lable at ScienceDirect

Manual Therapy 19 (2014) 311e318

Contents lists avai

Manual Therapy

journal homepage: www.elsevier .com/math

Original article

Sensory characteristics of chronic non-specific low back pain: Asubgroup investigationq

Peter O’Sullivan*, Robert Waller, Anthony Wright, Joseph Gardner, Richard Johnston,Carly Payne, Aedin Shannon, Brendan Ware, Anne SmithSchool of Physiotherapy & Exercise Science, Curtin University, GPO Box 1987, Perth, WA 6845, Australia

a r t i c l e i n f o

Article history:Received 22 May 2013Received in revised form6 March 2014Accepted 14 March 2014

Keywords:Pain sensitivityChronic non-specific low back painBiopsychosocialClassification

q Ethical approval for this study was granted by tResearch Ethics Committee (PT0180).* Corresponding author. Tel.: þ61 8 9266 3629; fax

E-mail address: p.osullivan@curtin.edu.au (P. O’Su

http://dx.doi.org/10.1016/j.math.2014.03.0061356-689X/� 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

It has been proposed that patients with chronic non-specific low back pain (CNSLBP) can be broadlyclassified based on clinical features that represent either predominantly a mechanical pain (MP) or non-mechanical pain (NMP) profile. The aim of this study was to establish if patients with CNSLBP who reportfeatures of NMP demonstrate differences in pain thresholds compared to those who report MP char-acteristics and pain-free controls. This study was a cross-sectional design investigating whether pressurepain threshold (PPT) and/or cold pain threshold (CPT) at three anatomical locations differed betweenpatients with mechanical CNSLBP (n ¼ 17) versus non-mechanical CNSLBP (n ¼ 19 and healthy controls(n ¼ 19) whilst controlling for confounders. The results of this study provide evidence of increased CPT atthe wrist in the NMP profile group compared to both the MP profile and control subjects, when con-trolling for gender, sleep and depression (NMP versus MP group Odds Ratio (OR): 18.4, 95% confidenceinterval (CI): 2.5e133.1, p ¼ 0.004). There was no evidence of lowered PPT at any site after adjustment forconfounding factors. Those with an MP profile had similar pain thresholds to pain-free controls, whereasthe NMP profile group demonstrated elevated CPT’s consistent with central amplification of pain. Thesefindings may represent different pain mechanisms associated with these patient profiles and may haveimplications for targeted management.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Patients with chronic non-specific low back pain (CNSLBP) posea complex diagnostic and management challenge. Classificationsystems (CS) that identify mechanisms that underlie the pain dis-order have been advocated in clinical practice in order to bettertarget interventions (O’Sullivan, 2012a, 2005; Woolf, 2011). TheQuebec Task Force CS (Spitzer, 1987) whilst differentiating specificpathology and radicular pain from CNSLBP, does not furtherdifferentiate subjects with CNSLBP (Dankaerts et al., 2006). A recentreview of clinical CS’s for CNSLBP concluded that a limitation of themajority of CS’s is that they do not consider underlying painmechanisms and focus largely on biomechanical assessment(Karayannis et al., 2012).

A multidimensional CS system for LBP has been proposed that atthe first level triages people with LBP to identify red flag disorders

he Curtin University Human

: þ61 8 9266 3699.llivan).

and specific pathology from non-specific LBP (Fig. 1). Once identi-fied, CNSLBP disorders are further differentiated on the basis oftheir pain characteristic’s reflecting a spectrum from either ‘me-chanical pain’ (MP) to ‘non-mechanical pain’ (NMP) (Fig. 1). This isbased on routine clinical examination of the patient’s reported paincharacteristics linked to aggravating and easing factors and painresponses to movement and loading tests (O’Sullivan, 2005, 2012b;Vibe Fersum et al., 2009, 2012). While it is acknowledged that forsome patients there may be a mixed pain profile for others theclinical distinction is clear. It is postulated that these groups mayhave different underlying neurophysiological mechanisms, wherepain in the MP group is related to processes of peripheral sensiti-sation and some degree of activity dependent central sensitisation,whereas pain in the NMP group is related to more extensivechanges in central pain processing. Other dimensions such as paintype, psychosocial, lifestyle, and movement related factors as wellas pain comorbidities are also considered in the CS (O’Sullivan,2005, 2012b; Vibe Fersum et al., 2009). Although this CS has pre-viously been shown to have good inter-rater reliability for identi-fication of aspects of the CS related to movement and psychologicalprofiles (Vibe Fersum et al., 2009), no pain sensitivity (PS) testing

Fig. 1. Multidimensional classification of LBP disorders adapted from O’Sullivan, 2005, 2012b; Vibe Fersum et al., 2009, 2012.

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318312

has been conducted to quantify the sensory profiles associated withthese pain characteristic profiles.

Both cold hyperalgesia and widespread pressure hyperalgesiaare believed to be indicative of central hyperexcitability (Woolf,2011). Pain Sensitivity testing is used to assess sensory pre-sentations in various pain disorders (Rolke et al., 2006a) howeverlittle research has investigated PS in CNSLBP disorders and con-troversy exists regarding its value in understanding these disorders(Hubsher et al., 2013). A recent narrative review of available liter-ature in CLBP concluded that currently the available researchdemonstrates mixed results, with some studies documentingreduced pain thresholds suggestive of widespread or extra-segmental hyperalgesia, other studies observe only segmentalhyperalgesia and others reporting no hyperalgesia at all (Rousselet al., 2013). Another recent systematic review investigating therelationship between pain thresholds and pain intensity anddisability levels in LBP and neck pain patients, concluded that painthresholds are a poor marker for patients pain and disability levels(Hubsher et al., 2013). The apparent conflict between these findings

may reflect the heterogeneity of subjects in the different studies,with the potential for different pain phenotypes in the CNSCLPpopulation unaccounted for by study design (Giesecke et al., 2004;Roussel et al., 2013).

Both sensory perception and sensory testing are potentiallyinfluenced by a number of factors other than pain, such as gender,age, genetics, body composition, sleep and psychosocial factors(Dunn, 1997; O’Sullivan et al., 2008; Leboeuf-Yde et al., 2009;Heffner et al., 2011; Woolf, 2011), highlighting the need to considerthese factors when conducting research into PS. While there islimited research investigating whether the presence of CNSLBP isassociated with PS changes independent of these factors, a recentstudy reported that pressure pain threshold (PPT) was most pre-dictive of CNSLBP independent of age, gender, body compositionand psychological factors (Neziri et al., 2012). Therefore the primaryaim of this study was to investigate whether patients with CNSLBPwho report features of NMP demonstrate differences in cold painthreshold (CPT) and PPT compared to those who report MP char-acteristics and pain-free controls.

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318 313

2. Materials and methods

2.1. Study design

A cross-sectional study design was used.

2.2. Participants

A total of 53 participants were included in the study; 36 par-ticipants with CNSLBP (13 males and 23 females with a mean age of40.7 (standard deviation (SD) � 14.0)) were recruited from localprivate physiotherapy clinics in the greater Perth area, and 19 pain-free controls (8 males and 11 females with a mean age of 41.9(SD� 13.9)) recruited from the same district. Pain participants wereincluded if they had experienced pain for a minimum of 3 months,reported pain intensity on a Visual Analogue Scale (VAS) of 3 orgreater on the day of testing and LBP was their primary complaint(from T12 to gluteal fold). Control subjects were included on thebasis that they had not reported LBP or any other pain disorder inthe previous 6 months. Individuals were excluded if they had beendiagnosed with specific spinal pathology or medical causes of lowback pain, were pregnant or less than 6 months post-partum orsuffered from peripheral neuropathy. In all groups, subjects wereexcluded if they did not perceive pressure pain below 1000 kPaduring PPT testing, or they did not perceive a change in coldsensation during CPT testing. A-priori power calculation deter-mined 18 participants in each group would provide 85% power todetect pairwise differences of at least one standard deviation inmean CPT or PPT assuming a lognormal distribution, at a statisticalsignificance level of 0.05. Ethical approval for this study wasgranted by the Curtin University Human Research Ethics Commit-tee (PT0180).

2.3. Participants classification

The CNSLBP participants, identified following a triage process toexclude red flag and specific pathology, were divided into twogroups based on clinical criteria (Fig. 1). Participants in the MPgroup were included on the basis of: localised and anatomicallydefined LBP associated with reports of specific and consistentmechanical aggravating and easing factors (LBP that was moreintermittent in nature and demonstrated a proportionate painprovocation and easing response to specific postures, activities andmovements). Participants in the NMP group were included on thebasis of: LBP was more widespread and ill defined, LBP being moreconstant, non-remitting, spontaneous and where minor mechani-cal loading factors (such as simple spinal movements) resulted inexaggerated (severe) or prolonged (lasting hours) pain responses(O’Sullivan, 2005). The decision to classify was based on a combi-nation of patient report and response to routine clinical examina-tion. Pain sensitivity testing was not part of this decision makingprocess.

Recruitment of the CNSLBP participants occurred across anumber of Physiotherapy practices, and consecutive patients wereinvited to participate if they fulfilled the inclusion criteria. Furtherscreening was performed by RW (Musculoskeletal Physiotherapistwith 23 years clinical experience) and POS (Specialist Musculo-skeletal Physiotherapist and the developer of the CS who has 25years clinical experience) both of whom are trained in the CS toensure the patients fitted the clinical subgroups. A total of 3 par-ticipants who agreed to participate were excluded as they failed tomeet all the inclusion criteria. One was excluded due to a lack ofpain response to pressure, and two had a VAS of less than 3/10 onthe day of testing.

2.4. Procedures

On the day of testing all participants completed two question-naires, the Pittsburgh Sleep Quality Index (PSQI) and the Depres-sion Anxiety and Stress Scale (DASS 21), which have establishedreliability and validity (Buysse et al., 1989; Lovibond and Lovibond,1995) and were used as covariates to control for the potentialconfounding effect of poor sleep quality and stress on PS. Age, waistand hip girth were also recorded. Upon agreeing to take part in thestudy, participants were not asked to stop any of their regularmedications. A list of current medications taken over the weekprior to testing was documented.

Participants with NSCLBP were also asked to complete thefollowing to provide a clinical profile. The pain intensity level oftheir LBP was measured using the VAS (Huskisson, 1974), and painareas were recorded using a body chart to provide total areas ofpain using the Widespread Pain Index (Wolfe et al., 2010). TheRoland Morris Disability Questionnaire (RMDQ) was used to assessfunctional disability levels and is valid and reliable (Roland andMorris, 1983; Roland and Fairbank, 2000). The StarT Backscreening tool (SBST) was used to assess risk profile (Hill et al.,2008). The PainDETECT Questionnaire was used as a validatedself-report tool to identify neuropathic pain features. It is anestablished questionnaire with high sensitivity and specificity(Freynhagen et al., 2006).

2.5. Sensory testing

For participants with CNSLBP, the most painful side was tested.The right side was used for those where there was no pain domi-nant side and for the pain-free controls. Three test sites, the dorsalaspect of the wrist joint line, the L5/SI interspinous space and thelateral calcaneus, were tested in a standardised order and location(Jones, 2007). Each site was tested 4 times with the first test actingas familiarisation with the testing procedure (Wright et al., 1994;Lewis et al., 2010). The testing protocol was strictly followed tolimit tester error (Rolke et al., 2006b). Participants were allocated totesters according to time and location of testing, tester allocationwas distributed evenly between the three groups, and testers wereblinded to pain group allocation.

2.6. Pressure pain thresholds

PPT was defined as the moment the sensation of pressure be-comes one of pressure and pain (Jones, 2007). The PPT was testedusing an algometer (Somedic AB, Sweden) with a contact area of1 cm2 which was applied perpendicularly to the skin. The pressureincreased from 0 kPa at a constant rate of 40 kPa/s until PPT or amaximum of 1000 kPa (Chien and Sterling, 2010) was reached. Thestandardised instructions were, “Pressure will be applied at agradual rate. Allow the pressure to increase until it reaches a pointwhere it first feels uncomfortable and then press the button.”Testing was performed by one of two testers (CP, BW). Prior to PPTtesting, consistency for PPT measurement between testers wasensured.

2.7. Cold pain thresholds

An MSA Thermal stimulator (Somedic AB, Sweden) was used toobtain the CPT. Before assessing CPT a cold detection threshold wasobtained for each site to confirm the participant’s ability to detectcold (Mosek et al., 2001). Each test began at a baseline temperatureof 32 �C, and decreased at 1�C/s until reaching CPT or the automaticminimum cut-off of 5 �C (Carli et al., 2002). The standardised in-structions were, “The temperature probe will gradually get cooler.

Table 1Clinical profile of CNSLBP participants.

Instrument(max score)

MechanicalCNSLBP

Non-mechanicalCNSLBP

p-Value

Median (inter-quartile range), minemaxVAS (10) 4 (4), 8e17 6 (3), 10e19 0.018a

Widespread PainIndex (19)

2 (2), 1e7 3 (3), 1e9 0.014a

RMDQ (24) 3 (6), 1e15 11 (11), 2e20 0.004a

PainDETECT (39) Number (percentage of pain group)Nociceptive 14 of 17 (82%) 8 of 19 (42%)Unclear 3 of 17 (18%) 6 of 19 (32%)

b

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318314

Allow the temperature to drop until it reaches a point where it firstfeels uncomfortably cold, and then press the button.” Following CPTtesting the subjects were asked “Did you feel a sensation other thancold and if yes, how would you describe it?” These responses weredivided into ‘cold’ or non-noxious (pressure, nice, cold, tingling,pleasant and numb) and ‘non-cold’ or noxious (burning, ice, sharp,sting, gnawing and freezing) descriptors for further statisticalanalysis. Previous studies have reported the reliability of CPTmeasurement (Zwart and Trond, 2002; Wasner and Brock, 2008;Moloney et al., 2012). Testing was performed by one of two tes-ters (AS, BW).

Neuropathic 0 5 of 19 (26%) 0.010StarT Back score (9)Risk category Number (percentage of pain group)Low 11 of 17 (65%) 2 of 19 (11%)Medium 6 of 17 (35%) 9 of 19 (47%)High 0 of 17 (0%) 8 of 19 (42%) 0.001b

Medication use Number (percentage of pain group)Non-opioid 1 of 17 (6%) 6 of 19 (32%) 0.052b

NSAID 3 of 17 (18%) 8 of 19 (42%) 0.112b

Opioid 0 of 17 (0%) 5 of 19 (26%) 0.023b

Centrally acting 2 of 17 (12%) 6 of 19 (32%) 0.153b

For each questionnaire, the maximum score is given in brackets. RMDQ, the RolandMorris Disability Questionnaire; VAS, a Visual Analogue Scale for pain on the day oftesting; CNSLBP¼ chronic non-specific low back pain.

a Statistical test for group differences is ManneWhitney U test.b Statistical test for group differences is Fisher’s exact test.

2.8. Statistical analysis

The average of 3 trials at each site was used for statisticalanalysis (Slater et al., 2005). CNSLBP subgroups were examined fordifferences in clinical profile using chi-squared tests, Fisher’s exacttest, ManneWhitney U or KruskaleWallis tests as appropriate. Theassociation between sensory threshold measures and variablesconsidered as covariates (sex, age, waist/hip girth, DASS and PSQI)were examined using chi-squared tests, analysis of variance,ManneWhitney U or KruskaleWallis test as appropriate. Variableswith evidence for imbalance among pain groups (p < 0.200) wereincluded in multivariable models.

CPT values were suggestive of an underlying bimodal distribu-tion of this measure in the population (see Fig. 2), and all trans-formations including logarithmic failed to normalise the data. Forfurther analysis we created a dichotomous variable based uponvisual examination of the distribution of data for the CPT measurewhich supported a cut-off point of 15 �C as clearly separating twogroups in the data (�15 �C, >15 �C, see Fig. 2). This dichotomisationwas further supported by k-means cluster analysis, for which a two-cluster solution produced two clusters of individuals, with indi-vidual CPT measures below and above 15 �C. Descriptive statisticsand chi-squared tests were used to compare differences in pro-portions of participants with high CPT at each site between groups.Three binary logistic regressions with high/low CPT at each of thethree sites as the outcome variable were used to assess pain groupdifferences adjusting for covariates gender, DASS and PSQI. Differ-ences in frequency of use of non-cold descriptors of sensationexperienced during testing between groups were tested using achi-squared test.

PPT measures were log transformed to correct for positive skew.General linear regression models with log transformed PPT mea-sures as the outcome variablewere used to assess group differencesunadjusted and adjusted for covariates gender, DASS and PSQI(three models for three sites).

95% Confidence intervals with associated p-values are presentedfor all regression coefficients. All data were analysed using theStatistical Package for Social Sciences (SPSS) student version 18.0

Fig. 2. Untransformed ind

(SPSS Inc., Chicago, IL, USA). Data were inputted by one researcher(CP) and cross-checked by a second researcher (AS).

3. Results

Nineteen of the CNSLBP participants displayed NMP character-istics (4 males and 15 females with a mean age 42.6 (SD � 14.8))and 17 displayed MP characteristics (9 male and 8 females with amean age of 39.4 (SD� 14.2)). All subjects were screened for healthcomplaints and none reported other co-existing pain conditions,diabetes, endocrine disorders, nervous system disorders or psy-chiatric disorders. Clinical characteristics of the pain groups arereported in Table 1. The NMP group was characterised by higherpain levels, more pain areas, a larger proportion of neuropathicpain as classified by PainDETECT scores, greater disability, higherrisk rating on the SBST and greater frequency of medication use.

Fig. 2 presents the untransformed individual values for CPT.Initial univariable analyses provided evidence of group differencesin CPT at wrist, lumbar spine and heel sites, and PPT at the lumbarspine site (Table 2). The proportions of subjects with elevated CPTs(>15 �C) at the wrist were; control group 26%, MP group 24% andNMP group 84%. There was evidence of some imbalance betweengroups in DASS total, PSQI and gender, but not age or waist-hip ratio(Table 3), and of various associations between sex, DASS total and

ividual values for CPT.

Table 2Cold pain threshold (CPT) and pressure pain threshold (PPT) measures by participant group.

Subgroup

Control (n ¼ 19) Mechanical (n ¼ 17) Non-mechanical (n ¼ 19) p-Value

CPT (n (%)>15 �C)Wrist 5 (26.3) 4 (23.5) 16 (84.2) <0.001b

Lumbar spine 9 (47.4) 8 (47.1) 16 (84.2) 0.029b

Heel 6 (31.6) 9 (52.9) 14 (73.7) 0.034b

PPT (median (IQR), mm(Hg)) and Ln(PPT)a (mean (SD))Wrist (untransformed) 301.3 (141.7) 302.0 (177.3) 239.7 (167.7)Ln(PPT) 5.73 (0.27) 5.66 (0.40) 5.59 (0.31) 0.416c

Lumbar spine (untransformed) 352.7 (222.3) 288.7 (289.0) 183.0 (115.3)Ln(PPT) 5.84 (0.40) 5.72 (0.60) 5.14 (0.71) 0.001c

Heel (untransformed) 309.3 (151.0) 315.0 (159.0) 270.3 (109.3)Ln(PPT) 5.76 (0.36) 5.78 (0.40) 5.58 (0.34) 0.055c

Bold represent significant findings based on alpha of 0.05.a Natural log transformation.b Statistical test for group differences is chi-squared test.c Statistical test for group differences is analysis of variance test.

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318 315

PSQI and CPT measures, and between waist:hip ratio and PPTmeasures (Table 4). Therefore, DASS total, PSQI and sex wereincluded in multivariable models as potential confounders.

The results of the multivariable logistic regression modeladjusting for sex, DASS total and PSQI for CPT at the wrist showedstatistical evidence for group differences (Table 5). It was estimatedthat those patients in the NMP group had 18.4 (95% CI: 2.5e133.1,p¼0.004) times theodds of having anelevated (>15 �C)CPT to thosein the MP group. This estimate was similar to the unadjusted oddsratio (OR) of 17.3 (p¼ 0.001), meaning that sex, DASS total and PSQIwere not important confounders of the association between groupand CPT. CPT at the lumbar spine and heel sites was not statisticallysignificantly different between NMP and MP groups after adjust-ment for covariates. At the lumbar spine, patients in the NMP groupwere estimated to have 5.9 (95% CI: 0.9e38.4, p ¼ 0.064) times theodds of having an elevated (>15 �C) CPT to those in the MP groupafter adjustment for sex, DASS total and PSQI, with the adjusted ORwas similar in magnitude to the unadjusted estimate (6.0). At theheel, patients in the NMPhad 6.3 (95% CI: 0.9e41.5, p¼ 0.058) timesthe odds of having an elevated (>15 �C) CPTcompared to those in theMP group adjusting for sex, DASS total and PSQI. At this site theadjusted OR (6.3) was larger than the unadjusted OR (2.5) whichindicates the likely presence of negative confounding by covariates.At all sites there was no evidence that the MP group had greater orlesser odds than the control group for elevated CPT thresholds.

The results of the linear regression models for PPT provided noevidence for group differences at any site (Table 5). Although therewas some evidence that the NMP group had lower PPT than the MPgroup at the lumbar spine for the univariable model(difference: �0.58, 95% CI: �0.97 to �0.19, p ¼ 0.004), the modeladjusted for sex, DASS total and PSQI did not confirm a differenceexisted independently of these covariates (difference: �0.37, 95%CI: �0.83 to 0.09, p ¼ 0.117).

There were significant differences in the frequency of reportingof non-cold descriptors (at CPT) between groups at all three sites.Table 6 shows that the NMP group had the highest frequency ofnon-cold descriptors and controls the lowest. For the wrist andback sites the NMP group had a higher frequency than the MPgroup.

4. Discussion

This study lends support to the presence of differences in PSprofile between clinically determined subgroups of CNSLBP

participants based on their pain characteristics, whilst adjusting forpotential confounding factors known to influence sensory thresh-olds. The NMP group was estimated to have at least 2.5 times theodds of having cold hypersensitivity, as defined by a CPT greaterthan 15 �C at the wrist, when compared to the MP CNSLBP groupand the pain-free control group (95% CI for OR: 2.5e133.1,p ¼ 0.004), although the sample size was small and consequentlyconfidence intervals for group differences were wide. Estimates ofelevated CPT at the lumbar spine and heel were not statisticallysignificant meaning that the null hypothesis of no difference be-tween groups cannot be rejected, however the pattern of largerodds of having cold hypersensitivity in the NMP group is consistentacross all three sites, and it is possible that the lack of statisticalsignificance is due to the low power of the study to detect possiblysmaller effects at the lumbar spine and heel.

Whilst a lower PPT in the NMP group at the lumbar spine wasalso detected, interestingly there was no statistical evidence for anindependent group difference between the MP and control groupafter controlling for sex, sleep and psychological factors. Thesefindings suggest that the changes in PPTobserved in the NMP groupmay be mediated via gender differences, sleep deficits and/orpsychological distress highlighting the multidimensional nature ofPS. They also suggest that changes in PPT were limited to thelumbar test site. These findings however are at odds with previousreports where PPT was shown to be the best PS measure todistinguish a group of 40 patients with CNSLBP from pain-freecontrols after adjusting for age, gender, body compositions andpsychological factors (Neziri et al., 2012). The differences in thefindings may again reflect different patient profiles and methodo-logical differences.

The findings of our study may explain some of the conflictingand variable findings in the previous PS research into CNSLBP dis-orders (Lewis et al., 2010; Attal et al., 2011; Blumenstiel et al., 2011;O’Neill et al., 2011; Hubsher et al., 2013; Neziri et al., 2012; Rousselet al., 2013), suggesting that NSCLNP is not a homogeneous groupand that patient classification is one means by which to deal withthis problem. Other authors have also proposed the need to classifyNSCLBP patients based on neurophysiological mechanisms (Nijset al., 2010; Smart et al., 2010; Woolf, 2011). Smart et al. (2010)also described a group of CNSLBP patients with ‘central sensitisa-tion’, defined by pain that is diffuse, lacks clear proportionatemechanical characteristics and present with associated psycho-logical factors. They defined a ‘nociceptive’ CNSLBP group by painthat is more intermittent, localised and responds to clear

Table 3Association between participant group membership and sex, age, waist:hip ratio, DASS and PSQI scores.

Subgroup

Control (n ¼ 19) Mechanical (n ¼ 17) Non-mechanical (n ¼ 19) p-Value

Female sex (n (%)) 11 (57.9) 8 (47.1) 15 (79.0) 0.132b

Age (mean (SD)) 42.6 (14.9) 39.4 (14.2) 41.9 (13.9) 0.788c

Waist:hip ratio (mean (SD)) 0.83 (0.11) 0.85 (0.09) 0.85 (0.10) 0.696c

DASS total (0e126a) (median (IQR)) 10 (14) 20 (18) 30 (34) <0.001d

PSQI (0e21a) (mean (SD)) 4.3 (2.8) 7.7 (3.5) 11.0 (3.4) <0.001c

Bold represent significant findings based on alpha of 0.05.a Minimum to maximum score possible.b Statistical test for group differences is chi-squared.c Statistical test for group differences is analysis of variance.d Statistical test for group differences is KruskaleWallis test.

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318316

aggravating and easing factors (Smart et al., 2010). Although notpreviously investigated against PS measures, these profiles aresimilar to the NMP and MP CNSLBP groups described.

4.1. Possible pain mechanisms

It is proposed that central amplification of pain may be associ-ated with a number of changes within the central nervous system(CNS). These include neuronal hyperexcitability (Scott et al., 2005),enlarged receptor fields (Kasch et al., 2005), lowered thresholds ofsecond order neurons (Kasch et al., 2005), reduced recruitment ofpain modulating control systems (Campbell and Edwards, 2009),temporal summation (wind up) (Meeus and Nijs, 2007) andneuronal reorganisation (Woolf andMannion,1999). These changesmay be associated with a combination of factors including: sleepdysfunction, psychosocial factors and environmental/genetic in-teractions, influencing the immune-endocrine system and theneuromatrix, highlighting the complex multidimensional nature ofpersistent pain (O’Sullivan, 2012a). While peripheral sensitisationand some degree of segmental activity dependent central sensiti-sation might be anticipated for all patients with CNSLBP thedevelopment of widespread extra-segmental pressure hyperalgesiaor the development of multi-modality sensitisationwith either coldor heat hyperalgesia implies much more extensive changes in painprocessing within the CNS. These changes are likely to involvemany of the processes outlined above. These central amplificationprocesses may also be linked to the presence of more constant andpersistent pain.

Interestingly there was also a bimodal distribution for CPT in thepain-free control group demonstrating that a number of controlsubjects (26%) had elevated CPT, although their descriptors were

Table 4Associations between cold (�15 �C, >15 �C) and pressure pain (ln(PPT)) thresholdmeasures and sex, age, waist:hip ratio, DASS and PSQI scores.

CPT PPT

Wrist Lx Heel Wrist Lx Heel

Sex 0.116d 0.122d 0.080d �0.166c �0.236c �0.094c

Age 0.006c �0.133c L0.365c** 0.240b 0.254b 0.278b*Waist:hip

ratio0.001c �0.071c �0.040c 0.285a* 0.069a 0.273a*

DASS total 0.312c* 0.177c 0.038c 0.042b L0.394b** �0.110b

PSQI 0.336c* 0.290c* 0.141c �0.031a L0.400a** 0.550a

Bold represent significant findings based on alpha of 0.05.Lx ¼ lumbar spine.*p < 0.05; **p < 0.01.

a Measure of association is Pearson correlation coefficient.b Measure of association is Spearman correlation coefficient.c Measure of association is Point-biserial correlation coefficient.d Measure of association is Phi correlation coefficient.

different to the centrally amplified group. Whether this findingrepresents vulnerability in these subjects to future pain is notknown but has been hypothesised previously (Woolf, 2011).

The absence of PS differences between theMP and control groupmay reflect that MP participants’ spinal structures are sensitised tomovement and/or load, but not local pressure (O’Sullivan, 2005;Dankaerts et al., 2009), or that the site where the PPT testing wasapplied was not specific to their pain location. Participants in theMP CNSLBP group also displayed lower levels of psychosocial fac-tors and sleep disturbance (Meeus and Nijs, 2007). Further researchis required to determine the relationship between pain character-istics and PS to determine the role that central amplification has onpatient clinical profiles.

4.2. Clinical relevance

The classification of CNSLBP based on an underlyingmechanismhas been proposed to enhance targeted management (Deyo et al.,2009; O’Sullivan, 2012a; Woolf, 2011). While formal reliabilitytesting of clinicians ability to discriminate these groups was notcarried out, the results suggest that experienced physiotherapistswere able to identify patients with different PS profiles, based onroutine clinical examination (O’Sullivan, 2005; 2012b). Evaluationof cold hyperalgesia at the wrist may be combined with clinicalassessment of patients to provide quantitative confirmation forpatients who exhibit some degree of central amplification of theirpain. Recent research in neck pain subjects indicates a painresponse > 5/10 with the application of ice indicated a 90% likeli-hood of laboratory measured CPT being �13 �C (Maxwell andSterling, 2013). These findings may have clinical application tothe CNSLBP population described in this study, although furtherresearch is required to confirm this.

Although speculative it would be interesting to investigatewhether patients with MP respond better to locally targetedtreatments, whereas patients with NMP require management ap-proaches that more specifically address central pain mechanisms.Clearly further research is required in larger populations of patientswith CNSLBP to determine the validity of the clinical profiles anddetermine their predictive validity in relation to different targetedinterventions as well as the ability of less experienced clinicians toreliably identify these patients.

4.3. Methodological considerations/limitations

This study was only powered to detect large effect sizes, andconsequently may have failed to detect smaller but still clinicallymeaningful differences between groups. As a consequence of thesmall sample size, the confidence intervals for the elevated odds ofcold hypersensitivity in the NMP group were very wide, limiting

Table 5Adjusted and unadjusted parameter estimates for group membership from univariable and multivariable binary logistic regression (CPT: �15 �C, >15 �C) and general linear(ln(PPT)) models (multivariable models adjusted for gender, DASS total and PSQI).

Cold pain threshold Unadjusted Adjusted

Odds ratioa (95% CI) p-Value Odds ratio (95% CI) p-Value

Wrist Non-mechanical 17.3 (3.3, 91.7) 0.001 18.4 (2.5, 133.1) 0.004Control 1.2 (0.3, 5.3) 0.849 1.3 (0.2, 7.0) 0.759Mechanical REF REF

Lx Non-mechanical 6.0 (1.3, 28.5) 0.024 5.9 (0.9, 38.4) 0.064Control 1.0 (0.3, 3.8) 0.985 1.4 (0.3, 6.4) 0.637Mechanical REF REF

Heel Non-mechanical 2.5 (0.6, 10.1) 0.201 6.3 (0.9, 41.5) 0.058Control 0.4 (0.1, 1.6) 0.198 0.3 (0.1, 1.4) 0.134Mechanical REF REF

Pressure pain threshold b (95% CI)b p-Value b (95% CI) p-Value

Wrist Non-mechanical �0.07 (�0.29, 0.15) 0.551 �0.11 (�0.37, 0.15) 0.407Control 0.08 (�0.14, 0.30) 0.489 0.11 (�0.13, 0.35) 0.347Mechanical REF REF

Lx Non-mechanical �0.58 (�0.97, -0.19) 0.004 �0.37 (�0.83, 0.09) 0.117Control 0.12 (�0.27, 0.51) 0.547 0.04 (�0.38, 0.47) 0.838Mechanical REF REF

Heel Non-mechanical �0.19 (�0.43, 0.06) 0.133 �0.22 (�0.52, 0.08) 0.142Control �0.01 (�0.26, 0.23) 0.911 0.01 (�0.27, 0.28) 0.954Mechanical REF REF

Bold represent significant findings based on alpha of 0.05.a The ratio of the odds for cold pain threshold > 15 �C for the test group (non-mechanical or control) to the reference group (mechanical pain group).b Coefficient represents difference in the natural log of pressure pain threshold between test groups (non-mechanical or control) and the reference group (mechanical pain

group, positive value indicates higher value in test group).

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318 317

precise estimation of this association in the population understudy. Furthermore, for PPT, testing at a generic site at the lumbarspine may not detect local hyperalgesia. Previous authors havesuccessfully tested PPT in patients with CNSLBP at the site of mostsevere pain and found it to be predictive with pain, independent ofpotential confounders (Neziri et al., 2012). Pain medication usewas only reported in CNSLBP participants, which precluded use ofthis variable in the multivariable models presented. However,group differences in medication use were not large (Table 6) andcurrent pain medication use was not associated with CPT or PPT.Other variables only assessed in CNSLBP participants and associ-ated with pain group membership, such as pain intensity, numberof pain areas, RMDQ and SBST, were not considered as con-founders of the group membership/pain threshold association inthis study as they represented part of the common clinical profileof these groups and thus consequences of differential sensoryprocessing. The inter-therapist reliability of the ability of Physio-therapists to differentiate NP from NMP requires further investi-gation and larger studies are required to verify these results.

5. Conclusion

This study provides preliminary evidence that two patientgroups with CNSLBP identified clinically, can be distinguishedbased on their PS profile. When used in conjunction with sound

Table 6Frequency of non-cold descriptorsa cited during the cold pain testing.

Controls(n ¼ 19)

MechanicalCNSLBP (n ¼ 17)

Non-mechanicalCNSLBP (n ¼ 19)

p-Valueb

N (% of group)CPT wrist 3 (15.8) 6 (35.3) 12 (63.2) 0.010CPT L5/S1 7 (36.8) 12 (70.6) 18 (94.7) 0.001CPT heel 3 3 (15.8) 9 (52.9) 11 (57.9) 0.017

a Non-cold descriptors used were ‘burning, ice, sharp, sting, gnawing andfreezing’. Any other descriptions were not included as non-cold descriptors. Thesewere ‘pressure, nice, cold, tingling, pleasant and numb’.

b Chi-squared test.

clinical reasoning, these profiles may help clinicians more accu-rately identify mechanisms that underlie CNSLBP and better targetinterventions. While these pain profiles have yet to be furthervalidated, they provide a framework for future research.

Author contributions

All authors discussed the results and contributed to themanuscript.

References

Attal N, Perrot S, Fermanian J, Bouhassira D. The neuropathic components of chroniclow back pain: a prospective multicenter study using the DN4 questionnaire.J Pain 2011;12:1080e7.

Blumenstiel KMD, Gerhardt AMA, Rolke RMD, Bieber CMD, Tesarz JMD,Friederich H-CMD, et al. Quantitative sensory testing profiles in chronic backpain are distinct from those in fibromyalgia. Clin J Pain 2011;27:682e90.

Buysse D, Reynolds C, Monk T, Berman S, Kupfer D. The Pittsburgh sleep qualityindex: a new instrument for psychiatric practice and research. Psychiatry Res1989;28:193e213.

Campbell C, Edwards R. Mindebody interactions in pain: the neurophysiology ofanxious and catastrophic pain-related thoughts. Transl Res 2009;153:97e101.

Carli G, Suman AL, Biasi G, Marcolongo R. Reactivity to superficial and deep stimuliin patients with chronic musculoskeletal pain. Pain 2002;100:259e69.

Chien A, Sterling M. Sensory hypoaesthesia is a feature of chronic whiplash but notchronic idiopathic neck pain. Man Ther 2010;15:48e53.

Dankaerts W, O’Sullivan P, Burnett A, Straker L. Differences in sitting postures areassociated with nonspecific chronic low back pain disorders when patients aresubclassified. Spine 2006;31:698e704.

Dankaerts W, O’Sullivan P, Burnett A, Straker L, Davey P, Gupta R. Discriminatinghealthy controls and two clinical subgroups of nonspecific chronic low backpain patients using trunk muscle activation and lumbosacral kinematics ofpostures and movements. Spine 2009;34:1610e8.

Deyo RA, Mirza SK, Turner JA, Martin BI. Overtreating chronic back pain: time toback off? J Am Board Fam Med 2009;22:62e8.

Dunn W. The impact of sensory processing abilities on the daily lives of youngchildren and their families: a conceptual model. Infants Young Child 1997;9:23e35.

Freynhagen R, Baron R, Gockel U, Tolle T. painDETECT: a new screening question-naire to identify neuropathic components in patients with back pain. Curr MedRes Opin 2006;22:1911e20.

Giesecke T, Gracely RH, Grant MAB, Nachemson AL, Petzke F, Williams DA, et al.Evidence of augmented central pain processing in idiopathic chronic low backpain. Arthritis Rheum 2004;50:613e23.

P. O’Sullivan et al. / Manual Therapy 19 (2014) 311e318318

Heffner K, France C, Trost Z, Ng H, Pigeon W. Chronic low back pain, sleep distur-bance, and interleukin-6. Clin J Pain 2011;27:35e41.

Hill JC, Dunn KM, Lewis M, Mullis R, Main CJ, Foster NE, et al. A primary care backpain screening tool: identifying patient subgroups for initial treatment.Arthritis Rheum 2008;59:632e41.

Hubsher M, Moloney N, Leaver A, Rebbeck T, McAuley J, Refshauge K. Relationshipbetween quantitative sensory testing and pain or disability in people withspinal pain e a systematic review and meta-analysis. Pain 2013;154:1497e504.

Huskisson EC. Measurement of pain. Lancet 1974;2:1127e31.Jones D. Testeretest reliability of pressure pain threshold measurements of the

upper limb and torso in young healthy women. J Pain 2007;8:650e6.Karayannis NV, Jull GA, Hodges PW. Physiotherapy movement based classification

approaches to low back pain: comparison of subgroups through review anddeveloper/expert survey. BMC Musculoskelet Disord 2012;13:24.

Kasch H, Qerama E, Bach FW, Jensen TS. Reduced cold pressor pain tolerance in non-recovered whiplash patients: a 1-year prospective study. Eur J Pain 2005;9:561e9.

Leboeuf-Yde C, Nielsen J, Kyvik K, Fejer R, Hartvigsen J. Pain in the lumbar,thoracic and cervical regions: do age and gender matter? A population-basedstudy of 34,902 Danish twins 20e71 years of age. BMC Musculoskelet Disord2009;10.

Lewis C, Khan A, Souvlis T, Sterling M. Sensory characteristics of tender points in thelower back. Man Ther 2010;15:536e41.

Lovibond S, Lovibond P. Manual for the depression anxiety stress scales; 1995.Maxwell S, Sterling M. An investigation of the use of a numeric pain rating scale

with ice application to the neck to determine cold hyperalgesia. Man Ther2013;18:172e4.

Meeus M, Nijs J. Central sensitization: a biopsychosocial explanation for chronicwidespread pain in patients with fibromyalgia and chronic fatigue syndrome.Clin Rheumatol 2007;26:465e73.

Moloney NA, Hall TM, Doody CM. Reliability of thermal quantitative sensorytesting: a systematic review. J Rehabil Res Dev 2012;49:191e207.

Mosek A, Yarnitsky D, Korczyn AD, Niv D. The assessment of radiating low back painby thermal sensory testing. Eur J Pain 2001;5:347e51.

Neziri A, Curatolo M, Limacher A, Nüesch E, Radanov B, Andersen O, et al. Rankingof parameters of pain hypersensitivity according to their discriminative abilityin chronic low back pain. Pain 2012;153:2083e91.

Nijs J, Van Houdenhove B, Oostendorp RAB. Recognition of central sensitization inpatients with musculoskeletal pain: application of pain neurophysiology inmanual therapy practice. Man Ther 2010;15:135e41.

O’Sullivan P. Diagnosis and classification of chronic low back pain disorders: mal-adaptive movement and motor control impairments as underlying mechanism.Man Ther 2005;10:242e55.

O’Sullivan P, Straker L, Smith A, Perry M, Kendall G. Carer experience of back pain isassociated with adolescent back pain experience even when controlling forother carer and family factors. Clin J Pain 2008;24:226e31.

O’Sullivan P. It’s time for change in the management of non-specific chronic lowback pain. Br J Sports Med 2012a;46:224e7.

O’Sullivan PB. A classification-based cognitive functional approach for the man-agement of back pain. J Orthop Sports Phys Ther 2012b;42:A17e21.

O’Neill S, Kjær P, Graven-Nielsen T, Manniche C, Arendt-Nielsen L. Low pressurepain thresholds are associated with, but does not predispose for, low back pain.Eur Spine J 2011;20:2120e5.

Roland M, Fairbank J. The RolandeMorris disability questionnaire and the Oswestrydisability questionnaire. Spine 2000;25:3115e24.

Roland M, Morris R. A study of the natural history of back pain. Spine 1983;8:141e4.Rolke R, Baron R, Maier C, Tolle TR, Treede RD, Beyer A, et al. Quantitative sensory

testing in the German research network on neuropathic pain (DFNS): stan-dardized protocol and reference values. Pain 2006a;123:231e43.

Rolke R, Baron R, Maier C, Tölle TR, Treede RD, Beyer A, et al. Quantitative sensorytesting in the German research network on neuropathic pain (DFNS): stan-dardized protocol and reference values. Pain 2006b;123:231e43.

Roussel NAPMPT, Nijs JPMPT, Meeus MPPT, Mylius VPMD, Fayt CPMD,Oostendorp RPMPTPT. Central sensitization and altered central pain processingin chronic low back pain: fact or myth? Clin J Pain 2013;29:625e38.

Scott D, Jull G, Sterling M. Widespread sensory hypersensitivity is a feature ofchronic whiplash-associated disorder but not chronic idiopathic neck pain. ClinJ Pain 2005;21:175e81.

Slater H, Arendtnielsen L, Wright A, Gravennielsen T. Sensory and motor effects ofexperimental muscle pain in patients with lateral epicondylalgia and controlswith delayed onset muscle soreness. Pain 2005;114:118e30.

Smart K, Blake C, Staines A, Doody C. Clinical indicators of ‘nociceptive’, ‘peripheralneuropathic’ and ‘central’mechanisms of musculoskeletal pain. A Delphi surveyof expert clinicians. Man Ther 2010;15:80e7.

Spitzer WO. Scientific approach to the assessment and management of activity-related spinal disorders: a monograph for clinicians. Report of the Quebectask force on spinal disorders. Spine 1987;12:1e59.

Vibe Fersum K, O’Sullivan PB, Kvåle A, Skouen JS. Inter-examiner reliability of aclassification system for patients with non-specific low back pain. Man Ther2009;14:555e61.

Vibe Fersum K, O’Sullivan P, Skouen JS, Smith A, Kvåle A. Efficacy of classificationbased ‘cognitive functional therapy’ in patients with Non Specific Chronic LowBack Pain – a randomized controlled trial. Eur Pain J 2012;17(6):916e28.

Wasner GL, Brock JA. Determinants of thermal pain thresholds in normal subjects.Clin Neurophysiol 2008;119:2389e95.

Wolfe F, Clauw D, Fitzcharles M, Goldenberg D, Katz R, Mease P, et al. The AmericanCollege of Rheumatology preliminary diagnostic criteria for fibromyalgia andmeasurement of symptom severity. Arthritis Care Res 2010;62:600e10.

Woolf C, Mannion R. Neuropathic pain: aetiology, symptoms, mechanisms, andmanagement. Lancet 1999;353:1959e64.

Woolf CJ. Central sensitization: implications for the diagnosis and treatment of pain.Pain 2011;152:S2e15.

Wright A, Thurnwald P, O’Callaghan J, Smith J, Vicenzino B. Hyperalgesia in tenniselbow patients. J Musculoskelet Pain 1994;2:83e97.

Zwart J-A, Trond S. Repeatability of dermatomal warm and cold sensory thresholdsin patients with sciatica. Eur Spine J 2002;11:441e6.