Ò
PAIN 154 (2013) 1509–1518
www.elsevier.com/locate/pain
Research papers
Hypnotic susceptibility modulates brain activity related to experimental placebo analgesia Alexa Huber ⇑, Fausta Lui, Carlo Adolfo Porro Department of Biomedical, Metabolic and Neural Sciences, University of Modena and Reggio Emilia, Modena I-41125, Italy
Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.
a r t i c l e
i n f o
Article history: Received 17 September 2012 Received in revised form 8 February 2013 Accepted 22 March 2013
Keywords: Hypnotisability Placebo fMRI Pain anticipation Pain Humans
a b s t r a c t Identifying personality traits and neural signatures that predict placebo responsiveness is important, both on theoretical and practical grounds. In the present functional magnetic resonance imaging (fMRI) study, we performed multiple-regression interaction analysis to investigate whether hypnotic susceptibility (HS), a cognitive trait referring to the responsiveness to suggestions, explains interindividual differences in the neural mechanisms related to conditioned placebo analgesia in healthy volunteers. HS was not related to the overall strength of placebo analgesia. However, we found several HS-related differences in the patterns of fMRI activity and seed-based functional connectivity that accompanied placebo analgesia. Specifically, in subjects with higher HS, the placebo response was related to increased anticipatory activity in a right dorsolateral prefrontal cortex focus, and to reduced functional connectivity of that focus with brain regions related to emotional and evaluative pain processing (anterior mid-cingulate cortex/ medial prefrontal cortex); an opposite pattern of fMRI activity and functional connectivity was found in subjects with lower HS. During pain perception, activity in the regions reflecting attention/arousal (bilateral anterior thalamus/left caudate) and self-related processing (left precuneus and bilateral posterior temporal foci) was negatively related to the strength of the analgesic placebo response in subjects with higher HS, but not in subjects with lower HS. These findings highlight HS influences on brain circuits related to the placebo analgesic effects. More generally, they demonstrate that different neural mechanisms can be involved in placebo responsiveness, depending on individual cognitive traits. Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
1. Introduction Recently, there has been growing interest in identifying personality traits that predict good placebo responsiveness [11,16,21, 22,27,48,65]. Hypnotic susceptibility (HS), or hypnotisability [25], is a cognitive trait that refers to the generalised tendency to respond to hypnotic suggestions [23], including those for analgesia [30,47]. HS also predicts the efficacy of suggestions administered during normal wakefulness – termed ‘‘imaginative suggestions’’ [43], which might be relevant for the placebo effect. The HS trait is associated with attentional absorption [67,79], imagery vividness [9,32], and fantasy-proneness [37]. It is in part heritable [63] and can be reliably measured with standardised scales [17,53]. It has been claimed that HS should predict good placebo responsiveness (eg, by [31,63]), as both placebo effects [3,55,59] ⇑ Corresponding author. Address: Dipartimento di Scienze Biomediche, Metaboliche e Neuroscienze, Sezione Fisiologia e Neuroscienze, Università di Modena e Reggio Emilia, Via Campi, 287, I-41125 Modena, Italy. Tel.: +39 059 205 5684; fax: +39 059 205 5363. E-mail address:
[email protected] (A. Huber).
and analgesia related to hypnotic or imaginative suggestions are at least in part mediated by expectancy [20,28,31,44]. However, previous behavioural studies failed to demonstrate a significant association between HS and placebo response [39,66]. Therefore, the relationship between HS and placebo responsiveness may be more complex than previously assumed. One possible hypothesis is that different neurocognitive mechanisms underlie placebo effects, depending on the individual level of HS, even if the overall magnitude of placebo response is not affected. Indeed, the placebo analgesic effect is not a unitary phenomenon. Rather, it has been demonstrated that placebo analgesia can be mediated by different mechanisms related to conditioning, expectancy, reduced anxiety, and reward anticipation [16]. The aim of the present study was to investigate, using functional magnetic resonance imaging (fMRI), whether HS is associated with differences in the neural mechanisms underlying the placebo analgesic response in healthy volunteers, in a conditioned placebo protocol. Specifically, we investigated whether neural activity and functional connectivity can be explained in terms of an interaction between HS and behavioural placebo effects.
0304-3959/$36.00 Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.pain.2013.03.031
1510
Ò
A. Huber et al. / PAIN 154 (2013) 1509–1518
Previous studies on HS have been criticised for including only subjects with high (‘‘Highs’’) or low (‘‘Lows’’) HS score, thereby ignoring about one half of the population, which falls in the medium range of HS (‘‘Mediums’’) [46,58]. In the present study, we did not preselect the subjects, and we did not divide them into subgroups. Instead, we used the individual HS score as a linear regressor to assess relationships based on the whole naturally occurring distribution of HS. Findings obtained during the fMRI session of the study, but not related to HS, have been published in Lui et al. [36].
2. Materials and methods 2.1. Subjects We investigated 36 healthy volunteers without any history of neurological or psychiatric illness, who were not under medications at the time of the study. Eight subjects were excluded from the analysis because of excessive head motion (n = 2) or technical problems during the MR session (n = 3), or because they did not return for the HS session (n = 3). The remaining 28 subjects (11 male; mean age 22.4 years; 23 right-handed, 4 ambidextrous, and 1 lefthanded) entered the final analysis. Handedness was assessed using the Edinburgh Inventory [51]. 2.2. Experimental procedures All experimental procedures were conducted in conformance with the politics and principles contained in the Declaration of Helsinki, and all subjects gave their written informed consent to take part in the study. The study included 2 sessions – the first for the fMRI experiment and the second for the HS measurement. During the first session, subjects were told that the aim of the study was to evaluate both the efficacy and the brain correlates of a new analgesic procedure; to this end, they would receive noxious cutaneous heat stimuli on the top of one foot while in the MR scanner and, in some trials, the painful stimulus would be accompanied by a sub-threshold electric stimulation at the ankle, which could induce analgesia. After the completion of the fMRI experiments, each subject was debriefed, then was invited to a second session for the assessment of some personality characteristics, during which HS was assessed individually by one of the authors (A.H.); HS assessment was double-blind for placebo response, and vice versa. Subjects received a small remuneration for taking part in the study. 2.2.1. Session 1: placebo fMRI experiment Full details of the experimental protocol and of fMRI data acquisition can be found in Lui et al. [36] and will only be summarised here. Radiant heat pulses, generated by an infrared solid-state laser stimulator (Electronic Engineering, Florence, Italy) with a wavelength of 1.34 lm and a laser beam approximately 10 mm in diameter, were used as cutaneous stimuli. To avoid nociceptor fatigue or sensitisation, the laser beam was moved randomly after each stimulus over a 3 5 cm skin area. During the fMRI experiment, stimulus intensity was set either at an individually defined level inducing a nonpainful (N) warm sensation, just below pain threshold, or at a level inducing pain (P) of moderate intensity. The side of stimulation was randomised across subjects (the right foot was always stimulated in 13, the left foot always stimulated in 15 subjects).The experiment was performed during a normal alert state. Two conditioning runs and one test run, each including 12 trials, were carried out for each subject while fMRI images were acquired.
Each trial lasted 51 seconds (see Fig. 1). At time 0, subjects received a visual warning cue: the black screen they were looking at turned either Red (‘‘Red’’ trials) or Green (‘‘Green’’ trials). Twelve seconds later the screen turned black again, and immediately afterwards the subject received a stimulation on the top of one foot. The subjects had been informed that the Red cue would be followed by a brief painful laser stimulus, whereas the Green cue would be followed by an identical painful stimulus associated with a subthreshold electric shock, which could induce analgesia (the placebo manipulation). To this purpose, 2 electrodes were pasted above the ankle; however, the electrodes were not connected to any pulse generator, and no electric shock was ever delivered. In the 2 conditioning runs, the laser stimulus intensity after the Green cue was milder (nonpainful stimulus, N) with respect to the stimulus intensity following the Red cue (painful stimulus, P). In contrast, in the test run the stimulus following the Green cue had the same power as the stimulus following the Red cue (painful stimulus, P). In each run, ‘‘Red’’ (n = 6) and ‘‘Green’’ (n = 6) trials were pseudo-randomly alternated. Seventeen seconds after the stimulus, subjects had to rate the perceived pain intensity by rotating a knob, which moved a cursor on a computerised visual analogue scale, anchored at 0 = no pain, and 100 = worst imaginable pain. The differences in pain ratings between the ‘‘Red’’ and the ‘‘Green’’ trials in the test run, expressed as a t-score, were taken as a measure of the behavioural placebo response of the subject. Functional images were acquired with a Philips Intera MR system at 3T and a gradient-echo echo-planar sequence (repetition time [TR] = 3000 ms; echo time [TE] = 35 ms; field of view [FOV] = 240 mm; acquisition matrix 80 80, reconstructed at 128 128; 30 axial slices; voxel size 1.9 1.9 3.5 mm; 0.5-mm interslice gap). High-resolution T1-weighted anatomical images were acquired for each subject to allow anatomical localisation (TR = 9.9 ms; TE = 4.6 ms; 170 sagittal slices; voxel size 1 1 1 mm). 2.2.2. Session 2: hypnotic susceptibility and self-report measures Fear of pain was assessed using an ad hoc Italian translation of the Fear of Pain Questionnaire (FPQ-III) [40], which is specifically designed for healthy people. The FPQ-III consists of 30 items and has shown good internal consistency and test-retest reliability. Subjects report how fearful they are about experiencing pain associated with various situations (eg, ‘‘Biting your tongue while eating,’’ ‘‘Having a tooth pulled’’). The Tellegen Absorption Scale (TAS) [67] consists of 34 items and assesses the tendency to be involved in one’s own mental images, a cognitive trait related to HS. HS was assessed using the Italian version of the Stanford Hypnotic Susceptibility Scale – Form A [76], which yields a total HS score ranging between 0 and 12. Scores 0-3 are considered ‘‘low,’’ scores 4-8 ‘‘medium,’’ and scores 9-12 ‘‘high.’’ The Stanford Hypnotic Susceptibility Scale is the most widely used measure of hypnotisability and yields reliable and stable results [10,53].
Fig. 1. Time course of events in a single trial during the functional magnetic resonance imaging placebo experiment.
Ò
A. Huber et al. / PAIN 154 (2013) 1509–1518
2.3. Data analysis
1511
2.3.1. Statistical analysis of behavioural results To assess associations between HS and pain ratings in the fMRI experiment, we performed an analysis of covariance including one covariate (HS) and the following 3 repeated-measures factors: cue (Red vs Green), run (1-3), and trial (1-6). Pearson correlation coefficients were used to assess correlations between HS, behavioural placebo response, and questionnaire results. We also performed 4 multiple regression analyses to assess whether HS and behavioural placebo response interacted in explaining 1) TAS score, 2) FPQ score, and pain ratings in the 3) ‘‘red’’ and 4) ‘‘green’’ trials of the test run, respectively, using the interaction approach described in detail in the next session. All analyses were carried out using the SPSS software package (SPSS Inc, Chicago, IL, USA) and a significance level set at P < 0.05.
interaction variable indicates that the behavioural placebo response is less positively (or more negatively) correlated with the BOLD signal in the more hypnotisable subjects. In summary, a significant interaction effect would confirm our hypothesis that specific brain regions are more or less relevant for the placebo effect, depending on the subjects’ level of HS. Note that the division into low vs medium-high hypnotisable subjects shown in the Figures in the Results section serves only to illustrate the interaction – all statistical analyses were actually performed on the whole sample using multiple regression, without creating any subgroups. A double statistical threshold (voxel intensity and spatial extent) was adopted to achieve a combined significance level, corrected for multiple comparisons, of a < 0.05, as assessed by AlphaSim (Alpha Simulation) with 1000 Monte Carlo simulations (http://afni.nimh.nih.gov/afni/doc/manual/AlphaSim).
2.3.2. fMRI data analysis Data analysis was performed using the MATLAB 7.12 (The MathWorks, Inc, Natick, MA, USA) and SPM5 (Wellcome Department of Imaging Neuroscience, London, UK) software packages. For each subject, all functional volumes were realigned to the first volume acquired, slice time corrected, normalised to the MNI (Montreal Neurological Institute) template, and smoothed with a 4 4 8 mm full width at half maximum (FWHM) Gaussian kernel. The statistical analysis of fMRI data was performed under the framework of the generalised linear model (GLM), including separate regressors related to the anticipation phase (block of 4 TRs), the perception phase (modelled as a brief event starting at stimulus onset), and the rating phase – see [33]. Only results related to the anticipation and perception phase for the test run are described in the present study. The 6 motion parameters obtained during image realignment were included as nuisance regressors to account for possible residual movement-related signal changes. In each subject, conditionspecific effects were compared using linear contrasts. The individual contrast images were then submitted to a second-level random-effects regression analysis, which included either HS, or behavioural placebo response (as revealed by t-scores obtained by comparing pain ratings between ‘‘Red’’ and ‘‘Green’’ trials in the test run), or both, as the main regressor(s). These analyses aimed to investigate, first, whether HS explains increases in fMRI blood oxygen level dependent (BOLD) signals in the ‘‘Green’’ compared to the ‘‘Red’’ trials, and whether it did so independently of the strength of the behavioural placebo response. Second, in order to evaluate a possible link between HS and overall differences in cortical activity not related to the specific cue, for example, due to differences in attention or anxiety during the experiment, we assessed the correlation between HS and BOLD signals for ‘‘Red’’ and ‘‘Green’’ trials taken together. Results associated with HS, but independent of the placebo effect, are not presented here because they were outside the aim of the study. Third, we used interaction analysis [8] to assess whether HS and placebo responsiveness interacted in explaining BOLD signals in specific brain areas. To this end, we created a new regressor, representing the interaction between HS and the behavioural placebo response, by subtracting the mean from these 2 variables, and multiplying them element by element with each other. As described in Friston et al. [19], this analysis looks for a difference between subjects with higher vs lower HS scores in the slope of regression describing the relationship between the behavioural placebo response and BOLD signal in a specific brain region. A significant positive correlation of this new HS-by-placebo interaction variable with the BOLD signal indicates that the behavioural placebo response is more positively correlated with the BOLD signal in subjects with higher HS. Conversely, a negative correlation with the
2.3.3. Seed-based functional connectivity analysis As described in the Results section, we found a significant HSby-placebo interaction in the right dorsolateral prefrontal cortex (DLPFC). To better understand the meaning of this interaction, we performed seed-based functional connectivity (FC) analysis using the AFNI software package (http://afni.nimh.nih.gov/afni). Seed-based FC analysis examines the correlations in BOLD signal fluctuations over time between a region of interest (‘‘seed’’) and all other brain voxels, thus providing information on the coherence of fMRI activity between different brain regions [18]. Only the test run was included in the FC analysis. For each subject, all functional volumes were slice time corrected, realigned to the last volume acquired, globally scaled by dividing the BOLD signal in each voxel by the total mean to make the data more comparable between subjects, normalised to the Talairach template, and smoothed with a 6-mm FWHM Gaussian kernel. The seed signal was obtained by averaging the BOLD signal within a sphere of 5-mm radius around the Talairach coordinates x = 37, y = 17, z = 28, referring to the peak voxel of the interaction cluster in the right DLPFC obtained by repeating the GLM analysis (described in the previous section) in AFNI. Following [64], no global signal regressor was included in the analysis to avoid biasing of the correlation results. Several different GLMs were constructed for each subject to explore the FC with the seed in various phases of the test run, namely: 1) during the entire run; 2) only during the 12-second anticipation phase of each trial (including both ‘‘Red’’ and ‘‘Green’’ trials); and 3) only during the perception phase of each trial, defined for the purpose of the FC analysis as the 12-second period starting with the laser stimulation (including both ‘‘Red’’ and ‘‘Green’’ trials). Two additional GLMs were constructed following a psychophysiological interaction model approach as described in [19] to assess the change in FC with the seed: 4) during the ‘‘Green’’ compared to the ‘‘Red’’ anticipation phases, and 5) during the ‘‘Green’’ compared to the ‘‘Red’’ perception phases. Average white matter signal, average lateral ventricle signal, the 6 motion parameters obtained during image realignment, and the various task-related regressors (‘‘Red’’/ ‘‘Green’’ anticipation, stimulation and motor response – see previous section) were included as nuisance regressors in all models. As the final output of these models for each subject, the FC of each voxel with the seed region was expressed as a Pearson correlation coefficient, transformed to a z-score using Fisher’s transformation. These individual z-score maps were then submitted to a second-level random-effects regression analysis, which included HS, the behavioural placebo response (t-score), and the HS-by-placebo interaction variable as regressors of interest. As for the original GLM analysis, a double statistical threshold was adopted to achieve a combined corrected significance level of a < 0.05 as tested using AlphaSim.
1512
Ò
A. Huber et al. / PAIN 154 (2013) 1509–1518
Material. The main result was that HS was negatively correlated with overall pain intensity in the test run, but independently of the cue and, thus, of the placebo effect (see Fig. 2). TAS score (mental absorption) was significantly correlated with HS score, as expected (Pearson r = 0.41, P = 0.029). FPQ score (fear of pain) showed no significant correlation with HS or placebo response score. Multiple regression interaction analysis showed that HS and behavioural placebo responsiveness did not interact in explaining either TAS score, FPQ score, or pain ratings in the ‘‘Red’’ or ‘‘Green’’ trials of the test run. 3.2. fMRI results, anticipation phase
Fig. 2. Relationship of pain intensity ratings (within-run average of 6 trials) with behavioural placebo response (t-score; panel A) and hypnotic susceptibility (HS) score (panel B) in the test run. Separate regression lines and Pearson correlation coefficients (r) are shown for each cue (red/green). VAS, visual analogue scale. ⁄ P < 0.05.
3. Results 3.1. Behavioural results A significant placebo response (as revealed by a t-test comparing pain ratings between ‘‘Red’’ and ‘‘Green’’ trials in the test run) was found in 43% of participants. HS scores ranged from 0 to 11 (mean ± SD, 3.4 ± 3.3). Seventeen subjects had a low (7 of them were placebo responders), 9 a medium (4 of them placebo responders), and 2 a high HS score (one of them a placebo responder). There was no significant correlation between HS score and placebo response t-score (Pearson r = 0.03, P = 0.897). Fig. 2 shows the relationship of pain ratings in the test run with placebo response (top panel) and with HS (bottom panel). The results of the analysis of covariance assessing the impact of HS on pain ratings over the 3 runs are reported in the Supplementary
The placebo-related brain activity changes, without considering HS, have been described in Lui et al. [36]. A large focus in the right DLPFC (Brodmann area [BA] 46) showed signal increases in the test run during ‘‘Green’’ vs ‘‘Red’’ anticipation that were significantly correlated with the strength of the behavioural placebo response (DLPFC ‘‘placebo’’ focus). This focus remained significant when fear-of-pain (FPQ score) was included in the regression model. Activity changes in this region were not correlated with HS. In fact, HS showed no significant correlations with BOLD signal changes in any region in the anticipation or perception phase. The results regarding the interaction between HS and placebo responsiveness in the anticipation phase are shown in Table 1A. Two foci, in the right DLPFC and parietal regions, showed a significant positive interaction effect (considering ‘‘Green’’ and ‘‘Red’’ trials together). Fig. 3 illustrates this interaction effect for the DLPFC focus. The behavioural placebo response was associated with increased BOLD signal in the more hypnotisable subjects (orange lines in Fig. 3, panel B), but with decreased BOLD signal in the less hypnotisable subjects (black lines). Within the DLPFC focus, the interaction effect reached the cluster-wise significance threshold also in the ‘‘Green’’ trials alone. This DLPFC ‘‘interaction’’ focus is adjacent to, and partly overlaps, the DLPFC ‘‘placebo’’ focus, as shown in Fig. 3, panel C. In summary, a strong behavioural placebo response was associated with 1) higher activity in the right DLPFC during ‘‘Green’’ compared to ‘‘Red’’ anticipation, in the whole experimental population; and 2) higher total activity in this region in subjects with higher HS as compared to less hypnotisable subjects (see Fig. 3, panel C). 3.3. fMRI results, perception phase Overall signal changes (‘‘Green’’ and ‘‘Red’’ trials taken together) were significantly negatively correlated with the HS-by-placebo
Table 1 BOLD signal changes related to the HS-by-placebo interaction in the test run. Regions
BA
Cluster level
Voxel level
MNI coordinates (peak)
Talairach coordinates (peak)
P (corr)
t
x
z
k
Z
y
z
y
z
A. Anticipation phase: Regions showing a positive correlation between behavioural placebo response and signal changes during anticipation (‘‘Red’’ and ‘‘Green’’ trials taken together), but only in subjects with higher HS: R supramarginal/angular gyrus, inferior parietal lobule 40 0.010 160 5.32 4.28 34 54 30 34 51 30 R middle frontal gyrus 0.039 129 5.02 4.11 36 24 22 36 24 19 B. Perception phase: Regions showing a negative correlation between behavioural placebo response and signal changes during pain perception (‘‘Red’’ and ‘‘Green’’ trials taken together), but only in subjects with higher HS: L middle/inferior temporal gyrus 37, 19 0.004 167 6.94 5.09 56 66 6 55 64 2 R middle temporal gyrus 21 0.002 214 6.33 4.81 52 36 10 51 35 7 R middle/inferior temporal gyrus, R middle occipital gyrus 39, 37, 19 0.001 265 6.11 4.70 52 72 8 51 70 3 L/R thalamus, L caudate 0.009 154 4.65 3.89 2 8 14 2 7 13 L precuneus, L superior parietal lobule 7 0.009 156 4.21 3.61 26 68 46 26 64 46 BOLD, blood oxygen level dependent; HS, hypnotic susceptibility; BA, Brodmann area; k, number of voxels; MNI, Montreal Neurological Institute; R, right; L, left.
Ò
A. Huber et al. / PAIN 154 (2013) 1509–1518
1513
temporal gyrus. The other 2 foci were in the left precuneus/superior parietal lobule, and in a subcortical region including the thalamus and left caudate nucleus (Fig. 4A). The details of these interactions are illustrated by the scatter plots in Fig. 4, panel B. Among the more hypnotisable subjects, the behavioural placebo response was associated with lower signal in these regions during the perception phase; in contrast, in subjects with lower HS, the association was either positive (in the case of thalamus and of right posterior middle temporal gyrus) or nonsignificant. The same pattern of correlations was present when either ‘‘Green’’ or ‘‘Red’’ trials were considered separately (not shown). In addition to the interaction effect, all foci except for the precuneus also showed a main effect of increased activity during the perception phase (not shown). 3.4. Functional connectivity of the right DLPFC ‘‘interaction’’ focus To further explore the functional significance of the right DLPFC ‘‘interaction’’ focus described in Section 3.2, we assessed its FC during the test run, and explored its relationship with the HS-byplacebo interaction variable. The regions showing the strongest FC with the right DLPFC across all subjects (irrespective of HS and placebo) are illustrated in Fig. 5. The results regarding differences in FC related to the HS-by-placebo interaction are summarised in Table 2 and in Fig. 6. During the entire test run, HS and behavioural placebo response negatively interacted in explaining DLPFC connectivity with a large focus comprising left anterior cingulate cortex (ACC; BA 32) and left medial and superior frontal gyrus (BA 8, 9, 10) (Table 2A and Fig. 6, panel B). During the perception phase only, HS and placebo response negatively interacted in predicting DLPFC connectivity with 2 foci located in the right cerebellum and in the left inferior frontal gyrus (BA 47), as shown in Table 2B and Fig. 6, panel C. This means that among the more hypnotisable subjects, the placebo response was associated with reduced FC of these regions with the DLPFC, whereas subjects with lower HS showed the opposite pattern. 4. Discussion
Fig. 3. Blood oxygen level dependent (BOLD) signal changes in the dorsolateral prefrontal cortex (DLPFC) related to the hypnotic susceptibility (HS)-by-placebo interaction variable during the anticipation phase of the test run. A right DLPFC cluster (panel A) showed a positive correlation between the HS-by-placebo interaction variable and BOLD signal changes, when ‘‘Red’’ and ‘‘Green’’ trials were considered together. We refer to this focus as the DLPFC ‘‘interaction’’ focus. Regression lines and Pearson correlation coefficients (r) (panel B) depict the correlation of the behavioural placebo response t-score with cluster-averaged signal strength (beta values) during the anticipation phase (average across both ‘‘Red’’ and ‘‘Green’’ trials), separately for subjects with low (black) or medium-high HS score (dashed orange). Note that this division into subgroups serves only to illustrate the interaction – all interaction results are based on a statistical analysis performed on the whole sample using multiple regression, without creating any subgroups – see Methods. The interaction between placebo response and HS score is indicated by the difference in slope between the black and orange regression lines. Panel C: The DLPFC ‘‘interaction’’ focus (yellow) partially overlapped with another focus located in the right DLPFC (red), already described in [36],which showed a BOLD signal increase positively correlated with the behavioural placebo response during ‘‘Green’’ vs ‘‘Red’’ anticipation. ⁄P < 0.05; ⁄⁄P < 0.01; L = left hemisphere. y values represent anteroposterior coordinates (expressed in mm) in the Montreal Neurological Institute (MNI) space.
interaction variable in 5 regions, listed in Table 1B. These included 3 temporal foci: 2 symmetrically located near the occipital-temporalparietal junction and the third, more anterior, in the right middle
This study aimed at exploring HS-related differences in the neural mechanisms mediating placebo analgesia. The results can be summarised as follows. First, HS was not related to the overall strength of placebo analgesia. Second, HS and placebo responsiveness interacted in explaining activity in a right DLPFC focus, and its functional connectivity pattern both during the anticipation and the perception phase. Third, HS and placebo responsiveness also interacted in explaining activity in several brain regions during the perception phase. The present results confirm our hypothesis that differences in HS are associated with different placebo effects, that is, similar behavioural responses mediated by different neural mechanisms; to our knowledge, these findings also provide the first demonstration that different neurocognitive mechanisms can underlie placebo responses, depending on individual cognitive traits. 4.1. Hypnotic susceptibility is not related to the overall strength of the placebo analgesic effect We found no significant correlation between HS and the behavioural placebo response. It has been demonstrated that both scores on HS scales [20,77] and responses to suggestions of analgesia [31,44,45] are, in part, mediated by the subject’s expectancies; this would appear to imply a relationship between HS and the placebo effect [31,63]. However, our behavioural results are in line with those of the few previous studies investigating the association between HS and placebo analgesia, which failed to demonstrate it [39,66].
1514
Ò
A. Huber et al. / PAIN 154 (2013) 1509–1518
Fig. 4. Blood oxygen level dependent (BOLD) signal changes related to the hypnotic susceptibility (HS)-by-placebo interaction variable during the perception phase of the test run. Five brain regions (panel A) showed a negative correlation between the HS-by-placebo interaction variable and BOLD signal changes (‘‘Red’’ and ‘‘Green’’ trials taken together). Regression lines and Pearson correlation coefficients (r) depict the correlation of behavioural placebo response (t-score) with cluster-averaged signal strength (beta values) during the perception phase (panel B), separately for subjects with low (black) or medium-to-high HS score (dashed orange). Note that this division into subgroups serves only to illustrate the interaction – all interaction results are based on a statistical analysis performed on the whole sample using multiple regression, without creating any subgroups. The interaction between placebo response and HS score is indicated by the difference in slope between the black and orange regression lines. Post. MTG, posterior middle temporal gyrus; PRECUN, precuneus; THAL, thalamus; L, left side; R, right side. Coordinates (mm) are in the Montreal Neurological Institute (MNI) space. ⁄ P < 0.05; ⁄⁄P < 0.01; ⁄⁄⁄P < 0.001.
4.2. Hypnotic susceptibility and placebo responsiveness interact in explaining activity and functional connectivity of the right DLPFC The behavioural placebo response was associated with higher anticipatory activity in more hypnotisable subjects, as opposed to less hypnotisable subjects, in the right DLPFC and in a right inferior parietal lobule focus. The DLPFC and the parietal cortex are part of the frontoparietal control network, which may integrate information from the external environment with stored internal representations [42] to guide decisions and performance adjustments [70]; this is relevant for cue modulatory effects on pain. Increased anticipatory activity in the frontoparietal network (albeit in the left hemisphere in that study) was shown to predict placebo analgesia [73]. The DLPFC (especially in the right hemisphere) may exert active control on pain perception by initiating top-down pain inhibitory mechanisms [35,78]; its role in placebo-induced pain modulation has been repeatedly demonstrated [14,36,73–75] (see, for review, [4,41]). Moreover, activity in the right DLPFC is related to the perceived intensity of suggestion-induced pain under hypnosis [60]. Our results point to different DLPFC-related mechanisms underlying the placebo response, depending on the subjects’ level of HS. This hypothesis is strengthened by the results regarding the FC of the right DLPFC. We found that, among subjects with higher HS, the behavioural placebo response was associated with reduced DLPFC connectivity with 1) a large focus encompassing right anterior mid-cingulate cortex (aMCC)/perigenual ACC/medial PFC (BA 32 and 8-10) during the entire test run, and 2) with 2 foci in the left inferior frontal gyrus (IFG; BA 47) and the right cerebellum during the perception phase only. Subjects with lower HS showed the
opposite pattern. The aMCC/perigenual ACC region is involved in pain anticipation [56,57], mediates the emotion and fear-avoidance aspects of pain, and can trigger descending antinociceptive circuits [71]. The medial PFC shows increased activity during self-related tasks, and more generally during evaluative processing [34]. The left IFG plays a role in evaluative judgment [68,80] and has been shown to increase activity during pain anticipation [54] and noxious stimulation [2,36,50]. The cerebellum is active during noxious heat pain [12], possibly in relation to motor aspects [52]. On the basis of these findings, we suggest that, only in the more hypnotisable subjects, the placebo response may involve a decoupling, or dissociation, of (increased) DLPFC activity from that of frontal/limbic midline structures and the left lateral frontal cortex. In contrast, among the less hypnotisable subjects, the placebo response is associated with increased connectivity of the DLPFC with these areas, possibly reflecting increased top-down modulation [5,14,15]. The present results are in line with a study demonstrating reduced FC among cognitive control-related brain regions in highly hypnotisable subjects in the hypnotic state [13], and with studies proposing a higher neurocognitive flexibility for these individuals [24–26]. 4.3. The placebo response is associated with a deactivation in several brain regions during pain perception, but only in subjects with higher HS Only in subjects with higher HS, the behavioural placebo response was associated with a significant decrease in activity in the thalamus/basal ganglia, left precuneus, and bilateral temporal
Ò
1515
A. Huber et al. / PAIN 154 (2013) 1509–1518
Fig. 5. Regions showing strong functional connectivity (FC) with the seed region located in the right dorsolateral prefrontal cortex (indicated by the green circle) throughout the test run, irrespective of hypnotic susceptibility (HS) and placebo response. Clusters containing at least 50 voxels satisfying a voxel-wise significance threshold of P < 1.9 10 7 are shown. Red colours indicate positive FC, blue colours negative FC. Coordinates (mm) are in the Montreal Neurological Institute (MNI) space.
Table 2 Differences in functional connectivity of the right DLPFC during the test run related to the HS-by-placebo interaction. Regions
BA
A. Region whose FC with the right DLPFC during the entire test run correlates negatively with the behavioural placebo response, but only in subjects with higher HS (subjects with lower HS show the opposite pattern): L medial/superior frontal gyrus, L anterior cingulate, L cingulate gyrus B. Regions whose FC with the right DLPFC during pain perception (‘‘Red’’ and ‘‘Green’’ trials taken together) correlates negatively with the behavioural placebo response, but only in subjects with higher HS (subjects with lower HS show the opposite pattern): R cerebellum L inferior frontal gyrus
9, 32, 10, 8
47
Cluster level
Talairach coordinates (peak)
MNI coordinates (peak)
P (corr)
k
x
