1. Introduction
Acute pain signals tissue damage and potential threat to an organism's integrity. It is therefore essential to differentiate between situations or cues predicting harm and those predicting safety.64 In chronic pain, however, imminent physical threat is often missing, and an imbalance between threat and safety learning might promote maladaptive behavioral and emotional responses (eg, excessive avoidance behavior and pain-related fear). Over time, these can lead to physical deconditioning and affective distress, which in turn can drive the development or maintenance of chronic pain, as, for instance, described in the fear-avoidance model of pain.65
While pain-related and fear learning have been thoroughly investigated in healthy participants,26,44,46,59 mechanisms underlying their role in chronic pain is still insufficiently understood. Pain-related or fear learning is usually investigated using neutral stimuli as conditioned stimuli (CS), which predict the delivery (CS+) or absence (CS−) of aversive stimuli (= unconditioned stimuli [US]). Behavioral and physiological responses to CS were collected to quantify acquisition and extinction of conditioned responses.43 First studies in patients with chronic pain have suggested alterations in pain-related learning.22,46 For instance, impaired differential fear and contingency learning has been found in several pain conditions, such as fibromyalgia, chronic hand pain, or chronic neck pain.23,30,47,48 Of interest, these patient groups showed not only impaired threat learning but also reduced safety learning. Patients with irritable bowel syndrome, on the other hand, displayed enhanced safety learning compared with healthy volunteers.28
Although chronic back pain (CBP) is amongst the most common types of chronic pain, surprisingly few studies have investigated pain-related learning in this patient group so far. Two studies reported enhanced conditioned muscular responses in patients with CBP compared with those in healthy volunteers, which either persisted57 or disappeared34 after extinction training. A recent study found reduced differential learning in patients with persistent neck pain, with lower CS+ and higher CS− pain expectancies compared with those in healthy individuals, but comparable extinction rates.23 One reason for these inconsistent findings might be the sample sizes of these studies, which were rather small (ie, N = 30 or lower) for an investigation in a highly heterogeneous clinical population.
In this study, we examined potential alterations in pain-related learning in a sample of 62 patients with nonspecific CBP in comparison with 61 age- and sex-matched healthy volunteers (= healthy controls [HCs]). We used a differential conditioning paradigm to investigate group differences in acquisition and extinction of associations between visual cues as CS and phasic heat pain stimuli serving as US. To elucidate both, emotional and cognitive aspects of pain-related threat and safety learning, we obtained behavioral data (CS valence ratings, US-CS contingency ratings), as well as physiological data (electrodermal responses). We hypothesized that patients with CBP show impaired pain-related differential learning in comparison with HCs. More specifically, we assumed that this entails alterations in both pain-related threat (CS+) and safety (CS−) learning. Furthermore, we conducted exploratory analyses on the effects of psychological trait and state variables and pain-related parameters on learning.
2. Methods
2.1. Participants
Based on effect sizes reported in previous patient studies investigating pain-related learning,47,49 we calculated a required sample size of n = 63 per group to detect group differences in differential acquisition learning. Calculations were performed using the pwr package in R12 with the following parameters: d = 0.5, α = 0.05, and 1 − β = 0.80.
Sixty-seven patients with nonspecific CBP and 74 healthy volunteers (HCs) participated in a 2-day differential conditioning paradigm. Participants were recruited using local advertisements or through the local back pain center at the University Hospital Essen (U.B.) and a structured telephone screening. General inclusion criteria for both groups were as follows: age older than 18 and younger than 80 years, normal or corrected-to-normal vision, no acute infection, no participation in trials using investigational medicinal products within the last 3 months, and no alcohol consumption within the past 24 hours (assessed through self-report). Patients interested in study participation were screened for eligibility through telephone screening by trained study personnel. Further on-site screening of eligible patients was performed by physicians specialized in pain medicine (U.B. and J.K.-B.) through medical history and clinical examination. Patients experiencing nonspecific CBP (ie, exclusion of specific spinal pathologies, nerve root, postsurgical, or post-traumatic pain) were included in the study. In accordance with the European guidelines on CBP, the definition of CBP was given as remitting or persistent pain >12 weeks.1 Further exclusion criteria for the patients with CBP comprised a history of malignant diseases within the past 5 years, severe mental disorders (eg, major depression, psychosis, and schizophrenia), as well as opioid treatment with >100 mg of morphine equivalent per day. Any treatment had to be kept stable in the period of 3 weeks before study participation. Exclusion criteria for HCs comprised actual or history of internal, neurological, mental, pain-related, or dermatological diseases or cancer, regular consumption of recreational drugs, or intake of pain medication within the past 24 hours, all based on self-report. Five patients and 13 HCs had to be excluded from the analysis. In 3 patients with CBP and 2 HCs, pain intensity ratings did not reach the envisaged baseline level during calibration. In 2 patients with CBP and 1 HC, the mean pain intensity ratings remained below 30 on a 0 to 100 Visual Analogue Scale (VAS) during acquisition training. Two HCs showed clinically relevant levels of anxiety or depression, as assessed with the Depression Anxiety Stress Scale questionnaire (cutoff values: anxiety = 6, stress = 10, and depression = 10). Furthermore, 8 HCs were excluded because of protocol violation (eligibility after inclusion). The final data analysis was therefore based on 62 patients with CBP (18 male patients, age: 34.56 ± 13.37 (M ± SD) years) and 61 HCs (21 male individuals, age: 33.80 ± 11.83 years). Demographic information of the analyzed sample and pain-related patient characteristics are listed in Table 1.
Table 1 -
Demographic information and patient characteristics.
|
CBP group (N = 62) |
HC group (N = 61) |
| Demographic data |
|
|
| Age in years, M ± SD [range] |
34.6 ± 13.4 [16–69] |
33.8 ± 11.8 [19–70] |
| Sex |
|
|
| Female/male, n (%) |
44/18 (71.0/29.0) |
40/21 (65.6/34.4) |
| Education, n (%) |
|
|
| Basic (≤10 years) |
12 (19.4) |
7 (11.5) |
| High school (13 years) |
29 (46.7) |
40 (65.5) |
| University (>13 years) |
21 (33.9) |
14 (23.0) |
| Pain-related data, M ± SD [range] |
|
|
| Pain duration in years |
9.79 ± 8.73 [1–38] |
— |
| Mean back pain intensity (last 4 weeks), 1–10 NRS |
4.99 ± 1.56 [2–8] |
— |
| Maximum back pain intensity (last 4 weeks), 1–10 NRS |
7.48 ± 1.22 [5–10] |
— |
| Current back pain intensity on study day, 1–10 NRS |
3.38 ± 1.98 [0–8] |
— |
| Pain severity*, n (%) |
|
|
| Grade I (low pain intensity and disability) |
23 (37.1) |
— |
| Grade II (high pain intensity, low disability) |
27 (43.5) |
— |
| Grade III (high pain intensity and disability, moderately limiting) |
8 (12.9) |
— |
| Grade IV (high pain intensity and disability, severely limiting) |
4 (6.4) |
— |
| Type of medication, n (%) |
|
|
| Antidepressants |
2 (3.2) |
— |
| Nonopioid analgesics |
1 (1.6) |
1 (1.6)† |
| Pregabalin |
1 (1.6) |
— |
| Others‡ |
10 (16.1) |
10 (16.4) |
*Pain grading according to Von Korff et al., 1992.
†Daily dose: ASS 100 mg.
‡Other medication includes nonsteroidal anti-inflammatory drugs (NSAIDs), antipsychotics, antihistamines, antidiabetic medication, levothyroxine, HIV medication, asthma medication, bronchodilators, statins, COX-2 inhibitors, proton pump inhibitors, angiotensin-converting-enzyme (ACE) inhibitors, beta-blockers, angiotensin II type 1 (AT1) receptor antagonists, and calcium channel blockers. None of the patients with CBP took benzodiazepines, NSAID, or opioids (>100 mg of morphine equivalent/day according to our inclusion criteria).
CBP, chronic back pain; HC, healthy control; NRS, numeric rating scale.
The study was approved by the local Ethics Committee of the Medical Faculty of the University of Duisburg-Essen (University of Duisburg-Essen, Germany; 16-7248-BO). All participants gave their informed oral and written consent to participate, including their consent to publish, and were free to withdraw from the study at any time. Participants received a small monetary reimbursement for their study participation.
2.2. Experimental paradigm and procedures
The study was performed on 2 consecutive days. On the first day, participants filled in questionnaires assessing demographic information, anxiety, depression, stress, and pain-related psychological processing (see self-report questionnaires). We then determined the individual heat pain threshold at the location of stimulus application (left volar forearm, approximately 12 cm proximally from the wrist) and performed a calibration procedure to determine the temperature level that induced a pain intensity of 70 on a 0 to 100 VAS (“How painful was this stimulus?”, anchors: 0 = “not painful at all” and 100 = “unbearably painful”) using established procedures.19The study was performed on 2 consecutive days. On the first day, participants filled in questionnaires assessing demographic information, anxiety, depression, stress, and pain-related psychological processing (see self-report questionnaires). We then determined the individual heat pain threshold at the location of stimulus application (left volar forearm, approximately 12 cm proximally from the wrist) and performed a calibration procedure to determine the temperature level that induced a pain intensity of 70 on a 0 to 100 VAS (“How painful was this stimulus?”, anchors: 0 = “not painful at all” and 100 = “unbearably painful”) using established procedures.19
To acquire skin conductance responses (SCRs), 2 electrodes were attached to the thenar and hypothenar surfaces of the participant's nondominant hand (for details, see SCRs). Participants were then asked to rate their level of arousal and pain-related fear on a 0 to 100 VAS (arousal: “How tense are you right now?” anchors: 0 = “not tense at all” and 100 = “very tense”; pain-related fear: “How fearful are you regarding the upcoming pain stimulus?” anchors: 0 = “not fearful at all” and 100 = “extremely fearful”).
Please note that data were acquired as part of a larger pharmacological trial with a double-blind hydrocortisone/placebo intervention on the second examination day to investigate the effects of pharmacologically induced stress on the recall and reinstatement of former acquired threat and safety associations. Results pertaining to the effect of the pharmacological intervention (day 2) will be reported elsewhere. Within this article, we report only the results of the first examination day.
2.3. Differential conditioning paradigm
We used an established differential conditioning paradigm56 comprising 3 experimental phases, ie, habituation phase, acquisition training, and extinction training (Fig. 1). During acquisition training, 2 geometrical figures served as cues predicting the delivery (CS+) or absence (CS−) of a painful heat stimulus (US). In the consecutive extinction training, only CS were presented to examine extinction learning. Physiological responses to the CS and US were monitored using continuous skin conductance recordings.
;)
Figure 1.: Differential conditioning paradigm. The experimental paradigm consisted of 3 experimental phases, ie, habituation phase, acquisition training, and extinction training. Assignment of the geometrical figures to experimental conditions was randomized (example shown here). Habituation phase: 6 CS, no US, valence ratings for each CS. Acquisition training: 32 CS, 12 US. Valence (CS+/CS−) and pain intensity ratings (US) after each fourth CS+/CS−/US. Extinction training: 24 CS, no US. Valence ratings (CS+/CS−) after each fourth CS+/CS−. Each CS was displayed for 9 seconds; US presentation was set to 2.5 seconds; CS+-US overlap was set to 1 second. VAS was displayed for 7.5 seconds (valence rating, pain intensity rating) or 15 seconds (contingency rating). Intertrial interval was jittered between 6 and 11 seconds. CS, conditioned stimuli; US, unconditioned stimuli.
2.3.1. Habituation phase/familiarization
In the initial habituation phase, 6 CS (3 CS+, 3 CS−; duration: 9 seconds), but no US were presented. After each CS presentation, participants provided a valence rating (see below for details).
2.3.2. Acquisition training
After that, participants were informed about a potential association of CS and US but not about the exact contingencies between specific cues and painful stimulation, or experimental phases (exact instructions can be found in Supplementary Methods, available at https://links.lww.com/PAIN/B544). During acquisition training, a total of 32 CS were presented (16 CS+, 16 CS−). CS+ were paired with the US at a reinforcement rate of 75% (ie, 12 CS+ + US), whereas the CS− were never followed by US (differential conditioning). By choosing a partial reinforcement rate of 75%, we aimed to ensure the initial acquisition of CS-US associations in both groups and to prolong the process of extinction learning (for review, see Lonsdorf et al.43). CS+ were presented exclusively for 8 seconds before US onset (delay conditioning, 1 second overlap), which lasted 2.5 seconds in total. Furthermore, the first and last CS+ were always reinforced. To track individual learning and extinction rates of CS-induced emotional responses over the course of the experiment, participants were repeatedly asked to provide valence ratings for the CS+ and CS− using VAS during every fourth CS+/CS− presentation, resulting in 4 valence ratings per CS type. After every fourth US presentation, participants rated the pain intensity of the US, resulting in 3 pain intensity ratings. At the end of acquisition training, participants rated the contingency of CS-US coupling (see below for details).
2.3.3. Extinction training
Subsequently, extinction training followed without further instructions. A total of 24 CS (12 CS+, 12 CS−) were presented without US following. Participants provided a valence rating during every fourth CS+/CS− presentation (ie, 3 valence ratings per CS type). At the end of extinction training, participants again provided contingency ratings regarding CS-US coupling.
During acquisition and extinction training, CS types were presented in a pseudorandomized order with no more than 2 trials of the same CS presented consecutively. The intertrial interval was jittered between 6 and 11 seconds.
2.4. Stimuli
The Presentation software, version 18.0 (Neurobehavioral Systems, Inc, Berkeley, CA, https://www.neurobs.com), was used to present visual (including the VAS) and thermal stimuli and to record behavioral data. Geometrical figures with softened edges (color: RBG code 142, 180, 227) on a black background (rectangle: visual angle 8.3 × 3.14°, square: visual angle 4.99 × 4.99°, and rhombus: visual angle 7.38 × 5.36°) were presented on a computer screen positioned in front of the participant and served as CS. Heat pain stimuli, which served as US, were applied using a thermal device (PATHWAY system, model CHEPS, 27 mm diameter; Medoc, Israel), which was attached to the left volar forearm by an elastic tape. Baseline temperature was set to 35°C. Rates for heating and cooling were set to maximum (70 and 40°C/s, respectively). Total stimulation time was 2.5 seconds.
2.5. Skin conductance responses
Skin conductance responses were continuously recorded throughout all experimental phases by a constant voltage system (0.5 V) using a BIOPAC MP150 device (BIOPAC Systems, Inc, Goleta, CA) in combination with AcqKnowledge 5.0.2 software. Single-use, radiotranslucent dry electrodes (EL509, BIOPAC Systems, Inc) and a conductive electrode cream (SYNAPSE; Kustomer Kinetics) were applied to the thenar and hypothenar eminences of the participants' nondominant (left) hand. The sampling rate was set to 2 kHz. External triggers were recorded to mark the exact onset of an event.
2.6. Self-report questionnaires
Psychological trait and state variables are known to modulate pain perception, pain chronification, and pain-related learning.51 To thoroughly characterize the patient sample and to explore the potential role of maladaptive pain-related cognitions in modulating pain-related learning, all participants completed the German version of the following questionnaires: (1) State-Trait-Anxiety-Depression-Inventory41; (2) Depression Anxiety Stress Scales52; (3) Center for Epidemiological Studies-Depression Scale: ADS-K25; (4) Pain Catastrophizing Scale: PCS40; (5) Pain Anxiety Symptom Scale: PASS-D68; (6) Trier Inventory of Chronic Stress: TICS58; (7) Questionnaire for Experiences of Attention Deficits70; and (8) Perceived Stress Questionnaire: PSQ20.17 All questionnaires were analyzed according to their respective manuals.
2.7. Outcome measures
Emotional and affective aspects of pain might shape pain perception particularly in patients with chronic pain.45,60,67 We thus focused on the emotional aspects of pain-related learning and assessed cue valence ratings repeatedly throughout the different experimental phases. This allowed us to capture the temporal dynamics of pain-related threat and safety learning.56 Changes in CS valence were assessed during every fourth CS+/CS− presentation through a −50 to +50 VAS with the question “How do you perceive this geometric figure?” (anchors: −50 = “very pleasant,” 0 = “neutral,” and +50 = “very unpleasant”). Thus, positive CS valence is indexed by negative VAS ratings, whereas negative CS valence is indexed by positive VAS ratings. To also capture cognitive aspects of pain-related learning, we assessed contingency awareness at the end of the acquisition and extinction training. Therefore, participants answered the following question using a 0 to 100 VAS separately for both CS: “How often was this geometric figure followed by a painful stimulus?” (anchors: 0 = “never,” 50 = “50% pain,” and 100 = “100% pain”). In addition, pain intensity ratings were collected to control for comparable pain perception between groups. Subjective pain intensity was assessed after every fourth US presentation using a 0 to 100 VAS: “How painful was this stimulus?” (anchors: 0 = “not painful at all” and 100 = “unbearably painful”).
For all ratings, the scale was positioned below the CS with the cursor placed at a random starting position between VAS 25 and VAS 75. Except for contingency ratings (15 seconds), VAS were presented for 7.5 seconds for CS valence and US intensity ratings.
2.8. Statistical analysis
All statistical analyses were performed using the R software.55 Group differences in pain-related variables (pain threshold, temperature level corresponding to VAS70, and mean pain intensity ratings) and person-related variables (age, arousal, pain-related cognitive variables, such as fear of pain, pain catastrophizing, pain-related anxiety, state and trait anxiety, depression, perceived stress, chronic stress, and perceived deficits in attention) were tested using 2-sample t tests or the nonparametric 2-sample Wilcoxon rank sum test, where appropriate.
2.8.1. Valence ratings
To investigate whether healthy volunteers and patients with CBP differed in pain-related learning behavior, we followed a 2-step analysis strategy. First, we focused on differential learning, ie, the development of valence differences between CS+ and CS− (ΔCS valence = valence CS+ − valence CS−). To further explore the nature of any valence differences, we also examined nondifferential changes in valence ratings separately for both CS types to specify whether differences are due to altered learning about the aversive CS+, the safety signal (CS−), or both. Differential and nondifferential learning (ie, threat and safety learning) were investigated using separate linear mixed models (LMMs) on valence ratings, as implemented in the R package lme4.5
2.8.1.1. Model estimation for acquisition and extinction training
First, LMM analyses were performed separately for acquisition and extinction training to test for changes in differential CS valence. The mean ΔCS valence ratings of the habituation phase were included as a baseline in the analysis of acquisition training. Likewise, the extinction training model included the last ΔCS valence rating of acquisition training as a baseline rating.56
To compare ΔCS valence ratings and their changes over time between patients with CBP and HCs, the factors time and group (HCs and patients with CBP), as well as interactions of these factors, were included as fixed effects into the models. The factor time was included as a continuous factor to account for increases or decreases of ΔCS valences during the experiment.56 We tested whether including a random intercept for each participant (ie, allowing for interindividual differences in baseline valence ratings) and allowing variation for the factors time, group, and participants by adding random effects for these factors improved model fit, which would indicate a better prediction of data when accounting for random variance in changes over time and differences between groups and participants.
Model selection was based on the Akaike information criterion (AIC) provided by the anova (analysis of variance) function in R, which computes χ2 between each model to detect improvements in model fit, ie, explained variance (maximum likelihood method). The model comparison showed that including random slopes for each subject and the factor time into the models best predicted the data (LMM acquisition training: Δ AIC = −70.6, P < 0.001; LMM extinction training: Δ AIC = −75.5, P < 0.001). Final (ie, best-fitting) models were estimated according to the restricted maximum likelihood method. Effects with P values < 0.05 were considered statistically significant. Cohen's d values, as calculated using the R package EMAtools,33 were given as effect sizes.
Nondifferential analyses on valence ratings were performed analogously to the differential analyses including the additional factor CS type (CS+, CS−). As determined by model comparison, the model including random effects for each participant and the factors time and CS type best predicted the data (LMM acquisition training: Δ AIC = −685.5, P < 0.001; LMM extinction training: Δ AIC = −629.6, P < 0.001).
2.8.2. Contingency ratings
To investigate differences in contingency ratings of CS-US coupling between CS types and groups and to test for changes between phases, LMM analyses were performed.
2.8.2.1. Model estimation
The calculated model comprised the factors phase, CS type, and group and their interactions as fixed effects. We again tested whether including a random intercept for each participant and accounting for variation of the factors CS type, phase, group and participants by adding random slopes for these factors improved model fit. The factor phase was included as a categorical factor. The model including random slopes for the participants and the factors CS type and phase best predicted the data (Δ AIC = −34.7, P < 0.001).
2.8.3. Modulatory influence of pain duration, maladaptive cognitions, and disease-related variables on pain-related threat and safety learning
For patients with CBP particularly, we were interested in exploring whether potential impairments in pain-related threat and safety learning and extinction scaled with pain duration. Thus, we included pain duration (in years) as a covariate of interest while controlling for age. Furthermore, we performed exploratory analyses on person-related and pain-related variables aiming to explore whether those covariates differentially modulated threat and safety learning (CS valence) or contingency awareness.
2.8.4. Analysis of skin conductance response data
Note that 3 participants (n = 2 patients with CBP, n = 1 HC) had to be excluded from the SCR analysis because of technical issues during acquisition training. Skin conductance response analyses were thus based on 60 patients with CBP and 60 HCs. Skin conductance response data were processed and analyzed using the R software. In a first step, data were down-sampled to 20 Hz and smoothed using a low-pass filter with a cutoff frequency of 2 Hz. Local minima and maxima of electrodermal activity were automatically detected. To calculate the amplitude of stimulus-related SCRs, the local minimum at the onset of the first SCR after stimulus onset was subtracted from the maximum peak.54 The maximum amplitude was analyzed within a time window of 1 to 4 seconds after CS onset (first-interval response, FIR8,31). For the US-related SCR, a time window of 0.5 to 7 seconds after US onset was chosen. The minimum amplitude criterion was set to 0.01 μS, and responses below were scored as 0 µS.
Data were transformed with the natural logarithm to reduce the skew of the amplitude distribution and attain a normal distribution.29 Trials in which VAS ratings had to be provided were excluded to avoid a contamination of CS/US-related SCRs with movement-induced signal changes. Skin conductance responses that deviated more than 3 SDs from the individual mean were treated as outliers and removed from the analyses (n = 2 SCRs in total). Skin conductance responses between valence ratings were pooled for 3 consecutive trials each, resulting in 4 pooled responses for acquisition training and 3 pooled responses for extinction training.
2.8.4.1. Model calculation
Linear mixed model analyses were performed on CS-induced SCR amplitudes for each experimental phase separately to test for changes in SCR amplitudes over time (ie, SCR increase or decrease) and differences between CS types and groups according to the analyses of valence ratings. To analyze CS-related SCRs, we included the factors CS type, time, and group and the interactions of these factors as fixed effects into the models. Furthermore, we tested whether a random intercept for each participant and allowing variation for the factors CS type, time, group, and participants by adding random slopes for these factors improved model fit. The LMM on CS-related SCR including random slopes for each participant and the factors time, CS type, and group best predicted the data for acquisition training (Δ AIC = −89.4, P < 0.001). For extinction training, the model including random slopes for each participant and the factor CS type best predicted the data (Δ AIC = −49.6, P < 0.001). For LMM analyses on US-related SCR, including random slopes for each participant and the factor time into the model best predicted the data (Δ AIC = −70.9, P < 0.001).
3. Results
3.1. Pain-related variables and self-report measures
Patients with CBP and HCs did not show differences in heat pain thresholds, individually calibrated temperature levels corresponding to VAS 70, arousal ratings, and ratings of pain-related fear (Table 2). Of importance, as intended, pain intensity ratings were moderate to high and comparable between patients with CBP and HCs during acquisition training. Patients with CBP and HCs showed significant difference in all psychological state and trait variables, as well as pain-related cognitive variables. However, for all the assessed psychological state and trait variables, most of the patients with CBP showed values in a normal range.
Table 2 -
Group differences in heat pain–related data, fear and arousal ratings, and self-report questionnaires.
|
CBP group (M ± SD) |
HC group (M ± SD) |
Statistics |
P
|
| Heat pain–related data |
|
|
|
|
| Heat pain thresholds (°C) |
43.8 ± 2.46 |
43.8 ± 1.63 |
W = 1809.50 |
0.68 |
| Temperature US (°C) |
47.5 ± 1.84 |
47.7 ± 1.48 |
W = 2020.00 |
0.52 |
| Pain intensity rating during acquisition training (0-100 VAS) |
62.5 ± 12.8 |
64.4 ± 10.4 |
t(121) = 0.91 |
0.37 |
| Fear and arousal ratings |
|
|
|
|
| Pain-related fear (0-100 VAS) |
25.2 ± 23.1 |
19.1 ± 15.4 |
W = 1225.50 |
0.41 |
| Arousal (0-100 VAS) |
31.0 ± 21.4 |
23.8 ± 19.6 |
W = 632.50 |
0.08 |
| Questionnaire data |
|
|
|
|
| STADI |
|
|
|
|
| State anxiety |
18.00 ± 5.48 |
15.20 ± 3.02 |
W = 1318.50 |
0.005
|
| State depression |
18.10 ± 4.51 |
15.80 ± 2.88 |
W = 1203.00 |
<0.001
|
| Trait anxiety |
21.20 ± 6.19 |
17.00 ± 4.37 |
W = 1097.50 |
<0.001
|
| Trait depression |
17.80 ± 4.29 |
15.50 ± 3.23 |
W = 1255.50 |
0.002
|
| CES-D* |
9.68 ± 6.85 |
6.22 ± 5.28 |
W = 1265.00 |
0.002
|
| PCS |
20.40 ± 9.16 |
11.80 ± 10.10 |
W = 971.50 |
<0.001
|
| PASS 20-D |
31.90 ± 12.40 |
26.30 ± 14.40 |
t(120) = −2.30 |
0.023
|
| DASS |
|
|
|
|
| Depression† |
3.97 ± 3.85 |
1.30 ± 1.60 |
W = 947.00 |
<0.001
|
| Anxiety‡ |
3.39 ± 3.07 |
1.11 ± 1.39 |
W = 970.00 |
<0.001
|
| Stress§ |
7.15 ± 4.65 |
2.26 ± 2.59 |
W = 619.50 |
<0.001
|
| TICS |
20.90 ± 8.89 |
16.00 ± 7.71 |
t(120) = −3.23 |
0.002
|
| PSQ20 |
55.5 ± 18.9 |
43.9 ± 17.2 |
t(120) = −3.57 |
<0.001
|
| FEDA |
100.00 ± 13.10 |
107.00 ± 10.00 |
W = 2384.00 |
0.007
|
Note that questionnaire data (except for DASS) of 1 HC is missing due to technical problems. Bold entries indicate significant p-values (p<0.05).
*Nine patients with CBP (14.5%) scored above the cutoff of 16.
†Four patients with CBP (6.5%) scored above the cutoff of 10.
‡Eleven patients with CBP (17.7%) scored above the cutoff of 6.
§Forteen patients with CBP (22.6%) scored above the cutoff of 10.
CBP, chronic back pain; HC, healthy control; CES-D, Center for Epidemiologic Studies Depression Scale; DASS, Depression Anxiety Stress Scales; FEDA, Questionnaire for Experiences of Attention Deficits; PASS 20-D, Pain Anxiety Symptom Scale; PCS, Pain Catastrophizing Scale; PSQ20, Perceived Stress Questionnaire; STADI, State-Trait-Anxiety-Depression-Inventory; TICS, Trier Inventory of Chronic Stress; US, unconditioned stimuli.
3.2. Valence ratings
Figure 2 shows valence ratings of patients with CBP and HCs during the habituation phase, acquisition training, and extinction training for both CS. During the habituation phase, valence ratings for CS+ and CS− were comparable within and between groups indicating that the affective connotation of the visual stimuli did not differ before conditioning (see Supplementary Table 2 and Supplementary Results, available at https://links.lww.com/PAIN/B544, for statistics).
Figure 2.: Reduced pain-related threat and safety learning (acquisition) in patients with chronic back pain compared with pain-free healthy volunteers: The mean valence ratings (±SEM) of patients with nonspecific chronic back pain (CBP, orange) and healthy volunteers (HCs, blue) for CS− (empty circles) and CS+ (filled circles) during the habituation phase (Hab), acquisition training (Acq 1–Acq 4), and extinction training (Ext 1–Ext 3). Experimental phases are separated by dashed lines. Note that positive CS valence is indexed by negative VAS ratings, and negative CS valence is indexed by positive VAS ratings. CBP, chronic back pain; CS, conditioned stimuli; HC, healthy control.
3.2.1. Acquisition training
3.2.1.1. Differential learning
As expected, both groups showed differential learning (∆CS valence) throughout acquisition training (patients with CBP: β = 3.45 ± 0.84; t(121.98) = 4.11, P < 0.001, d = 0.74; HCs: β = 6.02 ± 0.84; t(119.75) = 7.15, P < 0.001, d = 1.31). However, differential learning was significantly weaker in patients with CBP than in HCs (interaction (IA) time × group for ∆CS valence: ∆β = −2.57 ± 1.19; t(120.86) = −2.17, P = 0.03, d = −0.39).
3.2.1.2. Threat and safety learning
We further tested whether group differences in differential learning were due to altered responses to the CS+, the CS−, or both. While both groups showed significant increases in negative valence for the CS+ (patients with CBP: β = 2.54 ± 0.45; t(315.43) = 5.62, P < 0.001, d = 0.63; HCs: β = 3.81 ± 0.45; t(312.62) = 8.37, P < 0.001, d = 0.95) and significant increases in positive valence for the CS− over time (patients with CBP: β = −0.88 ± 0.45; t(313.07) = −1.94, P = 0.05, d = −0.22; HCs: β = −2.27 ± 0.45; t(307.97) = −5.02,P < 0.001, d = −0.57), both these effects were less pronounced in patients with CBP than in HCs (IA time × group, CS+: ∆β = −1.26 ± 0.64; t(314.02) = −1.96, P = 0.05, d = −0.22; CS−: ∆β = 1.40 ± 0.64; t(310.50) = 2.18, P = 0.03, d = 0.25).
3.2.2. Extinction training
3.2.2.1. Differential learning
Both groups showed extinction learning as indicated by a significant decrease of ΔCS valences over the course of extinction training (patients with CBP: β = −4.96 ± 0.98; t(121.48) = −5.06, P < 0.001, d = −0.92; HCs: β = −7.06 ± 0.98; t(119.59) = −7.17, P < 0.001, d = −1.31). In patients with CBP, ΔCS valence ratings returned to baseline level (ie, ratings provided in the habituation phase) in the last extinction trial (β = 1.41 ± 2.23; t(116.32) = 0.63, P = 0.53, d = 0.12). By contrast, ΔCS valence ratings were still elevated in HCs (β = 5.20 ± 2.24; t(115.77) = 2.33, P = 0.02, d = 0.43). Because slopes did not differ between groups (patients with CBP vs HCs: Δβ = 2.10 ± 1.39; t(120.53) = 1.51, P = 0.13, d = 0.27), this difference is likely due to the stronger conditioning effect in HCs during acquisition training, which may have resulted in a floor effect in the patient group.
3.2.2.2. Threat and safety learning
Both groups showed a significant decrease of negative valence for the CS+ (patients with CBP: β = −4.20 ± 0.60; t(246.98) = −7.06, P < 0.001, d = −0.90; HCs: β = −6.02 ± 0.60; t(247.00) = −10.04, P < 0.001, d = −1.28), reaching baseline (ie, habituation) level at the end of extinction training (patients with CBP: β = 0.15 ± 2.04; t(204.46) = 0.07, P = 0.94, d = 0.01; HCs: β = −2.24 ± 2.10; t(204.49) = −1.09, P = 0.28, d = −0.15). The decrease of negative CS+ valence was significantly less pronounced in patients with CBP when compared with HCs (IA time × group: ∆β = 1.82 ± 0.84; t(246.99) = 2.15, P = 0.03, d = 0.27), which might again be reflective of the stronger conditioning effect in HCs. Of interest, CS− valence remained at a stable (positive) level across extinction training in both groups (patients with CBP: β = 0.87 ± 0.60; t(251.47) = 1.44, P = 0.15, d = 0.18; HCs: β = 0.90 ± 0.60; t(244.58) = 1.50, P = 0.13, d = 0.19). Of note, CS− valence assessed at the end of the extinction training phase was significantly more positive than during the habituation phase in HCs, whereas patients with CBP showed no significant difference (patients with CBP: β = −0.97 ± 2.05; t(206.38) = −0.47, P = 0.64, d = −0.07; HCs: β = −7.44 ± 2.05; t(204.53) = −3.62, P < 0.001, d = −0.51).
3.3. Contingency ratings
In general, participants correctly identified the CS to predict the delivery or absence of the US (Fig. 3). Contingency ratings for the CS+ were significantly higher than those for the CS− in both experimental phases and both groups (patients with CBP, acquisition training: β = 47.41 ± 3.97; t(227.71) = 11.94, P < 0.001, d = 1.58; HCs, acquisition training: β = 58.00 ± 4.02; t(229.34) = 14.44, P < 0.001, d = 1.91; patients with CBP, extinction training: β = 14.38 ± 4.00; t(232.53) = 3.60, P < 0.001, d = 0.47; HCs, extinction training: β = 21.24 ± 4.02; t(230.69) = 5.28, P < 0.001, d = 0.69). Furthermore, contingency ratings for both CS were significantly lower after extinction training than after acquisition training (patients with CBP, CS+: β = −43.98 ± 3.28; t(242.67) = −13.42, P < 0.001, d = −1.72; HCs, CS+: β = −46.68 ± 3.33; t(243.95) = −14.04, P < 0.001, d = −1.80; patients with CBP, CS−: β = −10.96 ± 3.34; t(247.14) = −3.29, P = 0.001, d = −0.42; HCs, CS−: β = −9.92 ± 3.34; t(244.81) = −2.97, P < 0.001, d = −0.38). CS-specific contingency ratings did not differ significantly between patients with CBP and HCs, neither after acquisition nor after extinction training (no significant IA group × CS type) and did not change differentially over time (no significant IA group × time × CS type).
Figure 3.: Comparable contingency awareness in patients with chronic back pain and healthy volunteers. The mean contingency ratings (with SEM) of patients with nonspecific chronic back pain (CBP, orange) and healthy participants (HCs, blue) provided for CS+ (filled circles) and CS− (empty circles) after acquisition (Acq) and extinction training (Ext). CBP, chronic back pain; CS, conditioned stimuli; HC, healthy control.
3.4. The modulatory influence of pain duration, maladaptive cognitions, and pain-related variables on pain-related threat and safety learning
To test whether pain duration significantly modulates the acquisition or extinction of CS+ or CS− valence ratings in patients with CBP, we added pain duration as a covariate of interest, while controlling for age. We found that during acquisition training, valence ratings for the CS+ showed a slower increase (ie, reduced slope) with increasing pain duration (IA time × pain duration: β = −0.09 ± 0.05; t(179.08) = −1.84, P = 0.067, d = −0.27). Moreover, during extinction training, the decrease of CS+ valence ratings was significantly slower (ie, less extinction) with increasing pain duration (IA time × pain duration: β = 0.17 ± 0.06; t(114.76) = 2.84, P = 0.005, d = 0.53). Valence ratings for the CS− were not significantly modulated by pain duration. Similar results were found when testing an association between contingency awareness and pain duration, while controlling for the modulatory effects of age. Longer pain duration was associated with significantly lower contingency awareness after acquisition training for the CS+ (β = −1.53 ± 0.32; t(97.25) = −4.79, P < 0.001, d = −0.97), whereas no significant effect was found for the CS−. Contingency awareness assessed after extinction training was not modulated by pain duration. The decline in CS+-related contingency ratings from post acquisition to post extinction training was significantly less pronounced for patients with CBP with longer pain duration (IA time × pain duration: β = 1.15 ± 0.38; t(117.15) = 3.00, P = 0.003, d = 0.56).
Note that exploratory analyses of the modulatory influence of maladaptive cognitions and other pain-related variables will only be briefly described in this article. Detailed results on covariates and statistics are given in the Supplementary Results, available at https://links.lww.com/PAIN/B544. The level of pain chronicity (according to the von Korff scale37) and back pain intensity at the day of study participation did not modulate threat and safety learning during acquisition and extinction training in patients with CBP. However, covariate analyses revealed a significant influence of maladaptive cognitions in both groups. Higher levels of state anxiety, for instance, were linked to increased threat learning in patients with CBP and enhanced safety learning in healthy participants (valence ratings). In patients with CBP only, enhanced levels of state anxiety and depression were associated with an overestimation of the association between the safety cue and the US (contingency ratings). Trait anxiety and depression did not significantly modulate learning behavior. Neither threat nor safety learning (valence and contingency ratings) was modulated by the individual level of US-related fear. However, differential learning was significantly enhanced in more aroused HCs—an effect that was driven by more pronounced threat learning with higher levels of arousal. In patients with CBP with high pain anxiety and pain catastrophizing scores, threat learning and extinction were enhanced. This effect was not reflected in contingency ratings.
3.5. Skin conductance responses
3.5.1. Unconditioned responses
US-related SCR amplitudes significantly decreased during acquisition training in both groups (patients with CBP: β = −0.003 ± 0.001; t(118.60) = −4.97, P > 0.001, d = −0.91; HCs: β = −0.003 ± 0.001; t(117.40) = −3.57, P < 0.001, d = −0.66). However, the groups did not differ in US-related SCR amplitudes or amplitude changes over time.
3.5.2. Conditioned responses
During acquisition training, we observed a general decrease in CS-related SCR amplitudes over time in both groups (patients with CBP, CS+: β = −0.008 ± 0.002; t(209.40) = −4.93, P < 0.001, d = −0.68; HCs, CS+: β = −0.003 ± 0.002; t(219.50) = −2.06, P = 0.04, d = −0.28; patients with CBP, CS−: β = −0.005 ± 0.002; t(252.00) = −3.08, P = 0.002, d = −0.39; HCs, CS−: β = −0.005 ± 0.002; t(245.50) = −2.81, P = 0.005, d = −0.36; see Supplement, Figure S1, available at https://links.lww.com/PAIN/B544). Of interest, the decrease of CS+-induced SCR amplitudes was more pronounced in patients with CBP when compared with HCs (∆β = −0.004 ± 0.002; t(214.90) = −1.97, P = 0.05, d = −0.27). There were no significant differences between CS types (all P > 0.05). No significant effects were found in extinction training.
4. Discussion
In this study, we examined potential differences in pain-related acquisition and extinction learning in a large sample of patients with nonspecific CBP and age- and sex-matched healthy volunteers (HCs). In a differential conditioning paradigm with painful heat stimuli (US), valence ratings indicated less differential learning during acquisition training in patients with CBP. Of importance, this effect was driven by both reduced threat (CS+) and safety (CS−) learning. Furthermore, both groups showed a decrease of negative valence ratings for the CS+ but no significant change in CS− valuation in the subsequent extinction training phase.
4.1. Successful differential acquisition and extinction in patients with chronic back pain and healthy volunteers
In both groups, CS+ ratings increased in negative valence whereas CS− ratings increased in positive valence during acquisition training, indicating that, in general, stimulus valence was modulated because of the conditioning procedure. Contingency ratings after acquisition training confirmed that both groups were equally able to differentiate between CS types. Furthermore, during extinction training, CS+ ratings decreased in negative valence in both groups. These findings are in line with previous reports of successful pain-related differential learning, as well as extinction learning in other pain conditions.18,23,28,48,49
4.2. Impaired threat and safety learning in patients with nonspecific chronic back pain
Although valence ratings indicated differential learning in both groups, differences between CS+ and CS− were smaller in the patient group, indicating weaker learning, which corroborates earlier findings in patients with chronic pain.22,46 In principle, impaired differential learning can be caused by a deficit in threat learning, safety learning, or both. Our data suggest that in patients with CBP, both threat and safety learning are affected, which is in line with reports of diminished threat learning23 and safety learning7,23,47 in various chronic pain conditions. Diminished differential learning has previously been found in healthy individuals under conditions of high uncertainty about probabilistic cue–outcome contingency.9,69 Given that, in our study, contingencies were identical in both groups, such uncertainty is unlikely to be the direct effect of outcome probabilities but might reflect the degree of participants' certainty about the reinforcement schedule. In this study, CS+ was followed by an unexpected outcome (ie, omission of US after CS+ presentation) in a quarter of CS+ trials. The degree to which this deviant information is taken into account during acquisition training may have differed between groups. Less extreme valence ratings in the patient group could be reflective of more emphasis on the unexpected outcome than in HCs. Another reason for the impaired discrimination between CS+ and CS− might be diminished perceptual discrimination in patients with CBP, as suggested by Catley et al.10 However, in our sample, proxies of pain-related sensory discrimination (eg, pain thresholds and pain intensity ratings during acquisition training) did not differ between groups, which does not support this view. Rather, patients may have adopted a “better safe than sorry” strategy on an affective-motivational level. Moreover, impaired discrimination learning could contribute to increased avoidance behavior when patients are not able to properly distinguish the safety from the threat stimulus. Whether other processes such as generalization and avoidance behavior also played a role here23,47,49 was not addressed in this study. Another explanation for our findings might be a difference in US salience or emotional relevance. Although the US was rated as equally painful in both groups, its threat value or emotional relevance might have differed between the groups. Predictable, short-lasting heat stimuli may not be as salient and emotionally relevant for patients with CBP as for healthy individuals because their chronic pain with its oftentimes unpredictable symptom fluctuations is far more bothersome. In line with this, the threat value or salience of US has been shown to shape the acquisition and extinction of conditioned responses, especially in the context of several aversive stimuli.35,56
Aberrant pain-related learning and extinction in chronic pain states may be further related to structural and functional changes in emotion regulation networks3,24,32 and networks underlying higher cognitions.11,32,61 Especially, safety learning deficits might be the result of functional and neurochemical changes in the mesolimbic dopamine circuitry in chronic pain.36,42,60
4.3. Differences in extinction learning between patients with chronic back pain and healthy volunteers
Studies investigating extinction processes in patients with chronic pain are scarce, and results differ between pain syndromes.18,23,28,49 In this study, patients with CBP showed less decrease in negative CS+ valence during extinction training than HCs. However, these findings may indicate a floor effect because patients with CBP had shown weaker threat and safety learning during acquisition training.23 More robust conclusions regarding extinction learning deficits would ideally require similarly strong US-CS associations after acquisition training. Of importance, the difference between CS+ and CS− valence ratings assessed at the end of extinction training was comparable with the difference at baseline, indicating successful extinction in patients. Hence, our data do not support the assumption of a general deficit in extinction learning in patients. In HCs, after extinction training, ratings for the CS+ were still significantly higher than for the CS−, which suggests a lingering differential conditioning effect. Of interest, CS− valence ratings did not change significantly during extinction training in either group, as has previously been discussed for evaluative conditioning.15,21,27
4.4. Modulatory influence of pain duration and trait and state variables on pain-related learning
Of importance, our study is the first to show that longer pain duration is associated with reduced threat learning in patients with CBP both on the emotional (CS valence) and cognitive (contingency awareness) level. These findings may suggest that alterations in threat learning might develop gradually (at least in patients with CBP) and could thereby contribute to the maintenance of chronic pain. However, longitudinal studies are needed to delineate whether reduced threat learning is already present at onset or before pain chronification and gradually worsens as the pain duration progresses. The increasing relative differences in salience of the clinical pain and the experimental stimulus may contribute to this effect.
Safety learning, on the other hand, was not significantly modulated by pain duration. Whether safety learning deficits precede and potentially predispose to chronic pain or, alternatively, develop at an early phase of chronic pain cannot be answered, given our experimental design and patient sample. Neural changes in emotional learning circuits and functional connectivity within these circuits4 have been described to be altered in chronic pain and to contribute to the transition from subacute to chronic pain.2,39,50 Given the known role of learning mechanisms in treatment responses,71 especially in chronic pain,13 impaired safety learning might hamper the response to therapeutic approaches in which features of the treatment can serve as (safety) cues predicting pain relief.
Chronic pain and mental disorders such as depression or anxiety disorders commonly co-occur,63 and these conditions have been shown to be related to altered learning and extinction processes.6,16,53 Although modulatory effects of trait anxiety were not evident in either group,62 we found significant modulations of learning with state anxiety. Of interest, state anxiety showed opposing effects on patients' and healthy participants' learning behavior. Whereas state anxiety was associated with enhanced threat learning in patients, higher state anxiety increased safety learning in HCs but did not influence threat learning. This dissociation could be reflective of an adaptive mechanism in healthy individuals with an attentional focus on cues and situations signaling safety and a maladaptive mechanism in patients with chronic pain with an attentional focus on cues signaling potential threat (“better safe than sorry”). Previous results pertaining the influence of state anxiety on threat and safety learning are scarce and inconsistent.14,20,38,66 Of interest, in our patient sample, other maladaptive pain-related cognitions (pain catastrophizing and pain anxiety) boosted both threat acquisition and extinction learning. The finding of enhanced threat and extinction learning in patients with CBP with higher pain anxiety and pain catastrophizing levels was not expected here. However, it has to be noted that threat acquisition in those patients was also enhanced, which could have driven the effect in extinction learning potentially in the sense of unequal starting points of US-CS associations before extinction training. To investigate the effects of pain anxiety and pain catastrophizing on the speed of extinction learning properly, one would ideally need similar strength of US-CS associations after acquisition training.
4.5. Limitations and implications for future studies
Our study results have to be interpreted in the light of several limitations. While in our controlled experimental setting, both conditions were relatively easy to distinguish (pain vs no pain), learning to predict phases of pain exacerbation or alleviation from clinical pain with its natural fluctuations (instead of complete relief) is undoubtedly more challenging. Although not explicitly tested in this study, there is reason to assume that the perceived threat value of the US in this study differed substantially between groups. Moreover, the patients with CBP investigated in this study were rather homogeneous regarding pain duration and pain-related impairments (low chronic pain grades), were younger when compared with the average patients with chronic pain, and took rather few medications.23,34,47,49 In addition, patients with CBP did not show clinically relevant psychological comorbidities, which are expected to be more pronounced in more severely affected patients. However, we observed reduced acquisition and extinction of threat associations, as well as lower contingency awareness in patients with CBP with increased pain duration. Therefore, our findings might underestimate clinical effects in more severely affected patients. Future studies investigating a more heterogeneous patient sample with varying degrees of pain-related impairments or, ideally, longitudinal studies are needed to explore threat and safety learning and their modulation by psychological variables in different stages of chronic pain.
5. Conclusion
Our study presents evidence that patients with CBP differentiate less between threat and safety cues than pain-free individuals. We therefore conclude that CBP is associated with altered threat and safety learning. This ambiguity to emotionally evaluate threat and safety information might lead to overly protective behavior and thus contribute to the maintenance of chronic pain. Intriguingly, state anxiety had opposing effects on threat and safety learning in patients with CBP and healthy participants. Our data also provide first evidence for reduced threat learning with pain duration. However, the functional relevance of this effect needs to be explored in longitudinal studies to discern whether learning deficits are the cause or consequence of chronic pain and when such learning deficits can be best addressed therapeutically.
Conflict of interest statement
The authors have no conflicts of interest to declare.
Appendix A. Supplemental digital content
Supplemental digital content associated with this article can be found online at https://links.lww.com/PAIN/B544.
Acknowledgements
The authors thank Jaspreet Kaur, Vanessa Dobischat, and Ashtar Hashim for their help in recruiting participants and data acquisition, Harald Engler and Alexandra Kornowski for analyzing saliva samples, and Diana Müßgens for editing the manuscript. Author J. Kleine-Borgmann was supported as a Clinician Scientist within the University Medicine Essen Academy (UMEA) program, funded by the German Research Foundation (DFG; grant FU356/12-1) and the Faculty of Medicine, University of Duisburg-Essen. This work was supported by a grant from the German Research Foundation (DFG; project number 316803389–SFB 1280 given to U.B., K.F., S.E., C.J.M., and O.T.W.).
Data availability: Raw data are available from the corresponding author on request.
Code availability: The R code used for data analyses in this study is available from the corresponding author on request.
Preregistration statement: The conducted research was not preregistered.
References
[1]. Airaksinen O, Brox JI, Cedraschi C, Hildebrandt J, Klaber-Moffett J, Kovacs F, Mannion AF, Reis S, Staal JB, Ursin H, Zanoli G. Chapter 4: European guidelines for the management of chronic nonspecific low back pain. Eur Spine J 2006;15:192–300.
[2]. Apkarian AV, Baliki MN, Farmer MA. Predicting transition to chronic pain. Curr Opin Neurol 2013;26:360–7.
[3]. Baliki MN, Chialvo DR, Geha PY, Levy RM, Harden RN, Parrish TB, Apkarian AV. Chronic pain and the emotional brain: specific brain activity associated with spontaneous fluctuations of intensity of chronic back pain. J Neurosci 2006;26:12165–73.
[4]. Baliki MN, Petre B, Torbey S, Herrmann KM, Huang L, Schnitzer TJ, Fields HL, Apkarian AV. Corticostriatal functional connectivity predicts transition to chronic back pain. Nat Neurosci 2012;15:1117–19.
[5]. Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using {lme4}. J Stat Softw 2015;67:1–48.
[6]. Blechert J, Michael T, Vriends N, Margraf J, Wilhelm FH. Fear conditioning in posttraumatic stress disorder: evidence for delayed extinction of autonomic, experiential, and behavioural responses. Behav Res Ther 2007;45:2019–33.
[7]. Both S, Brauer M, Weijenborg P, Laan E. Effects of aversive classical conditioning on sexual response in women with dyspareunia and sexually functional controls. J Sex Med 2017;14:687–701.
[8]. Boucsein W. Methods of electrodermal recording. electrodermal activity. New York: Springer US, 2012:87–258.
[9]. Brown CA, Seymour B, Boyle Y, El-Deredy W, Jones AKP. Modulation of pain ratings by expectation and uncertainty: behavioral characteristics and anticipatory neural correlates. PAIN 2008;135:240–50.
[10]. Catley MJ, O'Connell NE, Berryman C, Ayhan FF, Moseley GL. Is tactile acuity altered in people with chronic pain? A systematic review and meta-analysis. J Pain 2014;15:985–1000.
[11]. Cauda F, Palermo S, Costa T, Torta R, Duca S, Vercelli U, Geminiani G, Torta DME. Gray matter alterations in chronic pain: a network-oriented meta-analytic approach. Neuroimage Clin 2014;4:676–86.
[12]. Champely S, Ekstrom C, Dalgaard P, Gill J, Weibelzahl S, Anandkumar A, Ford C, Volcic R, De Rosario H. Basic functions for power analysis. 2020. Available at:
https://cran.r-project.org/package=pwr. Accessed March 24, 2021.
[13]. Cordier L, Diers M. Learning and unlearning of pain. Biomedicines 2018;6:1–7.
[14]. Dibbets P, Evers EAT. The influence of state anxiety on fear discrimination and extinction in females. Front Psychol 2017;8:347.
[15]. Dirikx T, Hermans D, Vansteenwegen D, Baeyens F, Eelen P. Reinstatement of extinguished conditioned responses and negative stimulus valence as a pathway to return of fear in humans. Learn Mem 2004;11:549–54.
[16]. Duits P, Cath DC, Lissek S, Hox JJ, Hamm AO, Engelhard IM, Van Den Hout MA, Baas JMP. Updated meta-analysis of classical fear conditioning in the anxiety disorders. Depress Anxiety 2015;32:239–53.
[17]. Fliege H, Rose M, Arck P, Levenstein S, Klapp BF. PSQ—Perceived Stress Questionnaire. Trier, Germany: ZPID, 2001.
[18]. Flor H, Knost B, Birbaumer N. The role of operant conditioning in chronic pain: an experimental investigation. PAIN 2002;95:111–18.
[19]. Forkmann K, Wiech K, Ritter C, Sommer T, Rose M, Bingel U. Pain-specific modulation of hippocampal activity and functional connectivity during visual encoding. J Neurosci 2013;33:2571–81.
[20]. Glotzbach-Schoon E, Andreatta M, Mühlberger A, Pauli P. Reinstatement of contextual anxiety in humans: effects of state anxiety. Int J Psychophysiol 2015;98:557–66.
[21]. Haaker J, Golkar A, Hermans D, Lonsdorf TB. A review on human reinstatement studies: an overview and methodological challenges. Learn Mem 2014;21:424–40.
[22]. Harvie DS, Moseley GL, Hillier SL, Meulders A. Classical conditioning differences associated with chronic pain: a systematic review. J Pain 2017;18:889–98.
[23]. Harvie DS, Weermeijer JD, Olthof NA, Meulders A. Learning to predict pain: differences in people with persistent neck pain and pain-free controls. PeerJ 2020:2020.
[24]. Hashmi JA, Baliki MN, Huang L, Baria AT, Torbey S, Hermann KM, Schnitzer TJ, Apkarian AV. Shape shifting pain: chronification of back pain shifts brain representation from nociceptive to emotional circuits. Brain 2013;136:2751–68.
[25]. Hautzinger M, Bailer M. Allgemeine depressions-skala: testmappe. Weinheim: Beltz, 1993.
[26]. Hermans D, Craske MG, Mineka S, Lovibond PF. Extinction in human fear conditioning. Biol Psychiatry 2006;60:361–8.
[27]. Hofmann W, De Houwer J, Perugini M, Baeyens F, Crombez G. Evaluative conditioning in humans: a meta-analysis. Psychol Bull 2010;136:390–421.
[28]. Icenhour A, Langhorst J, Benson S, Schlamann M, Hampel S, Engler H, Forsting M, Elsenbruch S. Neural circuitry of abdominal pain-related fear learning and reinstatement in irritable bowel syndrome. Neurogastroenterol Motil 2015;27:114–27.
[29]. Ishihara K, Miyata Y. Skin potential activity under two electrode electrolytes. Shinrigaku Kenkyu 1980;51:291–4. Available at:
http://www.ncbi.nlm.nih.gov/pubmed/7052370. Accessed February 24, 2020.
[30]. Jenewein J, Moergeli H, Sprott H, Honegger D, Brunner L, Ettlin D, Grillon C, Bloch K, Brügger M, Schwegler K, Schumacher S, Hasler G. Fear-learning deficits in subjects with fibromyalgia syndrome?. Eur J Pain (United Kingdom) 2013;17:1374–84.
[31]. Jentsch VL, Wolf OT, Merz CJ. Temporal dynamics of conditioned skin conductance and pupillary responses during fear acquisition and extinction. Int J Psychophysiol 2020;147:93–9.
[32]. Kang D, McAuley JH, Kassem MS, Gatt JM, Gustin SM. What does the grey matter decrease in the medial prefrontal cortex reflect in people with chronic pain? Eur J Pain 2019;23:203–19.
[33]. Kleiman E EMAtools: data management tools for real-time monitoring/ecological momentary assessment data. 2017. Available at:
https://cran.r-project.org/package=EMAtools. Accessed March 26, 2020.
[34]. Klinger R, Matter N, Kothe R, Dahme B, Hofmann UG, Krug F. Unconditioned and conditioned muscular responses in patients with chronic back pain and chronic tension-type headaches and in healthy controls. PAIN 2010;150:66–74.
[35]. Koenen LR, Icenhour A, Forkmann K, Theysohn N, Forsting M, Bingel U, Elsenbruch S. From anticipation to the experience of pain: the importance of visceral versus somatic pain modality in neural and behavioral responses to pain-predictive cues. Psychosom Med 2018;80:826–35.
[36]. Kong E, Monje FJ, Hirsch J, Pollak DD. Learning not to fear: neural correlates of learned safety. Neuropsychopharmacology 2014;39:515–27.
[37]. Von Korff M, Ormel J, Keefe FJ, Dworkin SF. Grading the severity of chronic pain. PAIN 1992;50:133–49.
[38]. Kuhn M, Mertens G, Lonsdorf TB. State anxiety modulates the return of fear. Int J Psychophysiol 2016;110:194–9.
[39]. Kuner R, Flor H. Erratum: structural plasticity and reorganisation in chronic pain. Nat Rev Neurosci 2017;18:113.
[40]. Lautenbacher S, Huber C, Kunz M, Parthum A, Weber PG, Griessinger N, Sittl R. Hypervigilance as predictor of postoperative acute pain: its predictive potency compared with experimental pain sensitivity, cortisol reactivity, and affective state. Clin J Pain 2009;25:92–100.
[41]. Laux L, Hock M, Bergner-Köther R, Hodapp V. Stadi: state-trait-angst-depressions-inventar. Hogrefe Verlag, 2013. Available at:
https://fis.uni-bamberg.de/handle/uniba/41475. Accessed March 11, 2020.
[42]. Lloyd K, Dayan P. Safety out of control: dopamine and defence. Behav Brain Funct 2016;12:15.
[43]. Lonsdorf TB, Menz MM, Andreatta M, Fullana MA, Golkar A, Haaker J, Heitland I, Hermann A, Kuhn M, Kruse O, Meir Drexler S, Meulders A, Nees F, Pittig A, Richter J, Römer S, Shiban Y, Schmitz A, Straube B, Vervliet B, Wendt J, Baas JMP, Merz CJ. Don't fear “fear conditioning”: methodological considerations for the design and analysis of studies on human fear acquisition, extinction, and return of fear. Neurosci Biobehav Rev 2017;77:247–85.
[44]. Lonsdorf TB, Merz CJ. More than just noise: inter-individual differences in fear acquisition, extinction and return of fear in humans—biological, experiential, temperamental factors, and methodological pitfalls. Neurosci Biobehav Rev 2017;80:703–28.
[45]. Lumley MA, Cohen JL, Borszcz GS, Cano A, Radcliffe AM, Porter LS, Schubiner H, Keefe FJ. Pain and emotion: a biopsychosocial review of recent research. J Clin Psychol 2011;67:942–68.
[46]. Meulders A. Fear in the context of pain: lessons learned from 100 years of fear conditioning research. Behav Res Ther 2020;131:103635.
[47]. Meulders A, Harvie DS, Bowering JK, Caragianis S, Vlaeyen JWS, Moseley GL. Contingency learning deficits and generalization in chronic unilateral hand pain patients. J Pain 2014;15:1046–56.
[48]. Meulders A, Jans A, Vlaeyen JWS. Differences in pain-related fear acquisition and generalization: an experimental study comparing patients with fibromyalgia and healthy controls. PAIN 2015;156:108–22.
[49]. Meulders A, Meulders M, Stouten I, De Bie J, Vlaeyen JWS. Extinction of fear generalization: a comparison between fibromyalgia patients and healthy control participants. J Pain 2017;18:79–95.
[50]. Navratilova E, Porreca F. Reward and motivation in pain and pain relief. Nat Neurosci 2014;17:1304–12.
[51]. Nees F, Becker S. Psychological processes in chronic pain: influences of reward and fear learning as key mechanisms—behavioral evidence, neural circuits, and maladaptive changes. Neuroscience 2018;387:72–84.
[52]. Nilges P, Essau C. Die Depressions-Angst-Stress-Skalen: der DASS – ein Screeningverfahren nicht nur für Schmerzpatienten. Obere Extrem 2015;10:649–57.
[53]. Otto MW, Moshier SJ, Kinner DG, Simon NM, Pollack MH, Orr SP. De novo fear conditioning across diagnostic groups in the affective disorders: evidence for learning impairments. Behav Ther 2014;45:619–29.
[54]. Prokasy WF, Ebel HC. Three components of the classically conditioned gsr in human subjects. J Exp Psychol 1967;73:247–56.
[55]. RStudio Team. RStudio: integrated development environment for R. 2016. Available at:
http://www.rstudio.com/. Accessed March 16, 2020.
[56]. Schmidt K, Forkmann K, Elsenbruch S, Bingel U. Enhanced pain-related conditioning for face compared to hand pain. PLoS One 2020;15:e0234160.
[57]. Schneider C, Palomba D, Flor H. Pavlovian conditioning of muscular responses in chronic pain patients: central and peripheral correlates. PAIN 2004;112:239–47.
[58]. Schulz P, Schlotz W, Becker P. Trierer inventar zum chronischen stress (TICS). Göttingen, Germany: Hogrefe, 2004.
[59]. Sehlmeyer C, Schöning S, Zwitserlood P, Pfleiderer B, Kircher T, Arolt V, Konrad C. Human fear conditioning and extinction in neuroimaging: a systematic review. PLoS One 2009;4:e5865.
[60]. Serafini RA, Pryce KD, Zachariou V. The mesolimbic dopamine system in chronic pain and associated affective comorbidities. Biol Psychiatry 2020;87:64–73.
[61]. Shi HC, Yuan CH, Dai ZY, Ma HR, Sheng LQ. Gray matter abnormalities associated with fibromyalgia: a meta-analysis of voxel-based morphometric studies. Semin Arthritis Rheum 2016;46:330–7.
[62]. Torrents-Rodas D, Fullana MA, Bonillo A, Caseras X, Andión O, Torrubia R. No effect of trait anxiety on differential fear conditioning or fear generalization. Biol Psychol 2013;92:185–90.
[63]. Tunks ER, Crook J, Weir R. Epidemiology of chronic pain with psychological comorbidity: prevalence, risk, course, and prognosis. Can J Psychiatry 2008;53:224–34.
[64]. Vlaeyen JWS. Learning to predict and control harmful events: chronic pain and conditioning. PAIN 2015;156(Suppl):S86–93.
[65]. Vlaeyen JWS, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. PAIN 2000;85:317–32.
[66]. Vriends N, Michael T, Blechert J, Meyer AH, Margraf J, Wilhelm FH. The influence of state anxiety on the acquisition and extinction of fear. J Behav Ther Exp Psychiatry 2011;42:46–53.
[67]. Walker S. Psychological aspects of chronic pain: a literature review. South Afr J Anaesth Analg 2005;11:139–45.
[68]. Walter B, Hampe D, Wild J, Vaitl D. Die Erfassung der Angst vor Schmerzen: eine modifizierte deutsche Version der Pain Anxiety Symptom Scale (PASS-D). Der Schmerz 2002;16:83.
[69]. Zaman J, Wiech K, Vlaeyen JWS. Perceptual decision parameters and their relation to self-reported pain: a drift diffusion account. J Pain 2020;21:324–33.
[70]. Zimmermann P, Messner C, Poser U, Sedelmeier P. Ein Fragebogen erlebter Defizite der Aufmerksamkeit (FEDA). Universität Freiburg: Institut für Psychologie, 1991.
[71]. Zunhammer M, Ploner M, Engelbrecht C, Bock J, Kessner SS, Bingel U. The effects of treatment failure generalize across different routes of drug administration. Sci Transl Med 2017;9:eaal2999.
