Abstract
Objective: The authors hypothesized that central factors may underlie sensory deficits in patients with nondermatomal somatosensory deficits (NDSD) and that functional brain imaging would reveal altered responses in supraspinal nuclei.
Background: Patients with chronic pain frequently present with NDSD, ranging from hypoesthesia to complete anesthesia in the absence of substantial pathology and often in association with motor weakness and occasional paralysis. Patients with pain and such pseudoneurologic symptoms can be classified as having both a pain disorder and a conversion disorder (Diagnostic and Statistical Manual of Mental Disorders–IV classification).
Methods: The authors tested their hypothesis with functional MRI (fMRI) of brush and noxious stimulation-evoked brain responses in four patients with chronic pain and NDSD.
Results: The fMRI findings revealed altered somatosensory-evoked responses in specific forebrain areas. Unperceived stimuli failed to activate areas that were activated with perceived touch and pain: notably, the thalamus, posterior region of the anterior cingulate cortex (ACC), and Brodmann area 44/45. Furthermore, unperceived stimuli were associated with deactivations in primary and secondary somatosensory cortex (S1, S2), posterior parietal cortex, and prefrontal cortex. Finally, unperceived (but not perceived) stimuli activated the rostral ACC.
Conclusions: Diminished perception of innocuous and noxious stimuli is associated with altered activity in many parts of the somatosensory pathway or other supraspinal areas. The cortical findings indicate a neurobiological component for at least part of the symptoms in patients presenting with nondermatomal somatosensory deficits.
Nondermatomal somatosensory deficits (NDSD) to various cutaneous sensory modalities (touch, pinprick, cold) ranging from hypoesthesia to complete anesthesia are often considered functional or psychogenic when they occur in the absence of substantial pathology. They are often accompanied by unexplainable reduction or loss of vibration sense at the same side.1 In chronic pain populations the prevalence of NDSD varies between 25 and 40%1,2⇓ but may be higher in patients with compensable injuries.2 NDSD are often associated with variable motor abnormalities ranging from loss of dexterity to complete motor paralysis.1
Inability to perceive stimuli could result from abnormalities within the central pathways mediating touch or pain. The cortical targets of the touch and pain pathways include the primary and secondary somatosensory cortex (S1, S2), insula, and anterior cingulate cortex (ACC). However, additional cortical regions associated with attention, such as the posterior parietal cortex (PPC), prefrontal cortex (PFC), and the temporoparietal junction (TPJ), can also impact on or be influenced by somatosensory processing. Furthermore, attentional state can modulate sensory-evoked responses.3 There is also interdependence of somatosensory and motor systems, as sensory stimuli, including touch and pain, can have an effect on the function of multiple motor structures, including the primary motor cortex (M1), supplemental motor area (SMA), and basal ganglia.4,5⇓ There are also somatosensory inputs to circuits involved in the processing of emotional or other aspects of psychosocial behavior that may then feed back to somatosensory or motor circuits.6,7⇓
Recent neuroimaging studies have identified abnormal responses to sensory stimuli or motor challenge in patients with conversion disorders.8-10⇓⇓ These studies focused on motor abnormalities, even if some patients also presented with chronic pain.11 In the current study, we hypothesized that if there is a neurobiologic substrate associated with NDSD in patients with chronic pain, it can be detected with sensory-evoked functional MRI (fMRI). We predicted that unperceived somatosensory stimuli would fail to activate vital nodes of the somatosensory pathway.
Methods.
General patient characteristics.
Four patients with chronic pain entered the study 15 months to 9 years after the onset of pain and sensory deficits (table 1; supplementary appendix e1, containing patient histories, is available on the Neurology Web site at www.neurology.org). Detailed sensorimotor and musculoskeletal examination was performed on all patients repeatedly. The sensory examination during clinical visits involved the use of a soft brush, a pinwheel, and a cold roller for documentation of sensation to touch, pinprick, and cold. Vibration sense was tested with a 128-Hz tuning fork. Deep pain was tested with a pressure algometer (Pain Diagnostics, Great Neck, NY). All patients had an area of complete anesthesia to all cutaneous modalities (including loss of vibration sense and sensitivity to deep pain) in a limb that was painful. In most cases, the area of pain was much larger than the anesthetic area, with the exception of Patient 4 who had proximal pain and distal anesthesia. Two patients (1 and 4) had two limbs involved, but on contralateral sides of the body and opposite quadrants. In Patient 1, the right anesthetic arm was painless, whereas Patient 4 had proximal pain and distal anesthesia in both involved extremities. Patients 1, 2, and 3 displayed mild reduction in dexterity or mobility of the painful limb despite the presence of anesthesia, whereas 4 had complete anesthesia and motor paralysis in the affected limbs. The patients were thoroughly investigated from the psychological, neurologic, and musculoskeletal point of view and had been followed by the pain team for months before and after completion of the fMRI study. Head MRI in Patients 2, 3, and 4 failed to disclose any anatomic abnormalities. Lumbosacral spine gadolinium MRI disclosed limited left L5 root perineural fibrosis in Patient 1, which, however, could not explain the dense sensory loss of the extremity. Electromyographic and nerve conduction studies for all four patients were negative. Somatosensory evoked potentials (SSEP) from the upper and lower extremities in all four patients were normal with normal cortical latencies (N20 and N45 components) with the exception of some delay in left leg SSEP for Patient 4 (PF or P40 component only). This was due to technical factors because of substantial swelling of the leg and profound induration of the skin resulting from dependency and complete lack of mobility. Shoulder x-rays revealed mild degenerative changes in Patients 3 and 4.
Table 1. Patient characteristics
Detailed psychological assessments and behavioral observations suggested that psychosocial factors were involved in the onset, maintenance, exacerbation, or severity of pain and related problems. Psychological factors considered important included the following: 1) aspects of the developmental psychosocial history (e.g., parental perceptions), 2) dependence-independence conflicts, 3) psychosocial stressors leading up to onset of medical problems or post-traumatic reactions associated with high anxiety at time of trauma, 4) dissociative experiences or phenomena either before onset or subsequent to pain, and 5) discrepancies between aspects of behavior and reported pain severity. The latter was manifested especially by la belle indifference type of presentation, in which very high pain severity ratings were given despite very comfortable, happy, or jovial demeanor. Psychometrics, including the Minnesota Multiphasic Personality Inventory (MMPI-2) and Millon Clinical Multiaxial Inventory (MCMI-III), were suggestive of a conversion of psychological conflicts or stress into somatic symptoms. All subjects were involved in litigation or compensable issues. However, at no time did the suspicion of malingering arise, as the patients had been repeatedly seen by the pain team with consistent and reproducible findings over the period of several months or years. Administration of sodium amytal, a medium action barbiturate, transiently but substantially reduced pain and sensory abnormalities in Patients 1, 3, and 4, suggestive of lack of structural damage.1,12⇓ Sodium amytal infusion was preceded by infusion of normal saline, to which all patients failed to respond. All patients were blinded to the nature of the infusions and were given neutral instructions; i.e., they were told that any or both drugs could produce pain relief, have no effect on pain, or even make the pain worse.
On the day of imaging, all patients rated their pain intensity as 7 to 8 on a verbal analogue scale from 0 to 10. Also, on the day of imaging, stimulation with innocuous brush or intense mechanical stimuli (via a pressure algometer and von Frey probes with a contact area of ∼1 mm delivering forces from 75 to 280 g) within the patients’ anesthetic regions were not perceived and did not evoke dysesthetic feelings.
Functional MRI.
All patients provided informed consent before the functional imaging session. Patients underwent fMRI on a 1.5 T Echospeed MRI system (GE Medical Systems, Milwaukee, WI) fitted with a standard quadrature head coil to investigate activations evoked by innocuous brushing and noxious mechanical stimulation. Experimental runs consisted of alternating 32-second blocks of rest and left- and right-sided stimulation repeated six times. Stimuli consisted of either repeated noxious (intensity based on sensitivity on nonanesthetic limb) mechanical stimulation with a von Frey probe or innocuous brushing delivered at ∼2 Hz. Stimuli were applied to the upper (Patients 2, 3) or lower extremity (Patients 1, 4). Patients were instructed to close their eyes throughout the scan so that they could not be alerted to the presence or absence of stimulation.
A high-resolution three-dimensional (3D) anatomic scan of the whole head (124 sagittal slices; 256 × 256 matrix, 24 × 24 cm field of view, 1.5 × 1.17 × 1.17 mm voxels) was obtained using a T1-weighted 3D spoiled gradient echo (SPGR) sequence (flip angle = 45°, echo time [TE] = 5 msec, repetition time [TR] = 25 msec). Whole brain functional imaging used two-shot gradient echo imaging utilizing a spiral trajectory through k-space from 25 axial slices (T2*-weighted images; flip angle = 85°, TE = 40 msec, TR = 4000 msec, 128 × 128 matrix, 20 × 20 cm field of view, 1.56 × 1.56 × 4 mm voxels). A total of 152 functional volumes (i.e., frames) were acquired, of which the first three were discarded to allow for signal equilibration.
Brain Voyager 2000 software (version 4.4; Brain Innovation B.V., Maastricht, the Netherlands) was used for preprocessing and statistical analysis. Details of the imaging, preprocessing, and statistical methodology for thresholding have been described in our previous studies.13 Briefly, preprocessing included resampling the anatomic images to 1 × 1 × 1 mm using sinc interpolation,correcting functional data interslice differences based on the time of acquisition, 3D motion correction with sinc interpolation, and resampling images at 3 × 3 × 3 mm. Images were spatially normalized into a common stereotaxic space.14 Data from the patient with right-sided anesthesia were left-right flipped to correspond to the other patients with left-sided anesthesia. Global signal intensity differences were corrected for by proportionally scaling the data to a common mean. Linear trends were removed separately for each pixel using a least squares standard method. Data were highpass filtered to remove slow drifts in signal intensity with a period greater than twice the total duration of the three stimulus condition blocks. Spatial smoothing using a Gaussian kernel with 6 mm at full-width half maximum was also performed.
Separate group analyses were performed for noxious and innocuous data using the general linear model. Each condition (perceived, not perceived, rest) was modeled as a reference waveform (i.e., boxcar function convolved with a gamma variate hemodynamic response function) and was treated as a separate predictor for each subject. Two contrasts of interest were examined resulting in four t-statistic maps (i.e., noxious perceived vs rest, noxious unperceived vs rest, innocuous perceived vs rest, innocuous unperceived vs rest). All maps were thresholded at p < 0.0001 (uncorrected, two-tailed) and a minimum cluster size of 150 mm3. Given that the volume of each statistical map was 1,502,673 mm3, a threshold of p < 0.0001 would be expected to show activation in ∼150 of the 1 mm3 voxels due to type 1 errors. The cluster size criterion was therefore used as a conservative measure to minimize false positive activation (see reference 13).
Results.
The locations of key activations and deactivations in the group analysis are graphically depicted in figures 1 and 2⇓ (additional material in tabular form can be found on the Neurology Web site; go to www.neurology.org). A summary of the key findings can be found in table 2.
Figure 1. Regions of activation related to the brush stimuli applied to the control limb (A, brush perceived condition) and to the affected limb (B, brush unperceived condition). Arrows in brain images show the key areas of activation. Images are displayed so that the hemisphere contralateral to side of the body stimulated is indicated by the “c” and shown on the right. Significance is indicated by the z-score color bar.
Figure 2. Regions of activation related to the noxious mechanical stimuli applied to the control limb (A, pain perceived condition) and to the affected limb (B, pain unperceived condition). Arrows in brain images show the key areas of activation. Images are displayed so that the hemisphere contralateral to side of the body stimulated is indicated by the “c” and shown on the right. Significance is indicated by the z-score color bar.
Table 2. Summary of group results
All patients perceived the brush and noxious stimuli when applied to the control limb, and activated as expected the brain regions typically associated with the perception of somatosensory stimuli (i.e., perceived condition). Both the innocuous brushing and the painful stimuli activated the S1, S2, PPC, anterior insula, thalamus, and regions within the PFC, TPJ, and the M1 and SMA (see figures 1 and 2⇑). The topography of the S1 activations reflects the fact that stimulation was applied to the leg in two patients and hand in the other two patients. Additionally, the posterior region of the ACC was activated by the perceived painful stimuli, but not by the perceived brush stimuli. There was a reduction in signal (i.e., deactivation) within a small region of the ipsilateral S1 during painful stimulation of the control limb. No other deactivations were detected during brush or pain stimuli of the control limb.
In contrast, all subjects failed to report any sensation when the brush or noxious stimuli were applied to the affected limb. These unperceived stimuli evoked abnormal responses within many cortical regions as compared to those evoked by the perceived stimuli. These abnormalities included 1) absence of activation, 2) deactivations, and 3) previously unidentified activations (see table 2 and figures 1 and 2⇑). Several regions that had been activated by the perceived stimuli were not activated with the unperceived stimuli. These areas include the anterior insula and thalamus (unperceived brush condition) and the posterior ACC, BA 44/45, and thalamus (unperceived noxious condition). Clusters of deactivation associated with the unperceived brush stimuli were identified in S1, S2, and BA 9 and 45 (an example of S1 deactivation in an individual subject can be found on the Neurology Web site; go to www.neurology.org). During the unperceived noxious stimulation, deactivations were identified in S1, S2, PPC, M1, and BA10. Finally, the unperceived stimuli resulted in several activations that had not been detected in the perceived condition; namely, in the anterior ACC during unperceived brush and in the anterior and perigenual ACC during unperceived noxious stimulation.
It is possible that the pronounced activations detected in the group results were due to a large contribution from a subset of subjects (i.e., fixed effect error). To test this possibility we performed a conjunction analysis (methodology previously described by us13) to detect activations present in all subjects. The results of the conjunction analysis confirmed that the main pain-related cortical effects detected in the general linear model (GLM) analysis were consistent across all patients, although the activations were somewhat smaller in volume. All brush-related cortical effects also were confirmed in the conjunction analysis, with the exception of some of the IFG and SMA regions.
Discussion.
Our patients had chronic intractable pain associated with anesthesia in the symptomatic limb and variable motor deficits, and based on their medical and psychological assessment they fit the classification of conversion disorder in association with a pain disorder as per Diagnostic and Statistical Manual of Mental Disorders–IV definition. The fMRI findings indicate that in these patients diminished perception is associated with altered processing within the somatosensory system.
In normal subjects, innocuous somatosensory stimuli typically activate a cortical network that includes the S1, S2, PPC, and insula.15-17⇓⇓ Painful stimuli additionally can activate the ACC, PFC, and motor cortical areas.3,15⇓ Perception of somatosensory stimuli depends upon the integrity of this network. However, whether any one particular node of the network is critical for conscious perception remains a contentious topic. In our study, fMRI during stimulation of the symptomatic limb revealed prominent abnormalities in somatosensory areas; namely, lack of activations, novel activations, and stimulus-related deactivations in the S1, S2, and PPC cortex. This cortical dysfunction indicates a neurobiological component associated with at least part of the symptoms in such patients in accordance with our original hypothesis.
Our findings of abnormalities in key somatosensory and frontal/cingulate areas share some similarities with other neuroimaging studies that have demonstrated reduced or absent supraspinal responses in patients with hysterical motor paralysis9-11,18⇓⇓⇓ or hysterical sensorimotor abnormalities.8 Interestingly, suppression of higher order processing may also underlie hysterical deafness.19 Similarly, in a case of conversive anesthesia, there were normal evoked potentials relating to sensory and perceptual processing of both innocuous and noxious stimuli, but an abnormal P300 (cognitive) component from the anesthetic right hand. Interestingly, the P300 generated from the healthy left hand of the patient as well as from a healthy subject instructed to feign sensory loss (a malingerer) was normal.20 Furthermore, a PET study of patients with tactile extinction (with an intact S1) found an absence of or reduced activation of S1 contralateral to the extinguished stimuli.21
In the current study, we did not detect any lateral thalamic activation in either the perceived or unperceived condition. This negative finding may have been due to technical issues because we used a larger voxel size and spatial filtering across a group analysis, compared to our previous study that was able to detect discrete sensory-evoked lateral thalamic activation.22 Furthermore, the stimulus-related cortical effects indicate that some information must have been transmitted through the thalamus to the cortex. Thus, it might be the case that in the unperceived condition, a much smaller region of the thalamus was activated that could not be detected in our analysis. Furthermore, the thalamic activation that was detected in the current study was in the nonsomatotopically organized dorsomedial thalamus, possibly related to an attentional or cognitive response to the stimuli.
Our results of cortical deactivation during unperceived stimulation are clear and not likely due to technical artifactual issues. The source of a fMRI blood oxygen level dependent (BOLD) deactivation signal is unknown but is an active research topic.23-27⇓⇓⇓⇓ One possibility is that negative BOLD signals represent hemodynamic steal from an adjacent active area. This is unlikely in the current study because the deactivations were spatially distant from the activations. A more likely possibility is that the negative BOLD signal is due to reduced or suppressed neuronal activity that attenuates blood flow locally.
In addition to the deactivations, activations were found during unperceived stimulation that were not detected in the perceived condition. One notable novel activation during unperceived noxious stimulation was found in the rostral and perigenual ACC, in contrast to perceived noxious stimulation that activated the posterior ACC only. This is an interesting finding because the posterior ACC is consistently activated during acute and chronic pain states.3,28⇓ However, the more rostral regions of the ACC, including the perigenual ventral portions, are thought to be involved more generally in cognitive processes and emotion.29 These findings may be significant in our patients, as psychological factors were believed to be contributory to the onset, exacerbation, severity, or maintenance of the NDSD. Similarly, several frontal regions that were activated by the perceived stimuli were either nonresponsive or deactivated by the unperceived stimuli. Once again, these findings suggest abnormal cognitive or attentional cortical processing during the unperceived stimuli.
The cause of altered cortical responsiveness in patients with chronic pain with NDSD is unknown, but as suggested by the transient re-establishment of normal sensory perception observed during IV administration of the barbiturate sodium amytal,30 it likely reflects a central mechanism. In most patients with NDSD, sodium amytal infusions have a dramatic effect with normalization or remarkable improvement of all sensorimotor abnormalities in tandem with pain relief.1,31⇓ This response clearly indicates that the sensory and motor deficits are not structural (anatomic), but functional (dynamic). Of particular interest to subjects with pain are two specific actions of barbiturates: enhancement of GABA inhibition (GABAmimetic effects) in multiple brain and spinal cord sites, and ionotropic AMPA, kainate, and NMDA receptor noncompetitive antagonistic effects. The later may be responsible for the dramatic reduction in allodynia (touch-evoked pain) to infusions of sodium amytal.12,32,33⇓⇓ A neuro-psycho-biologic theory of conversion symptoms with increased inhibition of sensory and motor functions from corticofugal tracts has been proposed.34 Barbiturates, therefore, may serve to produce “inhibition of inhibition”—i.e., disinhibition and release of motor and sensory functions.
We have speculated before1,30⇓ that the phenomena of NDSD are of dynamic nature. In the presence of intractable pain, we suspect that NDSD constitute an unsuccessful attempt of the CNS to shut down or inhibit all peripheral inputs originating in or associated with the painful limb in an effort to control pain. Given the persistence of pain despite suppression of the cutaneous and often deep sensation, as well as the variable motor deficits, one cannot help but compare such patients with those with structural deafferentation. An example is patients with brachial plexus avulsion who continue to perceive severe pain in the absence of any sensory or motor function. NDSD then can be considered examples of functional deafferentation as opposed to structural deafferentation, and may result from maladaptive supraspinal neuroplasticity. However, in some long-standing cases, sodium amytal may remove pain but not the sensory deficit.12 In this case, the defects may have become fixed, not as a result of peripheral pathology, but as the result of permanent maladaptive plastic changes at cortical or subcortical levels.
Given the involvement of psychological/psychosocial factors as outlined in our case reports, it is conceivable that dynamic aberrations of brain function can occur under a multiplicity of emotionally charged conditions or certain personality organizations, where the individual utilizes specific mechanisms to avoid unpleasant physical or emotional events. The magnitude of original trauma or inciting event and the duration of actual nociception may be insignificant, but serve as a trigger of underlying central mechanisms in emotionally charged personal or psychosocial situations. We suspect that there may be an interaction between peripherally generated nociceptive or neuropathic pain and psychological vulnerability factors, which interaction is mediated by supraspinal mechanisms. Such patients with enhanced psychological vulnerability may be at risk to develop a pain disorder, NDSD, or other pathologic condition with relatively little peripherally generated nociceptive or neuropathic pain. Once such a central process develops at supraspinal levels, it may become independent from any actual peripheral inputs.
One wonders if this maladaptive neuroplasticity could be due to an attentional switch, with the patients directing attention toward the ongoing pain, which in turn could attenuate stimulus-evoked activation. In support of this concept are our findings of rostral and perigenual ACC activation during unperceived brush or noxious stimulation. Furthermore, neuroimaging studies suggest that pain-evoked responses within S1, ACC, PFC, and PPC are modulated by attention.16,35⇓ In the current study, the patients kept their eyes closed throughout imaging to minimize cues about the stimulus. However, we cannot rule out entirely that the patients’ attention varied during the stimulation of the affected vs unaffected limb.
Regarding the observed lack of activations in cortical regions representing the painful body part (e.g., S1), another possibility is that tonic activity in cortical neurons related to the state of ongoing pain precludes further significant increases in neuronal activity. However, if this is true, it is not clear why such a ceiling effect in neuronal firing does not create a situation of reduced rather than the typical heightened sensory sensitivity (e.g., allodynia, hyperalgesia) that accompanies most chronic pain states. Nevertheless, if tonic activity in cortical neurons indeed creates a ceiling effect in terms of neuronal firing, it may also be responsible for a hemodynamic ceiling effect and a negative BOLD signal. The effect of increased baseline levels on the BOLD signal in fMRI is currently being investigated as a cause of task-related deactivations.23,24,36⇓⇓
The normal clinical SSEP (except for some slowing in Patient 4) may seem at odds with the clear finding of abnormal fMRI in S1. However, anatomic or functional factors likely contribute to these seemingly contradictory findings. It is generally accepted that very early SSEP are generated from volleys in the dorsal columns and medial lemniscus and potentials around 20 msec involve S1 and possibly S2.37,38⇓ Although there is evidence that an anatomically intact S1 is essential for normal N20 SSEP,39 a recent report with the use of modified oddball task with rare stimuli applied to an anesthetic right hand in a case of conversive anesthesia shows that a particular component (P300) associated with cognitive processing could not be generated.20 Functional considerations may contribute to the dissociation between normal SSEP and abnormal fMRI results in our patients. A recent magnetoencephalographic study40 recorded normal short latency S1 responses to unperceived tactile stimulation in patients with posterior parietal tumors. Despite intact 40 msec responses, these patients lacked the typical longer responses at 60 msec and 150 msec when the stimuli were applied to their anesthetic side. The authors attributed the absence of tactile sensation in part to the lack of longer latency S1 responses. Furthermore, SSEP are quite resistant to changes in functional states (e.g., during anesthesia) and are likely to be normal unless there is clear structural pathology.41 The difficulty of interpreting clinical SSEP in patients with hysteria has also been extensively studied.42
Acknowledgments
K.D.D. is a Canada Research Chair in Brain and Behavior. Funds provided by this program contributed to this study.
Footnotes
-
See also page 1410
-
Additional material related to this article can be found on the Neurology Web site. Go to www.neurology.org and scroll down the Table of Contents for the May 13 issue to find the title link for this article.
- Received June 17, 2002.
- Accepted November 29, 2002.
References
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.