Paradoxical heat sensation as a manifestation of thermal hypesthesia: a study of 1090 patients with lesions of the somatosensory system

Sensory profiles of 1090 patients reveal loss of small thermosensory fibre function to be associated with paradoxical heat sensation.


Introduction
Paradoxical heat sensation (PHS) occurs when cooling of the skin is perceived as warmth instead of cold.Typically, single cold stimuli are perceived accurately, whereas PHS is experienced during temporal alternation of warming and cooling on the same area of the skin.Paradoxical heat sensation is rare in healthy described as a painless sensation, some studies have reported that PHS can also be perceived as painful or burning. 14ltered sensation such as PHS can be evoked in healthy individuals by applying nerve compression blocks.Classic studies showed that when the conduction of myelinated fibres is blocked by ischaemic or nerve compression, pain induced by single cold stimuli is perceived at higher temperatures and is typically described as burning hot. 11,38,39These results suggest that the paradoxical sensation of heat and burning elicited by cold stimuli is mediated by unmyelinated fibres (ie, C fibres) that are not affected by the compression.This interpretation is consistent with conduction velocity findings demonstrating that PHS is peripherally transmitted through C fibres. 30Mechanistic hypotheses have suggested that PHS is mediated through a specific class of C fibres, known as C2, normally responsive to cooling, heating, and menthol in humans 6 or by disinhibition of the heat-pinchcold (HPC) pathway in the peripheral or central nervous system. 6,14The HPC disinhibition model is a common explanation of perceptual changes with high similarity to PHS, such as the thermal grill illusion or A-fibre block. 7,11,38,39owever, these human experimental models are generally linked to a painful, burning heat, rather than painless warmth as mostly reported for PHS as perceived by patients.In this article, we analysed PHS in relation to aetiology (the specific injury or disease of the somatosensory system), presence or absence of pain, somatosensory profiles, descriptors of spontaneous pain, and abnormal sensations in 1090 patients with painful or painless polyneuropathy, unilateral peripheral nerve damage, or central lesions.Patients were part of the prospectively collected (N) European database and published previously in part. 2,19he objective of this study was to gain a better understanding of the clinical relevance of PHS in neuropathy by evaluating the relationship between paradoxical heat sensations, aetiology, presence or absence of pain, somatosensory profile, and pain symptoms.Furthermore, we aimed to explore putative mechanisms of this paradoxical phenomenon.
First, we compared the frequency of PHS across patient populations.Second, we tested whether an increase in PHS is associated with changes in the somatosensory profile, especially regarding sensory loss of cold and warm detection, combined with sensory gain of cold pain thresholds, equating PHS to the paradoxical sensations induced by an experimental A-fibre block. 11Finally, we analysed whether PHS could be related to specific neuropathic pain symptoms as captured by the Neuropathic Pain Symptom Inventory.

Standard protocol approvals, registrations, and patient consents
The study was approved by the local ethics committees at each participating centre.All patients have given written informed consent for their data to be collected and analysed in the database.All procedures were conducted in accordance with the Declaration of Helsinki.

Study design
All patients were tested in a single setting, using the QST protocol of the German Research Network on Neuropathic Pain (DFNS), 26 during which presence or absence of paradoxical heat sensation is recorded.Patients filled in the questionnaire before assessment.Fifteen pain facilities across Europe participated in gathering data in 3 consortia (German Research Network on Neuropathic Pain (DFNS), IMI (Innovative Medicines Initiative) EUROPAIN, and NEUROPAIN).All centres underwent a strict quality assessment. 18,37Data were collected in a central database in Bochum, Germany.An analysis of heterogeneity between the participating centres showed a high degree of homogeneity between the different centres, making it possible to analyse the database as a homogenous group. 33Subsets of this database as well as inclusion and exclusion have been published previously. 2,8,10,12,19,22,34,36

Inclusion or exclusion criteria
We included patients with confirmed peripheral neuropathic pain 9,31 due to polyneuropathy or unilateral peripheral lesions (peripheral nerve injury, postherpetic neuralgia, or radiculopathy); patients with central neuropathic pain due to spinal cord injury, syringomyelia, or stroke; and patients with similar painless conditions. 10,19Exclusion criteria were age , 18 years, insufficient language skills or other communication problems, pain treatment by topical local anaesthetics for $7 days in the last 4 months or by topical capsaicin in the last 6 months, comorbidities treated by anticonvulsants or antidepressants, other pain locations with pain intensities $6/10 on $15 days/month, spinal canal stenosis, peripheral vascular disease (Fontaine stage II or higher), pending litigation, and major cognitive or psychiatric disorders.Patients who were unable to detect any temperature change during thermal sensory limen (TSL) due to extensive sensory loss to thermal stimuli were excluded because PHS could not be determined (see below).We also excluded patients with unilateral syndromes who were also affected by contralateral neuropathy or painful conditions of the contralateral limb, as well as patients with incomplete records (eg, no precise diagnosis available, more than 2 missing variables of the QST in the affected area, and no information about age, sex, or other demographic data).

Quantitative sensory testing protocol and paradoxical heat sensation assessment
We used the QST protocol established by the DFNS. 25,26Patients suffering from polyneuropathy were tested on the dorsum of the foot, whereas patients suffering from other aetiologies were tested on the most affected area.In addition to PHS, QST according to the DFNS protocol assesses 12 other parameters: cold and warm detection thresholds (CDT and WDT), thermal sensory limen, cold and heat pain thresholds (CPT and HPT), mechanical pain threshold and sensitivity (MPT and MPS), dynamic mechanical allodynia (DMA), pressure pain threshold (PPT), wind-up ratio (WUR), tactile (mechanical) detection threshold (MDT), and vibration detection threshold. 25Thermal detection and pain thresholds were measured using either a TSA 2001-II (MEDOC, Israel) or a MSA (SOMEDIC SALES AB, Sweden) thermode at a rate of temperature change of 1˚C/second. 25The number of PHS (if any) was recorded during assessment of the TSL.In this task, patients were instructed that the testing device would either cool or warm their skin several times, and they should press a stop button every time they felt a new temperature change.Furthermore, patients were instructed to describe the quality of the temperature change (eg, cold, warm, hot, or painfully hot) after each button press.This procedure consisted of alternated warming and cooling of the skin for a total of 6 consecutive temperature changes.The TSL procedure always started with warming of the skin, but no preheating to noxious temperatures was applied. 13The number of PHSs (ie, report of a warm sensation while the skin was cooled) varied between 0 and 3.According to the DFNS normative data, if patients were female and older than 50 years or male and older than 40 years and they were tested on the feet, abnormal frequency of PHS is defined as at least 2 PHSs during the 3 cooling temperature changes.In younger ages, any number of PHSs was considered abnormal.Henceforth, we will use the term "no PHS" to refer to "no elevated reports of PHS compared with normative data."Patients who were unable to detect any temperature change during TSL were excluded because PHS could not be determined.
An initial assessment of data from 180 healthy individuals in the DFNS reference database revealed that all parameters, except PHS and DMA, were normally distributed or could be transformed to a standard normal distribution using a log transformation. 18,24,25uantitative sensory testing raw or log-transformed results were normalised using a z-transformation, based on normative material defined for 5 different age decades, sex, and 4 testing sites (Magerl  et al., 2010).If the patient's most prominent pain was in an upper limb, the values were z-transformed using the reference data for hands; if the most painful area was at a lower limb, the values were z-transformed using the reference data for feet.Assessments at the torso were z-transformed using the reference data for trunk, whereas all assessments at the face were z-transformed using the reference values for the cheek.Vibration detection threshold was compared with normative data assessed at the ulnar styloid for hand and arm, malleolus for foot and leg, costal arch for trunk, and zygomatic process for face.As a consequence of normalization, all measures had a mean 5 0 and a standard deviation 5 1 and were comparable between patients irrespective of their demographics or affected body area (eg, face or feet).Abnormal values were defined as values beyond the 95% confidence interval, corresponding to z values ,21.96 or .1.96.

The Neuropathic Pain Symptom Inventory questionnaire
The NPSI was developed to identify different dimensions of neuropathic pain. 5It comprises 10 items, including 5 descriptors of ongoing pain quality (burning, squeezing, or electric shocks), 3 descriptors of evoked pain (pain evoked by light touch, thermal stimuli, or light pressure or cold), and 2 reports of tingling and pins and needles.For each of the 10 items, patients rated the intensity on an 11-point rating scale, ranging from "zero or no [...]" to "10 or worst [...] imaginable."

Statistics
We applied unpaired t-tests (2-tailed, P , 0.05) to investigate whether PHS could discriminate differences between the groups of patients with different aetiologies or with and without pain.Using multivariate analyses, we tested 11 normally distributed QST parameters and 10 NPSI items, with PHS (elevated or normal), aetiology (polyneuropathy, unilateral peripheral, or central), duration of disease (less than a year, 1 to 5 years, over 5 years), and presence or absence of pain as fixed main effects, and PHS 3 aetiology, PHS 3 duration, and PHS 3 pain as interaction effects.To avoid overinterpretation of false positive results due to alphacumulation and multiple testing, we applied a Benjamini-Hochberg correction to control the false discovery rate and only discuss effects that are significant models on a corrected P , 0.05 level.

Data access and availability
All data and materials are in the database hosted by the German Research Network of Neuropathic Pain (DFNS) and accessible to the first author (J.V.).Data access for purpose of replication can be requested through submitting an informal data request to the DFNS and will be granted by the DFNS board after review.

Study population
A total of 1090 patients from 15 European centres were included in the analysis (polyneuropathy: 438 painful and 61 painless, unilateral peripheral nerve injury: 411 painful and 40 painless, and CNS lesion: 119 painful and 21 painless).The NPSI was not collected in all centres, and only for patients with painful conditions, and therefore was available from 404 patients only (polyneuropathy: 165, unilateral peripheral: 173, and CNS lesion: 66).Patients' demographics are presented in Table 1.Patients with elevated PHS were on average slightly older (59 years vs 57 years, P , 0.001), within patients with unilateral peripheral lesion more likely to appear after a radicular lesion, and within patients with central lesion less likely to appear after a stroke.There were no differences between painful and nonpainful conditions and neither in sex, duration of disease, or aetiology of polyneuropathy.

Frequency of paradoxical heat sensation
Abnormal PHS was reported by 30% of all included patients.Notably, although the percentage was higher in patients with loss of thermal detection (35%) compared with those with preserved thermal detection (29%), in absolute numbers, most patients with PHS had preserved thermal detection (n 5 248) compared with those with loss of thermal detection (n 5 82).This result is significantly different from the frequency of PHS in the DFNS reference data for healthy participants, 25 which found that only 2% of healthy participants reported PHS in lower limbs of older participants (P , 0.001).Paradoxical heat sensation was more frequent in polyneuropathy (42%) in comparison to unilateral peripheral nervous lesions (20%) and central nervous lesions (24%) (P , 0.001), whereas there was no difference between unilateral peripheral and central nervous lesions (P 5 0.33).Regarding painful vs painless conditions, we found no differences in the frequency of PHS, neither overall (P 5 0.67) nor within aetiologies (polyneuropathy: P 5 0.69, unilateral peripheral nervous lesions: P 5 0.38, and central nervous lesions: P 5 0.98).

Paradoxical heat sensation in relation to sensory profiles
Figure 1 displays the sensory profile for all patient groups in relation to PHS, and all results of the multivariate analysis can be found in Table 2.In models corrected for aetiology, duration of disease, and presence or absence of pain, PHS had a significant effect on all thermal QST parameters, both thermal detection (CDT, WDT, and TSL) and pain thresholds (CPT and HPT).There was a significant interaction of PHS and aetiology for each of the thermal QST parameters, indicating that thermal thresholds were elevated in the presence of PHS for patients with unilateral nerve lesions and central lesions but not polyneuropathy.In addition, there was a significant interaction effect of PHS and presence of pain for thermal detection thresholds and MPT, indicating more loss of function associated with PHS in patients without pain.Finally, there was a significant interaction between PHS and duration of disease for CPT, PPT, and MPS.In the presence of PHS, more loss of function for CPT was detected.This was, however, not the case in patients with a disease duration of over 5 years.In patients with a disease duration of under 1 year there was a gain of function for PPT and MPS in the presence of PHS.With longer disease duration, the opposite (ie, a loss of function for PPT and MPS) could be shown.Neuropathic Pain Symptom Inventory profiles of all patient etiologies can be found in Figure 2. In the multivariate NPSI models, there was a significant effect of PHS on burning pain and electric shock-like pain, both being less frequent in patients with PHS.In addition, there was a significant interaction between PHS and duration of disease for burning pain, indicating more burning sensation for patients with presence of PHS and a disease duration below 1 year, and less burning sensation for patients with elevated PHS and a duration over a year.

Discussion
In this large clinical data set of patients experiencing peripheral and central nervous system diseases with and without neuropathic pain, abnormal PHS was frequently reported by all patient groups irrespective of whether they suffered from painful or painless lesion or disease of the somatosensory system (30% each).We found that PHS was much more frequent in patients with polyneuropathy (39%-42%) compared with unilateral peripheral or central lesions.Paradoxical heat sensation was associated with reduced thermosensitive small fibre function.The presence of PHS was not related to mechanical detection or pain thresholds, suggesting that PHS is not a sign of a general, but rather a thermosensory-specific small fibre dysfunction.Finally, we found an association between the presence of abnormal PHS and lower NPSI scores for electric shock-like and burning pain quality.

Paradoxical heat sensation is more frequent in polyneuropathy
Paradoxical heat sensation was significantly more frequent in polyneuropathy compared with other conditions with peripheral (ie, unilateral nerve damage) or central lesions of the somatosensory system, with 42% of patients having abnormal increased frequency of PHS compared with healthy individuals.The QST profiles were different for patients with polyneuropathy compared with patients with other neuropathies.Patients with polyneuropathy showed on average a general sensory loss on all sensory parameters, whereas the other groups also exhibited sensory gain, for example, lower pressure pain thresholds.Interestingly, we found no differences between QST profiles of patients with polyneuropathy with and without PHS.This result is in contrast to previous findings indicating that patients with uremic polyneuropathy experiencing PHS also exhibited less cold sensitivity, 40 and patients with chemotherapy-induced neuropathy experiencing PHS had less warm sensitivity. 32These findings were not reproduced in our population, which covers polyneuropathy of diabetic, HIV, idiopathic, chemotherapy-induced, but not uremic aetiology.A previous study on diabetic children also did not find an association of PHS with sensory loss to cooling. 4 These differences across studies may be related to symptoms severity.In our data, all patients with polyneuropathy already had major thermosensory deficits, suggesting potentially more severe Patient demographics.

Paradoxical heat sensation is less frequent and related to thermal sensory loss in patients with unilateral nerve damage and central lesions
In patients with unilateral nerve damage or central lesions, PHS was similarly frequent across groups and associated with thermal detection deficits.This result suggests that PHS reflects loss of both C-fibre and Ad-fibre function or their associated pathways in the central nervous system, although the relative importance of individual pathways cannot be determined from this study because the nervous lesions in these patients are rarely selective for single thermal pathways.Thus, the prediction from experimental A-fibre block studies in healthy individuals 11,38,39 that selective loss of cold detection is linked to PHS is difficult to assess in this study.However, our finding that PHS is related to functional losses in thermal, but not mechanical, modalities suggests that PHS reflects a specific deficit in the function of small fibres involved in thermosensation, rather than a general sensory deficit.

Association between paradoxical heat sensation and Neuropathic Pain Symptom Inventory ratings
In the NPSI ratings, the presence of burning pain was similar in frequency in patients with and without PHS.Burning sensations are believed to be associated with loss of small fibre function 17 and to be a result of a disinhibition of warm detection neurons caused by rarefication of cold-related fibres. 7In our analysis, spontaneous burning sensation reported by patients was lower for patients reporting PHS.Furthermore, PHS was associated with lower intensity of electric shock-like pain, which is considered to reflect small, myelinated fibre-related pain.These data suggest that impaired or unimpaired function of Ad fibres and C fibres is relevant.

Limitations
In this study, PHS was assessed during 6 alternating warming and cooling stimuli, and in most cases, a single warm report during cooling of the skin was sufficient to classify the patient as reporting PHS.25,27 However, other methods to assess PHS have been described previously.Yosipovitch and colleagues applied a series of 8 alternating warming and cooling stimuli and considered patients reporting PHS as those individuals who reported at least 3 or 4 warming sensations during cooling of the skin. 40By contrast, in our protocol, patients reporting PHS were patients who experienced 1 or 2 of 3 warm sensations during cooling.
Another limitation was the selection of the body site where QST was performed.This selection was not standardised for most patients' aetiologies, except for polyneuropathy.The body region for QST assessment was selected based on where the patient reported the most prominent pain or strongest somatosensory dysfunction in case of painless conditions.Consequently, while the tested body region could have been anywhere on the body, the reference values are only defined for hand, feet, trunk, and face.For each patient, we used reference values that were the closest approximation to the tested area.Finally, temporal and spatial summation effects may be underestimated 1 because the battery of thermal and mechanical measures included in the QST according to the DFNS protocol only captures short-term effects (ie, each measure is obtained within a limited amount of time), rather than long-term effects, which may be more closely related to spontaneous pain symptoms.

Conclusions
We found that PHS was common not only in patients with polyneuropathy but also in patients with central lesions or unilateral peripheral injury.In patients with central or unilateral peripheral lesions, the presence of PHS was associated with reduced thermosensitive small fibre function, most likely including thinly myelinated Ad fibres and unmyelinated C fibres.Our findings suggest that PHS is not a general sign of nerve damage, but rather a thermosensory-specific small fibre dysfunction, the mechanisms of which include a disinhibition phenomenon.In healthy subjects, the thermoreceptive and nociceptive circuitries include a pathway by which mild cooling can induce a percept of warmth or heat, and this circuitry is normally suppressed by thermosensory small fibre pathways.By contrast, our patient data indicate that the association between the presence of elevated PHS and lower NPSI scores for electric shock-like and burning pain quality plays an important role for Ad than C fibres.Mechanistically, the inhibitory interactions between different small fibre pathways deserve more experimental studies; our findings suggest that there likely are additional inhibitory mechanisms distinct from the known gate control or the brainstem descending modulatory system.Paradoxical heat sensation is an interesting sensory phenomenon that can be easily detected and is reported spontaneously by patients.This makes it an interesting parameter for clinical assessment of neuropathy.We argue that PHS should be included in the list of sensory signs of small fibre functional sensory loss, 15 although at face value it seems to be a positive phenomenon.It may be an easy way to clinically screen for loss of small fibre function with a dichotomous outcome (PHS yes or no) and more sensitive than assessing abnormal thermal detection thresholds using DFNS reference values.
Funding source: The EUROPAIN project is a public-private partnership and has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement n˚115007, resources for which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations (EFPIA) companies' in-kind contribution.
The NEUROPAIN project is an investigator-initiated European multicentre study with Prof. Dr. Ralf Baron as the principal investigator and 10 co-investigator sites, supported by an independent investigator-initiated research grant from Pfizer Ltd.The funding source had no role in study design, data collection and analysis, or writing of the manuscript.

Figure 1 .
Figure 1.Quantitative sensory testing z profile of patients with elevated frequency of paradoxical heat sensation (PHS) in comparison to those without PHS for painful (1) and painless (2) lesion or disease of the somatosensory system.Z values between 21.96 and 11.96 indicate the 95% confidence interval of values found in healthy subjects.(A) all patients, (B) polyneuropathy, (C) unilateral peripheral nerve injury, and (D) central lesion.All values presented as mean 6 95% CI of the mean.

Figure 2 .
Figure 2. Neuropathic Pain Symptom Inventory profile of patients with elevated frequency and without paradoxical heat sensation.(A) all patients with neuropathic pain, (B) painful polyneuropathy, (C) unilateral peripheral neuropathic pain, and (D) central neuropathic pain.All values presented as mean 6 95% CI of the mean.