pain intensity	26.10	<0.001	1, 69.4
Stress × group → subsequent pain intensity	0.56	0.576	2, 79
Stress → subsequent pain binary*	2.87	0.093	1, 123
Valence → subsequent pain intensity	18.26	<0.001	1, 47.5
Valence × group → subsequent pain intensity	2.34	0.105	2, 66.3
Anxiety → subsequent pain intensity	9.62	0.003	1, 49
Anxiety × group → subsequent pain intensity	0.89	0.417	2, 59
Rumination → subsequent pain intensity	5.27	0.027	1, 38
Rumination × group → subsequent pain intensity	0.23	0.796	2, 55.4

Degrees of freedom are provided as numerator, denominator.

^*On the prompt level, pain frequency is a binary outcome variable (being in pain or not); thus, a generalized mixed linear model of the logit family was used here.

Bold indicates P < 0.05.

3.3. Amygdala neurofeedback

Having found converging evidence of a distinctive overlap of ER deficits and associated clinical symptoms in patients with MDD and FMS compared to HC, we assessed a potential shared neurobiological signature of ER deficits in patients with MDD and FMS. We found significant group differences between HC and patients in the ability to downregulate brain activation within a bilateral amygdala mask after fMRI neurofeedback training compared to baseline ability. Specifically, the statistical parametric map in the contrast of interest (pre–post neurofeedback training effect in “regulate” > “view” conditions) showed higher T-values in the left amygdala for the HC group compared to the MDD and FMS patient groups (Figs. 6A and B, Montreal Neurological Institute (MNI) space coordinates of the peak voxel: x = −18, y = −2, z = −12, t = 3.13, pFWE = 0.039, FWE-corrected for voxels in the bilateral amygdala using the AAL3 atlas).⁷¹ Since here, higher T-values signify more successful downregulation, this indicates that HC were better able to learn emotion regulation after neurofeedback training than patients were. Exploratory whole-brain analysis did not reveal any other areas of significantly increased downregulation in HC compared to patients.

Figure 6.

Neurobiological marker of emotion regulation capacity and its correlation with daily life psychological well-being. (A) Brain area located in the left amygdala which HC downregulate to greater degree than patients after NFB training. (B) Peak voxel estimates of individual activation within left amygdala, by group. Higher contrast estimates indicate more downregulation. (C–E) Correlations of left amygdala downregulation with daily life valence, anxiety, and rumination. Note that for visualization purposes, daily life average values were plotted, and statistical analyses were conducted using appropriate multilevel models. (F) Brain area in the right amygdala which patients with MDD downregulate to a greater degree than patients with FMS. (G) Peak voxel estimates of individual activation differences within right amygdala. (H) Correlations of right amygdala downregulation with daily life pain spread. FMS, fibromyalgia syndrome; HC, healthy controls; NFB, neurofeedback training; MDD, major depressive disorder.

To ensure the validity of our paradigm, we conducted additional region of interest (ROI) analyses to confirm expected amygdala activation patterns: we contrasted View > Regulate conditions over the entire experiment and observed increased amygdala activation in all groups, as expected (FMS: T = 3.38, PFWE = 0.020; MDD: T = 3.98, PFWE = 0.003; HC: T = 3.40, PFWE = 0.019. All P-values FWE-corrected for voxels in the bilateral amygdala using the AAL3 atlas).⁷¹ Similarly, a comparison of all VIEW conditions with the implicit baseline (rest periods and breaks with no stimuli shown) revealed significantly increased amygdala activation across all groups as was expected (FMS: T = 6.87, PFWE = 2.37E-08; MDD: T = 7.85, PFWE = 2.03E-10; HC: T = 7.38, PFWE = 2.24E-07. All P-values FWE-corrected for voxels in the bilateral amygdala using the AAL3 atlas).⁷¹ Details of these analyses and the results are presented in SI 10.2, http://links.lww.com/PAIN/C379.

To rule out that the observed learning effect was caused by preexisting baseline differences, we assessed all pairwise group comparisons of amygdala activation for baseline regulate trials, baseline view trials, and the baseline regulate > view contrast using the same approach as described above. No significant differences in amygdala activation were found during regulate or view trials. For the regulate > view contrast, both patients with MDD and HC showed a trend toward higher amygdala activation compared to patients with FMS (T = 3.00, PFWE = 0.055 and T = 2.78, PFWE = 0.093, respectively, all P-values FWE-corrected for voxels in the bilateral amygdala using the AAL3 atlas).⁷¹ No other group comparisons revealed any differences, and we found no evidence that any group exhibited higher amygdala activation during regulate compared to view at baseline. Further analyses pooling all patients and comparing them to healthy controls did not reveal any baseline amygdala activation differences. For more details, see SI 10.3, http://links.lww.com/PAIN/C379.

Next, we combined individual beta weights of left amygdala regulation success with more ecologically valid ambulatory assessment data: the ability to learn to downregulate the left amygdala after neurofeedback training correlated significantly and positively with momentary valence and calmness (P = 0.040, uncorrected and P = 0.042, uncorrected, respectively) and significantly and negatively with momentary anxiety over all 3 groups in daily life (P = 0.036, uncorrected), even after statistically controlling effects of group, age and sex. For a visualization using the average daily life values of valence and anxiety, see Figures 6C and D, and for the full models of all tests, see SI 9, http://links.lww.com/PAIN/C379. The ability to learn to downregulate the amygdala after neurofeedback training was also a significant predictor of rumination in daily life across all 3 groups (P = 0.023, uncorrected). Here, we additionally found a significant interaction effect for group (P = 0.030), with the relationship between the neurobiological marker and daily life rumination being more pronounced for patients with FMS (P = 0.015, FWE-corrected for 3 pairwise group comparisons) and MDD (P = 0.031, FWE-corrected for 3 pairwise group comparisons) than for HC (see Fig. 6E for a visualization using the average daily life values of rumination and SI 9, http://links.lww.com/PAIN/C379). No other psychological daily life measure assessed here (momentary stress and energetic arousal) was significantly linked to regulation success within the left amygdala (P = 0.166, and P = 0.207, uncorrected, respectively). After false discovery rate correction for performing 6 tests (for momentary valence, calmness, anxiety, rumination, stress, and energetic arousal) in total, all results remain marginally significant at P = 0.063.

To capture potential differences between patients with FMS and MDD in the neurofeedback task, we defined 2 additional comparisons on the second level, FMS > MDD and MDD > FMS in a follow-up analysis (all other parameters unchanged). We found a significant difference in downregulation in the right amygdala (MNI space coordinates of peak voxel at x = 24, y = 2, z = −24, T = 3.91, PFWE = 0.004, corrected for voxels in the bilateral amygdala using the AAL3 atlas)⁷¹ in the MDD > FMS comparison of the contrast of interest (pre–post neurofeedback training effect in “regulate” > “view” conditions). This indicates that patients with MDD were better able to downregulate brain activation in the right amygdala after neurofeedback training than patients with FMS (Figs. 6F and G). Exploratory whole-brain analyses revealed several additional clusters of comparatively better downregulation in patients with MDD, but these did not survive whole-brain FWE correction for multiple testing. There were no significant differences in the reverse contrast (FMS > MDD) on either whole-brain level or in the amygdala ROI analysis, indicating that there were no brain regions in which patients with FMS downregulated brain activation significantly more strongly than patients with MDD. Patients' ability to downregulate the right amygdala after neurofeedback training correlated significantly and negatively with pain spread in daily life (P = 0.011, uncorrected, see Fig. 6H, SI 9, http://links.lww.com/PAIN/C379), even after statistically controlling effects of group, age and sex. We did not find any association with pain intensity, unpleasantness, or interference (all P values > 0.05, uncorrected). After false discovery rate correction for performing 4 tests in total (for pain spread, intensity, unpleasantness, and interference), the negative correlation with pain spread remained significant at P = 0.044, FDR-corrected.

To further explore these observed group differences in fMRI neurofeedback-assisted amygdala regulation learning, we examined the contrast of interest for each participant group separately. We found significantly larger differences in amygdala activation between regulate and view trials after neurofeedback compared to baseline for HC (T = 3.27, P = 0.027, corrected for voxels in the bilateral amygdala using the AAL3 atlas)⁷¹ and marginally larger differences for patients with MDD (T = 2.93, P = 0.065, corrected for voxels in the bilateral amygdala using the AAL3 atlas).⁷¹ For patients with FMS, we did not find any statistically significant amygdala downregulation in regulate vs view trials after neurofeedback training compared to baseline, see also SI 10.4, http://links.lww.com/PAIN/C379.

4. Discussion

We assessed affective dysregulation using psychometric questionnaires, ecological momentary assessment, and an fMRI neurofeedback task to investigate a potential transdiagnostic risk phenotype in patients with FMS and MDD compared to HC. Consistent with our hypothesis, we found a pronounced overlap of pain and affective pathology in both FMS and MDD: FMS reported consistently more pain specific symptoms, yet MDD did not report higher levels of psychopathology. This pattern of similar psychopathology and different pain pathology was mirrored in daily life. Time-lagged analysis further showed that in all 3 groups, subjective stress predicted higher subsequent pain intensity and had detrimental psychological effects. Finally, we found shared (in patients compared to controls) and specific (in FMS compared to MDD) deficits in amygdala downregulation, which correlated with daily life psychopathological measures and pain spread, respectively.

4.1. Clinical observations

Given the lack of similar prior research, comparisons with existing findings are limited to studies that compare patients with FMS or MDD separately to HC or assessed other types of chronic pain. Therein, our findings confirmed previous research: clinical symptom overlap in MDD and FMS is well-established and has been previously discussed.^6,56,67,74 Deficits in ER of patients with FMS compared to HC, including alexithymia and use of maladaptive strategies, such as catastrophizing and rumination have previously been shown.^{13,17,18,35,51,57} Similarly, a robust body of research indicates ER deficits in MDD compared to HC, see Reference 48 for a review and meta-analysis. One study comparing patients with MDD to those with chronic musculoskeletal pain found higher catastrophizing in the MDD population.⁵⁸ In our data, only rumination and self-blame were more pronounced in MDD, whereas both catastrophizing and depression scores were similar in FMS and MDD; these findings support our hypothesis of a transdiagnostic psychological risk phenotype.

4.2. Temporal dynamics and relations between stress, pain, and affect

We found that in all participants, antecedent subjective stress predicted higher subsequent pain intensity to the same degree. One previous study found a comparable effect of momentary stress on pain intensity for patients with FMS,²² our data extend these findings by suggesting that this effect is probably not specific to chronic pain. Notably, stress did not predict the likelihood of being in pain in our study; a distinction that might not have been possible or obvious in other studies. One interpretation would be that, independent of diagnosis, stress exacerbates preexisting pain without initially causing it, the mechanism, therefore, might be most relevant for those who are already in pain.

For all participants, subjective stress also predicted worse subsequent affective states, including lower valence, higher anxiety, and more rumination. These effects, although apparent in all 3 groups, were significantly more pronounced in patients with FMS and MDD than in HC. Our findings are in line with another study reporting greater emotional reactivity to stress in patients with chronic pain compared to those without pain.⁴¹ Similarly, patients with MDD experience greater negative affective reactions to momentary stress than HC,^36,76 and Wichers et al.⁸⁹ suggested that genetic liability to depression is partly expressed as negative affective stress reactivity in daily life. Our results extend these observations by showing that not only valence but also anxiety and the tendency to ruminate are adversely affected by prior stress in daily life.

Finally, valence, anxiety, and rumination in turn each predicted higher subsequent pain intensity in our study, highlighting the importance of affective dysregulation in daily life experience independent of diagnosis and suggesting dimensional and transdiagnostic relationships. Our findings provide supporting evidence within the framework of recent models of an indirect pathway of ER deficits onto pain symptoms via emotional distress.^24,97

4.3. Neural substrates of impaired emotion regulation capacities in fibromyalgia syndrome and major depressive disorder

Regarding the ability to quickly adapt ER capabilities to a novel context, we found that, consistent with our hypothesis, both patients with FMS and MDD experienced difficulties in using amygdala neurofeedback for emotion regulation while viewing aversive pictures, reflected in lower individual downregulation of left amygdala activation. Consistent with a greater involvement of left amygdala in negative emotional processing,⁸⁸ especially for sustained stimulus evaluations,⁹³ this lack of downregulation correlated with daily life valence, anxiety, and rumination. In addition, patients with FMS exhibited reduced capacity to downregulate right amygdala compared to patients with MDD. This might reflect a predominantly right-lateralized role of the amygdala in nociceptive processing, as previously reported.² The observation that relatively preserved right amygdala downregulation correlated with less widespread pain is in line with the idea of a shift in the top-down control of the balance between habituation and sensitization in central nociceptive pathways via brainstem regions including the periaqueductal grey and rostroventromedial medulla.⁸¹ Such a shift has long been thought to underlie widespread pain syndromes^28,29 and is in line with observations in a rat model of stress-induced functional pain suggesting that stress-related kappa opioid signaling in the right central amygdala might shift the balance of descending noxious inhibitory controls towards increased facilitation.⁶⁰

Regarding neuromechanistic studies of ER using amygdala neurofeedback, there is no prior research comparing FMS and MDD against HC. One group used a comparable amygdala neurofeedback paradigm in separate studies of HC and MDD and also observed greater improvements in amygdala downregulation in the HC group.⁹⁶ No fMRI neurofeedback research comparing FMS and HC has been published to date.

Despite the relative difficulties patients with FMS and MDD experienced in using neurofeedback for emotion regulation, HC might simply have adapted faster in our experiment, thus retaining the beneficial effects after only one session of neurofeedback training. It is important to note that both FMS and MDD can successfully use neurofeedback to improve ER: potential therapeutic interventions based on amygdala neurofeedback are currently studied both for MDD⁹⁵ and FMS³¹ with randomized controlled trials showing promising results. In fact, the group interaction effect we observed in the relationship between the neural signature of ER deficits and daily life rumination suggests that neurofeedback training may be most beneficial not for HC, but for patients with FMS and MDD.

4.4. Strengths and limitations

This study has several general limitations. First, we recruited a study population reflecting the clinical reality of less readily separable diagnostic groups: patients with MDD with and without pain were included as well as patients with FMS scoring high on depression scales. Compared with an earlier Norwegian study,⁴³ we thus had a strong representation of FMS with lifetime depressive disorder; however, the fact that daily life stress effects were significant across all participant groups suggests that improved stress reactivity through ER should also be beneficial in patients with FMS without lifetime depressive disorder. Recruitment through different, disorder-specific channels was necessary, but might have introduced confounding effects. Second, although our time-lagged EMA models serve as micro-longitudinal observations and imply some temporal directionality, it is important to stress that these analyses cannot prove “true” causality and our study cannot fully disentangle the relationships between stable, personality-based risk factors, modifiable ER behaviors, and clinical psychological and pain symptoms. In fact, living with a chronic pain condition is a major burden and animal models have shown that sustained nociceptive processing may itself induce plastic cortical changes that lead to anxiety.⁹⁹ Third, the sample size of our fMRI tasks was limited because strict MRI safety criteria restricted inclusion, particularly of patients with FMS. Fourth, no direct behavioral indices of regulation success have been acquired during the session. Finally, although we focused on affective dysregulation as a transdiagnostic process, other frameworks like the approach/avoidance framework could provide further valuable insights.

4.5. Conclusions and outlook

Our study offers insights into the shared and unique characteristics of ER deficits in FMS and MDD and is the first study to directly compare clinical symptoms and ER deficits in patients with FMS and MDD using EMA and fMRI neurofeedback. Our results provide mechanistic evidence across modalities for transdiagnostic affective dysregulation processes in FMS and MDD. At the same time, we show differences in specific measures that appear to be disorder-specific, such as the extent of pain spread in FMS or rumination in MDD. The comprehensive EMA assessments across diverse daily experiences enhance the ecological validity of our findings and provide some level of evidence for the temporal dependencies of effects. In addition, we report neural markers of both shared and FMS-specific ER deficits in the amygdala during emotion regulation that may correspond to previously hypothesized alterations in the top-down control of the balance between habituation and sensitization in central nociceptive pathways via brainstem regions. In conclusion, our results suggest that addressing the underlying affective dysregulation phenotype could potentially benefit both patients with MDD and FMS. Future research should aim to identify individual predictive signatures leveraging key symptom and risk dimensions to distinguish distinct patient subgroups and ultimately guide tailored treatment strategies.

Conflict of interest statement

A.M.-L. has received consultant fees from the Daimler and Benz Foundation, EPFL Brain Mind Institute, Fondation FondaMental, Hector II Foundation, Invisio, Janssen-Cilag GmbH, Lundbeck A/S, Lundbeckfonden, Lundbeck Neuroscience Foundation, Neurotorium, MedinCell, The LOOP Zürich, University Medical Center Utrecht, University of Washington, the Mental Wellbeing Association, and the von Behring-Röntgen Foundation; speaker fees from Ärztekammer Nordrhein, Caritas, Clarivate, the German Society for Neuroscientific Assessment, Gentner Verlag, the State Medical Association Baden-Württemberg, LWL Bochum, Northwell Health, Ruhr University Bochum, Penn State University, the Society of Biological Psychiatry, the University Prague, and Vitos Klinik Rheingau; and editorial or author fees from the American Association for the Advancement of Science, the European College of Neuropsychopharmacology, Servier Int. and Thieme Verlag. R.D.-T. reports grants from Deutsche Forschungsgemeinschaft during the conduct of the study; grants from Innovative Medicines Initiative EU and EFPIA, from Esteve, TEVA, personal fees from Bayer, Grünenthal, GSK, Merz, Saluda Medical, Sanofi, Cered, and Vertex outside the submitted work. All other authors do not declare any conflicts of interest.

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C379.

Supplementary Material

jop-166-e788-s001.pdf ^{(2.2MB, pdf)}