Modulation of attention to pain by goal-directed action: a somatosensory evoked potentials approach

Participants

Forty-five healthy, pain-free volunteers (35 women and 10 men, mean age 22.47 years (range 18–44)) gave their written informed consent and participated in the experiment. Sample size calculation was performed with GPower software (G*power 3.1 (Faul et al., 2007) to obtain 95% power to detect medium effect size (based on Cohen’s effect size convention f = 0.25; Faul et al., 2007)). Sample size calculation was also based on the effect sizes on previous studies with a similar paradigm in which medium-sized effects were observed (Clauwaert et al., 2018, 2020). Students were recruited via the SONA system of Maastricht University, while members of the general population were reached via social media platforms, flyers and word-of-mouth. After participating in the study, of approximately 2 h, participants choose to receive either two course credits or a gift voucher worth 15 Euros.

The exclusion criteria included pregnancy, left-handedness, chronic pain anywhere in the body, acute pain anywhere between the right shoulder and right hand, uncorrected hearing and/or vision problems, current or history of cardiovascular disease, neurological disorder, current or history of a psychiatric disorder (e.g., depression, panic/anxiety disorder), any other serious medical condition and an electronic implant (e.g., cardiac pacemaker) or having been advised by a general practitioner to avoid stressful situations. Participants were informed about the potential discomfort of pain stimuli before signing the informed consent. The study was conducted in accordance with the principle of the Declaration of Helsinki and the Ethics Review Committee Psychology and Neuroscience (ERCPN) of Maastricht University (number ERCPN- 233_18_02_2021_A1). One participant was excluded from the analysis because of the bad quality of the EEG data that could not be pre-processed.

Pain stimuli

A commercial constant-current square wave electrical pulse stimulator (DS7A) delivered pain stimuli, consisting of brief electrocutaneous stimuli (duration 2 ms). These stimuli were administered via two reusable stainless steel disk electrodes (8 mm diameter with 30 mm spacing; Digitimer, Welwyn Garden City, UK). The electrodes were filled with K–Y gel (Reckitt Benckiser, Slough, UK), and attached to the right wrist or triceps tendon (counterbalanced across participants) of the right (dominant) arm using a Velcro strap (Meulders et al., 2016). The stimulus intensity was determined for each participant individually using a standardized calibration procedure (see Pain Calibration under Procedure).

Tactile stimuli

Two resonant-type tactile probes (C-2 TACTOR; Engineering Acoustics, Inc., Casselberry, FL, USA; Van Hulle et al., 2013) were used to administer vibrotactile stimuli. These stimuli lasted 200 ms, their frequency was 300 Hz and their intensity 0.04 watts; these parameters were set via a self-developed software program. The probes were attached with double-sided tape rings to the skin of the participants in the pain-relevant location (proximal to electrodes used for the pain stimulation) and another one in a pain-irrelevant location (distal from the electrodes used for the pain stimulation, depending on the counterbalancing the dominant wrist or triceps tendon).

Robotic arm task

A three-degrees of freedom robotic arm (The HapticMaster) was used to perform the motor task. This robotic arm allows for a range of movements restrained to 0.36 m depth, 0.40 m height and 1 m width (FCS Robotics; Moog Inc, East Aurora, NY, USA). The movements were confined to a two-dimensional, horizontal movement plane. As the participants moved the robotic arm, they could see their movements on the screen represented by a green ball. The task included different reaching movements passing through one of three arches positioned mid-way the movement plane. Movements were executed using the right hand by holding the sensor at the extremity of the robotic arm. Participants seated in front of an LCD screen, which was used to visualize the task set-up, the participants’ movements, and the ratings scales for self-reports. A Windows 10 compatible triple foot switch (USB–3FS-2; Tokyo, Japan) was used to answer questions presented occasionally during the task. The left pedal was used to scroll to the left, the middle to scroll to the right on the rating scale, and the right foot pedal was used to confirm the participant’s answer. While participants responded to these questions, the robotic arm was blocked.

A green ball indicated the position of the robotic arm sensor representing participants’ own movement on-screen. During each trial, a movement initiation cue (coloring of the trajectory arch in yellow) was presented for 1,000 ms. This cue instructed participants through which of the three arches they were required to move. Participants were told to start moving as soon as the trajectory arch turned black again (movement execution cue). A red ball was shown as a tracing cue, meaning that it moved ahead of the green ball and participants received the instruction to accurately follow the trajectory of the tracing cue. The three possible trajectories were equivalent in terms of force needed to execute the movement. The robotic arm only allowed movement during the movement execution window (during other phases it was haptically blocked). Although all arches were displayed on the screen, only the instructed movement was possible (i.e., the movement through the arch that was colored in yellow during movement preparation phase). This meant that the movement of the robotic arm through those arches was not possible (if participants tried to go through the wrong arch the robotic arm would block). Boundaries were set along the trajectory and the robotic arm was haptically blocked beyond the boundaries so that participants had limited movement space around the defined trajectories. The boundaries were not visible on the movement plane. This set of boundaries determined the trajectory that the tracing cue would follow through each of the arches. This way, it was made sure participants stayed within the limit of the trajectory of the tracing cue. Each trial ended once the green ball reached the end of the movement plane where a target arch was presented. The robotic arm was then automatically repositioned to the start location (in 750 ms). As the participants moved the robotic arm, aiming for the indicated arch, the other two non-available arches were colored grey (for an illustration of the trials structure see Fig. 1).

Each of the three possible movements was paired with a specific outcome. One movement was followed by a pain stimulus (threat movement), one movement was accompanied by pain and reward (conflicting movement, see Reward manipulation). In the neutral movement, there were no consequences. The order by which the movement trajectories were displayed on the screen was counterbalanced across participants so that the trajectories serving as threat, conflicting, and neutral were located on different positions to avoid confounding effects due to their location.

Reward manipulation

To increase the motivational value of the movement, a reward manipulation was used. At the beginning of the experiment, participants were asked to choose a prize worth 100 Euros from a list. During the conflicting movement trials only, a lottery ticket image was presented on the right upper corner of the movement plane. Participants were told that the image of the lottery ticket indicated that their motor accuracy was being measured and that their motor accuracy at these trials would determine the number of lottery tickets that they could receive. This was done to increase the motivational value of performing the movement accurately. Moreover, at the end of each block, they received feedback regarding their motor accuracy together with the cumulative number of lottery tickets collected up to that point. However, although we did measure participants’ motor accuracy, the feedback presented consisted of a bogus percentage accuracy meter (programmed to show a fixed value for all participants). This was done to ensure that all participants received the same reward. The sequence of lottery tickets gained was: 12 and 14 (during acquisition), 15, 16 and 18 (during test phase) and 20 after the free choice phase. An exemplary picture of the bogus percentage accuracy meter is shown in the last screen of Fig. 1.

Procedure

The experiment started with the pain calibration procedure. Subsequently, the experiment consisted of four experimental phases, namely a practice, acquisition, test, and free choice phase.

Pain calibration

For the calibration, a series of electrocutaneous stimuli of increasing intensity were manually administered following a standardized procedure based on previous studies (Gatzounis & Meulders, 2020; Glogan et al., 2020; Meulders et al., 2016). Hence, data were collected as described previously (Vandael, 2022). Specifically, after each stimulus participants rated how painful it was on a 0-to-10-point rating scale, where 0 is labelled as “I feel nothing”; 1 as “I feel something, but this is not unpleasant; it is only a sensation”, 2 as “the stimulus is not yet painful, but is beginning to be unpleasant”, 3 as “the stimulus starts being painful”; and 10 as “this is the worst pain I can imagine”. The participants were encouraged to select a stimulus that was “significantly painful and demanding some effort to tolerate”, roughly representing a stimulus intensity of 7 or 8, on this calibration scale. Participants’ physical intensity of the electrocutaneous stimulus selected with the calibration procedure was 31.44 mA (SD = 1.44) and the mean self-reported stimulus intensity was 7.67 (SD = 0.56).

Practice phase

A practice phase was included to make sure that participants understood the task and knew how to answer the self-report questions by using the triple foot switch. For this purpose, each of the three movements (conflicting, threat, and neutral) was performed twice (six trials in total) and no pain or tactile stimuli were presented. After the practice phase, participants were trained on how to use the triple foot switch to answer the questions. The corresponding arch was colored green as soon as the question was presented on the screen to indicate to which trajectory the question pertained and remained colored until the question was answered. The following questions were presented on the screen above the movement plane after completion of the practice trials: “To what extent would you expect a stimulus when moving through the green arch?” (Pain expectancy), “How afraid would you be to move through the green arch?” (Fear of movement) and “How painful was the electrical stimulus you received when you moved through the green arch?” (Pain intensity). Since the neutral movement was not paired with a pain stimulus, the latter question was not presented for this trajectory. Participants were asked to answer using the 0–100 Visual Analogue Scale (VAS; 0 = “not at all” and 100 = “very much”) presented on the screen.

Acquisition phase

This phase consisted of two blocks of 12 trials, with four trials per movement type per block (24 trials in total). During this phase, participants learned that each movement was paired with different outcome. The threat movement was paired with a pain stimulus in 75% of the trials. The conflicting movement was paired with pain in 75% of the trials, but also with a reward (See Reward manipulation for more details). The neutral trials were not paired with any consequences. For the threat and conflicting movements, a 75% reinforcement rate was used—instead of a 100%—since anticipating uncertain threat elicits stronger conditioning responses (Bennett, Dickmann & Larson, 2018; Morriss, Zuj & Mertens, 2021). At the end of each block, participants answered the three questions (pain-related fear, pain expectancy and pain intensity) for each trajectory (for the full text of these questions see Practice section).

Test phase

The test phase consisted of three blocks of 96 trials (32 trials of each movement type per block), for a total of 288 trials. In addition, tactile stimuli were administered in 75% of the trials. Pain stimuli were administered in the remaining 25% of the trials during the threat and conflicting movement (this was done to counteract quick extinction; Glogan et al., 2020). In Table 1, we illustrate the study design and the number of trials per each condition. The stimuli (pain and tactile) were administered between 500 and 550 ms from the onset of the movement. Furthermore, the same questions as in the practice and acquisition phase were presented. Throughout the test phase, the EEG signal was recorded (see EEG recording and analysis for more details).

Free choice

Finally, the last phase was a free choice phase in which participants could freely decide through which trajectory they could move. This block was included as a manipulation check to measure how motivated participants were to perform each movement. For 12 free choice trials of this phase, an auditory start cue (1,000 ms) replaced the visual movement initiation cue (i.e., the yellow coloring of the arch). No arch color was changed, and the tracing cue was not displayed. Participants were simply asked, on each trial, to freely choose one arch to move through with the green ball. The trial ended once the movement was completed after which the robotic arm was repositioned to the start location and fixed until the start of the next trial. While no tactile stimuli were administered, pain stimuli were again administered in 25% of trials between 500 and 550 ms from movement initiation (only for threat and conflicting trials).

Two additional questions were asked assessing the participants’ personal importance of reward and avoidance respectively with the questions: “How important was it for you to gain more lottery tickets?” (Reward importance) and “How important was it for you to avoid the pain stimuli?” (Avoidance importance) using a 0–100 Visual Analogue Scale (VAS; 0 = “not at all” and 100 = “very much”).

Outcomes

Our primary outcome, namely pain-related somatosensory attention, was operationalized as the mean amplitude of the N120 and P200 components evoked by the tactile stimuli at the pain-relevant and pain-irrelevant location. In addition, pain-related somatosensory attention was also assessed via the mean amplitude of the SEPs (N120 and P200) elicited by the pain stimuli (during the conflicting and threat trials only). The latter (pain-evoked) SEPs were analyzed as secondary outcomes. Note that the pain stimuli were implemented in this study for conditioning purposes, which is pairing a motor action with painful or conflicting outcomes (as reported in our pre-registration osf.io/vbgru). We also included manipulation check measures of pain intensity, fear of movement, pain expectancy (per movement type) and self-reported measures of questions asked at the end of the free choice phase (avoidance and reward importance). Moreover, we obtained measures of individual preferences during the free choice phase (number of trials participants selected each of the three trajectories when they could freely choose). Finally, we measured motor performance accuracy per movement type (operationalized as the number of times the boundaries were hit).

EEG recording and analysis

EEG was recorded with a 32-channel portable system (EEGO Sports; ANT neuro system, Hengelo, Netherlands). The active electrodes were placed according to the international 10–20 setting system. The ground electrode was located in the active-shield cap fronto-centrally between the FPz and the Fz electrode. Impedances were kept <10 kΩ, using an electroconductive paste. The sampling rate was set at 2,000 Hz. No online filters were applied. Data were pre-processed off-line by using a Matlab open-source toolbox for EEG analysis: Letswave 7 (www.letswave.org/).

The continuous EEG data were re-referenced offline to an average of all electrodes. Data were filtered with a 4th order Butterworth 0.3–30 Hz band pass filter. Independent component analysis (ICA) was performed to correct artefacts (muscular, bad channel and ocular) and blinks. The EEG signal was epoched in segments from −200 to 500 ms relative to the onset of the tactile stimulation. An automatic artefact rejection was applied to these segments based on a minimum/maximum amplitude check (−75 µV and 75 µV respectively; Clauwaert et al., 2020). Next, a baseline correction (−200 ms, 0 from the onset of the stimulus) was carried out and epochs were averaged per each movement condition and stimulus location (the latter for the tactile stimuli only). As a result of the averaging procedure, we obtained the following waveforms (based on movement type and stimulus location): (a) Threat movement—Pain-irrelevant location, (b) Threat movement—Pain-relevant location, (c) Neutral movement—Pain-irrelevant location, (d) Neutral movement—Pain-relevant location, (e) Conflicting movement—Pain-irrelevant location, (f) Conflicting movement—Pain-relevant location. Additionally, we obtained waveforms elicited by the pain stimuli in the threat movement and in the conflicting movement (secondary outcome). Note that the neutral movement was never paired with a pain stimulus, therefore no SEPs were obtained for this movement type. Similarly, for the pain stimuli, the averaging procedure only resulted in SEPs from two conditions, namely conflicting and threat movements.

A collapsed localizer (grand-grand average across conditions) was used to identify the waveforms peaks and their scalp distribution (Luck & Gaspelin, 2017). With this procedure and consistently with previous research (Clauwaert et al., 2018, 2020), the following peaks were identified: a negative peak around 130 ms (for the electrodes Fz, Fc1, Fc2 and Cz) and a positive one around 214 ms (for the electrodes Cz Fc1, Fc2, Cp2 and Cp1) from stimulus (tactile) onset. Similarly, for the pain stimuli, a negative peak was identified at 118 ms (for the electrodes Fz, Fc1, Fc2 and Cz) and a positive peak at 220 ms (for the electrodes Cz Fc1, Fc2, Cp2 and Cp1) from the onset of the pain stimulus. We selected 50 and 100 ms intervals around averaged waveform peaks of all participants and all conditions for the N120 and the P200 component respectively (Clauwaert et al., 2018, 2020). With this purpose, we computed the waveform mean area (amplitude by time). Statistical analysis was performed on the pooled (averaged) channels where the peaks were identified.

Data analysis

Statistical analyses were performed in JASP (2020 Version 0.13.1; JASP Team, 2020).

To test whether the threat of pain (hypothesis 1.a) or nonpain motor goal pursuit (hypothesis 2.a) modulates somatosensory attention (operationalized as the amplitude of the N120 component elicited by tactile stimuli applied to the pain-relevant location), we used a 2 (stimulus location: pain-relevant vs pain-irrelevant) × 3 (movement type: threat, neutral, conflicting) repeated measures analysis of variance (RM ANOVA). For significant main results or interactions, we used post-hoc tests. For multiple comparisons, we used a Bonferroni correction (i.e., we divided the critical p value (α) by the number of comparisons) to test for significant differences in the SEPs amplitude between each pair of conditions. Specifically, we compared threat and neutral trials for hypothesis 1.a and threat and conflicting trials for hypothesis 2.a.

Similarly, to test whether the threat of pain (hypothesis 1.b) or nonpain motor goal pursuit (hypothesis 2.b) modulates general non site specific somatosensory attentional processes (operationalized as the amplitude of the P200 component), we used a 2 (stimulus location: pain-relevant vs pain-irrelevant) × 3 (movement type: threat, neutral, conflicting) repeated measures analysis of variance (RM ANOVA). For significant main results or interactions, we used post-hoc tests (with Bonferroni correction for multiple comparisons) to test for significant differences in the SEPs amplitude between each pair of conditions. Specifically, we compared threat and neutral trials for hypothesis 1.b and threat and conflicting trials for hypothesis 2.b. Note that for this component, we expected a location unspecific effect.

As a secondary analysis, we explored whether the amplitude of the pain-evoked SEPs was decreased in the conflicting compared to the threat movement. For this purpose, we used a t-test for both the N120 and the P200. Cohen’s d or partial eta squared is reported to calculate effect size (respectively, for the t-tests and ANOVAs). For the analyses in which Mauchly’s test of sphericity indicated a violation of the assumption of sphericity, a Greenhouse-Geisser sphericity correction was applied to the degrees of freedom. To test whether the conditioning was successful, two different one-way ANOVAs (within factor: movement type) were performed on the mean score of the manipulation check items (i.e., fear of movement, pain expectancy). Differences in pain intensity were checked using a t-test since there were only two movements paired with pain stimuli (no painful stimuli were administered during the neutral movement). To test for significant differences in the free choices a one-way ANOVA was used with the factor movement type. Furthermore, to test whether the type of movement affected the motor task performance a one-way ANOVA (within factor: movement type) was performed on the motor performance score.

EEG somatosensory evoked potentials

Primary outcomes

Tactile N120. The 2 × 3 RM ANOVA (stimulus location (pain-irrelevant vs pain-relevant) × movement type (threat, neutral and conflicting)) on the tactile N120 SEPs did not show significant effects of movement type, F_(1.514,43) = 1.488, P = 0.23, η_p² = 0.03, or stimulus location, F_(1,43) = 0.002, P = 0.96, η_p² = 0.0005, or interactions, F_(1.582,43) = 0.289, P = 0.698, η_p² = 0.007. Contrary to our hypothesis, threat of pain and nonpain motor goal pursuit (hypothesis 1.a and 2.a respectively) did not impact pain-related somatosensory attention, as the amplitude of this component was not influenced by the type of movement or by the stimulus location.

Tactile P200. The 2 × 3 RM ANOVA (stimulus location (pain-irrelevant vs pain-relevant) × movement type (threat, neutral and conflicting)) on the tactile P200 component showed a main effect of movement type, F_(1.73,43) = 17.20, P =< 0.001, η_p² = 0.29, but no main effect of stimulus location, F_(1,43) = 1.41, P = 0.32, η_p² = 0.02, nor interaction, F_(2,43) = 0.46, P = 0.63, η_p² = 0.01. The post-hoc follow up analysis revealed that the P200 component was significantly larger when the tactile stimuli were administered during a conflicting movement (M = 6.71, SD = 4.08) compared to neutral (M = 5.55, SD = 3.23; t₍₄₃₎ = 5.75, P_bonf =< 0.001, d = 0.87), but not compared to threat movements (M = 6.33, SD = 3.72; t₍₄₃₎ = 1.86, P_bonf = 0.20, d = 0.28) and larger during the threat movement compared to neutral t₍₄₃₎ = 3.88, P_bonf < 0.001, d = 0.59. As predicted (hypothesis 1.b), the threat of pain induced an unspecific increase of this component. This was reflected by a larger P200 in the threat movements as compared to neutral movements. However, against our prediction, our results suggest that nonpain motor goal pursuit did not affect pain-related somatosensory attention (hypothesis 2.b), as reflected by the lack of difference in amplitude between the threat and the conflicting components.

A graphic representation of the N120 and P200 components is showed in Fig. 2.