Neural correlates of conversion disorder: overview and meta-analysis of neuroimaging studies on motor conversion disorder

The largest area of the meta-analysis comparing patients to the control population is derived from two studies [27, 43]. This cluster shows increased activity in the amygdala of patients with MCD in comparison to healthy controls. The amygdala is known to be involved in autonomic responses, including freezing behaviour, attention, vigilance and arousal [44]. Changes in this function might be an important factor for the occurrence of MCD, as patients with MCD show increased responses to startling responses [43, 45] as well as complications during the habituation to positive and negative emotional stimuli [27]. Additionally, increased functional activity in patients with MCD in comparison to healthy controls in the amygdala [27] might correlate with activity in the supplementary motor area (SMA). Voon et al. [43] suggest that the increased connectivity and activity in the SMA-amygdala motor complex facilitates the expression of previously learned conversion motor representations. This aberrant activation of prefrontal areas is also supported by our study.

Prefrontal hyper-activity are involved in clusters 6, 8, 11, 12, and 13, based on the frontal cortex (BA 9) [46, 47] anterior (BA 10) [47] as well as the dorsolateral prefrontal cortex (BA 46) [47]. Elzinga and colleagues [46] discuss the increased activation in the left anterior prefrontal cortex, left dorsolateral prefrontal cortex and left parietal lobe in correspondence to high loads and performance of the working memory in MCD patients. Aybek and colleagues [47] discuss that the increased activation in the dlPFC are based on active memory suppression during the recall of unwanted memories. Increased activation in patients with MCD was also reported during working memory studies in frontal cortex, anterior as well as dorsolateral prefrontal cortex [46]. The dorsolateral prefrontal cortex is also often dysfunctional in patients with other neuropsychiatric disorders that affect volition [20, 48, 49]. In MCD the top-down regulation of motor intention by prefrontal areas might have a crucial influence on its occurrence [43] as the decreased prefrontal activation might be correlated with impaired control of motor execution [50].

Voon et al. [43] propose the increased activation of the insula in the context of potential motor-limbic network, whereby the insula is involved in the subjective representation of internal body and feeling states during motor-selection. The increased activity of the insula found in our study might correlate with this limbic function as well as hyper-activity in the retrosplenial area, i.e. ventral posterior cingulate cortex represented in clusters 3, 9, and 10 in the analysis. This portion of the cingulate cortex is discussed to be important in the evaluation of emotional objects and memories of the past for self-relevance [43, 51].

In addition to these areas cluster 7 is centered at the red nucleus based on the study from Aybek and colleagus [52]. In their paper they do not confer the red nucleus as the most important area, but the periaqueductal grey (PAG), and hypothesize PAG to be a key region in the “freeze response” [52]. Especially the interaction between PAG and the amygdala seems to be important for autonomic fear responses and might via hyper activation result in a threat induced “freeze response” [52, 53].

The superior temporal area is another area represented in the found clusters [43]. This area incorporates the Broca area known for its contribution to language production and possibly understanding [54]. The difference between the superior temporal lobe activation in patients and controls might be based on differences in the processing of study instructions based on increased internal verbalizations of patients [55]. Additionally the temporal lobe is discussed to be a critical site in the network dealing with emotional trauma [56], whereby resolving or repressing emotional traumas seems to be partially mediated by temporal structures.

Our meta-analysis shows reduced activity in the thalamus [43]. Vuilleumier et al. [22] suggest that striatothalamocortical circuits controlling voluntary motor and sensorimotor conversion are crucial for functional disorders like conversion. Hypofunction of thalamus during conversion disorder resolved after recovery [22]. This effect might be based on the function of the thalamus as the main hub system to cortical areas from sensory and motor signals; thus it plays a crucial role in generating intentional movement and learning adaptive motor action [22, 57, 58]. Reduced activity of the primary motor cortex [43] supported by our results, might correlate with the mentioned hypo-activity of the thalamus. Additionally, it might also represent reduced motor activity during motor conversion disorder.

The defect in motor action is additionally enhanced by problems with the anterior cingulate cortex (ACC; BA 32) found in cluster 5 when comparing increased activation of affected versus unaffected sides [50]. The ACC has one of its various functions in motor preparation, specifically selection of action, and conflict monitoring [59]. While Czarnecki et al. [60] report decreased activation in the ACC when comparing patients versus healthy controls, other studies report increased activation in MCD patients [43, 50, 52]. Similar to the hypothesis proposed by Voon et al. [43] in relation to increased activation of the amygdala, it might be that cingulate hyper-activation is related to emotional responses to motor action planning that inhibit motor execution especially when movement in the affected side is occurring. A study by van Beilen et al. [50] suggests that the over-activation in the cingulate cortex, especially when occurring in posterior parts, is related to alterations in functioning of the internal selection of movement that were previously described by Picard et al. [61] and might be especially pronounced when comparing movements of the affected versus unaffected side.

Increased activation of the primary somatosensory cortex [16] of the affected side might be related to increased cognitions about motor planning and sensory input. Still, especially stimulation of the affected side was associated with a decrease of the primary somatosensory cortex in sensory conversion disorder [62, 63]. Similarly to differences between patients and healthy controls, frontal and prefrontal areas are repeatedly identified within our meta-analysis when comparing increased activation in affected versus unaffected sides. The superior frontal gyrus [16, 64] and the anterior prefrontal cortex [16, 64] are significantly increased in the affected side. Increased prefrontal and frontal areas show increased activity in CP patients trying to move the affected body part [16, 17]. The increase in activity due to motor preparation of the affected side seems also associated with increased self-monitoring [16, 64‐66]. Studies reporting this difference are mainly based on studies on imagination of motor initiation.

The anterior insula was significantly increased in activity when the affected side was tested [16, 64]. The region, which is continuous with the primary gustatory cortex, is involved in the experience of emotions, particularly disgust [67‐69]. Additionally, it is an important integrator of multimodal stimuli responsible for interfacing internal motivational states and external information. [70‐72]. Both functions seem to be increased in movement preparation in the affected compared to the unaffected side.

The dorsal temporal lobe [16] and the temporal cortex [16] are further areas showing increased activation. Similar to differences between patients and healthy controls, the temporal lobe may be involved in the network for dealing with emotional trauma, especially when resolving or repressing them [56]. The temporal region has furthermore been identified to be important for cognitive processes including implied and executed movement [73].

Decreases in the activity of medial prefrontal cortex for the affected side have been discussed to be based on impaired willed action [50, 74]. This effect is contrary to other findings, where prefrontal activation is increased [16, 64]. This difference might be task dependent, so that some tasks like motor imagination of the affected side result in increased activation of frontal areas, while others like movement execution show decreased activation of the affected side.

The supramarginal gyrus [50] exhibited decreased activation in the affected side compared to the unaffected side. The supramarginal gyrus has been revealed to be functionally coupled in conversion disorder with the dorsolateral prefrontal cortex [21]. This connection of the prefrontal area with the sensorimotor system are involved in generating and planning motor action [75]. The decreased activation might result in abnormal movement initiation processes [50].

When all experiments are analysed and therefore differences between patients and healthy controls as well as differences between affected and unaffected sides are calculated within the same ALE irrespective of activation or deactivation, the following areas of interest were calculated, namely the dorsolateral and medial prefrontal cortex, the superior frontal gyrus, the insula as well as the amygdala and the dorsal anterior cingulate cortex. All areas have been identified in the sub analyses of the data. Even though no differentiation between activation or deactivation can be interpreted from the overall discussion, these areas seem to be repeatedly identified as dysfunctional in MCD and might be the core network for MCD.

Our results show that emotional, motor planning, and inhibitory processes are involved in MCD. Instead of single miss-functioning of a specific neuroanatomical area, a complete network of areas seems to influence the presentation of MCD symptoms. Patients with MCD seem to primarily differentiate from healthy controls in the frontal and prefrontal cortices, ACC, and amygdala relevant for motor-planning and -selection, intentional behaviour, volition, and autonomic responses. This effect, as well as the results from all included experiments, is similar to the emotional unawareness theory of Perez and colleagues [7, 33], which states that the large-scale brain network mediates emotional and cognitive mechanisms and is modulated by experience-dependent neuroplasticity. Our meta-analysis gives strong indications and supports the differences proposed in ACC, prefrontal areas, the dlPFC, and the amygdala. Our meta-analysis does not show strong evidence for changes in the posterior parietal cortex as proposed by Perez et al. [7]. In the more recent discussion of functional unawareness by Perez et al. [33] all described areas are supported by our meta-analysis. Still, strong evidence is only supported for frontal and prefrontal areas, Insula, ACC, and amygdala. In order to be able to support the suggested functional-unawareness neural circuit framework more studies have to be conducted on MCD.

In summary, our results support the perspective that specific functions, which are discussed by Perez and colleagues [7, 33] are disrupted. We found substantive backing for the disturbance of intentional capacities with our analysis based on the repeated malfunctioning of frontal and prefrontal areas [20, 21]. Changes in affective functions are highly plausible based on the increased activity of the amygdala and ACC in MCD [25‐27]. Support of dysfunctional inhibitory abilities is provided by the increased activity of the affected side [16‐19]. Attentional defects are partially substantiated by decreased activity of the thalamus in patients with MCD, the reduced activity of the supramarginal gyrus and increased activity of the ACC in the affected side [18, 22‐24]. Alterations of action authorship in MCD is partially supported by our meta-analysis based on the changes in tempoparietal junction, somatosensory cortex, ACC, parital cortex and the temporal lobe [18, 22‐24]. When considering the overall analysis, most support is provided for changes in intentional and affective functions in patients with MCD. While conversion symptoms might correlate with failures of normal neurocognitive functioning, personal experience of patients of conversion symptoms perceived as disruptive, a loss of needed information, discontinuity of experience, or recurrent, jarring, involuntary intrusions into executive functioning and sense of self should not be lost sight of [14, 76].

There is growing evidence for a relation between dissociation and trauma: on the one hand dissociative symptoms often occur in patients with PTSD [15], on the other hand depersonalization and derealisation are quintessential responses to acute trauma [14]. For the dissociative subtype of PTSD, increased activation of frontal structures is consistent with hyper inhibition of those same limbic regions in states of pathological emotional over-modulation [77]. Studies support that a PTSD dissociative subtype should be included in DSM-5 [78]. It is not surprising that a meta-analysis of neuroimaging studies of PTSD patients [15] showed an overlap in neuronal activation with the current results of MCD patients. However, some of the abnormal activations in PTSD appear to be stress related, while other activations seem to be disease related [15]. This could also apply for CD; acute dissociation is primarily related to traumatic and/or overwhelming experiences, but dissociative symptoms in life-long presentations such as Dissociative Identity Disorder may also occur in circumstances that are unrelated to trauma or overwhelming circumstances [14]. Additionally, similarities to other functional disorders might exist [79, 80]. Differences between resulting areas in the two sub-analyses for patients versus healthy controls and affected versus unaffected side might be based on the fact that we calculate inter- and intra- individual differences. Still both comparisons shed light on the development and perpetuation of the disorder.

This meta-analysis represents a first approach to combine the imaging results of MCD from various studies and is based on studies using differing imaging modalities and paradigms. Even though this approach might summarize various phenomena, the study of MCD is in need of a meta-analysis to more thoroughly examine the findings. The whole sample as well as the sub-analyses are sufficiently large enough to provide a first meta-analytical approach. Due to a low prevalence of MCD, the high costs of imaging, and the long data collection periods required, studies on MCD are rare. Ideally, a subcategory ALE analysis for each imaging modality and paradigm would be conducted. Due to the sparse imaging studies conducted on MCD, this was not feasible for the current study. Studies using hypnotization as a control are more prevalent, but might be explaining unintended phenomena within this meta-analysis and were therefore excluded. However, some of the mechanisms in hypnosis [81] and active feigning [19, 20] might share some neural mechanisms with MCD. Additionally, it is not clear whether the data presented in the various studies are based on recently developed conversions or on chronic conversion patients. Functional, as well as structural anatomy might change in the course of chronification. In the presented cases, not all patient characteristics are listed within the sample description, and medication as well as comorbidities might have skewed the data accordingly. Thus, it would be highly advantageous to understand functional and structural influences of the different disorders, as well as other factors influencing the variance, such as substance abuse, imaging parameters, software for analysis, experimental design etc. Future work should control for these differences, but at this time, the presented data can help present a starting point to the understanding of the neural correlates of MCD and CD in general. Still, our results are generalizable to limited extend on CD other than MCD. In order to better understand CD and specific forms of CD, more neurobiological research has to be conducted on disorders listed in Table 1, so that comparative ALEs can be calculated.