Freeze-like responses to pain in humans and its modulation by social context

Participants

Thirty-nine healthy, pain-free individuals (31 females; mean age ± SD = 22.79 ± 3.07; range = 18–33) voluntarily participated in this study. Participants were recruited by means of flyers and the departmental experiment management system (EMS; Sona Systems Ltd.). The majority of the participants were students (n = 34; 87%) and ten (25.6%) participants were working. Of the 39 participants, 34 were living alone (87.2%) and five (12.8%) coinhabiting with someone else. Regarding highest education, 16 (41%) participants had completed middle school, 10 (25.6%) participants high school, and 13 (33.3%) participants university education. The exclusion criteria were pregnancy, current or history of cardiovascular disease, chronic or acute respiratory disease (e.g., asthma, bronchitis), neurological diseases (e.g., epilepsy), any current or past psychiatric disorders, acute or chronic pain, uncorrected hearing problems, cardiac pacemaker or the presence of any other electronic, medical devices, impaired, uncorrected vision, or use of anxiolytics or antidepressants. All participants received a financial compensation of €15 for their participation. Power calculations were run using GPower (Faul et al., 2007), assuming a medium effect size of Cohen’s f = .25, a (conservative) between-measurement correlation of .05, an alpha level of .05, and aiming for a power of 90%, resulting in a required sample of 36 participants. A medium effect size was based on the range of effect sizes found in the original study by Roelofs, Hagenaars & Stins (2010).

Ethical approval

The experimental protocol was approved by the Social and Societal Ethics Committee of the KU Leuven (Belgium) (registration number = S-56678). All participants provided written informed consent prior to participation. It was emphasized that participation was completely voluntary and that participants were allowed to stop the experiment at any time without any negative consequences.

Apparatus and experimental stimuli

Experimental setting

The experiment took place in a dimmed experimental room. During the experiment participants stood on the stabilometric force platform for the majority of the time while the experimenter was present in the same room but out of sight.

Procedure

The experimental session lasted approximately 120 minutes and consisted of a preparation and calibration phase, an acquisition phase, and a generalization phase. Figure 1 provides a more detailed overview of the acquisition and generalization phases, which we subsequently refer to as the experimental phases. A within-subject design was employed, meaning that all participants ran through both threatening and safe contexts. Note that the generalization phase was exploratory. Considering the high drop out in this phase the findings are unreliable and thus should be interpreted with caution. We decided to include a full description of the methods and results of the generalization phase in the supplementary material only rather than in the main manuscript to facilitate clarity (see Supplemental Information 3).

Outcome measures

Behavioral outcomes

Statistical analyses and data reduction

Paired-samples t-tests were carried out to compare the level of arousal, control and valence after hearing the auditory cue (CS) before and after it was paired with the painful stimulus. As a manipulation check for the social context manipulation, paired-samples t-tests were conducted to compare the valence, arousal, control, threat, and pleasantness ratings in response to the happy and angry facial stimuli. To this end, all ratings were averaged separately for the safe and threatening blocks.

Pain-US expectancy ratings and body sway were analyzed using repeated measures (RM) ANOVAs with 2 [Stimulus type (CS+/context-alone)] × 2 [Context (threat/safe)] × 3 [Block (A1, A2, A3)] as within-subject factors. Separate 2 [Context (threat/safe)] × 3 [Block (A1, A2, A3)] RM ANOVAs were run to examine the effects of social context on heart rate, facial pain expression and pain ratings (pain intensity and pain unpleasantness). In the analyses of heart rate, painful facial expression and body sway, centered STAI-T scores were included as a covariate.

To deal with frequent violations of the assumption of sphericity, multivariate analyses were run. Pillai’s trace estimates and the effect size indication ${\mathrm{\eta }}_p^2$ are reported. Planned comparisons were carried out to test our a priori hypotheses. Holm Bonferroni corrections were used to correct for multiple testing (Holm, 1979). All statistical analyses were run using IBM SPSS 20 (SPSS Inc, Chicago, IL). An alpha level of .05 was used for all statistical tests. We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study (Simmons, Nelson & Simonsohn, 2012).

Dropout

The experimental procedure was quite demanding for the participants and 6 participants stopped their participation before the end of the experiment because they indicated that they started to feel unwell and as a precaution the experimenter decided to terminate the experiment. From the initial 39 participants, 33 participants (84.6%) completed the full experiment including the generalization phase, and 36 participants (92.3%) completed only the acquisition phase which is reported in this manuscript. The three (7.7%) remaining participants stopped their participation during the acquisition blocks. All subsequent analyses are run on the 36 participants who completed the acquisition phase unless otherwise stated.

Manipulation checks

Pain-US expectancy ratings

We found support for differential acquisition: Participants increasingly expected the pain-US following the presentation of the CS, but not when the context was presented alone, Stimulus type x Block, F (2, 34) = 8.79, p < .01, ${\mathrm{\eta }}_p^2$ = .34, (see Fig. 2). However, the magnitude of the differential acquisition effect did not differ between the two contexts, Stimulus type ×Block ×Context, F (2, 34) = 1.17, p = .32, ${\mathrm{\eta }}_p^2$ = .07. In contrast, pain-US expectancy ratings were overall higher in the threatening context, M _threat = 6.70, SD _threat = 1.85, compared to the safe context, M _safe = 6.01, SD _safe = .2.21, main effect of Context: F (1, 35) = 5.87, p = .02, ${\mathrm{\eta }}_p^2$ = .14.

Pain intensity and unpleasantness ratings

Pain intensity ratings did not differ across blocks, main effect of Block: F (2, 34) < 1, p = .64, ${\mathrm{\eta }}_p^2$ = .03, or between contexts, main effect of Context: F (1, 35) < 1, p = .68, ${\mathrm{\eta }}_p^2$ < .01, and there was also no interaction between block and context, Block ×Context interaction: F (2, 34) = 1.03, p = .37, ${\mathrm{\eta }}_p^2$ = .06. The same was true for pain unpleasantness ratings, Block: F (2, 34) < 1, p = .42, ${\mathrm{\eta }}_p^2$ = .05; Context: F (1, 35) < 1, p = .56, ${\mathrm{\eta }}_p^2$ = .01, Block ×Context interaction: F (2, 34) < 1, p = .96, ${\mathrm{\eta }}_p^2$ < .01.

Hypothesis 1: does social threat modulate body sway in response to pain?

Two participants were identified as outliers because of their excessive movements, as reflected in Z-scores greater than 4 on the body sway measures.¹ Consequently, we decided to remove these participants from the analyses. Thus, the following analyses were run on 34 participants. The four-way interaction was not significant, Context × Block × Stimulus × Trait anxiety, F (2, 30) < 1, p = .74, ${\mathrm{\eta }}_p^2$ = .02, and neither were any of the other interaction or main effects. There was a slight trend showing less body sway during CS trials compared to context-alone, Stimulus, F (1, 31) = 3.03, p = .09, ${\mathrm{\eta }}_p^2$ = .09, but contrary to our hypothesis we found no evidence for any effect of social context or painful stimuli on body sway.

Hypothesis 2: does social threat reduce heart rate?

Because of technical difficulties with the heart-rate monitor, recordings of three participants were missing and hence the following analyses are run on 33 participants. With regard to heart rate, a significant three-way interaction emerged, Block × Context × Trait anxiety, F (2, 30) = 3.68, p = .04, ${\mathrm{\eta }}_p^2$ = .20 (see Fig. 3). We then evaluated context differences per block for high (mean STAI-T + 1SD) and low levels of trait anxiety (mean STAI-T –1SD) using planned comparisons. Note that this analysis models the interaction of block and context at different levels of the covariate (+1 SD and −1 SD) and thereby avoids the power loss associated with a median split. The model is thus based on the whole sample [see 27 for use of a similar analysis].

While there were no differences between a threatening and a safe social context in any of the blocks at low anxiety levels, F (1, 31) < 1, p = .89, ${\mathrm{\eta }}_p^2$ < .01, there was a difference in heart rate at high anxiety levels. Specifically, the threatening social context, M _threat = 645.47, SD _threat = 141.49, led to reduced heart rate in the very last block compared to a safe social context, M_safe = 659.49, SD_safe = 147.64, F (1, 31) = 9.92, p < .01, ${\mathrm{\eta }}_p^2$ = .24.

Hypothesis 3: does social threat reduce facial pain expression?

Because of technical difficulties recordings of three participants were missing and hence the following analyses are run on 33 participants. Regarding facial pain expression, no significant three-way interaction emerged, Block × Context × Trait anxiety, F (2, 30) = 1.26, p = .30, ${\mathrm{\eta }}_p^2$ = .08. As expected, we found a significant effect of social context that was moderated by trait anxiety, Context ×  Trait anxiety, F (1, 31) = 5.35, p = .03, ${\mathrm{\eta }}_p^2$ = .15, (see Fig. 4). Subsequently, we evaluated simple effects of social context at high anxiety levels (1 SD above mean STAI-T score) and low anxiety levels (1 SD below mean STAI-T score). In line with our hypotheses, we found that there is less facial pain expressions in the threatening context at high anxiety levels, M_threat = 2.62, SD_threat = 5.27, compared to the safe context, M_safe = 3.55, SD_safe = 5.56, F (1, 31) = 4.89, p = .04, ${\mathrm{\eta }}_p^2$ = .14. In contrast, there was no difference between the two contexts at low anxiety levels, M_threat = 3.56, SD_threat = 4.91; M_safe = 3.09, SD_safe = 5.21, F (1, 31) = 1.24, p = .27, ${\mathrm{\eta }}_p^2$ = .04.

Conclusions and future directions

In conclusion, the current study provides preliminary evidence that social context modulates pain-relevant processes, especially in highly anxious individuals. Social threat increased overall expectation of pain and reduced facial pain expression. In addition, increased bradycardia in the threatening social context was restricted to a single block and only present in highly anxious participants. In contrast, we did not find freezing in response to painful stimuli and also no effect of social context on body sway. This research relates to earlier studies showing impaired safety learning and excessive avoidance behavior in highly anxious individuals, and might add to our understanding of how chronic pain complaints develop and are maintained (Meulders, Meulders & Vlaeyen, 2014; Meulders & Vlaeyen, 2013). In addition, the present research further highlights the importance of social context in the study and understanding of pain (Karos et al., 2018a).

There are several possible avenues for future research. First, more research is needed on the effects of acute and chronic pain on freezing parameters. To this end, the effect of experimental pain induction methods and anticipation of pain on freezing responses in humans should be investigated in isolation. In addition, freezing responses in patients with chronic pain should be compared to pain-free controls to investigate possible differences in defensive responding. There is some evidence for maladaptive freezing responses in other psychopathologies such as PTSD (Niermann, Figner & Roelofs, 2017) and in rats with chronic pain (Lamana et al., 2018), but research in humans is absent. It is well established that patients move differently when they are in pain (Hodges & Tucker, 2011) and even the anticipation of pain changes movement (Karos et al., 2017). If chronic pain patients do show signs of maladaptive freezing responses, as they do with avoidance behavior, longitudinal research is needed to evaluate the role of this process in the etiology of chronic pain complaints. Second, more research into the modulation of defensive behaviors (e.g., freezing, pain-related avoidance behavior) by social context is warranted. One possibility is the use of more dynamic and ecologically valid social manipulations such as virtual reality, video materials or actual social interactions using confederates (Karos et al., 2019; Karos et al., 2018).