ORIGINAL ARTICLE
Impact of etiology and duration of pain on pharmacological
treatment effects in painful polyneuropathy
S. H. Sindrup1, J. Holbech1, D. Demant1, N. B. Finnerup2, F. W. Bach3, T. S. Jensen2
1 Department of Neurology, Odense University Hospital, Denmark
2 Danish Pain Research Center, Aarhus University Hospital, Denmark
3 Department of Neurology, Aarhus University Hospital, Denmark
Correspondence
Søren H. Sindrup
E-mail: soeren.sindrup@rsyd.dk
Funding sources
Research reported in this publication is part
of the International Diabetic Neuropathy
Consortium IDNC research programme,
which is supported by a Novo Nordisk Foun-
dation Challenge Programme grant (Grant
number NNF140C0011633).
Conflicts of interest
F. W. Bach reports to have been compen-
sated as an investigator in clinical trial spon-
sored by Pfizer and Gru €nenthal and personal
fees from Pfizer. N. B. Finnerup reports per-
sonal fees from Pfizer, Astellas and Nor-
pharma, grants from EU/EFPIA, and grants
and personal fees from Gru €nenthal. T. S.
Jensen reports to be on Advisory Board for
Pfizer, Gru€nenthal and Orion. J. Holbech
reports personal fees and study grants from
Pfizer. S. H. Sindrup reports study grants
from Pfizer and EU/EFPIA. D. Demant reports
to have no conflicts of interest to declare.
Accepted for publication
15 March 2017
doi:10.1002/ejp.1048
1. Introduction
Treatment options for painful polyneuropathy have
not changed much during the past two decades, and
in general performance of pharmacological treat-
© 2017 European Pain Federation - EFICâ
Abstract
Background: The pharmacological treatments for painful polyneuropa-
thy have not changed much for more than a decade, and less than half
of the patients obtain adequate pain relief with first line treatments.
Therefore, patient-specific factors which could predict drug response are
searched for.
Methods: We analysed data from four published, randomized,
controlled trials of drugs in painful polyneuropathy to see if diabetic
etiology and duration of neuropathic pain had an impact on drug
efficacy. The studies had a cross-over design, and had nearly similar
outcome recordings as well as a thorough baseline registration of
symptoms, signs and quantitative sensory testing. 244 patient records of
drug effect distributed over treatments with three antidepressants
(imipramine, venlafaxine, escitalopram) and two anticonvulsants
(pregabalin, oxcarbazepine) were analysed.
Results: Diabetes as etiology of polyneuropathy had no impact on the
effect of antidepressants (imipramine, venlafaxine, escitalopram), but
there was a significant interaction with treatment effect on
anticonvulsants with better effects in diabetics (0.86 NRS points,
p = 0.021) with most pronounced interaction for oxcarbazepine (1.47
NRS points, p = 0.032). There was an interaction between duration of
neuropathic pain and treatment with antidepressants with better effect
with duration less than 3 years (0.62 NRS points, p = 0.036), whereas
anticonvulsants tended to work best with duration of pain for more
than 3 years.
Conclusion: Despite the small sample size and limited number of drugs
included this study suggests that diabetic etiology of polyneuropathy
may impact on the efficacy of anticonvulsants, and duration of
neuropathic pain may impact on the efficacy of antidepressants.
Significance: This study found that duration of pain appears to have an
impact on the effect of antidepressants in neuropathic pain and that
diabetes as etiology for painful polyneuropathy appears to influence pain
relief obtained with anticonvulsants.
ments is disappointing (Sindrup and Jensen, 2000;
Finnerup et al., 2010, 2015). The mainstay of treat-
ment for painful neuropathy is still the antidepres-
sants of tricyclic and balanced serotonin and
Eur J Pain 21 (2017) 1443--1450 1443
Factors in drug response in neuropathic pain
noradrenaline reuptake inhibitor category as well as
the anticonvulsants gabapentin and pregabalin.
These treatment options provide adequate pain relief
in less than half of the patients.
It has for decades been discussed if mechanism-
based treatment could improve treatment outcome
in neuropathic pain after this concept was intro-
duced by various researchers (Max, 1990; Woolf and
Mannion, 1999; Holbech et al., 2016). A major
obstacle for implementing the concept in clinical
practice is that we do not have the tools for deter-
mining the molecular mechanisms of pain at play in
each patient. A range of different pain mechanisms
have been acknowledged mainly from experiments
in animals (Woolf and Mannion 1999), but their
individual contribution and interplay in each patient
are not easily determined. The pattern of symptoms
and quantitative sensory examination data has been
suggested as a tool for determining mechanisms in
neuropathic pain (Baron et al., 2012). Recently, it
was reported from a controlled clinical trial with
stratification according to sensory phenotype that
patients with peripheral neuropathic pain having a
phenotype termed irritable nociceptor had a better
pain relief from the sodium channel blocker oxcar-
bazepine than patients without this phenotype
(Demant et al., 2014). However, a retrospective
analysis of data from several clinical trials comprising
the major drug classes used in painful polyneuropa-
thy found no interaction between sensory phenotype
and drug efficacy in this pain condition (Holbech
et al., 2016). Challenges in this context are both
inability of sensory phenotyping to clearly reflect
mechanisms of pain and the fact that the major drug
classes tested do not have selective mechanism of
action targeting specific pain mechanisms.
Other factors than sensory phenotype could pre-
dict response of pharmacological treatment in
peripheral neuropathic pain. In painful polyneuropa-
thy, the etiology could be a factor with for example
pain in diabetic polyneuropathy being more respon-
sive to treatment than neuropathies of other etiolo-
gies, as has been indicated in one of our previous
trials (Sindrup et al., 2003). Furthermore, in experi-
mental models of neuropathic pain it has been
shown that inhibitory GABAergic interneurons in
the dorsal horn of the spinal cord may degenerate
after a relatively short time, i.e. within a few weeks
of pain (Moore et al., 2002). Thus, duration of neu-
ropathic pain could also be a factor, with long-term
neuropathic pain being less responsive to drug treat-
ment because of degeneration of neurons involved
in their mechanism of pain relief. Based on these
1444 Eur J Pain 21 (2017) 1443--1450
S. H. Sindrup et al.
observations, we retrospectively searched for a
possible influence of duration of pain and diabetes as
the etiology for polyneuropathy on the treatment
response to the major drug classes in peripheral neu-
ropathic pain as determined in a series of previously
conducted controlled clinical trials in painful
polyneuropathy (Sindrup et al., 2003; Otto et al.,
2008; Demant et al., 2014; Holbech et al., 2015).
2. Methods
2.1 Study design
This was a retrospective analysis of data from four
randomized, double-blind, placebo-controlled, cross-
over trials (Sindrup et al., 2003; Otto et al., 2008;
Demant et al., 2014; Holbech et al., 2015) of almost
similar design and in which there was a significant
effect of the study drug. The trials were performed at
Odense University Hospital, Aarhus University
Hospital and Aalborg University Hospital from 1999
to 2014. All the included trials met the CONSORT
Guidelines for randomized, controlled trials and have
been reported in detail previously.
2.2 Patients
Patients with painful diabetic and non-diabetic
polyneuropathy were included in the analysis.
Patients were male and female, aged > 18 years and
had symptoms compatible with polyneuropathy for
more than 6 months with distal symmetric pain
localisation and sensory disturbance in the area of
pain. Polyneuropathy diagnosis had to be confirmed
by clinical signs and electrophysiological tests or skin
biopsy. In three of the trials (Otto et al., 2008;
Demant et al., 2014; Holbech et al., 2015), the
patients had to have a median total pain rating of at
least 4 on an 0-10-point numeric rating scale
(0 = no pain, 10 = worst possible pain) (NRS) during
1 week without pain medication. In one trial (Sin-
drup et al., 2003), it was sufficient with at least 4 on
the NRS for one of four specific pain phenomena,
i.e. constant pain (burning, pressing or deep aching),
pain paroxysms, touch-evoked pain or pain on pres-
sure. To be included in the analysis, the patients
needed to have data from at least two treatment
periods (one being a placebo period). Therefore, in
the 3- and 4-armed studies (Sindrup et al., 2003;
Holbech et al., 2015), a number of patients were
included based on observations from only two
periods. The trials were designed to investigate the
independent effect of the drugs on peripheral
© 2017 European Pain Federation - EFICâ
S. H. Sindrup et al.
neuropathic pain. Hence, all concomitant treatments
for neuropathic pain used before inclusion were
either discontinued or patients on such treatments
excluded. Other causes of pain were exclusion crite-
ria in all 4 trials.
2.3 Randomization and treatment
For patients with ongoing medication for their neu-
ropathic pain, treatments were slowly discontinued
during a pre-study period of 1 week. After 1 further
week for baseline observations, the patients entered
the computer-generated, randomized treatment
sequence with treatment periods of 4-6 weeks’ dura-
tion. Treatment periods were separated by 1 or
2 week(s) for washout. All trials were double-blind
and trial drugs of identical appearance as placebo,
and trials with two active treatments employed dou-
ble-dummy technique. Drug doses and duration of
treatment periods are shown in Table 1. Paracetamol
in doses up to 3000 mg was used as escape medica-
tion, and in one trial (Demant et al., 2014), addi-
tional tramadol 50 mg daily was allowed as escape
medication.
2.4 Assessments
2.4.1 Symptoms
The data used in the analysis originated from ratings
in diaries. All ratings were entered in diaries in the
morning covering the symptoms experienced during
the preceding 24 h. Total pain intensity (three trials)
and specific pain symptoms (all trials) were rated
using the 0-10-point NRS throughout all baseline
and treatment periods. For the one trial (Sindrup
et al., 2003) in which total pain intensity was not
scored, the mean score of the four registered specific
Table 1 Studies, drugs, drug doses and duration of treatment periods.
Study Drug Doses (mg/day)
Holbech et al., 2015 Imipramine 75
25
25
Pregabalin 300
150
Sindrup et al., 2003 Imipramine 150a
Venlafaxine 225a
Otto et al., 2008 Escitalopram 20
Demant et al., 2014 Oxcarbazepine 2400
1800
a
Poor metabolisers excluded from study.
b
Dosages adjustable due to side effects, average dose 1684–1846 mg/day.
© 2017 European Pain Federation - EFICâ
Factors in drug response in neuropathic pain
pain symptoms (constant pain, paroxysms,
touch-evoked, pain on pressure) was used. As this
would lead to an unrealistically low total score for
some patients (e.g. a patient with 4 on constant pain
and 0 on the three other pain symptoms would have
a total pain intensity of 1), we used a correction fac-
tor, making the patient’s baseline score = 6 (average
baseline score of total pain in trials using this pain
measure). At baseline, the cause of polyneuropathy
was recorded as well as the duration of neuropathic
pain due to the polyneuropathy.
2.4.2 Bedside examination
A full clinical examination was performed by a neu-
rologist at the time of inclusion. The examination
included assessment of muscular function, deep ten-
don reflexes, vibration sensation, touch sensation
(hypoesthesia), pin prick sensation (hyperalgesia),
temperature sensation and dynamic mechanical allo-
dynia.
2.4.3 Quantitative sensory testing (QST)
The most painful site on each individual patient was
located, and QST was performed at this site at base-
line. The following was assessed: Pressure pain
thresholds with pressure algometer (Somedic AB,
Stockholm, Sweden) (all trials); mechanical allody-
nia by stroking skin with a sponge rubber brush
(Demant et al., 2014; Holbech et al., 2015), or stim-
ulation with an electronic toothbrush (Sindrup et al.,
2003); cold-allodynia with Thermoroller at 20 °C
(Somedic AB, Stockholm, Sweden) (Demant et al.,
2014; Holbech et al., 2015); pain by repetitive pin-
prick stimulation using von Frey hair, 2 Hz during
30 s (all trials); cold and warm detection thresholds
with a Thermotest (Somedic AB) (all trials). The
Duration treatment periods (weeks)
(patients aged < 70 years) 5
(patients aged ≥ 70 years)
(poor metabolisers)
(patients aged < 70 years)
(patients aged ≥ 70 years)
4
6
(patients aged < 70 years)b 6
(patients aged ≥ 70 years)b
Eur J Pain 21 (2017) 1443--1450 1445
Factors in drug response in neuropathic pain S. H. Sindrup et al.

oxcarbazepine trial differed by having used the stan- In the general linear statistical model including all
dardised QST protocol of the German Research Net- patients and all drugs the treatment effect was 0.91
work on Neuropathic Pain (Rolke et al., 2006). NRS points (95% CI 0.66-1.15 NRS points,
p < 0.001) and baseline severity had an impact of
2.5 Data analysis 0.18 NRS points (95% CI 0.02-0.34 NRS points,
p = 0.024).
Due to the retrospective nature of the study, the
The impact of diabetes as etiology for neuropathy
data analysis should mainly be regarded as explo-
on drug effect is detailed in Table 2 and illustrated in
rative. The overall effect of diabetes and duration of
(Fig. 1). For antidepressants, there were no differ-
neuropathic pain on drug effect on pain was anal-
ences in effect between diabetics and non-diabetics.
ysed using a general linear model with pain as
The anticonvulsants as a group reduced pain signifi-
dependent variable, drug and the interaction
cantly more in diabetics than in non-diabetics, and
between drug and diabetes or duration of pain as
for the individual anticonvulsants this was also the
factors, baseline pain as co-variate, period as fixed
case for oxcarbazepine, whereas the results for pre-
effect, and patient as random effect. Mean differ-
gabalin failed to reach statistical significance. The fre-
ences with 95% confidence intervals of pain reduc-
quency of lancinating pain was higher and the
tion on drugs between diabetics and non-diabetics as
frequency of gain of sensation, preserved thermal
well as between short and long duration of pain
sensation, pinprick hyperalgesia and pain summation
were determined. The drug effect was not corrected
on repetitive pinprick lower in diabetics as compared
for placebo effect, since the effect of placebo did not
to non-diabetics (Fig. 2).
differ systematically between the groups (diabetics
The impact of duration of neuropathic pain on
vs. non-diabetics: 0.15 NRS points, p = 0.580; short
drug effect is detailed in Table 3 and illustrated in
pain duration vs. long: 0.22 NRS points, p = 0.407).
Fig. 3. For the total group of antidepressants, there
Differences in frequency of patient characteristics
between the groups were tested by use of Fisher’s
exact test and p values presented are not corrected
for multiple comparisons. ALL DRUGS

ANTIDEPRESSANTS
Impramine
3. Results
Venlafaxine
The dataset comprised in total 244 observations of Escitalopram
patient characteristics and the effect of drugs and ANTICONVULSANTS
placebo in painful polyneuropathy with 86 being on Pregabalin
imipramine, 32 on venlafaxine, 41 on escitalopram, Oxcarbazepine
56 on pregabalin and 29 on oxcarbazepine. Diabetes
–4 –2 0 2 4
was the etiology for neuropathy in 81 patients and Difference DM vs. non-DM (mean, 95% CI)
88 had duration of neuropathic pain less than
3 years. Other causes of polyneuropathy were alco- Figure 1 Mean differences in pain reduction with 95% confidence
hol (n = 33), drug-induced (n = 12), idiopathic intervals between diabetics and non-diabetics for different drug
groups and drugs. Positive values indicate better effect in diabetics.
(n = 86), and other (n = 32).

Table 2 Interaction between drug effect and diabetes as etiological diagnosis of painful polyneuropathy.

Total n Diabetes n Drug Effect Δ NRS points (95% CI)a p Drug-DM-Interaction Δ NRS points (95% CI)a p

All drugs 244 81 0.86 (0.57:1.15) <0.001 0.15 ( 0.31:0.61) 0.521

Antidepresants 159 60 1.09 (0.71:1.48) <0.001 0.20 ( 0.78:0.38) 0.498
Imipramine 86 27 1.14 (0.67:1.61) <0.001 0.15 ( 0.93:0.62) 0.701
Venlafaxine 32 14 1.14 (0.05:2.23) 0.040 0.48 ( 2.01:1.06) 0.543
Escitalopram 41 19 0.90 (0.13:1.66) 0.020 0.07 ( 0.94:1.06) 0.904
Anticonvulsants 85 21 0.49 (0.09:0.89) 0.016 0.86 (0.13:1.58) 0.021
Pregabalin 56 14 0.41 ( 0.03:0.86) 0.069 0.78 ( 0.06:1.61) 0.067
Oxcarbazepine 29 7 0.58 ( 025:1.42) 0.169 1.47 (0.12:2.81) 0.032
a
Estimated by general linear model (GLM) with total numeric rating scale score (NRS) of pain as dependent variable, treatment and interaction
between treatment and diabetes etiology (DM) as factors, baseline pain as co-variate, and period as fixed effect, and patient as random effect.

1446 Eur J Pain 21 (2017) 1443--1450 © 2017 European Pain Federation - EFICâ
S. H. Sindrup et al. Factors in drug response in neuropathic pain

was a significantly better effect in patients with ALL DRUGS

duration of pain less than 3 years, but this was not ANTIDEPRESSANTS
significant for the individual antidepressants. Oxcar- Impramine
bazepine reduced pain less in patients with short Venlafaxine
duration of pain, whereas there was no difference Escitalopram
for pregabalin or for pregabalin and oxcarbaxepine
ANTICONVULSANTS
taken together. The frequency of preserved thermal
Pregabalin
sensation was higher in patients with short duration
Oxcarbazepine
of pain than in patients with longer duration of pain
–4 –2 0 2 4
(Fig. 4). None of the other characteristics differed
Pain duration < 3 years vs. > 3 year (mean, 95% CI)
between the two groups.
Figure 3 Mean differences in pain reduction with 95% confidence
intervals between patients with short duration of pain (<3 years) and
4. Discussion long duration of pain (≥ 3 years) for different drug groups and drugs.
This retrospective analysis of data from drug trials Positive values indicate better effect with short duration of pain.
found that the anticonvulsants oxcarbazepine and
100
100
80
80 *
*
60
* *
60
*
40
*
40
20
20
0
0
d

a
n

in

k
l a sia
n

es n
ed

ric
ni
ai
ve

ni
io

tio

pa
rv

dy
p

dy
at

np
er
a

lg
e

g
a
d

in

in

a
l a sia

k
es n

ns
d

ns
se on

lo
llo
es

ra
ric

pi
La nin

in
ni
ve

ni
io

ve

pa

al
se

e
i

at
y
p

dy
at

pe
pr
at

pr
np

n
er
er

s
od

r
g

su old

io
in
g

Bu
ns

ca
ns

llo
l

hy
es

n
a

pi
La nin

n
in

nc

at
l

ai

s
al

io
er

io

C
ni
se

at

Lo
pr
pr

m
G

ck
at
r

at
yp

ha
su old

io
in
Bu

ca

m
s
n

ns
se on

am pri
s
nc

at
h
s

ec
ai

io

en
C
ni
Lo

se
i

yn Pin
ck
G

at
at

ha

m
ls
m

in
ns

am pri
ns

n
ec

ic

Pa
a

io
se

m
yn Pin

at
in

er
n

br
al

ic

Pa

Th
io
m

Vi

D
at
er

br
Th

Vi

Figure 4 Frequency of different patient characteristics in patients

Figure 2 Frequency of different patient characteristics in diabetics with short duration of pain (black columns) and long duration of pain
(black columns) and non-diabetics (white columns). Statistically signifi- (white columns). Statistically significant difference (p = 0.017) marked
cant differences (p-value range from <0.001 to 0.006) marked with *. with *.

Table 3 Interaction between drug effect and duration of painful polyneuropathy.

Drug Effect Δ NRS Drug-Duration-Interaction Δ NRS

Total n Pain < 3 years n points (95% CI)a p points (95% CI)a p

All Drugs 244 88 0.82 (0.52:1,11) <0.001 0.24 ( 0.21:0.69) 0.299

Antidepresants 159 58 0.79 (0.41:1.17) <0.001 0.62 (0.04:1.20) 0.036
Imipramine 86 31 0.90 (0.42:1.38) <0.001 0.54 ( 021:1.28) 0.156
Venlafaxine 32 13 0.46 ( 0.58:1.50) 0.390 1.17 ( 0.35:2.69) 0.133
Escitalopram 41 14 0.77 (0.06:1.47) 0.033 0.47 ( 0.57:1.51) 0.377
ANTICONVULSANTS 85 30 0.88 (0.45:1.32) <0.001 0.51 ( 1.18:0.15) 0.131
Pregabalin 56 18 0.65 (0.17:1.13) 0.008 0.13 ( 0.92:0.66) 0.744
Oxcarbazepine 29 12 1.43 (0.56:2.30) 0.001 1.21 ( 2.36:-0.06) 0.040
a
Estimated by general linear model (GLM) with total numeric rating scale score (NRS) of pain as dependent variable, treatment, and interaction
between treatment and duration of pain as factors, baseline pain as co-variate, and period as fixed effect, and patient as random effect.

© 2017 European Pain Federation - EFICâ Eur J Pain 21 (2017) 1443--1450 1447
Factors in drug response in neuropathic pain
pregabalin either worked or tended to work better in
diabetics than in non-diabetics, and that antidepres-
sants had a better effect on patients with short than
longer duration of neuropathic pain. To our knowl-
edge, this is the first systematic analysis of such rela-
tions in painful polyneuropathy. Two previous post-
hoc analyses of data from controlled trials in painful
diabetic neuropathy found no impact of neuropathy
duration and the effect of duloxetine or pregabaline
(Ziegler et al., 2007; Marchettini et al., 2016). These
studies separated duration of neuropathy by 2 years
and focused on duration of neuropathy and not
directly on duration of neuropathic pain. Further-
more, the latter study mainly searched for differ-
ences in predictors of response between the
antidepressant duloxetine and the anticonvulsant
pregabalin. In line with our findings the latter study
reported that the treatment effect size was higher in
patients with more severe pain at baseline (Marchet-
tini et al., 2016).
The present study has several substantial limita-
tions. First of all, based on the retrospective character
of the study, it can only be hypothesis-generating.
Minor differences in study design and outcomes may
impact on the results and pooling the data, and ana-
lyzing them together may therefore be problematic.
Grouping of drugs as antidepressants and anticonvul-
sants is arbitrary, since with respect to mode of action
drugs may differ within groups and drugs may have
similarities between groups. The study also miss data
on the important first line drugs for neuropathic
pain, i.e. the antidepressant duloxetine and the anti-
convulsant gabapentin, and this may impact on con-
clusions for drug groups One of the trials (Holbech
et al., 2015) used relatively low drug doses, since this
was part of the design in this drug combination trial.
However, this may not have had a major impact on
the pertinent subgroup comparisons which are not
between drugs, but between groups based on patient
characteristics. The most important limitation of the
present study is that the number of observations is
small, which reduces the statistical power. This
becomes even more important when we analyze data
from single drugs. We may therefore overlook impor-
tant differences and there is also a risk of spurious
findings when the numbers are very small as for
example with oxcarbazepine in diabetics. Further-
more, the limited study population does not allow for
analysis with statistical models which include correc-
tion for co-variates. In all, the results should be inter-
preted cautiously. Taken together, the study
limitations calls for a very cautiously interpretation of
the results. The low number of observations that
1448 Eur J Pain 21 (2017) 1443--1450
S. H. Sindrup et al.
could be included in this analysis may also call for a
change of how clinical trial results are reported. Thus,
it could be requested that single patient data, which
are needed for doing analyses as the present, should
always be made available as an appendix to publica-
tions.
There is no obvious explanation for the present
findings of our retrospective analysis, but one possi-
bility is that the differences in effect are related to dif-
ferences in molecular or cellular mechanisms of pain
depending on diabetic etiology of neuropathic pain
and duration of pain. The difference in phenotype
profile with respect to symptoms and signs between
diabetics and non-diabetics found in this analysis
imply that this could indeed be the case. However, as
mentioned in the introduction, there is no well-
established relation between symptoms and signs and
molecular mechanisms of pain. Lancinating pain,
here seen with higher frequency in diabetics, has
been attributed to a sodium channel mechanism
(Baron et al., 2012), which could be targeted by the
sodium channel blocker oxcarbazepine and thus
explain why oxcarbazepine apparently works better
in diabetics. Tricyclic antidepressants, such as imipra-
mine, also have sodium channel-blocking effect, but
tricyclic antidepressants probably mainly work
through monoaminergic mechanisms, and this may
mask a difference in effect between diabetics and
non-diabetics due to its sodium channel-blocking
effect. It is in line with common thinking that pre-
served thermal sensation is more frequent in patients
with short duration of pain, but it is surprising that
we at the same time observed less effect of especially
oxcarbazepine. Preserved small fibre function has
been found to be a factor for the effects of oxcar-
bazepine in peripheral neuropathic pain (Demant
et al., 2014). Similarly, the better effect of antidepres-
sants with short duration of pain cannot easily be
explained in these terms, i.e. when we look at the
symptom and sensory sign profile differences
between patients with short versus long duration of
pain. In another retrospective analysis of trial data it
was found that imipramine reduces pain significantly
more in patients with gain of sensory function and
that pregabalin had a better effect in patients with
preserved large fibre function (Holbech et al., 2016).
In a post hoc analysis of data from a study of prega-
balin and duloxetine combination some inter-pheno-
type differences in response between treatments
were reported (Bouhassira et al., 2014).
The slightly better effect of antidepressants with
short duration of pain could be explained by the
mechanisms of pain and the action of
© 2017 European Pain Federation - EFICâ
S. H. Sindrup et al.
antidepressants. In experimental models of neuro-
pathic pain, it has been shown that there is a loss of
GABAergic neurones in the superficial layers of the
dorsal horn of the spinal cord (Moore et al., 2002).
These inhibitory interneurons are supposed to be
important as relay neurons for the descending mod-
ulation exerted by opioids, cannabinoids and
monoamines (Lau and Vaughan, 2014). GABAergic
neurones in the spinal cord do have monoaminergic
receptors (Wang et al., 2009). So, the less effect of
antidepressants with long duration of pain may
reflect that with time the effect of monoamines will
decrease due to loss of GABAergic neurones in the
dorsal horn of the spinal cord.
Taken together, it is suggested that the differences
in effect may be related to differences in pain mech-
anism or pain modulation, but the molecular or cel-
lular nature of these differences remains obscure.
Furthermore, the limitations of the present analysis
should be kept in mind.
What are the clinical implications of these find-
ings? The effect size of anticonvulsants in painful
polyneuropathy in the present analysis is smaller
than that of antidepressants, and if it is largely dri-
ven by effect in diabetics one could argue that
antidepressants could be the first choice in non-dia-
betic polyneuropathy. The impact of duration of pain
on effect of antidepressants should be kept in mind
and considered as a reason for low efficacy of these
drugs with long duration of pain. If the reason is
degeneration of inhibitory GABAergic interneurons
in the spinal cord, we may also have to consider
early treatment with drugs that silence the periph-
eral nociceptors, since the degeneration of the
GABAergic neurons is probably a result of continu-
ous over-activity in the nociceptors. Although this is
speculative, there could be a role for channel block-
ers here with sodium channel blockers for reducing
ectopic activity in the nociceptors and calcium chan-
nel blockers for inhibiting synaptic transmission of
the increased peripheral activity. Thus, early combi-
nation therapy with drugs with different mecha-
nisms of action or drugs with multiple mechanisms
action may be an attractive strategy. But again, the
limitations of this study should be taken into consid-
eration.
In clinical trials, it may be considered to study dia-
betics and non-diabetic polyneuropathies separately,
but it would also be feasible to size the studies to have
a sufficient number of patients with each etiology to
obtain adequate statistical power for analysis of sub-
group-specific effects. Similarly, duration of neuropa-
thy could be included in the model for statistical
© 2017 European Pain Federation - EFICâ
Factors in drug response in neuropathic pain
analysis of trial data or it could be considered to
restrict inclusion to short duration of pain when drugs
working via pain modulation are studied.
In conclusion, etiology of painful polyneuropathy
with respect to diabetes may impact on the pain-
relieving effect of some anticonvulsants, and dura-
tion of neuropathic pain may impact on the effect of
some antidepressants. This could have implications
for the choice of treatment strategy in clinical prac-
tice and may also be considered in the planning of
clinical drug trials.
Acknowledgements
The studies that provided data for this analysis were sup-
ported by grants from H. Lundbeck, Pfizer, Takeda Pharma
and the IMI EuroPain collaboration.
Author contributions
All authors contributed to the idea and concept of this
study and all authors contributed to one or more of the
clinical trials which formed the background for the present
analysis. SHS and JH performed the statistical analysis of
the data and SHS wrote the first version of the manuscript.
All authors commented on the analysis and the manu-
script, and approved the final version of the manuscript to
be published.
References
Baron, R., F€ orster, M., Binder, A. (2012). Subgrouping of patients with
neuropathic pain according to pain-related sensory abnormalities: A
first step to a stratified treatment approach. Lancet Neurol 11, 999–
1005.
Bouhassira, D., Wilhelm, S., Schacht, A., Perrot, S., Kosek, E. et al.
(2014). Neuropathic pain phenotyping as a predictor of treatment
response in painful diabetic neuropathy: Data from the randomized,
double-blind, COMBO-DN study. Pain 155, 2171–2179.
Demant, D., Lund, K., Vollert, J., Maier, C., Segerdahl, M€artha,
Finnerup, N., Jensen, T.S., Sindrup, S.H. (2014). The effect of
oxcarbazepine in peripheral neuropathic pain depends on pain
phenotype: A randomised, double-blind, placebo-controlled,
phenotype stratified study. Pain 155, 2263–2273.
Finnerup, N.B., Sindrup, S.H., Jensen, T.S. (2010). The evidence for
pharmacological treatment of neuropathic pain. Pain 150, 573–581.
Finnerup, N.B., Attal, N., Haroutounian, S., McNicol, E., Baron, R.
et al. (2015). Pharmacotherapy for neuropathic pain in adults: A
systematic review and meta-analysis. Lancet Neurol 14, 162–173.
Holbech, J.V., Bach, F.W., Finnerup, N.B., Brøsen, K., Jensen, T.S.,
Sindrup, S.H. (2015). Imipramine and pregabalin combination for
painful polyneuropathy: A randomized controlled trial. Pain 156,
958–966.
Holbech, J., Bach, F.W., Finnerup, N.B., Jensen, T.S., Sindrup, S.H.
(2016). Pain phenotype as a predictor for drug response in painful
polyneuropathy. A retrospective analysis of data from controlled
clinical trials. Eur J Pain 157, 1305–1313.
Lau, B.K., Vaughan, C.W. (2014). Descending modulation of pain: the
GABA disinhibition hypothesis of analgesia. Curr Opin Neurobiol 29,
159–164.
Eur J Pain 21 (2017) 1443--1450 1449
Factors in drug response in neuropathic pain
Marchettini, P., Wilhelm, S., Petto, H., Tesfaye, S., T€ olle, T. et al.
(2016). Are there different predictors of analgesic response between
antidepressants and anticonvulsants in painful diabetic neuropathy.
Eur J Pain 20, 472–482.
Max, M.B. (1990). Towards physiologically based treatment of patients
with neuropathic pain. Pain 42, 131–133.
Moore, K.A., Kohno, T., Karchewski, L.A., Scholz, J., Baba, H., Woolf,
C.J. (2002). Partial peripheral nerve injury promotes a selective loss
of GABAergic inhibition in the superficial dorsal horn of the spinal
cord. J Neurosci 22, 6724–6731.
Otto, M., Bach, F.W., Jensen, T.S., Brøsen, K., Sindrup, S.H. (2008).
Escitalopram in painful polyneuropathy: A randomized, placebo-
controlled, cross-over trial. Pain 139, 275–283.
Rolke, R., Baron, R., Maier, C., T€ olle, T.R., Treede, R.D. et al. (2006).
Quantitative sensory testing in the German Research Network on
Neuropathic Pain (DFNS): Standardized protocol and reference
values. Pain 123, 231–243.
1450 Eur J Pain 21 (2017) 1443--1450
S. H. Sindrup et al.
Sindrup, S.H., Jensen, T.S. (2000). Pharmacologic treatment of pain in
polyneuropathy. Neurology 55, 915–920.
Sindrup, S.H., Bach, F.W., Madsen, C., Gram, L.F., Jensen, T.S. (2003).
Venlafaxine versus imipramine in painful polyneuropathy: A
randomized controlled trial. Neurology 60, 1284–1289.
Wang, Y.Y., Wei, Y.Y., Huang, J., Wang, W., Tamamaki, N., Li, Y.Q.,
Wu, S.X. (2009). Expression pattern of 5-HT receptor subtypes 1A
and 2A on GABAergic neurons within the spinal dorsal horn of
GAD67-GFP knock-in mice. J Chem Neuroanat 38, 75–81.
Woolf, C.J., Mannion, R.J. (1999). Neuropathic pain: aetiology,
symptoms, mechanisms, and management. Lancet 353, 1959–1964.
Ziegler, D., Pritchett, Y.L., Wang, F., Desaiah, D., Robinson, M.J., Hall,
J.A., Chappell, A.S. (2007). Impact of disease characteristics on the
efficacy of duloxetine in diabetic peripheral neuropathic pain. Diabetes
Care 30, 664–669.
© 2017 European Pain Federation - EFICâ