Introduction

The skin is the largest physical organ of the human body, and represents a complex immunological environment which responds to pathogens as well as provides protection from physical insult and injury1. In order to remove pathogens from the surface of the skin, itch is an evolved mechanism which creates an urge to scratch1,2,3. Although itch is a normal response in most circumstances, it is pernicious in numerous clinical conditions, such as atopic dermatitis and psoriasis4,5, as well as maladaptive in major burn injuries, such as when itch arises around the burn, graft or donor site during the skin’s healing process6, or during phantom limb amputation7.

Previous work has outlined the immunological and neurophysiological mechanisms of itch3,8; however, this research has typically been conducted in model organisms8,9,10. In humans, individual studies have recently begun to examine the neural correlates of itch using techniques such as fMRI. Initial evidence has posited the recruitment of the insula, cingulate and prefrontal cortices as markers of acute itch, as well as ipsilateral motor cortices, potentially in the context of a planned scratch response11. Furthermore, research has also implicated reward circuitry, such as the striatum and midbrain, as important in the context of pleasurable sensations evoked from scratching an itch12. Whilst insightful, lacking to date is an understanding of the replicated brain regions and networks, reliably associated with itch as reported within individual neuroimaging studies. Identification of replicable mechanisms can afford the capacity to identify how neural markers of itch may be unique to certain clinical populations, such as those with atopic dermatitis and/or psoriasis1,9.

In order to assess these possibilities, we conducted a series of Activation-Likelihood Estimate (ALE) meta-analyses on published fMRI studies of itch. These analyses are empirically driven, and use brain activation coordinates across the published literature to identify similarities (i.e., clusters) of brain activation that are consistently associated with a particular process or experimental manipulation. A previous coordinates-based meta-analysis examined acute itch in healthy volunteers across 14 studies13, and identified consistent foci across regions of the Secondary Somatosensory Cortex, Bilateral Anterior Insula, Supplementary Motor Area, Inferior Parietal Cortex, and Inferior Frontal Gyrus. Further to this, a previous meta-analysis examined the similarities and differences in itch and pain, within studies that report comparisons for both itch and pain14. Here we build upon these initial studies, to offer an analysis of the overarching neural mechanisms associated with itch across the published literature, and report how this brain activation may differ for healthy participants and clinical groups. Furthermore, we highlight the promise of transdiagnostic psychological interventions such as cognitive behaviour therapy (CBT), mindfulness, and compassion-based approaches, that have shown much promise in reducing distress associated with chronic itch15,16,17,18. Indeed, to train individuals who experience chronic itch in strategies for regulating itch-related distress, may help to improve their quality of life and reduce potential concurrent presentations of depression, stress, anxiety, and disgust15,16,17,18, indirectly improving the ‘itch/scratch cycle’.

Results

Overview

From the twenty studies that passed the inclusion criteria19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38, we included all voxels that the authors explicitly interpreted as reflecting significant itch-related activity, such as a whole-brain contrast of itch and a neutral condition. If studies reported original voxels in Talairach space, these were transformed into MNI space using the tal2mni algorithm implemented within the ALE software. Our first ALE-analysis identified shared neural activation and deactivation associated with itch, pooled across healthy participants and clinical groups (Fig. 1, Tables 1 and 2). Our second ALE analysis explored how neural markers of itch may differ across healthy participants and clinical groups (Fig. 2, Tables 3 and 4). We also report details of each study’s stimuli and experimental design, as well as the populations sampled (Table 5). As shown below, we compared the coordinates of our ALE-activation with published data, reviews and meta-analyses, to provide convergent evidence for the labels applied automatically from the ALE software.

Figure 1
figure 1

Shared Brain Regions for Itch across Published fMRI Studies. This figure demonstrates significant shared activation and deactivation associated with itch. Figure legend indicates ALE activation map (orange) and ALE deactivation map (green). Colour bar indicates the Activation-Likelihood Estimate (ALE) test-statistic, which is the degree of convergence of activation across multiple studies within a given voxel. (a) These clusters show significant activation across Bilateral Thalamus and Basal Ganglia circuits, including Putamen. (b) These clusters show significant deactivations in Medial Frontal Gyrus and Paracentral Lobule. (c) This cluster shows significant activation across Basal Ganglia/Thalamus extending through the midbrain. (d) These clusters show activation across Bilateral Insula and Inferior Frontal Gyri, as well as Left Thalamus. These ALE maps were corrected at a cluster-level threshold of FWE, p < .05, with a voxel-level forming threshold at p < .001, uncorrected.

Full size image
Table 1 Significant brain activations associated with itch across 20 fMRI studies.
Full size table
Table 2 Significant brain deactivations associated with itch across 6 fMRI studies.
Full size table
Figure 2
figure 2

Brain Regions for Itch as a Function of Participant Groups. This figure demonstrates significant brain activation associated with itch across healthy participants and clinical groups. Figure legend indicates ALE map for healthy participants (blue) and ALE map for clinical groups (gold). Colour bar indicates the Activation-Likelihood Estimate (ALE) test-statistic, which is the degree of convergence of activation across multiple studies within a given voxel. Across both axial (a) and coronal (b) images, significant activation across Thalamus and Basal Ganglia circuits were observed for healthy participants, as opposed to activation of the Paracentral Lobule for clinical groups (a, b). These ALE maps were corrected at a cluster-level threshold of FWE, p < .05, with a voxel-level forming threshold at p < .001, uncorrected.

Full size image
Table 3 Significant ALE activation associated with itch across healthy participants (17 total studies).
Full size table
Table 4 Significant ALE activation associated with itch across clinical patient groups (6 total studies).
Full size table
Table 5 Paradigms, stimuli and sample characteristics across studies included in our meta-analysis.
Full size table

Activation and deactivation map

Our first ALE-analysis explored significant neural activation and deactivation associated with itch, pooled across healthy participants and clinical groups. Overall, this analysis identified 5 clusters for shared neural activation, and 2 clusters for shared neural deactivation. For the activation map, 40 experimental contrasts were included, that comprised 1008 foci from a total of n = 728 subjects. For the decrease map, 6 experimental contrasts were included that comprised 56 foci from a total of 131 subjects. For shared neural activation (as shown in Table 1 and Fig. 1), clusters identified from this ALE-analysis include nuclei of the Left Thalamus (cluster 1); Left Insula, Claustrum, Left Inferior Frontal Gyrus, and Left Insula (cluster 2); Right Insula, Right Putamen, Right Claustrum, and Right Precentral Gyrus (cluster 3); Right Thalamus (cluster 4) and Left Claustrum (cluster 5). For shared neural deactivation (as shown in Fig. 1 and Table 2), clusters identified from the ALE-analysis include Paracentral Lobule (cluster 1) and Medial Frontal Gyrus (cluster 2).

Increase map: cluster 1

The first cluster identified in the ‘activation’ map had a volume of 7840 mm3, spanning from (− 24, − 32, − 18) to (0, 0, 14) centered at (− 10, − 14.9, 2) with 6 peaks with a max value of 0.0369 ALE, 3.5330288E-9 P, 5.79 Z at (− 6, − 18, 10). Activation from these coordinates have been attributed to the Medial Dorsal Nucleus39 and Ventral Lateral Nucleus40 of the Left Thalamus41, as well as the Red Nucleus of the Left Brainstem/Midbrain42,43, as well as the Left Lentiform Nucleus44. Ten studies contributed foci to this cluster25,26,27,28,29,30,31,32,33,34. Note that one study34 included 2 foci for an itch rating, and 2 foci for a pain rating.

Increase map: cluster 2

The second cluster identified from the ‘activation’ map had a volume of 5472 mm3 from (− 52, − 4, − 16) to (− 26, 26, 8) centered at (− 37.9, 13.8, − 4.8) with 6 peaks with a max value of 0.035 ALE, 1.4005524E−8 P, 5.55 Z at (− 36, 18, − 8). Activation from these coordinates have been attributed to the Left Claustrum45,46, Left Inferior Frontal Gyrus47,48 and Left Insula49,50. Twelve studies contributed foci to this cluster25,26,27,28,29,30,32,34,35,36,37,38.

Increase map: cluster 3

The third cluster identified had a 5384 mm3 from (22, − 6, − 14) to (56, 26, 14) centered at (37, 9.8, 1.6) with 6 peaks with a max value of 0.0303 ALE, 3.218045E−7 P, 4.98 Z at (36, 24, 2). Eleven studies contributed foci to this cluster (25–27, 30–35, 37, 39). Note that one study34 included 2 foci for an itch rating, and 2 foci for a pain rating. Activation from these coordinates have been attributed to the Right Claustrum45,46 and Right Insula51, as well as the Right Lentiform Nucleus and Putamen, as situated within the Basal Ganglia52,53, as well as the Right Precentral Gyrus54,55.

Increase map: cluster 4 and 5

The fourth cluster had a volume of 1240 mm3 from (8, − 12, − 4) to (18, 2, 12) centered at (12.9, − 3.9, 2.4) with 2 peaks with a max value of 0.0243 ALE, 1.4436547E−5 P, 4.18 Z at (12, − 4, 0). Seven studies contributed foci to this cluster25,27,28,29,30,31,34. Note that one study34 included 2 foci for an itch rating, and 2 foci for a pain rating. Activation from these coordinates have been attributed to the Right Thalamus56 as well as the Left Claustrum57. The fifth cluster had a volume of 1072 mm3 from (− 40, − 26, − 8) to (− 34, − 8, 10) centered at (− 37.2, − 16.5, − 1.9) with 3 peaks with a max value of 0.0272 ALE, 2.4281858E−6 P, 4.57 Z at (− 38, − 14, − 4). Four studies contributed foci to this cluster27,30,32,39.

Decrease map: cluster 1 and 2

For the decrease map, cluster 1 included 816 mm3 from (− 2, − 28, 44) to (14, − 16, 54) centered at (4.7, − 22.1, 49.1) with 2 peaks with a max value of 0.0126 ALE, 4.1608578E−6 P, 4.46 Z at (4, − 20, 50). Two studies contributed foci to this cluster29,49. Cluster 2 included 656 mm3 from (− 2, 56, 8) to (8, 64, 16) centered at (3.1, 59.9, 11.7) with 1 peaks with a max value of 0.0181 ALE, 1.1981463E−8 P, 5.58 Z at (4, 60, 12). One study contributed foci to this cluster29. Activation from these coordinates have been attributed to the Paracentral Lobule58,59,60 as well as the Medial Frontal Gyrus61.

Healthy participants and clinical groups map

This ALE map explored neural activation associated with itch for healthy participants and clinical groups (Fig. 2, Tables 3 and 4). For the healthy participants map, 34 experimental contrasts were included, that comprised 748 foci from a total of n = 714 subjects. For the clinical groups map, 12 experimental contrasts were included that comprised 316 foci from a total of 145 subjects. Given that the shared neural activation for healthy participants comprise approximately the same clusters as identified for the above ‘activation’ map (i.e., Fig. 1, Table 1), this was not compared to previously published coordinates, yet details for this analysis are provided on the OSF (see “Methods” for a link to the repository). In contrast, the shared neural activation with clinical groups included 1 cluster of 976 mm3 from (− 6, − 16, 46) to (4, − 4, 56) centered at (− 0.8, − 9.4, 51.3) with 1 peaks with a max value of 0.0218 ALE, 4.1508642E−8 P, 5.36 Z at (0, − 8, 50). Three studies contributed foci to this cluster27,32,38. Activation from these coordinates have been attributed to the Left Paracentral Lobule57,58,59 as part of the Medial Frontal Gyrus.

Table of stimuli and paradigms

Table 5 outlines the stimuli and paradigms used within each fMRI experiment, as well as demographic information for the populations that were studied (i.e., healthy participants or clinical groups).

Discussion

We used a coordinates-based approach to synthesize pre-existing fMRI studies of itch, in order to identify replicable brain regions consistently reported to be associated with itch. We found evidence for a core “itch network” that comprised the Basal Ganglia/Thalamus, Bilateral Insula, and Claustrum, as well as the Inferior Frontal Gyrus and Precentral Gyrus. These brain regions are consistent with studies that have identified how histamine-induced itch can modulate regions of the Basal Ganglia/Thalamus, such as the Caudate Nucleus and Putamen62,63, as well as research that has shown how cowhage-induced itch can activate the Insula, Claustrum and Basal Ganglia/Thalamus64. The Thalamus and Claustrum have been attributed a key role in attention and multisensory integration65,66, and the Insula has been implicated in tasks that require attentional control and response inhibition48,49, particulary in contexts of emotional salience67,68.

In addition to our activation map, we also examined neural deactivation associated with itch. Although these results showed fewer foci, we identified significant neural deactivation in both the Paracentral Lobule and Medial Frontal Gyrus. Of note, the Paracentral Lobule has previously been associated with cognitive/executive control59 and response inhibition60, whereas the Medial Frontal Gyrus has previously been associated with social cognition and mentalising61,69. Thus, that regions associated with cognitive control and task demands (i.e., Paracentral Lobule)70,71,72 are shown to deactivate during itch, might suggest a putative marker for how the experience of itch may dysregulate executive control. However, given only three studies contributed foci to this ALE map, this finding should be interpreted with caution. Given the multifunctional role of these two brain regions, we acknowledge that alternative explanations unrelated to cognitive control, response inhibition or mentalising are possible. Future research is needed to examine these processes in relation to itch with carefully designed empirical experiments.

Unique neural activation for health controls and clinical groups

A further focus of this paper was to explore how neural activation during itch may be unique across healthy participants and clinical groups (i.e., individuals with atopic dermatitis). When examining neural markers of itch for health controls, we identified five clusters of activation (Fig. 2), across similar brain regions identified in the activation map reported above (Fig. 1). When examining neural markers of itch for the clinical groups, however, we identified significant recruitment of the Left Paracentral Lobule (Fig. 2), a brain region previously associated with cognitive control and response inhibition58,59,60. To examine the influence of cognitive control during the experience of itch in future experimental studies will provide precise evidence to inform how the urge to itch may be interrupted via executive function73.

Limitations and future directions

Within this paper we pooled across various clinical groups who report chronic itch due to the small number of studies with clinical groups (Table 5). Although this approach provided further insights as to the neurobiology of itch in clinical groups, it is likely that neural/immune mechanisms for itch could vary across specific clinical conditions (e.g., in the case of atopic dermatitis and psoriasis)1. Furthermore, given few studies have examined itch in clinical groups with fMRI, we did not have the statistical power to compare how neural activation for itch may differ between healthy participants versus clinical groups, or between clinical groups versus healthy participants. Indeed, our ALE activation of itch incorporated foci from only three studies, albeit with two distinct clusters. While this finding should be interpreted with relative caution given the small sample-size, future studies can help inform our understanding of this initial evidence that suggests the Left Paracentral Lobule and the Medial Frontal Gyrus as associated with itch for those who report chronic itch. It is also worth noting that the extracted contrasts used in the analysis for this paper are quite diverse. As an example, while most included studies focus on physically induced itch, there was also a study that used visually induced itch. We also included a contrast that examined an effect of sex (i.e., neural markers of itch for males versus females), as well as a correlation between pleasurability ratings and neural markers of itch. Given these diverse (but perhaps more inclusive) inputs in our analysis, it is unsurprising that certain brain regions previously associated with itch found in prior meta-analyses (i.e., Primary Somatosensory Cortex)11,13 were not identified.

It is important for future research to continue to examine the neural correlates of itch within clinical groups, to provide further support for how to best understand and treat these conditions. Indeed, there is a substantial burden of disease associated with living with chronic itch, evident through the downstream impact of these conditions on mental health15,16.

Experimental extensions

The “itch network” identified in this research has potential implications for ways to treat and mitigate itch in clinical settings. The identification of this interconnected network implies that a treatment targeting any node in this network has the potential for ripple on effects so to speak, to influence the entire system. To draw on parallel research within the area of pain, the historical “Gate Control Theory of Pain”—now known as the “Neuromatrix Model of Pain”74,75—revolutionised how pain was conceptualized and treated. This model was the first to recognize the interconnected role of areas such as the prefrontal cortex and amygdala (and others) in processing pain, which essentially broadened the traditional points of biomedical intervention to recognize the capacity for psychological interventions targeting such cognitive and emotion centres to harness modulatory processes. Now, the neuroscience of pain relief is included as a treatment rational in educational interventions76, cognitive-behavioural therapy77, and mindfulness-based cognitive therapy78, and emerging research is showing how changes in brain function during such treatment are driven by change in psychological processes (e.g.,79,80).

Linking this line of research back to itch, convergent evidence from clinical psychology has shown how training in mindfulness and compassion-based approaches can reduce the presentation of depression and/or anxiety associated with chronic itch81,82,83. Indeed, several randomised control trials (RCTs) have previously demonstrated the efficacy of cognitive behavioural therapy (CBT) for improving mental distress associated with atopic dermatitis81,82,83,84, and mindfulness and compassion-based approaches have been shown to improve emotional regulation, stress reduction and improved cognitive control85,86,87,88.

Possibly, these treatments are efficacious as they are engendering functional shifts within the neural “itch network” identified in this current study; it would be fascinating to explore such a possibility with experimental fMRI, particularly given the need to develop novel therapies for itch12, through the recognition of psychological factors implicated in the exacerbation or treatment of chronic itch. The findings from such future research could be harnessed in a similar way as the research on the neuroscience of pain and its relief has been used. That is, to develop targeted treatments with a clear rationale to engender positive outcome expectancies and “buy in” that psychological, non-pharmacological treatments are effective for itch, what has long been considered a purely “biomedical problem”.

Conclusions

Conditions such as atopic dermatitis comprise a large burden of disease89,90,91, with downstream consequences on mental health89,91,92. In this article, we explored the shared, consistent brain regions identified across the published literature for itch, and explored how these regions may differ for healthy participants and chronic itch populations. We provide evidence for a core “itch network” identified across the published literature to date, discuss the importance of executive control in inhibiting itch, and offer how training in mindfulness and/or compassion-based approaches may improve emotional regulation and executive control for chronic itch groups. We look forward to future research in compassion or mindfulness-based approaches to explore how these techniques might help individuals “put out the fire” of chronic itch.

Methods

The University of Queensland Low & Negligible Risk Ethics Sub-Committee approved the experimental protocol, and this project complies with the provisions contained in the National Statement of Ethical Conduct in Human Research and the regulations governing experimentation on humans.

Literature search and exclusion criteria

The PUBMED database was searched using the keywords ‘fMRI’ AND ‘itch’. The literature search was conducted in July 2020. Identification of studies occurred as follows:

  1. 1.

    Studies that mentioned ‘fMRI’ and ‘itch’ within the title, abstract and/or keywords were identified, to collect experimental work which either manipulated or measured the neural markers of itch. This search identified 216 published, peer-reviewed papers. After removal of 7 duplicate articles and removing 1 record for other reasons (this study did not report whole brain results for analysis and was written in Japanese), 208 studies remained for screening.

  2. 2.

    208 studies were screened, and 179 studies were excluded from further analysis as they were not relevant to the present study. This left 29 studies for retrieval.

  3. 3.

    We sought to retrieve 29 studies that were relevant to the present paper. From this number, we did not retrieve 5 studies as these employed other imaging techniques that did not report relevant whole-brain markers (i.e., Diffusion Tensor Imaging, Pharmacological Methods). This left 24 studies to assess for eligibility.

  4. 4.

    We assessed 24 studies for eligibility, and excluded 4 studies that did not report relevant whole-brain analyses.

  5. 5.

    This left a total of 20 studies that were included in the present review.

Activation Likelihood Estimation (ALE) overview

In order to identify regions of consistent brain activation across studies, we performed a series of ALE analyses using the GingerALE software93 (Version 3.0.2). Practically, ALE analyses report 3D coordinates of published brain data within a text file, which also lists subject size and the reference space the data were originally reported in (i.e., MNI or Talairach). An advantage of this current ALE version (3.0.2) over older versions of the algorithm is the use of an ALE method which incorporates random effects93.

An ALE analysis with GingerALE entails three steps. First, a 3D image for each group of foci are created using a mask, foci from the text file, and a Gaussian blur with a Full-Width Half-Maximum (FWHM) smoothing generated from the sample-size. These images are referred to as Modelled Activation (MA) maps which are created for each foci’s Gaussian profile. The ALE image is a union of each MA map. Second, GingerALE calculates the null distribution of the ALE test-statistic, which represents the degree of convergence of brain activation within a voxel across all included experiments94. Combining the probabilities of finding each value in a MA map creates a table of p-values. The ALE image is combined with a p-value table to create a 3D p-value image. Third, a significance threshold can then be set for the p-value ALE map (i.e., uncorrected, false-discovery rate, family-wise error, FWE) (Supplementary Information).

Following recommendations from simulated data95, we used a cluster-level threshold of FWE, p < 0.05, with a voxel-level forming threshold at p < 0.001, uncorrected, with 5000 permutations. A cluster-based threshold indicates significant clusters of voxels where a test-statistic is above a pre-specified threshold96. Statistical tests are used to control the estimated rate of false positives of each cluster, that is, the probability of violating family-wise error within the cluster96. For our data, thresholded ALE maps were overlayed on a Colin27 T1 structural scan (courtesy of Simon Eickhoff, http://www.brainmap.org/ale/) using Mango (http://ric.uthscsa.edu/mango/). Each ALE analysis provides a table of cluster information with corresponding anatomical labels (nearest grey matter within 5 mm) which we discuss in this manuscript.

Ethical standards

This research was approved by The University of Queensland Health and Behavioural Science Low & Negligible Risk Ethics subcommittee.

Informed consent

This is a meta-analysis with no primary data collection. Individual studies included in this meta-analysis received informed voluntary consent for participation.