The central role of psychological processes: emotion, cognition, and placebo effects in healthy volunteers
Villemure & Bushnell showed that pleasant odors, independent of attentional focus, induced positive mood changes and decreased pain unpleasantness, and pain-related activity in several key pain processing areas (Fig. 1) (MCC, thalamus, SI/SII). The effects of attentional state were less robust, with only the activity in anterior (a)IC showing possible attentional modulation. Moreover, they identified separate neuromodulatory circuits underlying emotional [lateral (l)PFC and lateral orbitofrontal cortex (lOFC)] and attentional (superior posterior parietal cortex) modulation of pain.9
Roy et al. conducted an fMRI study that allowed them to identify the distinct spinal (i.e., corticospinal, descending from brain to spinal cord) and supraspinal (i.e., cortico(sub)cortical, within brain) mechanisms underlying the modulatory effect of emotions on pain.10 The RIII reflex was used as an index of spinal nociception. Emotional modulation in subcortical (thalamus, amygdala, brainstem nuclei) as well as cortical regions [subgenual (s)ACC and medial (m)PFC] was found to be related to spinal processes, whereas modulation in the aIC was associated with supraspinal mechanisms, which is consistent with the proposed role of the insula in integrating interoceptive information with emotional state, leading to higher-order subjective feelings.11,12 Moreover, connectivity analyzes identified potential cerebral sources of emotional modulation in these regions, including orbitofrontal cortex (OFC), dorsolateral (dl)PFC, sACC, parahippocampal gyrus (PHG), and brainstem (rostral medulla). It should be noted that the PHG had been implicated in anxiety-driven pain modulation in healthy volunteers before,13 as well as in patients with somatoform pain disorder.14
The profound effect of placebo on pain perception suggests that cognitive/affective processes, including expectation, anxiety (reduction), and Pavlovian conditioning, modulate pain perception. It has been known for a long time that endogenous opioid mechanisms play a central role in placebo analgesia, as it is at least partially blocked by the opioid antagonist naloxone.15 A number of recent studies have elucidated the brain mechanisms underlying placebo analgesia. One of the first of these studies showed that placebo and opioid analgesia activate common pain modulatory brain regions, including the pregenual (p)ACC and the OFC; the pACC activation significantly covaried with the periaqueductal gray (PAG), suggesting involvement of the descending pain modulatory system in opioid and placebo analgesia.16 This has been confirmed by more recent findings,17 showing that this placebo-induced pACC – PAG coupling is disrupted by naloxone.18 However, prefrontal top-down (i.e., modulatory) influences [ventrolateral (vl)PFC, dlPFC] on the pACC are specifically associated with placebo.19,20 Positron emission tomographic studies using the radioligand [11C]carfentanil have directly demonstrated opioid release during placebo analgesia in brain regions including dlPFC, pACC, OFC, aIC, amygdala, nucleus accumbens and PAG21,22 (Fig. 1). Similarly, radioligand PET-studies have shown the involvement of the dopaminergic reward system (nucleus accumbens) in placebo analgesia.22
Taken together, these results indicate that, even in healthy volunteers, psychosocial context and normal psychological processes exert a profound influence on pain processing and – perception through complex reciprocal interactions between emotional and cognitive pain modulatory brain networks on the one hand and (supra)spinal pain processing (or, more broadly, ‘homeostatic-afferent’) networks on the other. This brings us closer to understanding functional specialization of different networks within the previously identified ‘pain neuromatrix’, which may be important for identifying target networks and neurotransmitter systems for the development of new pain treatments. These studies may also facilitate the development of more effective psychological treatments for pain and help identify the most active components of psychotherapeutical treatment programs. For more extensive excellent recent reviews on this topic, the reader is referred to.23–25
Depressive symptoms and pain processing
Schweinhardt et al. used fMRI to investigate the neurophysiological interactions between depressive symptoms and disease-relevant pain in rheumatoid arthritis (RA) patients. They found that depression scores correlated significantly with a clinical pain measure (the tender-to-swollen joint ratio, T/S) as well as mPFC activation during provoked joint pain; the association between depression scores and T/S was partially mediated by the mPFC activation. Furthermore, the mPFC activation covaried significantly with activity in emotional/cognitive pain modulatory areas (amygdala/hippocampus, dlPFC) (Fig. 1) and in areas that process self-relevant information [dorsomedial (dm)PFC, posterior cingulate cortex (PCC), precuneus]. These results suggest that the mPFC may play an important role in mediating the relationship between depressive symptoms and clinical pain severity in RA, possibly by engaging pain modulatory brain areas important for affective and self-relevant processing.26
Strigo et al. studied brain responses during anticipation and processing of heat pain in major depressive disorder (MDD) patients compared with healthy controls. Firstly, MDD patients showed higher activation mainly in pain processing regions including the right aIC and anterior (a)MCC but also the amygdala during anticipation of pain (Fig. 1). Secondly, the patient group was characterized by abnormal activation in pain modulatory regions (increased in the amygdala, decreased in PAG, pACC, and dlPFC) during the subsequent experience of pain (Fig. 1).27
In summary, these studies show that comorbid depressive symptoms (even at the subsyndromal level), profoundly influence pain perception through interfering with pain processing and modulatory brain mechanisms. This implicates that psychiatric comorbidity should be either excluded or registered and subsequently controlled for in (visceral) pain neuroimaging studies, especially when comparing healthy controls with FPS patients, as the latter are characterized by high prevalence of comorbid mood and anxiety disorders.
Imaging chronic pain in clinical populations rather than experimental acute pain in health
The vast majority of current somatic pain neuroimaging studies use acute superficial pain stimuli (e.g., brief thermal or laser stimulation of the skin) in healthy volunteers, as these are convenient for use in block- or event-related fMRI designs. However, intuition as well as increasing evidence suggests that the pain invoked in these experimental models may be quantitatively as well as qualitatively different from the chronic, mostly deep pain (e.g., muscular, neuropathic) experienced by patients and even from tonic deep pain in healthy volunteers. These differences may be attributable primarily to the affective and cognitive dimensions of the subjective pain experience.28–30 A growing number of somatic pain studies take these factors into account by measuring dynamic changes in brain activation during experimentally induced tonic pain in healthy volunteers or spontaneously fluctuating levels of pain in patient populations. Some of these studies use new fMRI techniques such as arterial spin labeling (ASL) that are better suited for longer stimulus durations compared to the classical blood oxygen level dependent (BOLD) signal.
Baliki et al. for example, showed in patients with chronic low back pain that continuous ratings of fluctuations of spontaneous pain could be separated into two components: high sustained pain and increasing pain. The former was strongly associated with activation of the pain modulatory mPFC/pACC (as well as atrophy of the dlPFC, suggesting impaired top-down control) (Fig. 1). The latter, on the contrary, transiently activated pain processing regions such as SI/SII, aIC, and MCC (Fig. 1), which were also activated during acute heat pain in the same patient sample as well as healthy controls. These findings demonstrate that results from acute superficial pain studies in healthy volunteers may not be readily translatable to chronic pain in clinical populations. While the comparison of superficial vs deep is not germane to visceral pain, the difference between acute and chronic is probably as important in visceral as in somatic pain. Therefore, brief balloon distensions of the GI tract may not be an appropriate model for visceral pain as experienced by FGID patients.
Structural brain imaging in chronic pain populations
Recent studies suggest that brain structure changes on a temporal scale that is faster than previously thought (weeks to months) and that chronic pain is associated not only with functional, but also structural abnormalities in the brain (i.e., altered gray and/or white matter volume/microstructure).
Differences in gray matter volume/density can be studied using high-resolution structural MRI imaging and an analysis technique called voxel-based morphometry (VBM).31 For example, regional decreases in gray matter density in both pain processing and modulatory regions (vmPFC/OFC, lPFC, ACC, insula and PHG) have recently been reported in patients with somatoform pain disorder, controlling for comorbid depression.32 Patients with FM also have reduced total gray matter, which is correlated with age and disease duration.33 Similar findings have been reported in chronic low back pain, different types of headache and phantom pain.34 While results differ among these different pain syndromes, differences in cingulate cortex, OFC, insula and dorsal pons were consistently observed.34 Unfortunately, these studies often do not control for comorbid psychiatric disorders or other patient characteristics that may influence brain structure. This point is illustrated by a recent study in a large FM sample, which included patients with and without comorbid affective disorders. Region of interest analysis revealed a significant difference in aIC volume, but this was entirely driven by the patients with affective comorbidity. In contrast, FM patients without affective comorbidity did not differ from healthy controls.35
Recent studies used another MRI-based method called diffusion tensor imaging (DTI), which allows in vivo mapping of brain tissue microstructure, more specifically integrity and anatomy of white matter tracts (i.e., anatomical connections) in the human brain.31 A first study in healthy controls visualized top-down (i.e., descending) pain modulatory pathways, including connections between amygdala, (hypo)thalamus and prefrontal subregions on the one hand and key pain modulatory areas in the brainstem, including the PAG, on the other 36 (Fig. 1). A second study, which combined VBM and DTI, showed significant differences in gray matter volume in pain processing [thalamus, SI, mid (m)IC] and pain modulatory (amygdala and hippocampus, pACC, dmPFC) in FM (Fig. 1). Differences in white matter tract integrity were found in the thalamocortical tract, pACC and dmPFC. Correlations with different patient characteristics (including pain, fatigue and anxiety levels) were found in dmPFC and pACC; no correlations were found with depression scores.37
In summary, there is increasing evidence that different FPS are not only characterized by functional, but also structural abnormalities in pain processing as well as modulatory regions and pathways, some but not all of these abnormalities overlap across studies.31,34 However, longitudinal studies are required to ascertain if these changes are the cause or the consequence of chronic pain. Moreover, the mechanisms underlying these changes as well as the exact nature of the relationship between gray matter structure and the MRI signal remain highly unclear.31 Furthermore, like in functional neuroimaging studies, controlling for psychiatric comorbidity should be applied systematically, but is currently often lacking. This once more illustrates the need to control for psychological factors and psychiatric comorbidity in pain imaging studies, which forms one of the key messages of the present article. Finally, these relatively new techniques may still suffer from methodological difficulties (e.g., registration difficulties in VBM, sensitivity to subject motion and resolution limitations in DTI) and heterogeneity, limiting comparison across studies.31 However, despite these limitations, they permit studying new important aspects of the neurobiology of (chronic) pain which could not be achieved with pre-existing MR techniques. Moreover, the general finding that chronic pain is not only associated with functional, but also structural abnormalities in the brain may be important, as this may have an impact on developing various future pain treatments targeting, for example, neuroplasticity in certain brain networks. Moreover, if these findings are confirmed, structural imaging of relevant brain networks may turn out to be a useful tool in the follow-up and evaluation of chronic pain treatment.