You are on page 1of 13

Research Paper

Downloaded from https://journals.lww.com/pain by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3pzrw1VmaZXSTmZg1JbZnxxQNR3HPszs+F9KYt8Zn/9enCcSDolwFxQ== on 03/23/2018

Pathophysiological mechanisms of neuropathic


pain: comparison of sensory phenotypes in patients
and human surrogate pain models
Jan Vollerta,b,*, Walter Magerlb, Ralf Baronc, Andreas Binderc, Elena K. Enax-Krumovaa,d, Gerd Geisslingere,f,
Janne Gierthmühlenc, Florian Henrichb, Philipp Hüllemannc, Thomas Kleinb, Jörn Lötsche, Christoph Maiera,
Bruno Oertelf, Sigrid Schuh-Hoferb, Thomas R. Tölleg, Rolf-Detlef Treedeb
Downloaded from https://journals.lww.com/pain by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3pzrw1VmaZXSTmZg1JbZnxxQNR3HPszs+F9KYt8Zn/9enCcSDolwFxQ== on 03/23/2018

Abstract
As an indirect approach to relate previously identified sensory phenotypes of patients suffering from peripheral neuropathic pain to
underlying mechanisms, we used a published sorting algorithm to estimate the prevalence of denervation, peripheral and central
sensitization in 657 healthy subjects undergoing experimental models of nerve block (NB) (compression block and topical lidocaine),
primary hyperalgesia (PH) (sunburn and topical capsaicin), or secondary hyperalgesia (intradermal capsaicin and electrical high-
frequency stimulation), and in 902 patients suffering from neuropathic pain. Some of the data have been previously published.
Randomized split-half analysis verified a good concordance with a priori mechanistic sensory profile assignment in the training (79%,
Cohen k 5 0.54, n 5 265) and the test set (81%, Cohen k 5 0.56, n 5 279). Nerve blocks were characterized by pronounced
thermal and mechanical sensory loss, but also mild pinprick hyperalgesia and paradoxical heat sensations. Primary hyperalgesia
was characterized by pronounced gain for heat, pressure and pinprick pain, and mild thermal sensory loss. Secondary hyperalgesia
was characterized by pronounced pinprick hyperalgesia and mild thermal sensory loss. Topical lidocaine plus topical capsaicin
induced a combined phenotype of NB plus PH. Topical menthol was the only model with significant cold hyperalgesia. Sorting of the
902 patients into these mechanistic phenotypes led to a similar distribution as the original heuristic clustering (65% identity, Cohen
k 5 0.44), but the denervation phenotype was more frequent than in heuristic clustering. These data suggest that sorting according
to human surrogate models may be useful for mechanism-based stratification of neuropathic pain patients for future clinical trials, as
encouraged by the European Medicines Agency.
Keywords: Quantitative sensory testing, German Research Network on Neuropathic Pain, Menthol, A-fiber block, UVB,
High-frequency electrical stimulation, Capsaicin, Lidocaine, Human surrogate pain models

1. Introduction loss (ie, hypoesthesia) and gain of function (ie, hyperalgesia and
Quantitative sensory testing (QST) in accordance with the DFNS allodynia).45 Quantitative sensory testing plays an important role for
(German Research Network on Neuropathic Pain) protocol diagnosing patients with neuropathic pain19,56 and can offer indirect
assesses the sensory function of afferent myelinated A-beta- and insights into underlying mechanisms of pathophysiology.7,15,18,27
unmyelinated and thinly myelinated C- and A-delta-fibers3 for both Peripheral neuropathic pain is induced by partial nerve
damage, leaving 2 classes of nociceptors as candidates to cause
the pain: damaged nociceptors and surviving undamaged
Sponsorships or competing interests that may be relevant to content are disclosed nociceptors.11 Damaged nociceptors are responsible for sensory
at the end of this article.
a
loss due to denervation, but possibly also for ongoing pain due to
Department of Pain Medicine, BG University Hospital Bergmannsheil GmbH,
ectopic activity either arising in the periphery or in denervated
Ruhr-University Bochum, Bochum, Germany, b Center of Biomedicine and Medical
Technology Mannheim (CBTM), Medical Faculty Mannheim, Heidelberg University, second-order neurons. Surviving nociceptors can be peripherally
Mannheim, Germany, c Division of Neurological Pain Research and Therapy, sensitized by inflammatory processes related to denervation and
Department of Neurology, Universitätsklinikum Schleswig-Holstein, Campus Kiel, reinnervation which may cause hyperalgesia. Input from both
Kiel, Germany, d Department of Neurology, BG University Hospital Bergmannsheil
GmbH, Ruhr-University Bochum, Bochum, Germany, e Institute of Clinical
damaged and undamaged nociceptors can induce central
Pharmacology, Pharmazentrum Frankfurt/ZAFES, University Hospital of Goethe- sensitization which may also cause hyperalgesia. We thus have
University, Frankfurt am Main, Germany, f Fraunhofer Institute for Molecular Biology 4 candidate mechanisms to model in human surrogate models of
and Applied Ecology (IME), Project Group Translational Medicine and Pharmacol- peripheral neuropathic pain: denervation, ectopic activity, pe-
ogy (TMP), Frankfurt am Main, Germany, g Department of Neurology, Klinikum
Rechts der Isar, Technische Universität München, Munich, Germany
ripheral sensitization, and central sensitization. Models of ectopic
*Corresponding author. Address: Neurophysiologie, Zentrum für Biomedizin und
activity relate to ongoing pain, whereas the other 3 mechanisms
Medizintechnik Mannheim, Medizinische Fakultät Mannheim, der Universität relate to altered perception of evoked pain.
Heidelberg, Ludolf-Krehl-Str.13-17, 68167 Mannheim, Germany. Tel.: 149 621/ There are no established human surrogate models of ectopic
383-9926; fax: 149 621/383-9921. E-mail address: Jan.Vollert@rub.de (J. Vollert). activity and ongoing pain,33 but human surrogate models for
PAIN 0 (2018) 1–13 denervation, peripheral sensitization, and central sensitization
© 2018 International Association for the Study of Pain have been used for pharmacological studies.8,33,42 The aim of
http://dx.doi.org/10.1097/j.pain.0000000000001190 this study was to estimate the prevalence of 3 distinct types of

Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 1

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
2
·
J. Vollert et al. 0 (2018) 1–13 PAIN®

predefined mechanisms (denervation, peripheral sensitization, (5) Intraepidermal capsaicin injection (unpublished data, methods
and central sensitization) in a large number of healthy subjects as in Refs. 36 and 43)
submitted to various surrogate models of experimental pain (both (6) Cutaneous electrical high-frequency stimulation28,32,37
previously published8,16,20,25,34,37,40,41,53,62 and unpublished (7) Muscle electrical high-frequency stimulation53
data) using a sorting algorithm developed in patients with (8) Topical application of capsaicin solution and lidocaine16
peripheral neuropathic pain. Each of the predefined mechanisms (9) Topical 40% menthol application9,62
was represented by a pair of established surrogate models. This
served the purpose of emphasizing those QST characteristic
2.3. Quantitative sensory testing
prototypical for the underlying mechanism and minimizing
idiosyncrasies of individual surrogate models: The standardized protocol of the DFNS was used for QST, which
(1) Transient functional denervation by nerve compression or assesses the function of small and large afferent fibers. The
topical lidocaine, targeting either myelinated A-fibers or protocol has been described in detail previously;52 in brief, the
unmyelinated C-fibers, or both. following parameters were tested: thermal detection thresholds
(2) Peripheral sensitization by inducing primary hyperalgesia (PH) by for the perception of cold (cold detection threshold [CDT]) and
topical application of capsaicin or by ultraviolet B (UVB) radiation; warmth (warm detection threshold), paradoxical heat sensations
these models commonly induce heat and pressure hyperalgesia. (PHSs) during assessment of thermal sensory limen of alternating
(3) Central sensitization by inducing secondary hyperalgesia (SH) warm and cold stimuli, thermal pain thresholds for cold and hot
by cutaneous electrical high-frequency stimulation (HFS) or stimuli, tactile mechanical detection threshold (MDT) and
intradermal injection of capsaicin; they induce mostly mechan- vibration detection threshold, pain summation to repetitive
ical hyperalgesia to pinprick. pinprick stimuli (wind-up ratio), mechanical pain sensitivity
We then used an algorithm that had been developed for (MPS) including thresholds for pinprick (mechanical pain thresh-
heuristic sorting of QST profiles of neuropathic pain patients to old) and blunt pressure (pressure pain threshold), and a stimulus–
sort sensory profiles of human surrogate models into mechanis- response–function for pinprick sensitivity (MPS) during which
tically defined groups,60 including 5 additional human surrogate dynamic mechanical allodynia (DMA) was assessed by stroking
models with less clearly defined, controversial, or mixed un- light touch stimuli. For all parameters, negative (loss of function)
derlying mechanisms. The final aim was to assign neuropathic as well as positive (gain of function) phenomena were assessed.
pain patients to presumed underlying mechanisms according to
their similarity to the mechanism-related profiles derived from the
2.4. z-Transformation
human surrogate models.
An initial analysis of 180 healthy subjects revealed that all QST
parameters are either normally or log-normally distributed (with
2. Methods the exception of DMA and PHS).51 To compare individual QST
2.1. German Research Network on Neuropathic Pain data between subjects regardless of sex-, age-, or body-site
differences, the DFNS proposed presenting QST values as z-
The German Research Network on Neuropathic Pain (DFNS, values.44,48,51,52 These values are normalized to the mean and
http://www.neuropathischer-schmerz.de) was established in SD of reference data obtained in healthy subjects so that a z-value
2002 to investigate mechanisms and treatments of neuropathic of zero represents the mean value of a healthy control cohort
pain. Quantitative sensory testing of human surrogate models of matched in age and sex, and the respective body re-
neuropathic pain using the complete DFNS QST protocol has gion.44,48,51,52 Z-scores above zero indicate gain of function
since been conducted at the following sites: Ruhr-University when the subject was more sensitive to the test stimuli compared
Bochum; Goethe-University Frankfurt; Medical Faculty Man- with the population mean (hyperesthesia or hyperalgesia),
nheim, Heidelberg University; University of Schleswig-Holstein whereas z-scores below zero indicate loss of function referring
Campus Kiel; and Technical University of Munich. Appropriate to a lower sensitivity of the subject (hypoesthesia or hypoalgesia).
ethics committee approval has been obtained at all sites. Although individual z-scores are considered as abnormal if
Subjects were included according to our developed guidelines,24 beyond 61.96,51 for groups of patients, z-scores of 61.0 have
and quality of the included QST data was assured by been shown to be of clinical significance, as they include
standardized training courses for each investigator of each a relevant number of patients with abnormal values.45
center and verified by regular audits.61 Each center and Dynamic mechanical allodynia and PHS, which do not normally
investigator used identical equipment and standardized instruc- occur in healthy subjects and do not fit any known distribution,
tions. A recent analysis of heterogeneity between centers has can thus not be z-transformed. They are usually presented as
shown that data of various centers can be analyzed as percentages of presence, but for use in statistical analysis, they
a homogenous data set, if these quality criteria are fulfilled.59 can be transformed to pseudo-z-values: DMA can be coded into
a 0/2/3-variable representing no DMA (coded as 0), DMA with
average pain ratings below 1 on a 0 to 100 numerical rating scale
2.2. Surrogate models
(coded as 12), and DMA with average pain ratings between 1 and
The following surrogate models were included in the analysis 100 (coded as 13). Paradoxical heat sensation can be trans-
(Table 1): formed to a binary 0/2-variable showing absence (coded as 0) or
(1) A-fiber-conduction blockade by sustained pressure on the presence (coded as 12) of pathological values.
nerve trunk (unpublished data, methods as in Ref. 65) As many surrogate models were applied at the volar lower arm
(2) Topical lidocaine 5% patch34 or upper thigh, which are not standardized reference areas, we
(3) Topical capsaicin application; using cream, watery solution, or also performed intraindividual comparisons either relative to the
patch16,40,41 same area before treatment, or, if these data were unavailable, to
(4) Ultraviolet B light irradiation at 3 times minimal erythema dose the contralateral untreated side using effect sizes (Cohen d13).
(published and unpublished data, methods as in Refs. 25 and 43) This measure normalizes changes in the mean value before and

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 3

Table 1
Human surrogate models and participating laboratories.
Model n Sex: n (%) female Age: mean (range) Participating centers Test area
NBs
A-fiber block 24 12 (50) 25 (21-39) Kiel, Mannheim Hand dorsum
Topical lidocaine 41 20 (49) 34 (19-69) Bochum Volar forearm
PH
Topical capsaicin 273 147 (53) 25 (15-75) Bochum, Frankfurt, Mannheim Volar forearm, dorsal hand, foot
UVB irradiation 158 51 (32) 24 (19-42) Frankfurt, Mannheim Volar forearm, upper thigh
SH
Capsaicin injection 36 19 (53) 32 (23-68) Kiel, Mannheim, Munich Volar forearm
Cutaneous HFS 12 3 (25) 36 (24-57) Mannheim Volar forearm
Mixed
Topical capsaicin, SH area 37 15 (41) 24 (19-39) Mannheim Volar forearm
UVB, SH area 22 0 (0) 24 (24-24) Mannheim Upper thigh
Muscular HFS 15 7 (47%) 24 (19-27) Mannheim Lower back
Topical menthol 11 0 (0) 25 (23-28) Kiel Hand dorsum
Topical lidocaine 1 topical capsaicin 28 17 (61) 30 (20-75) Bochum Volar forearm
HFS, high-frequency stimulation; NB, nerve block; PH, primary hyperalgesia; SH, secondary hyperalgesia; UVB, ultraviolet B.

after treatment to SD. There are no general interpretations of With i 5 one of 13 QST parameters, m 5 one of 3 prototypic
“good” effect sizes, but it is often considered that effect sizes sensory profiles and sim being the SD of the ith QST parameter for
below 0.3 can be considered small, above 0.5 medium, and the mth mechanism in the defining data set, mim being the mean z-
above 0.8 are commonly considered as large treatment effects.12 value of the same QST parameter and mechanism in the defining
data set, n 5 the nth participant in a set of participants, and finally
xin being the z-value found in the nth participant for the ith QST
2.5. Pattern-based sorting algorithm
parameter. By averaging the probability over the 13 QST
As sorting method, we applied a method developed for assigning parameters, we quantify the similarity of the individual subject’s
patients to sensory phenotypes developed in our previous work, QST profile to the mean profile of each of the 3 mechanisms.60
where we also showed general validity of the method.5,60 The same The resulting probability value is ranging from 0% to 100% and
method was applied here for 2 reasons: first, continuity to our work in can be calculated for m 5 3 mechanisms.
patients suffering from neuropathic pain would allow for higher As a simple way of categorizing patients into mechanisms, we
comparability between the hypothesis-driven approach in human suggest sorting each patient to the phenotype with the highest
surrogate models in the present article, and our earlier, hypothesis- probability value:
free work in patients suffering from neuropathic pain. Second, and (1) Calculate pi according to F(1) for each of the 13 QST
more importantly, the probabilistic version of our previously parameters. Use m and s from Table 2 for nerve block (NB).
published sorting algorithm allows for any given patient to be (2) Average the 13 probabilities. The resulting value is the
assigned to more than one sensory phenotype (or to present with probability for this patient to show the NB phenotype.
more than one underlying mechanism); our previous article (3) Repeat steps 1 and 2, using m and s from Table 2 for PH and
established a robust cut-off criterion for this type of sorting.60 In SH.
this study, the probabilistic sorting was used to assign individual (4) Allocate the patient to the phenotype with the highest
patients to appropriate mechanisms of pathophysiology. probability value.
Based on 6 surrogate models with a clearly described To show reproducibility of the QST profile of the subgroups
mechanism (A-fiber block by nerve compression or by topical found, a random number generator was used to split the data
lidocaine 5% for small fiber block, topical capsaicin or UVB set of the 6 defining surrogate models in half, using about 50%
radiation for PH, and intraepidermal capsaicin injection or of the data to define the prototypic sensory profiles of each of
cutaneous HFS for SH), we developed an algorithm determining the 3 mechanisms and the other half to verify correct
whether an individual QST profile is most similar to denervation, allocation.
PH, or SH. We used a method we recently established in patients
suffering from neuropathic pain,5,60 applying a standard normal 2.6. Healthy sensory profile and deterministic and
distribution, stripped from the normalization factor, so that the probabilistic algorithm
resulting value ranges on a scale from 0% to 100%.60
For each participant n, over the QST profile of the z-values of A fourth prototypical sensory profile is derived from the QST
i 5 13 parameters, a probability can be calculated, showing reference data sets in healthy subjects without any experimental
a percentage of similarity to the prototypic QST profile of fiber models. According to the definition of the z-transformation, this
block, PH, or SH: profile will have a mean z-value of zero and an SD of one for each
parameter. In our previous analysis,60 we showed that the
algorithm can separate patients suffering from neuropathic pain
0 11 0 from healthy subjects if a probability cut-off of 64% is applied. To
 2
B B xin 2 mim CC
13 this point, we use a deterministic approach, that is, each
+ pi
Fð1Þ: i ¼ 1 n;m
13
with pin;m 5exp@ 2 @ q ffiffiffiffiffiffiffiffiffi AA participant is sorted to exactly one mechanism. It is, however,
2s2im our belief that different mechanisms can coexist; thus, with the
cut-off determined for healthy subjects in our previous work

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
4 J. Vollert et al. 0 (2018) 1–13
· PAIN®

Mean QST z-values (m) and SDs (s, in brackets) for each of the 13 QST parameters separately for each of the mechanisms. Values for healthy subjects follow the definition of z-values: mean 5 0 and SD 5 1. PHS is coded as pseudonormally distributed with 0 5 absence and 2 5 presence, DMA is coded

CDT, cold detection threshold; CPT, cold pain threshold; DMA, dynamic mechanical allodynia; HPT, heat pain threshold; MDT, mechanical detection threshold; MPS, mechanical pain sensitivity; MPT, mechanical pain threshold; PH, primary hyperalgesia; PHS, paradoxical heat sensation; PNP, polyneuropathy,
transferred onto the surrogate models, we suggest 2 alternative

0.48 (0.86)
0.13 (0.49)
0.16 (0.54)
0.00 (1.00)
versions, namely a deterministic or a probabilistic algorithm. The
deterministic algorithm follows the steps below:

PHS
Calculate F(1) for each of the 13 QST parameters. Use m and s
from Table 2 for healthy.

0.06 (0.34)
0.81 (1.07)
1.12 (1.11)
0.00 (1.00)
(1) Average the 13 probabilities. The resulting value is the
DMA probability for this patient to show a healthy profile.
(2) Repeat steps 1 and 2, using m and s from Table 2 for NB, PH,
and SH.
0.22 (1.06)
20.23 (1.93)
21.08 (1.83)
0.00 (1.00)
(3) Allocate the participant to the phenotype with the highest
probability value.
By contrast, in the probabilistic algorithm, steps 1 to 3 remain
VDT

identical and step 4 is exchanged as follows:


23.24 (2.89)
20.32 (1.13)
20.80 (1.38)
0.00 (1.00)

(4) Sort the participant to all phenotypes with a probability


above 64%. If the only probability above this cut-off is for
being healthy or no phenotype reaches a probability above
MDT

this cut-off, the sensory profile of the subjects cannot be


distinguished from normal.
0.25 (1.29)
0.18 (1.01)
0.31 (1.15)
0.00 (1.00)

These 2 versions were used for all analyses below and are
PPT, pressure pain threshold; QST, quantitative sensory testing; SH, secondary hyperalgesia; TSL, temperature sensory limen; VDT, vibration detection threshold; WDT, warm detection threshold; WUR, wind-up ratio.

presented alongside, all QST profiles of all surrogate models


WUR

underwent both versions of the algorithm.


20.15 (1.38)
0.71 (1.21)
1.95 (1.19)
0.00 (1.00)

2.7. Comparison with sensory phenotypes in patients


Random split-half analysis of human surrogate models of nerve blocks, PH, and SH: training data set (n 5 265).
MPS

In our recent work,5,60 we have described 3 distinct sensory


phenotypes that appear in patients suffering from peripheral
0.86 (1.87)
1.54 (1.08)
3.12 (0.97)
0.00 (1.00)

neuropathic pain of various etiology or clinical origin. To analyze


how this previous hypothesis-free pattern searching method in
MPT

patients suffering from neuropathic pain may relate to our new


mechanistic sensory subgroup classification described above,
20.63 (0.97)
1.35 (1.66)
0.63 (1.45)
0.00 (1.00)

the n 5 902 patients from the initial cluster analysis study5


underwent the deterministic version of the new sorting algorithm
and the result was compared with each patient’s sensory
PPT

phenotype in the deterministic version of the previous sorting


algorithm developed for patients.60 Agreement between algo-
0.52 (1.23)
2.37 (1.14)
20.14 (1.24)
0.00 (1.00)

rithms was assessed using Cohen kappa.21 Cohen kappa is


a value that measures similarity between groups or concordance
HPT

of ratings on a scale ranging from 0 (no similarity) to 1 (perfect


identity). Although no universal guideline for interpreting Cohen
0.06 (1.13)
20.12 (1.31)
0.33 (1.16)
0.00 (1.00)

kappa exists, as a rule of thumb, it was suggested that values


pseudonormally distributed with 0 5 absence, 2 5 0 to 1 (on a 0-100 numerical rating scale), and 3 5 1 to 100.

below 0.4 would indicate poor similarity, between 0.4 and 0.75
CPT

good similarity, and above 0.75 excellent.21 We then applied the


new probabilistic sorting algorithm to estimate the prevalence of
22.55 (1.52)
21.03 (1.12)
22.28 (0.84)
0.00 (1.00)

the 3 predefined mechanisms in this cohort of patients; this allows


each patient to be assigned to more than one mechanism.
TSL

3. Results
21.36 (1.27)
20.23 (1.04)
21.87 (0.89)
0.00 (1.00)

The analysis comprised a total of n 5 657 healthy subjects who


WDT

participated in published and unpublished studies on human


surrogate models: 9 distinct models (2 of them including areas of
both PH and SH) at 5 centers (Table 1). Approximately 44% of
22.94 (2.36)
21.05 (1.59)
22.71 (1.07)
0.00 (1.00)

subjects were female (291/657); sex distribution was not


homogenous across models, as for the UVB, cutaneous HFS,
CDT

and menthol models, predominantly men were recruited. Age


ranges were mostly lower than in neuropathic pain populations.
Denervation (m [s]), n 5 33

These factors were accounted for by normalizing all QST data to


Healthy subjects (m [s])

sex-specific and age-specific reference data.


PH (m [s]), n 5 207
SH (m [s]), n 5 25

Figure 1 shows z-profiles of the 6 models that were chosen for


defining phenotype means. As human surrogate models of NBs,
both nerve compression and topical lidocaine led to substantial
Table 2

loss in thermal detection threshold and MDT (Fig. 1A). For CDT
and MDT in the A-fiber compression block, this loss was almost
complete, reaching a mean z-value beyond 25, that is, beyond 5

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 5

As human surrogate models of PH, topical capsaicin and UVB


sunburn both induced substantial heat and mechanical hyper-
algesia (Fig. 1B). Heat hyperalgesia was more pronounced for
capsaicin, whereas mechanical hyperalgesia was more pro-
nounced for UVB. Capsaicin also induced loss of cold detection
and cold pain.
As human surrogate models of SH, cutaneous HFS and
intradermal capsaicin injection led to mechanical hyperalgesia
and thermal sensory deficits (Fig. 1C). Mechanical hyperalgesia
and DMA were more pronounced in the capsaicin injection
model.
Table 3 compares mean z-scores normalized to published
reference data of healthy subjects with effect sizes (Cohen d) in
intraindividual comparison with untreated control areas. Intra-
individual comparisons mostly confirmed the patterns of negative
or positive sensory signs of z-values, but the loss of thermal
detection for intradermal capsaicin and cutaneous HFS (Fig. 1C)
seems to be overestimated in z-values due to the nonstandard
test area.
The QST profiles of the remaining models are presented in
Figure 2. Figure 2A shows the QST profiles of surrounding skin
from 2 models where SH is either controversial (UVB) or known to
be mild (topical capsaicin). Although the area of SH of topical
capsaicin displays sensory loss in the z-profile, these effects are
much smaller in the intraindividual comparisons with untreated
skin, again suggesting an overestimation in z-values due to the
nonstandard test area (Table 3). Topical menthol has been
introduced to induce cold hyperalgesia, and muscle HFS to
induce hyperalgesia of deep tissue (Fig. 2B); sensory changes in
both models were mild. The combined application of topical
capsaicin and lidocaine (Fig. 2C) shows the combined effects of
lidocaine and capsaicin-induced PH, that is, the capsaicin-
induced mechanical hyperalgesia is abolished, whereas the
thermal and tactile loss is exaggerated.

3.1. Sorting algorithm


By random number assignment, 49% of subjects from the 6
surrogate models with clearly defined mechanism (topical
lidocaine 5%, A-fiber block, topical capsaicin, UVB radiation,
capsaicin injection, and cutaneous HFS) were assigned to the
training data set (n 5 265), which defined mean values and SDs
for the sorting algorithm (Table 2). Individual allocation by the
deterministic sorting (DET) algorithm replicated the a priori
assignment of surrogate models in 79% of the cases for training
set and in 81% of the cases for the test set (remaining 279
Figure 1. Deep sensory profiling of human surrogate models of a priori subjects from the same surrogate models). Cohen kappa
mechanisms. (A) NBs, (B) PH, and (C) SH. Positive z-scores indicate positive coefficient of agreement (scale: 0 5 random classification, 1 5
sensory signs (hyperalgesia) and negative z-values indicate negative sensory perfect agreement between methods) was 0.54 (95% confidence
signs (hypoesthesia and hypoalgesia). Dashed lines: 95% CI for healthy
interval [CI]: 0.43-0.65) for the training set and 0.56 (95% CI: 0.46-
participants (21.96 , z , 11.96). Inserts show numeric pain ratings for DMA
on a logarithmic scale (0-100) and frequency of PHS (0-3). CDT, cold detection 0.67) for the test set; both values may be categorized as “good,”
threshold; CI, confidence interval; CPT, cold pain threshold; DMA, dynamic although no universal guideline for interpreting Cohen kappa
mechanical allodynia; HPT, heat pain threshold; MDT, mechanical detection exists.21 Most common shifts were PH or SH to NB (18% and
threshold; MPS, mechanical pain sensitivity; NB, nerve block; PH, primary 27%, respectively), and least common shifts were NB to SH and
hyperalgesia; PHS, paradoxical heat sensation; PNP, polyneuropathy; PPT,
pressure pain threshold; QST, quantitative sensory testing; SH, secondary
secondary to PH (both ,1%). Shifts between original
hyperalgesia; TSL, temperature sensory limen; VDT, vibration detection and algorithmic assignment are shown in Table 4.
threshold; WDT, warm detection threshold; WUR, wind-up ratio.

3.2. Deterministic and probabilistic sorting


SDs of normal detection, whereas for topical lidocaine, z-values The z-profiles of each sensory subgroup are shown in Figure 3.
were around 22 indicating highly significant but partial sensory Forty-one subjects showed a sensory profile that was most
loss. Unlike topical lidocaine, selective A-fiber block was also similar to untreated healthy skin, although part of a surrogate
associated with signs of sensory gain, namely the frequent model (Fig. 3A). The subjects sorted to the NB profile were mostly
occurrence of pinprick hyperalgesia and PHSs. characterized by loss of cold and mechanical detection (CDT and

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
6
·
J. Vollert et al. 0 (2018) 1–13 PAIN®

Table 3
z-scores and effect sizes (Cohen d) of human surrogate models.

Blue indicates loss of function and red indicates gain of function, for z-scores: boundaries 1.0 and 1.96, for Cohen d values: boundaries above 0.5 and 0.8.
CDT, cold detection threshold; CPT, cold pain threshold; HFS, high-frequency stimulation; HPT, heat pain threshold; MDT, mechanical detection threshold; MPS, mechanical pain sensitivity; MPT, mechanical pain threshold;
PPT, pressure pain threshold; TSL, temperature sensory limen; UVB, ultraviolet B; VDT, vibration detection threshold; WDT, warm detection threshold; WUR; wind-up ratio.

MDT, representing A-delta and A-beta fiber function, respec- frequency of the NB profile was found in cutaneous HFS and in
tively, Fig. 3B). Loss of warm detection was less pronounced the menthol model.
compared with cold detection because the topical lidocaine Quantitative sensory testing profiles compatible with PH
block was less effective in comparison with the A-fiber block. increased from 380 (58%) in the deterministic to 470 (72%) in
Loss of vibration detection was not detectable, which is due to the probabilistic algorithm. Beyond the defining models, this
a limitation of the A-fiber block, which only affects a small area, profile was frequent in the skin area surrounding the UVB
whereas vibration is poorly localized and can be sensed by rapidly radiation model (where profiles compatible with SH were rare),
adapting mechanoreceptors (Pacinian corpuscles) situated topical menthol, and muscular HFS.
beyond this area. Primary hyperalgesia (Fig. 3B) was character- Quantitative sensory testing profiles consistent with SH were
ized by pronounced heat hyperalgesia and moderate mechanical more than twice as frequent in the probabilistic sorting (n 5 134;
hyperalgesia, whereas SH presented most prominently by 20%) than in the DET (n 5 58; 9%). Beyond the defining capsaicin
mechanical hyperalgesia and some loss of thermal detection. injection and cutaneous HFS, this profile was only found in the
Figure 4 (6 defining models) and Figure 5 (5 additional models) skin area surrounding topical capsaicin in a relevant frequency
show results of deterministic and probabilistic sorting into the 3 (deterministic: 27% and probabilistic: 65%) confirming that SH
prototypic mechanistic profiles. In the probabilistic sorting occurs also in this model.
algorithm, the percentage of sensory profiles compatible with
being from normal skin increased to 187 (28%) vs 41 (6%) in the
deterministic version of the algorithm. The highest frequency of 3.3. Comparison with sensory phenotypes in patients
healthy profiles was found in muscular HFS and topical menthol, Figure 5F shows DET and probabilistic sorting results for 902
which also had the mildest sensory changes in their averaged neuropathic pain patients that were the basis of the original
QST profiles (Fig. 2B). cluster analysis (Baron et al. 2017), and Table 5 compares
Quantitative sensory testing profiles compatible with the NB mechanistic sorting results with original heuristic cluster alloca-
profile increased to 352 cases (54%) in the probabilistic algorithm tion. In 65% of the cases, patients from the sensory loss
vs 178 (27%) in the deterministic version. Apart from the defining phenotype were sorted to NB, patients from the thermal
models of A-fiber block and topical lidocaine, the highest hyperalgesia phenotype to the PH pattern, and patients from

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 7

agreement between methods) of 0.44 (95% CI: 0.28-0.60),


showing a substantial degree of agreement.21 The most
prominent shifts were from the 2 hyperalgesia phenotypes in
the heuristic patient clustering into the mechanistic phenotype
consistent with NB.
Interestingly, in the probabilistic sorting (Table 5), fewer
patients were sorted to the NB profile (512 vs 600 in DET),
whereas relatively more patients were sorted to PH (271 vs 186 in
DET) or SH (198 vs 116 in DET). But, the most prominent shift was
still towards the profiles consistent with NBs. Still, only one-third
of the patients could be uniquely assigned to 1 of the 3
mechanisms. Another third of the patient QST profiles was
consistent with multiple mechanisms (n 5 282), most frequently
a combination including the NB phenotype. The last third was not
sufficiently distinct from a normal skin QST phenotype to be
assigned to any mechanism (n 5 279).
Figure 6 illustrates the similarity in QST profiles grouped either
according to the original cluster analysis (Fig. 6A) or the
mechanistic sorting (Fig. 6B). Although the profiles of patients
sorted according to both algorithms are highly similar, the profiles
of patients and surrogate models in the mechanistic sorting
display some interesting differences (compare Fig. 3 with Fig.
6B): patients sorted to the NB exhibit hypoalgesia, whereas in
surrogate models, pain thresholds are normal or show mild gain.
Patients sorted to NB also have loss of vibration detection
threshold, which could not be modeled in the A-fiber block. For
PH, the models display almost no cold, but an isolated heat
hyperalgesia, whereas patients show both cold and heat hyper-
algesia. For SH, participants under surrogate models have more
gain in mechanical pain threshold than in MPS, which is reversed
in patients.

4. Discussion
These multicenter data confirmed principal properties of NB
(sensory loss), PH (heat hyperalgesia), and SH (pinprick hyper-
algesia) that have been already reported in human and animal
experiments.8,35,42,47,55 They also revealed additional sensory
alterations in these human surrogate models that had not been
previously described, that is, sensory gain in NBs and sensory
loss in hyperalgesia models. This probably reflects a selection or
reporting bias in previous studies that had not assessed these
sensory functions.
In a randomized split-half analysis, a sorting algorithm pre-
viously validated for sensory profiles of neuropathic pain
patients60 led to reproducible sorting of surrogate model sensory
profiles into patterns defined a priori according to known
mechanisms. Sorting of 902 neuropathic pain patients into these
Figure 2. Deep sensory profiling of additional human surrogate models. (A) SH
surrounding topical capsaicin or UVB-treated areas, (B) topical menthol (cold
mechanistic phenotypes led to a similar distribution as the original
hyperalgesia model) and muscle HFS (deep pain model), and (C) sequentially heuristic clustering,5,60 laying the basis for a mechanism-based
combined topical application of lidocaine and capsaicin. Positive z-scores treatment approach.64
indicate positive sensory signs (hyperalgesia) and negative z-values indicate
negative sensory signs (hypoesthesia and hypoalgesia). Dashed lines: 95% CI for
healthy participants (21.96 , z , 11.96). Inserts show numeric pain ratings for 4.1. Nerve blocks as human surrogate model of transient
DMA on a logarithmic scale (0-100) and frequency of PHS (0-3). CDT, cold functional denervation
detection threshold; CI, confidence interval; CPT, cold pain threshold; DMA,
dynamic mechanical allodynia; HFS, high-frequency stimulation; HPT, heat pain The sensory profile of human surrogate models for transient
threshold; MDT, mechanical detection threshold; MPS, mechanical pain functional denervation (ie, A-fiber compression block and topical
sensitivity; PHS, paradoxical heat sensation; PNP, polyneuropathy; PPT,
pressure pain threshold; QST, quantitative sensory testing; SH, secondary
lidocaine) was characterized by pronounced loss in thermal and
hyperalgesia; TSL, temperature sensory limen; UVB, ultraviolet B; VDT, vibration mechanical detection combined with PHS. These patterns have
detection threshold; WDT, warm detection threshold; WUR, wind-up ratio. been reported in previous single-center studies.34,65 Consistent
with its selective effect on myelinated nerve fibers, compression
the mechanical hyperalgesia phenotype to the SH QST pattern NB had larger effects on MDT and CDT than on warm detection
(Table 5). This corresponded to a Cohen kappa coefficient of threshold.28,65 Pinprick stimuli are often perceived as less painful
agreement (scale: 0 5 random classification, 1 5 perfect in complete A-fiber block,28,65 whereas mild pinprick

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
8
·
J. Vollert et al. 0 (2018) 1–13 PAIN®

Table 4
Cross-tabulation of dominant phenotype identified using the algorithm (columns) vs the original assignment (rows) for training
and test data set (before and after slash).
Algorithm
Original Nerve blocks Primary hyperalgesia Secondary hyperalgesia
Nerve blocks 29/30 4/2 0/0
Primary hyperalgesia 35/42 162/178 10/4
Secondary hyperalgesia 8/5 0/1 17/17
Overall, concordance of the algorithm to original assignment was strong (79% of the cases for the training and 81% for the test data set, note that the even higher concordance for the test data set compared with training data
indicates a very robust sorting).

hyperalgesia has been associated with preserved A-delta fiber probably due to sensitization of TRPV1 and TRPA1
function2,29 Topical lidocaine had mild effects compared with receptors.38,39,49,63
complete conduction block by regional anesthesia.22,33 Interest-
ingly, in contrast to the present data, intradermal lidocaine 0.01%
4.2. Primary hyperalgesia as human surrogate model of
and 0.1% was reported to induce also a heat hyperalgesia
peripheral sensitization
The profile of surrogate models of peripheral sensitization (topical
capsaicin or UVB radiation) was characterized by pronounced
hyperalgesia to heat, pressure, and pinprick pain. Primary
nociceptive afferents are easily sensitized to heat stimuli, but much
less so to von-Frey or pinprick stimuli.55 Peripheral sensitization to
heat may be explained by phosphorylation of the heat-gated cation
channel TRPV1,58 but has also been shown to be induced by
TRPA1 agonist allyl-isothiocyanate.1 Although peripheral sensiti-
zation to blunt pressure has been reported before,31 pinprick
hyperalgesia in these models may indicate additional central
sensitization induced by enhanced peripheral nociceptive input to
the spinal cord. Sensory loss occurred mostly in the topical
capsaicin model and was restricted to thermal detection thresh-
olds; this has been discussed previously10,16 and may reflect
desensitization by the TRPV1 agonist capsaicin, which is the
intended mode of action clinically.26

4.3. Secondary hyperalgesia as human surrogate model of


central sensitization
Human surrogate models of central sensitization (intradermal
capsaicin and cutaneous electrical HFS) were characterized by
pronounced pinprick hyperalgesia, but also pronounced thermal
sensory loss. Combined studies in monkeys and humans using
intradermal capsaicin had shown pronounced increases in both
wide dynamic-range and high-threshold spinal neuron output
despite unchanged A- and C-nociceptor input.6,54 Tactile
sensory loss has been reported in human surrogate models of
SH and also in patients with pinprick hyperalgesia.23 An inverse
spinal gate with small fiber input inhibiting processing of large fiber
input46,66 has been suggested as a mechanism. Thermal sensory
loss is a new finding, suggesting that sensory loss in chronic pain
patients does not necessarily have to be due to structural
Figure 3. Average profiles of 657 healthy subjects who had undergone human changes, but may also be a functional sign, and hence potentially
surrogate models stratified on prototypic mechanism-based sorting. (A)
Healthy profiles and (B) NBs vs PH vs SH profiles. Positive z-scores indicate
sensitive to analgesic treatment regimes.
positive sensory signs (hyperalgesia) and negative z-values indicate negative
sensory signs (hypoesthesia and hypoalgesia). Dashed lines: 95% CI for
healthy participants (21.96 , z , 11.96). Inserts show numeric pain ratings 4.4. Other human surrogate models
for DMA on a logarithmic scale (0-100) and frequency of PHS (0-3). CDT, cold
The presence of a zone of SH surrounding topical capsaicin and
detection threshold; CI, confidence interval; CPT, cold pain threshold; DMA,
dynamic mechanical allodynia; HPT, heat pain threshold; MDT, mechanical UVB is controversial in the published literature.8,33,40,41 Our data
detection threshold; MPS, mechanical pain sensitivity; NB, nerve block; PH, confirmed sensory profiles compatible with SH surrounding
primary hyperalgesia; PHS, paradoxical heat sensation; PNP, polyneuropathy, topical capsaicin but not UVB, which mostly had characteristics
PPT, pressure pain threshold; QST, quantitative sensory testing; SH, of PH or NB instead. These findings suggest that topical
secondary hyperalgesia; TSL, temperature sensory limen; VDT, vibration
detection threshold; WDT, warm detection threshold; WUR, wind-up ratio.
capsaicin is useful to study both PH and SH, whereas the UVB
model seems to be limited to PH.

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 9

Figure 5. Deterministic and probabilistic sorting of additional human


Figure 4. Deterministic and probabilistic sorting of the defining human surrogate models and patients suffering from peripheral neuropathic pain.
surrogate models. Deterministic sorting (DET): each profile is sorted to its best Deterministic sorting (DET): each profile is sorted to its best fit (this column
fit (this column always adds up to 100%); probabilistic sorting (the following 4 always adds up to 100%); probabilistic sorting (the following 4 columns,
columns, each profile can fall in multiple categories, so that these columns are each profile can fall in multiple categories, so that these columns are not
not additive): each profile is sorted to each mechanism for which it reaches additive): each profile is sorted to each mechanism for which it reaches
a probability above 64%, as determined as optimal cut-off in an ROC analysis a probability above 64%, as determined as optimal cut-off in an ROC
of patients suffering from neuropathic pain and healthy subjects in Vollert et al., analysis of patients suffering from neuropathic pain and healthy subjects in
2017. H (grey), normal healthy skin, NB (blue), nerve block; PH (red), primary Vollert et al., 2017. Patient data (F) are from Baron et al. 2017. H (grey),
hyperalgesia; SH (yellow), secondary hyperalgesia. ROC, receiver operating normal healthy skin; NB (blue), nerve block; PH (red), primary hyperalgesia;
curve. SH (yellow), secondary hyperalgesia.

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
10
·
J. Vollert et al. 0 (2018) 1–13 PAIN®

Table 5 The mechanical hyperalgesia patient phenotype shows a strong


Cross-tabulation of dominant phenotype identified using the overlap with surrogate models of SH, which supports an
deterministic algorithm developed in patients (columns) vs the interpretation of this phenotype to be a phenotype of reorganization
algorithm developed based on surrogate models (rows) for or central sensitization. Substantial thermal sensory loss suggests
n 5 902 patients. that also damaged nociceptors are involved, generating ongoing
pain and inducing central sensitization.4
Sensory Thermal Mechanical
loss hyperalgesia hyperalgesia The sensory loss patient phenotype shows a strong overlap
(n 5 (n 5 267) (n 5 279) with experimental NBs. These blocks were frequently used as
356) tools to identify normal sensory function of fiber classes (A vs C),
Deterministic but not yet widely recognized as mimicking aspects of
NB (n 5 600) 337 97 166 neuropathic pain.7,33 Both the clinical phenotype and the
PH (n 5 186) 6 156 24 surrogate models are dominated by loss of small and large fiber
SH (n 5 116) 13 14 89 functions. This supports an interpretation as denervation or
Probabilistic
deafferentation, where central neurons may develop denervation
NB (n 5 512) 202 130 180 super-sensitivity to other inputs.14
PH (n 5 271) 12 163 96 Roughly one-third of the patients were assigned a NB phenotype
SH (n 5 198) 38 26 134 according to the surrogate models vs one of the hyperalgesia
phenotypes in the heuristic patient cluster analysis. This systematic
Of the above: patients sorted to
a single mechanism (n 5 341) shift is consistent with the presence of partial nerve damage in most
NB, n 5 247 163 30 54 of the neuropathic pain conditions that may lead to coexistence of
PH, n 5 62 1 58 3 denervation and sensitization of the remaining pathways.11 These
SH, n 5 32 3 3 26 data cover, however, only peripheral neuropathic pain, and to this
Of the above: patients sorted to
point, we cannot make any extrapolations onto, for example,
multiple mechanisms central pain, nociceptive pain, or deep pains.
(n 5 282)
NB 1 PH (n 5 116) 4 84 28
4.6. Limitations and technical considerations
NB 1 SH (n 5 73) 28 2 43
PH 1 SH (n 5 17) 0 7 10 The validity of the prototypic mechanistic profiles is partly
NB 1 PH 1 SH (n 5 76) 7 14 55 demonstrated by the accuracy of allocation of the training set
Patients sorted to no 150 69 60 data to the appropriate a priori profile of roughly 80%. This shows
mechanism (n 5 279) general robustness, but also that the deterministic algorithm is
In the clear majority of the cases (65%), both algorithms concurred in the deterministic variant. not perfect and should be used with some caution on individual
NB, nerve block; PH, primary hyperalgesia; SH, secondary hyperalgesia. basis. For a better separation of mechanism-specific sensory
profiles from idiosyncrasies of individual experimental models,
Muscle HFS was introduced as a model for plasticity of deep more human surrogate models of PH and SH (eg, burn injury and
pain, and the overlying skin did not exhibit profiles compatible surgical incision) and of temporary functional denervation (eg,
with SH. Our data were from the lower back; more pronounced limb ischemia and capsaicin-induced defunctionalization) should
effects might occur when other muscles, for example, in the lower be studied using the methods of this article.
limb or face are stimulated. The models in this analysis do not explicitly cover the
Topical menthol had only mild effects with some character- endogenous pain-modulating systems30 nor ectopic activity.
istics of NBs and PH. Its main value is the pronounced cold Descending modulation may contribute to the SH phenotype, but
hyperalgesia, but the current data do not make a strong might also exhibit yet another sensory profile. Ectopic activity may
suggestion as to whether this is due to peripheral or central have been present in any of our models, but we have neither
sensitization or both. positive nor negative evidence about it.
The sensory profile of topical lidocaine plus topical capsaicin The percentage of sensory profiles compatible with normal
displayed an almost perfect combination profile of lidocaine block variability was higher in the surrogate models than in the patients
and capsaicin-induced PH, but it lacks the distinct mechanical (29% vs 20%). We think that this is due to the fact that for ethical
hyperalgesia characterizing SH. This finding suggests that reasons, all manipulations in humans were relatively minor (as
sensory profiling may be able to identify combined mechanisms compared to patients and animal models50).
and to distinguish between contributions of peripheral vs central
sensitization also in patients with neuropathic pain.
4.7. Summary and conclusions
With these prototypical sensory profiles for 3 predefined
4.5. Mechanistic significance of heuristically found mechanisms (denervation, peripheral, and central sensitization),
patient phenotypes
we provide a well-studied mechanistic background for our
The sorting according to surrogate model profiles was applied to previously described heuristic sensory phenotype clusters. Using
our previously published patient data, which leads to a similar the probabilistic sorting according to human surrogate model
distribution as the original heuristic clustering. This supports profiles, patients suffering from neuropathic pain can be
previous mechanistic interpretations of the clinically found tentatively stratified in future studies to presumed underlying
phenotypes: the thermal hyperalgesia patient phenotype shows mechanisms. The European Medicines Agency encourages such
strong overlap with surrogate models of PH. This supports an approach in its new guideline17 as a step towards increasing
previous interpretations as irritable nociceptor18 and peripheral response rates in clinical trials by a mechanism-based treatment
sensitization.57 Both evoked and ongoing pain is likely to be due approach towards neuropathic pain.64 It should be noted,
to surviving nociceptors in these patients. however, that the 3 classes of human surrogate models studied

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 11

which are composed of financial contribution from the European


Union’s seventh framework programme (FP7/2007–2013) and
European Federation of Pharmaceutical Industries and Associ-
ations (EFPIA) companies’ in kind contribution. The NEUROPAIN
project is an investigator-initiated European multicentre study
with R. Baron as principal investigator and 10 coinvestigator sites,
supported by an independent research grant from Pfizer Ltd.

Acknowledgements
The authors thank all consortia for building up the basis of this
study by patient and subject recruitment and assessment.

Article history:
Received 9 October 2017
Received in revised form 26 January 2018
Accepted 14 February 2018
Available online 24 February 2018

References
[1] Andersen HH, Lo Vecchio S, Gazerani P, Arendt-Nielsen L. A dose-
response study of topical allyl-isothiocyanate (mustard oil) as human
surrogate model of pain, hyperalgesia, and neurogenic inflammation.
PAIN 2017;158:1723–32.
[2] Andrew D, Greenspan JD. Peripheral coding of tonic mechanical
cutaneous pain: comparison of nociceptor activity in rat and human
psychophysics. J Neurophysiol 1999;82:2641–8.
[3] Backonja MM, Attal N, Baron R, Bouhassira D, Drangholt M, Dyck PJ,
Edwards RR, Freeman R, Gracely R, Haanpaa MH, Hansson P, Hatem
SM, Krumova EK, Jensen TS, Maier C, Mick G, Rice AS, Rolke R, Treede
RD, Serra J, Toelle T, Tugnoli V, Walk D, Walalce MS, Ware M, Yarnitsky
D, Ziegler D. Value of quantitative sensory testing in neurological and pain
Figure 6. Average profiles of 902 patients with peripheral neuropathic pain disorders: NeuPSIG consensus. PAIN 2013;154:1807–19.
sorted either according to the deterministic algorithm derived from heuristic [4] Baron R, Hans G, Dickenson AH. Peripheral input and its importance for
patient clusters (A) or according to the deterministic version of the algorithm central sensitization. Ann Neurol 2013;74:630–6.
derived from human surrogate models (B). Positive z-scores indicate positive [5] Baron R, Maier C, Attal N, Binder A, Bouhassira D, Cruccu G, Finnerup
sensory signs (hyperalgesia) and negative z-values indicate negative sensory NB, Haanpaa M, Hansson P, Hullemann P, Jensen TS, Freynhagen R,
signs (hypoesthesia and hypoalgesia). Dashed lines: 95% CI for healthy Kennedy JD, Magerl W, Mainka T, Reimer M, Rice ASC, Segerdahl M,
participants (21.96 , z , 11.96). Inserts show numeric pain ratings for DMA Serra J, Sindrup S, Sommer C, Tolle T, Vollert J, Treede RD. Peripheral
on a logarithmic scale (0-100) and frequency of PHS (0-3). CDT, cold detection neuropathic pain: a mechanism-related organizing principle based on
threshold; CI, confidence interval; CPT, cold pain threshold; DMA, dynamic sensory profiles. PAIN 2017;158:261–72.
mechanical allodynia; HPT, heat pain threshold; MDT, mechanical detection [6] Baumann TK, Simone DA, Shain CN, LaMotte RH. Neurogenic
threshold; MPS, mechanical pain sensitivity; PHS, paradoxical heat sensation; hyperalgesia: the search for the primary cutaneous afferent fibers that
PNP, polyneuropathy, PPT, pressure pain threshold; QST, quantitative sensory contribute to capsaicin-induced pain and hyperalgesia. J Neurophysiol
testing; TSL, temperature sensory limen; VDT, vibration detection threshold; 1991;66:212–27.
WDT, warm detection threshold; WUR, wind-up ratio. Patient data from Baron [7] Baumgartner U, Magerl W, Klein T, Hopf HC, Treede RD. Neurogenic
et al., 2017, heuristic sorting algorithm according to Vollert et al., 2017. hyperalgesia versus painful hypoalgesia: two distinct mechanisms of
neuropathic pain. PAIN 2002;96:141–51.
[8] Binder A. Human surrogate models of neuropathic pain: validity and
limitations. PAIN 2016;157(suppl 1):S48–52.
here likely represent combined rather than single mechanisms. [9] Binder A, Stengel M, Klebe O, Wasner G, Baron R. Topical high-
concentration (40%) menthol-somatosensory profile of a human
This, however, is likely true also for studies in awake behaving
surrogate pain model. J Pain 2011;12:764–73.
animals. Therefore, a reverse translation approach may be useful [10] Callsen MG, Moller AT, Sorensen K, Jensen TS, Finnerup NB. Cold
for developing novel analgesic medications: they should initially hyposensitivity after topical application of capsaicin in humans. Exp Brain
be tested in animal models of NB, PH, or SH. Medications Res 2008;191:447–52.
effective on these phenotypes can easily be validated in human [11] Campbell JN, Meyer RA. Mechanisms of neuropathic pain. Neuron 2006;
52:77–92.
surrogate models and then transferred to subgroups of neuro- [12] Cohen J. A power primer. Psychol Bull 1992;112:155–9.
pathic pain patients. [13] Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed.
Hoboken: Taylor and Francis, 2013.
[14] Colloca L, Ludman T, Bouhassira D, Baron R, Dickenson AH, Yarnitsky D,
Conflict of interest statement Freeman R, Truini A, Attal N, Finnerup NB, Eccleston C, Kalso E, Bennett
DL, Dworkin RH, Raja SN. Neuropathic pain. Nat Rev Dis Primers 2017;3:
The authors have no financial or other relationships that might 17002.
lead to a conflict of interest. [15] Edwards RR, Dworkin RH, Turk DC, Angst MS, Dionne R, Freeman R,
This study was supported by the German Research Network Hansson P, Haroutounian S, Arendt-Nielsen L, Attal N, Baron R, Brell J,
on Neuropathic Pain (DFNS). Bujanover S, Burke LB, Carr D, Chappell AS, Cowan P, Etropolski M,
Fillingim RB, Gewandter JS, Katz NP, Kopecky EA, Markman JD,
The EUROPAIN project is a public-private partnership and has Nomikos G, Porter L, Rappaport BA, Rice ASC, Scavone JM, Scholz J,
received support from the Innovative Medicines Initiative Joint Simon LS, Smith SM, Tobias J, Tockarshewsky T, Veasley C, Versavel M,
Undertaking under grant agreement no 115007, resources for Wasan AD, Wen W, Yarnitsky D. Patient phenotyping in clinical trials of

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
12
·
J. Vollert et al. 0 (2018) 1–13 PAIN®

chronic pain treatments: IMMPACT recommendations. PAIN 2016;157: sensitized by local anesthetics in rodent sensory neurons. J Clin Invest
1851–71. 2008;118:763–76.
[16] Enax-Krumova EK, Pohl S, Westermann A, Maier C. Ipsilateral and [39] Leffler A, Lattrell A, Kronewald S, Niedermirtl F, Nau C. Activation of
contralateral sensory changes in healthy subjects after experimentally TRPA1 by membrane permeable local anesthetics. Mol Pain 2011;7:62.
induced concomitant sensitization and hypoesthesia. BMC Neurol 2017; [40] Lotsch J, Dimova V, Hermens H, Zimmermann M, Geisslinger G, Oertel
17:60. BG, Ultsch A. Pattern of neuropathic pain induced by topical capsaicin
[17] European Medicines Agency. EMA/CHMP/970057/2011. Guideline on application in healthy subjects. PAIN 2015;156:405–14.
the clinical development of medicinal products intended for the treatment [41] Lotsch J, Dimova V, Ultsch A, Lieb I, Zimmermann M, Geisslinger G,
of pain, 2016. Available at: http://www.ema.europa.eu/docs/en_GB/ Oertel BG. A small yet comprehensive subset of human experimental pain
document_library/Scientific_guideline/2016/12/WC500219131.pdf. models emerging from correlation analysis with a clinical quantitative
Accessed December 15, 2016. sensory testing protocol in healthy subjects. Eur J Pain 2016;20:777–89.
[18] Fields HL, Rowbotham M, Baron R. Postherpetic neuralgia: irritable [42] Lotsch J, Oertel BG, Ultsch A. Human models of pain for the prediction of
nociceptors and deafferentation. Neurobiol Dis 1998;5:209–27. clinical analgesia. PAIN 2014;155:2014–21.
[19] Finnerup NB, Haroutounian S, Kamerman P, Baron R, Bennett DLH, [43] Magerl W, Fuchs PN, Meyer RA, Treede RD. Roles of capsaicin-
Bouhassira D, Cruccu G, Freeman R, Hansson P, Nurmikko T, Raja SN, insensitive nociceptors in cutaneous pain and secondary hyperalgesia.
Rice ASC, Serra J, Smith BH, Treede RD, Jensen TS. Neuropathic pain: Brain 2001;124:1754–64.
an updated grading system for research and clinical practice. PAIN 2016; [44] Magerl W, Krumova EK, Baron R, Tolle T, Treede RD, Maier C. Reference
157:1599–606. data for quantitative sensory testing (QST): refined stratification for age
[20] Fimer I, Klein T, Magerl W, Treede RD, Zahn PK, Pogatzki-Zahn EM. and a novel method for statistical comparison of group data. PAIN 2010;
Modality-specific somatosensory changes in a human surrogate model of 151:598–605.
postoperative pain. Anesthesiology 2011;115:387–97. [45] Maier C, Baron R, Tolle TR, Binder A, Birbaumer N, Birklein F,
[21] Fleiss JL, Levin B, Paik MC. Statistical methods for rates and proportions. Gierthmuhlen J, Flor H, Geber C, Huge V, Krumova EK,
3rd ed. Hoboken: Wiley-Interscience, 2003. Landwehrmeyer GB, Magerl W, Maihofner C, Richter H, Rolke R,
[22] Gandevia SC, Phegan CML. Perceptual distortions of the human body Scherens A, Schwarz A, Sommer C, Tronnier V, Uceyler N, Valet M,
image produced by local anaesthesia, pain and cutaneous stimulation. Wasner G, Treede RD. Quantitative sensory testing in the German
J Physiol 1999;514:609–16. Research Network on Neuropathic Pain (DFNS): somatosensory
[23] Geber C, Magerl W, Fondel R, Fechir M, Rolke R, Vogt T, Treede RD, abnormalities in 1236 patients with different neuropathic pain
Birklein F. Numbness in clinical and experimental pain–a cross-sectional syndromes. PAIN 2010;150:439–50.
study exploring the mechanisms of reduced tactile function. PAIN 2008; [46] Mendell LM. Constructing and deconstructing the gate theory of pain.
139:73–81. PAIN 2014;155:210–6.
[24] Gierthmuhlen J, Enax-Krumova EK, Attal N, Bouhassira D, Cruccu G, [47] Ochoa JL, Campero M, Serra J, Bostock H. Hyperexcitable polymodal
Finnerup NB, Haanpaa M, Hansson P, Jensen TS, Freynhagen R, and insensitive nociceptors in painful human neuropathy. Muscle Nerve
Kennedy JD, Mainka T, Rice ASC, Segerdahl M, Sindrup SH, Serra J, 2005;32:459–72.
Tolle T, Treede RD, Baron R, Maier C. Who is healthy? Aspects to [48] Pfau DB, Krumova EK, Treede RD, Baron R, Toelle T, Birklein F, Eich W,
consider when including healthy volunteers in QST–based studies- Geber C, Gerhardt A, Weiss T, Magerl W, Maier C. Quantitative sensory
a consensus statement by the EUROPAIN and NEUROPAIN consortia. testing in the German Research Network on Neuropathic Pain (DFNS):
PAIN 2015;156:2203–11. reference data for the trunk and application in patients with chronic
[25] Gustorff B, Sycha T, Lieba-Samal D, Rolke R, Treede RD, Magerl W. The postherpetic neuralgia. PAIN 2014;155:1002–15.
pattern and time course of somatosensory changes in the human UVB [49] Piao LH, Fujita T, Jiang CY, Liu T, Yue HY, Nakatsuka T, Kumamoto E.
sunburn model reveal the presence of peripheral and central TRPA1 activation by lidocaine in nerve terminals results in glutamate
sensitization. PAIN 2013;154:586–97. release increase. Biochem Biophys Res Commun 2009;379:980–4.
[26] Hayman M, Kam PCA. Capsaicin: a review of its pharmacology and [50] Reitz MC, Hrncic D, Treede RD, Caspani O. A comparative behavioural
clinical applications. Curr Anaesth Crit Care 2008;19:338–43. study of mechanical hypersensitivity in 2 pain models in rats and humans.
[27] von Hehn CA, Baron R, Woolf CJ. Deconstructing the neuropathic pain PAIN 2016;157:1248–58.
phenotype to reveal neural mechanisms. Neuron 2012;73:638–52. [51] Rolke R, Baron R, Maier C, Tolle TR, Treede RD, Beyer A, Binder A,
[28] Henrich F, Magerl W, Klein T, Greffrath W, Treede RD. Capsaicin- Birbaumer N, Birklein F, Botefur IC, Braune S, Flor H, Huge V, Klug R,
sensitive C- and A-fibre nociceptors control long-term potentiation-like Landwehrmeyer GB, Magerl W, Maihofner C, Rolko C, Schaub C,
pain amplification in humans. Brain 2015;138:2505–20. Scherens A, Sprenger T, Valet M, Wasserka B. Quantitative sensory
[29] Jørum E, Warncke T, Ziegler EA, Magerl W, Fuchs PN, Meyer R, Treede testing in the German Research Network on Neuropathic Pain
RD. Secondary hyperalgesia to punctate stimuli is mediated by A-fiber (DFNS): standardized protocol and reference values. PAIN 2006;
nociceptors. In: Devor M, Rowbotham M, Wiesenfeld-Hallin Z, editors. 123:231–43.
Proceedings of the 9th World Congress on Pain, Progr Pain Res [52] Rolke R, Magerl W, Campbell KA, Schalber C, Caspari S, Birklein F,
Management. Seattle: IASP Press, 2000. p. 215–223. Treede RD. Quantitative sensory testing: a comprehensive protocol for
[30] Kennedy DL, Kemp HI, Ridout D, Yarnitsky D, Rice ASC. Reliability of clinical trials. Eur J Pain 2006;10:77–88.
conditioned pain modulation: a systematic review. PAIN 2016;157: [53] Schilder A, Magerl W, Hoheisel U, Klein T, Treede RD. Electrical high-
2410–9. frequency stimulation of the human thoracolumbar fascia evokes long-
[31] Kilo S, Schmelz M, Koltzenburg M, Handwerker HO. Different patterns of term potentiation-like pain amplification. PAIN 2016;157:2309–17.
hyperalgesia induced by experimental inflammation in human skin. Brain [54] Simone DA, Sorkin LS, Oh U, Chung JM, Owens C, LaMotte RH, Willis
1994;117:385–96. WD. Neurogenic hyperalgesia: central neural correlates in responses of
[32] Klein T, Magerl W, Hopf HC, Sandkühler J, Treede RD. Perceptual spinothalamic tract neurons. J Neurophysiol 1991;66:228–46.
correlates of nociceptive long-term potentiation and long-term [55] Treede RD, Meyer RA, Raja SN, Campbell JN. Peripheral and central
depression in humans. J Neurosci 2004;24:964–71. mechanisms of cutaneous hyperalgesia. Prog Neurobiol 1992;38:
[33] Klein T, Magerl W, Rolke R, Treede RD. Human surrogate models of 397–421.
neuropathic pain. PAIN 2005;115:227–33. [56] Treede RD, Jensen TS, Campbell JN, Cruccu G, Dostrovsky JO, Griffin
[34] Krumova EK, Zeller M, Westermann A, Maier C. Lidocaine patch (5%) JW, Hansson P, Hughes R, Nurmikko T, Serra J. Neuropathic pain:
produces a selective, but incomplete block of Adelta and C fibers. PAIN redefinition and a grading system for clinical and research purposes.
2012;153:273–80. Neurology 2008;70:1630–5.
[35] LaMotte RH, Lundberg LE, Torebjörk HE. Pain, hyperalgesia and activity [57] Truini A, Biasiotta A, Di Stefano G, La Cesa S, Leone C, Cartoni C,
in nociceptive C units in humans after intradermal injection of capsaicin. Leonetti F, Casato M, Pergolini M, Petrucci MT, Cruccu G. Peripheral
J Physiol 1992;448:749–64. nociceptor sensitization mediates allodynia in patients with distal
[36] LaMotte RH, Shain CN, Simone DA, Tsai EF. Neurogenic hyperalgesia: symmetric polyneuropathy. J Neurol 2013;260:761–6.
psychophysical studies of underlying mechanisms. J Neurophysiol 1991; [58] Voets T, Droogmans G, Wissenbach U, Janssens A, Flockerzi V, Nilius B.
66:190–211. The principle of temperature-dependent gating in cold- and heat-
[37] Lang S, Klein T, Magerl W, Treede RD. Modality-specific sensory changes sensitive TRP channels. Nature 2004;430:748–54.
in humans after the induction of long-term potentiation (LTP) in cutaneous [59] Vollert J, Attal N, Baron R, Freynhagen R, Haanpaa M, Hansson P,
nociceptive pathways. PAIN 2007;128:254–63. Jensen TS, Rice ASC, Segerdahl M, Serra J, Sindrup SH, Tolle TR,
[38] Leffler A, Fischer MJ, Rehner D, Kienel S, Kistner K, Sauer SK, Gavva NR, Treede RD, Maier C. Quantitative sensory testing using DFNS protocol in
Reeh PW, Nau C. The vanilloid receptor TRPV1 is activated and Europe: an evaluation of heterogeneity across multiple centers in patients

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.
Month 2018
· Volume 0
· Number 0 www.painjournalonline.com 13

with peripheral neuropathic pain and healthy subjects. PAIN 2016;157: [62] Wasner G, Schattschneider J, Binder A, Baron R. Topical menthol–a
750–8. human model for cold pain by activation and sensitization of C
[60] Vollert J, Maier C, Attal N, Bennett DLH, Bouhassira D, Enax-Krumova nociceptors. Brain 2004;127:1159–71.
EK, Finnerup NB, Freynhagen R, Gierthmühlen JJ, Haanpää M, Hansson [63] Weinkauf B, Obreja O, Schmelz M, Rukwied R. Differential effects of
P, Hüllemann P, Jensen TS, Magerl W, Ramirez JD, Rice ASC, Schuh- lidocaine on nerve growth factor (NGF)-evoked heat- and mechanical
Hofer S, Segerdahl M, Serra J, Shillo PR, Sindrup S, Tesfaye S,
hyperalgesia in humans. Eur J Pain 2012;16:543–9.
Themistocleous AC, Tölle TRTR, Treede RD, Baron R. Stratifying patients
[64] Woolf CJ, Bennett GJ, Doherty M, Dubner R, Kidd B, Koltzenburg M,
with peripheral neuropathic pain based on sensory profiles. PAIN 2017;
Lipton R, Loeser JD, Payne R, Torebjork E. Towards a mechanism-based
158:1446–55.
[61] Vollert J, Mainka T, Baron R, Enax-Krumova EK, Hullemann P, Maier C, classification of pain? PAIN 1998;77:227–9.
Pfau DB, Tolle T, Treede RD. Quality assurance for Quantitative Sensory [65] Ziegler EA. Secondary hyperalgesia to punctate mechanical stimuli: central
Testing laboratories: development and validation of an automated sensitization to A-fibre nociceptor input. Brain 1999;122:2245–57.
evaluation tool for the analysis of declared healthy samples. PAIN 2015; [66] Zimmermann M. Dorsal root potentials after C-fiber stimulation. Science
156:2423–30. 1968;160:896–8.

Copyright Ó 2018 by the International Association for the Study of Pain. Unauthorized reproduction of this article is prohibited.

You might also like