ORIGINAL ARTICLE
A mechanism-based pain sensitivity index to characterize knee osteoarthritis patients with different disease stages and pain levels L. Arendt-Nielsen1, L.L. Egsgaard1,2, K.K. Petersen1, T.N. Eskehave1,2, T. Graven- Nielsen1, H.C. Hoeck2,3, O. Simonsen4 1 2 3 4
Center for Sensory-Motor Interaction, Aalborg University, Denmark CCBR, Aalborg, Denmark C4Pain, Aalborg, Denmark Sygehus Vendsyssel, Frederikshavn, Denmark
Correspondence Lars Arendt-Nielsen E-mail: [email protected] Funding sources The study was supported by The Danish National Advanced Technology Foundation, Aase and Ejnar Danielsens Foundation, Lions Club in Denmark, The Danish Council for Technology and Innovation (09-052174), The Bevica Foundation and The Danish Rheumatic Association. Conflicts of interest None declared. Accepted for publication 25 November 2014 doi:10.1002/ejp.651
Abstract Background: In a cohort of well-characterized patients with different degrees of knee osteoarthritis (OA) and pain, the aims were to utilize mechanism-based quantitative sensory testing (QST) to (1) characterize subgroups of patients; (2) analyse the associations between clinical characteristics and QST; and (3) develop and apply a QST-based knee OA composite pain sensitivity index for patient classification. Methods: Two hundred seventeen OA pain patients and 64 controls were included. Kellgren and Lawrence (KL) grading scores were obtained, and pressure pain thresholds (PPTs), temporal summation of pain to repeated painful pressure stimulation and conditioning pain modulation (CPM) were assessed. Associations between pain score/area/duration, radiological findings and QST-related parameters were analysed. A pain sensitivity index was developed and applied based on PPT, temporal summation and CPM. z-Score, as statistical tool, was calculated for statistically comparing the pain index of a single patient with a healthy control group. Results: High knee pain associated with low KL grade showed particular signs of pain sensitization. Patients showed significant associations between clinical knee pain intensity/duration and lowering of knee PPTs (p < 0.01), facilitation of temporal summation (p < 0.01), reduction of CPM function (p < 0.01) and high pain sensitivity index (p < 0.01). The index classified 27–38% of the OA patients and 3% of the controls as highly sensitive with no association to KL. The index increased for high knee pain intensities and long pain duration. Conclusions: Radiological scores, contrary to clinical pain intensity/ duration, were poorly associated with QST parameters. The pain sensitivity index could classify OA patients with different degrees of OA and pain.
1. Introduction Osteoarthritis (OA) is a common condition with increasing prevalence (Woolf and Pfleger, 2003; Jinks et al., 2007; Neogi, 2013) due to demographic and lifestyle changes. Pain and impaired function are the clinical representations of OA (Breivik et al., 2008; 1406 Eur J Pain 19 (2015) 1406--1417
Read and Dray, 2008; Neogi, 2013). Furthermore, psychosocial variables influence the pain and contribute to the variability in pain (Sluka et al., 2012). For the individual patient, severe knee joint damage may cause little pain and vice versa (Davis et al., 1992; Hannan et al., 2000; Neogi et al., 2009), but at a population level this discordance is less evident (Bedson © 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
What’s already known about this topic? • Pain is most often not associated with clinical features such as joint damage in osteoarthritis. • Pain sensitization seems to play an important role in osteoarthritis. What does this study add? • Use of quantitative sensory testing to characterize patient groups with knee osteoarthritis (OA). • Associate clinical features of OA with quantitative sensory testing. • Developing and applying a composite pain sensitivity index for profiling OA patients.
and Croft, 2008). To some degree this diversity in the clinical presentation of OA pain (Lanyon et al., 1998; Dieppe and Lohmander, 2005; Felson, 2005) has been associated with different degrees of pain sensitization in individual patients (Hannan et al., 2000; Dieppe and Lohmander, 2005; Finan et al., 2013; Schiphof et al., 2013). Pain intensity and OA duration are found to be associated with this development of sensitization (Fernández-de-las-Peñas et al., 2007, 2009; Arendt-Nielsen et al., 2010). If the peripheral nociceptive drive from the joint is reduced/terminated (e.g., knee replacement), the pain sensitization will normalize (Graven-Nielsen et al., 2012; Aranda-Villalobos et al., 2013), but unfortunately 10–34% of the patients develop chronic postoperative pain (Beswick et al., 2012) with continued pain sensitization (Petersen et al., 2015). Therefore, it would have major impact on management strategies and drug development if quantitative biomarkers could be developed for phenotyping OA pain patients to design optimal individualized management strategies. Quantitative sensory testing (QST) has the potential to be used for phenotyping of pain patients (Rolke et al., 2006a,b). In OA, pressure pain algometry has mainly been used to quantify localized muscle/joint pain sensitivity (Kosek and Ordeberg, 2000; Bajaj et al., 2001; Imamura et al., 2008; Arendt-Nielsen et al., 2010; Graven-Nielsen et al., 2012; Suokas et al., 2012; Skou et al., 2014). Recently, more advanced QST technologies have been developed and applied in OA to assess changes in the central pain mechanism such as temporal summation (TS) and descending pain inhibition (Arendt-Nielsen et al., 2010; Graven-Nielsen et al., 2012). Most QST studies in OA (1) are based on relatively small cohorts of patients with insufficient power to © 2014 European Pain Federation - EFICâ
grade the responses according to the pain status; (2) use only few QST parameters; and (3) do not utilize the complementary information that can be gained from combining a variety of mechanism-based QST tools. The aims of the present cohort study in patients with different degrees of knee OA and pain were to investigate how to utilize mechanism-based QST to (1) characterize subgroups of patients; (2) analyse the associations between clinical characteristics (pain intensity, pain area, pain duration, radiological scores) and mechanism-based QST [pressure pain, pain sensitivity maps, TS, conditioning pain modulation (CPM)]; and (3) develop and apply a QST-based knee OA composite pain sensitivity index for patient classification.
2. Methods 2.1 Participants A total of 217 patients (115 women) and 64 control subjects (32 women) were enrolled (Table 1). The patients and control subjects were recruited from the patient database at CCBR (https://dk.ccbr.com/ccbr-aalborg) and the clinical trial unit C4Pain (www.C4Pain.com), Aalborg, Denmark. The database consists of volunteers who are interested in participating in studies and clinical trials focusing on knee OA. All the patients in the database have verified knee OA and have been referred from GPs, physiotherapy clinics or recruited through newspaper advertisements. Patients from the database were invited to participate in the present study. Those responding were clinically evaluated and scored according to the American College of Rheumatology (ACR) OA criteria (Altman et al., 1986). If the diagnosis was confirmed and they fulfilled the ACR OA criteria, they were included in the study. If X-ray images were more than 6 months old, new X-ray images were taken and radiologically scored (see below). None of the participants included in the study had participated in other clinical trials for the last 12 months. The control subjects were mainly recruited among patients’ family members, and they were to fulfil the same exclusion criteria. Control subjects had no symptomatic OA with maximal pain less than 10 mm for the last 24 h on the visual analogue scale (VAS). The control subjects were invited to have X-ray images taken. The exclusion criteria comprised secondary OA, psychiatric conditions hindering the ability of the patient to participate in the study, pregnancy, previous/current drug(s) or alcohol abuse, previous neurological or concomitant musculoskeletal disorders, lack of collaboration ability and subjects unable to abstain from analgesic medication for at least 72 h prior to the knee pain assessment. The data collected from the cohort of patients participating in this study may provide the basis for additional studies. Eur J Pain 19 (2015) 1406--1417
1407
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 1 Demographic data from the low-pain control group [VAS (0–9), n = 64] and the three pain groups (n = 217). Groups
Controls (VAS 0–9)
VAS 10–39
VAS 40–69
VAS 70–100
Men (n) Women (n) Total (n) Age ± SD (years) BMI ± SD (kg/m2)
32 32 64 67.5 ± 4.4 27.1 ± 4.1
35 46 81 62.8 ± 8.3 27.5 ± 3.7
34 36 70 64.9 ± 7.6 28.0 ± 3.3
33 33 66 63.8 ± 8.1 29.6 ± 5.1
KL = 0 KL = 1 KL = 2 KL = 3 KL = 4
3 25 28 0 1
4 17 46 11 3
3 11 33 12 11
2 6 38 13 7
Men (n)
Women (n)
Age ± SD (years)
BMI ± SD (kg/m2)
8 28 66 19 10
4 31 79 17 12
60.1 ± 8.7 63.7 ± 8.0 64.7 ± 7.3 64.3 ± 7.1 67.8 ± 7.7
26.3 ± 3.3 27.0 ± 4.0 28.2 ± 3.8 29.3 ± 5.6 29.5 ± 3.9
KL = 0 KL = 1 KL = 2 KL = 3 KL = 4
Total 12 59 145 36 22
X-rays were not taken from seven control subjects with VAS = 0. The four groups are separated based on the clinical pain ratings (maximal pain the last 24 h). BMI, body mass index; KL, Kellgren and Lawrence; SD, standard deviation; VAS, visual analogue scale.
2.2 Patient characterization 2.2.1 Radiologic assessment of OA All participants had a bilateral weight bearing, 10° fixed flexion posteroanterior X-ray taken of the knees. The medial and lateral tibiofemoral joint space and tibial plateau were measured using a posteroanterior view. The superior and inferior osteophytes were evaluated. The X-ray beam was centred on the joint line of the knee in the popliteal space of a 10° caudal angulation. The degree of radiologic OA was evaluated using Kellgren and Lawrence grading scale (KL0-4) (Kellgren and Lawrence, 1957). The KL grading did not include the patellofemoral joint.
2.2.2 Clinical assessment of pain The maximal pain intensity during the past 24 h prior to the visit was assessed on a 100 mm VAS where 0 mm defined ‘no pain’ and 100 mm defined ‘maximal pain’. The participants were informed about the pain scale, instructed in detail how to use it and informed that the maximal knee intensity pain over the last 24 h could have occurred during rest, movement or during the night. The participants were grouped according to the maximal knee pain intensity as: VAS (0–9) (low-pain control group), VAS (10–39), VAS (40–69) and VAS (70–100). The knee in which OA was diagnosed and which showed the highest pain intensity was identified as the index knee. Clinical pain areas (individual conditions evoking pain, e.g., walking) and soreness areas (evoked by palpation) were mapped by the subjects on the anterior and posterior aspects of the index knee and later quantified in arbitrary units using a digitizer software (ACECAD, model 1408 Eur J Pain 19 (2015) 1406--1417
D9000, Taipei, Taiwan). The subjects were asked to indicate the duration (in years) during which they had had knee pain (in case they had any pain). This knee pain duration is often longer than the one of clinically verified OA. As all participants recruited were cognitively and adequately functioning, the instruction procedure could be highly standardized.
2.3 QST for characterizing subgroups of OA patients 2.3.1 Pressure pain threshold (PPT) assessment A computer-controlled pressure algometer (Center for Sensory-Motor Interaction, Aalborg University, Aalborg, Denmark) was used to assess the PPTs. The pressure algometer consisted of a round aluminium footplate with a padded contact surface of 1 cm2 fixed to the top of a piston. The pressure stimulation was feedback controlled via recordings of the actual force. The computer-controlled pressure algometer was placed perpendicularly to the test site and applied an increasing pressure of 60 kPa/s until the subjects defined the pressure as pain and pressed a stop button. For all experimental procedures, the subjects were asked to rest in a supine position with their target knee mildly raised and placed on a vacuum pillow in order to fixate the knee when assessing the PPTs. Ten test sites were located and marked: eight test-specific sites in the peripatellar region and two control points according to our previous recommendations (Arendt-Nielsen et al., 2010). The 10 points were defined as (1) 2 cm distal to the inferomedial edge of patella; (2) 2 cm distal to the inferolateral edge of patella; (3) 3 cm lateral to the centre of the lateral edge of patella; (4) 2 cm © 2014 European Pain Federation - EFICâ
L. Arendt-Nielsen et al.
proximal to the superolateral edge of patella; (5) 2 cm proximal to the superior edge of patella; (6) 2 cm proximal to the superomedial edge of patella; (7) 3 cm medial to the centre of the medial edge of patella; (8) on the centre of patella; (9) the belly of the tibialis anterior muscle; and (10) the belly of the extensor carpi radialis muscle. The PPTs were measured three times with an interval of minimum 25 s at each site, and the average of the three determinations was calculated for further analysis. The sequence for assessing the 10 individual locations was randomized.
2.3.2 Pressure pain sensitivity maps Pressure pain sensitivity maps (PPT maps) were constructed by interpolation of the mean PPT values across all subjects from the peripatellar sites using an inverse distanceweighted interpolation to compute PPT values of unknown locations by applying the known mean PPT values and locations (Arendt-Nielsen et al., 2010; Binderup et al., 2010). Those PPT maps were projected onto three-dimensional MRI-based contour models of a knee to obtain a visual impression of the pressure pain sensitivity distribution. The PPT maps were developed in relation to the clinical groupings based on the knee VAS pain scores and KL grading.
2.3.3 TS of pressure pain The computer-controlled pressure algometer (Center for Sensory-Motor Interaction, Aalborg University) was used to assess TS to repeated pressure-induced pain. The mean PPT values from the different locations were used to examine the extent of TS by applying 10 successive stimuli with an intensity at mean PPT and delivered at a frequency of 0.5 Hz (1 s stimulation and 1 s interval without stimulation) (Nie et al., 2005; Arendt-Nielsen et al., 2010). A constant contact pressure of 60 kPa was maintained between each stimulus, which was not painful. During repeated stimuli, the pain intensity was assessed by the patient on an electronic VAS ranging from 0 to 100 mm (Center for Sensory-Motor Interaction, Aalborg University) with end points defined as no pain and maximal pain, respectively. The electronic VAS ratings were interfaced to a computer using the LabViewbased Mr. Kick software (Center for Sensory-Motor Interaction, Aalborg University, Denmark). The VAS following each stimulus was extracted and the sum of the 10 increasing VAS scores (VAS sum) was used for analysis. This VAS sum was used as this measure was associated with other measures of summation such as the slope of increase in VAS or the ratio between the VAS from the 10th and the first stimulus. The mean VAS sum elicited from stimulation of sites 1–8 was used for statistical analysis.
2.3.4 CPM A computer-controlled cuff algometer (NociTech, Aalborg, Denmark) was used to evoke CPM (Yarnitsky et al., 2010). A © 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
24 in. tourniquet cuff was placed approximately 3 cm proximal to the cubital fossa on the contralateral arm to the target knee. The tourniquet applied a constant pressure of 36 kPa (approx. 270 mmHg) during which the subjects reported their pain on the electronic VAS. Two minutes after the subjects had reached a VAS score of 60 mm, the PPT recordings of the 10 points were commenced in randomized order. The PPTs were measured once at each of the 10 points. If needed, the volunteers could increase the pain intensity by performing grip movements around a soft rubber ball. For the knee, the mean effect assessed from points 1–8 was calculated. The mean ratio between PPT during cuff stimulation and PPT at baseline was defined as the CPM ratio.
2.4 Pain sensitivity index and z-score analysis Among all the relevant parameters to be included in a composite measure (the pain sensitivity index), an analysis was made to find the parameters that provided the most complementary information associated with pain sensitivity, i.e., the parameter correlating (Pearson’s correlation) least with the most sensitive PPT value measured from the knee. The PPT value from the knee was chosen as the reference for the analysis as this location is the main origin of the clinical pain problem. The rationale for this analysis was to identify which parameters in combination provided most entropy (obtaining maximal information from the data). The two parameters least correlated with the PPT value from the most sensitive location on the knee were the TS provoked for tibialis anterior (VASsumTA, Pearson correlation −0.059; two-tailed p-value 0.327) and the CPM ratio based on the test stimulus applied to tibialis anterior (CPMTA Pearson correlation −0.053; two-tailed p-value 0.379). The two parameters strongest correlated with the knee PPT value were PPT from the tibialis anterior (Pearson correlation 0.970; two-tailed p-value < 0.001) and the PPT assessed from the arm (Pearson correlation 0.967; two-tailed p-value < 0.001). Based on this information the pain sensitivity index was constructed. The index was used to calculate a z-score for all participants. Calculating the z-score from the index transforms the data into a standard normal distribution with zero mean and unit variance, i.e., z-values above 0 indicate a gain in index (sensitized), and z-values below 0 indicate a reduction in index (non-sensitized). The use of z-scores facilitates the comparison of the pain sensitivity of a single patient with a group of healthy control subjects (Rolke et al., 2006a,b). A principal component analysis (PCA) was also used as a tool to reduce the number of parameters to be included in the pain sensitivity index. The PCA is based on linear relationships like Pearson’s correlation. The PCA algorithm cannot take the a priori clinical considerations into account (e.g., we would like to use the PPTKneeMostSensitive as one component in the pain sensitivity index). The data from the Pearson’s correlation and the PCA were compared to ensure that the optimal set of data was selected for the calculation of the index. Eur J Pain 19 (2015) 1406--1417
1409
Pain sensitization in osteoarthritis
2.5 Statistical analysis Data are presented as mean and standard deviation (SD). The statistical analysis was conducted using SPSS 19.0 software (SPSS Inc., Chicago, IL, USA). In case of significant factors detected by the analysis of variance (ANOVA), Bonferroni post hoc test was performed to identify the pattern of significant differences. The level p < 0.05 was considered significant. The statistical analyses for the different parameters were performed according to the following plan.
L. Arendt-Nielsen et al.
2.8.1.1 Correlation between maximal pain intensity during the past 24 h and QST parameters Due to a linear distribution of the parameter, Pearson correlations were used for linear estimates of the correlation between maximal pain intensity during the past 24 h and the other QST parameters (PPT, TS, CPM).
2.8.1.2 Correlation between PPT measured from the most sensitive part of the knee and the other QST parameters
2.6 Patient characterization
A one-way ANOVA was performed for the demographic parameters (gender, age and BMI) to test for differences between groups [control subjects, VAS (0–9), VAS (10–39), VAS (40–69), VAS (70–100)].
To identify which QST parameters correlated least with the selected reference (PPT from the most sensitive point on the knee), a correlation matrix composed of Pearson correlations was used. This served as an analysis to identify which parameters should be included in the sensitization index. Pearson correlations were performed due to a linear distribution of the parameters.
2.6.2 Comparing OA groups with different pain ratings and KL scores
2.8.1.3 Correlation between clinical and QST parameters
The combination of different pain ratings and KL scores resulted in four groups (‘low pain’/‘low KL’; ‘low pain’/‘high KL’; ‘high pain’/‘low KL’; ‘high pain’/‘high KL’) where low pain was defined as VAS (0–51), and high pain was defined as VAS (52–100); low KL was defined as KL 0, 1 and 2, and high KL was defined as KL 3 and 4. A one-way ANOVA, adjusted for age, BMI and gender was performed for each assessment site (knee, arm, TA) on PPT, TS, CPM measurements and z-scores to test for differences between these pain/KL groups. The pressure pain sensitivity maps for the four groups were not analysed statistically.
Spearman’s correlations of clinical parameters (pain intensity, pain area, pain duration, radiological scores) and the mechanism-based QST responses were performed. Spearman’s correlations were performed due to non-linear distribution of some parameters.
2.6.1 Demographic data
2.7 QST for characterizing subgroups of OA patients
2.9 Pain sensitivity index The z-statistics used for classification of patients based on the pain sensitivity index as described in the sections above.
3. Results 3.1 Patient characterization
2.7.1 Group comparison of PPTs, TS and CPM
3.1.1 Demographic data
A one-way ANOVA was performed for each assessment site (knee, arm, TA) on PPT, TS and CPM measurements to test for differences between groups [control subjects; VAS (10– 39); VAS (40–69); VAS (70–100)]. The pressure pain sensitivity maps were not statistically analysed.
The demographic data from the study population are summarized in Table 1.
2.8 Association between clinical features and QST
The experimental pain measures for the four combinations (high pain, low pain, high KL, low KL) are given in Table 2. The high pain/low KL group showed significantly higher pain sensitivity and a higher proportion of sensitized subjects when compared with the low pain/low KL group as also reflected in the pressure pain sensitivity maps (Fig. 1A). The pain sensitivity index at group level (z-score; Table 2) was
2.8.1 Correlation analyses A variety of correlation analyses were used and the individual tests are described below. For all analyses the correlation coefficient (r), slopes and significance levels (p) were calculated. 1410 Eur J Pain 19 (2015) 1406--1417
3.1.2 Comparing OA groups with different pain ratings and KL scores
© 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 2 The experimental pain responses in four different OA patient groups (n = 217) based on high (VAS > 51), low (VAS < 51) last 24 h maximal VAS and on high KL (3 and 4) and low KL (1 and 2). Group/pain response
Mean PPT (1–8) (kPa)
Mean PPT (9–10) (kPa)
Mean CPM ratio (1–8)
Mean VAS sum (1–8)
z-score
% of sensitized
Low pain/low KL (n = 85) Low pain/high KL (n = 24) High pain/low KL (n = 76) High pain/high KL (n = 32)
448 ± 230 471 ± 222 379 ± 212 385 ± 147
373 ± 173 398 ± 179 320 ± 167 341 ± 134
1.277 ± 0.752 1.087 ± 0.603 0.927 ± 0.679 0.866 ± 0.448
25.203 ± 16.527 29.415 ± 20.921 43.283 ± 22.622 34.533 ± 18.923
1.6 ± 3.2 1.3 ± 2.2 2.9 ± 4.9 2.0 ± 2.9
30 25 39 34
Knee test points are 1–8 and the leg/arm test points 9–10. The z-score of the pain sensitivity index is calculated for the group and the proportions (% of patients) of sensitized (z-score >1.96) are indicated. The connecting lines indicate significant differences between groups (p < 0.02). The statistical analysis was adjusted for age, gender and body mass index. Mean ± standard deviation values are given. CPM, conditioning pain modulation; KL, Kellgren and Lawrence; OA, osteoarthritis; PPT, pressure pain threshold; VAS, visual analogue scale.
Figure 1 (A) The mean pressure pain threshold knee maps for four different patient group combinations based on high [visual analogue scale (VAS) ≥ 51], low (VAS < 51) maximal VAS during the last 24 h and on high Kellgren and Lawrence (KL) (3 and 4) and low KL (0, 1 and 2). (B) The top row shows the mean pressure pain threshold knee maps for the control subjects (VAS 0–9) and the three knee pain groups with different maximal VAS during the last 24 h. The bottom panel shows the mean pressure pain threshold knee maps for the different patient groups based on KL. KL 0 and KL 1 were combined.
© 2014 European Pain Federation - EFICâ
Eur J Pain 19 (2015) 1406--1417
1411
Pain sensitization in osteoarthritis
1412 Eur J Pain 19 (2015) 1406--1417
– r = −0.429** p = 0.000 r = −0.533** p = 0.000 –
–
–
r = 0.311** p = 0.000
r = 0.674** p = 0.000 r = 0.523** p = 0.000 r = 0.338** p = 0.000 r = −0.197** p = 0.001 r = −0.193** p = 0.001 r = 0.285** p = 0.000 –
r = 0.321** p = 0.000
r = 0.227** p = 0.000 r = −0.157** p = 0.008 r = 0.311** p = 0.000
r = 0.119* p = 0.046 –
r = 0.889** p = 0.000 – –
–
r = 0.437** p = 0.000 r = 0.274** p = 0.000 r = −0.160** p = 0.007 r = −0.124* p = 0.038 r = 0.243** p = 0.000 –
r = 0.285** p = 0.000 r = −0.131* p = 0.029 –
–
r = −0.301** p = 0.000 r = 0.142** p = 0.017
Mean VAS sum (1–8) Mean PPT (9–10) Mean PPT (1–8) Kellgren and Lawrence OA duration (years)
z-Score pain sensitivity index
Mean CPM ratio (1–8)
Mean VAS sum (1–8)
Mean PPT (9–10)
Mean PPT (1–8)
Kellgren and Lawrence
OA duration (years)
Soreness area
r = 0.544** p = 0.000 r = 0.520** p = 0.000 r = 0.562** p = 0.000 r = 0.303** p = 0.000 r = −0.342** p = 0.000 r = −0.302** p = 0.000 r = 0.482** p = 0.000 r = −0.345** p = 0.000 r = 0.412** p = 0.000
For the TS effect (mean of all 10 test points), the ANOVA showed a significant difference between groups (p < 0.001). The TS effect was significantly lower in low knee pain control subjects with VAS (0–9) compared with knee pain patients with VAS (40–69) (p < 0.001, 18.3 ± 14.2 vs. 38.7 ± 22.3) and knee pain patients with VAS (70–100) (p < 0.001, 18.3 ± 14.2 vs. 40.2 ± 20.5), respectively. Further, the TS effect was significantly lower for knee pain patients with VAS (10–39) compared with knee pain patients with VAS (40–69) (p < 0.001, 26.2 ± 18.2 vs. 38.7 ± 22.3) and knee pain patients with VAS (70– 100) (p < 0.001, 26.2 ± 18.2 vs. 40.2 ± 20.5), respectively. The TS correlated as the only experimental pain parameter with the KL score (r = 0.119, p = 0.046) (Table 3).
Pain area
3.2.1.2 TS to repeated pressure stimulation
Soreness area
For PPTs at knee and control points (arm, leg), the ANOVA showed a significant difference between groups (p < 0.001). The control group had significantly higher PPTs when comparing the average PPT from sites 1–8 and the averaged PPTs from control points [leg (9), arm (10)] with VAS (40–69) group (PPPT 1–8 = 0.004; PPPT 9–10 = 0.02) and VAS (70–100) group (PPPT 1–8 < 0.001; PPPT 9–10 < 0.001), respectively. When grouping subjects based on the KL grades, no differences in any of the PPT measures were found across groups with the following PPT (average points 1–8) values KL0: 440 ± 276 kPa, KL1: 477 ± 266 kPa, KL2: 427 ± 201 kPa, KL3: 427 ± 188 kPa and KL4: 364 ± 163 kPa. The mean pressure pain sensitivity maps were calculated as function of the pain intensity (control subjects and three pain groups) and KL grading (Fig. 1B). For KL grading 2 and 3, similar maps were seen with no correlation between KL and PPT (Table 3; Fig. 2). As few subjects had KL 0 (Table 1), the two groups 0 and 1 were combined.
Pain area
3.2.1.1 PPT
Maximal VAS least 24 h
3.2.1 Group comparison of PPTs, TS and CPM
Table 3 The table shows Spearman’s correlation coefficient (r) and p-value for mechanism-based pain parameters and clinical pain manifestations for the three knee pain groups.
3.2 QST for characterizing subgroups of OA patients
Mean CPM ratio (1–8)
significantly different between groups (ANOVA; p < 0.0143) with the high pain/low KL group significantly higher compared with the low pain/low KL group (p < 0.008).
Knee test points are 1–8 and the leg/arm test points 9–10. Non-significant findings are not listed (−). */** Significant correlation at p < 0.05/p < 0.01 level (two tailed). CPM, conditioning pain modulation; OA, osteoarthritis; PPT, pressure pain threshold; VAS, visual analogue scale.
L. Arendt-Nielsen et al.
© 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
3.2.1.3 CPM
For the CPM effect (mean of all 10 test points), the ANOVA showed a significant difference between the groups (p < 0.001). The CPM effect was significantly larger in low knee pain control subjects with VAS (0–9) compared with the knee pain patients with VAS (40– 69) (p = 0.001, 576.4 ± 170.4 vs. 445.5 ± 198.7 kPa) and the knee pain patients with VAS (70–100) (p < 0.001, 576.4 ± 170.4 vs. 366.8 ± 181.0 kPa), respectively. Further, the CPM effect was significantly larger in knee pain patients with VAS (10–39) compared with knee pain patients with VAS (70–100) (p < 0.001, 501.0 ± 227.7 vs. 366.8 ± 181.0 kPa).
3.3 Association between clinical features and QST The correlations of clinical characteristics and QST measures showed significant associations (Table 3). The maximal VAS during the last 24 h correlated significantly with all clinical and experimental measures. The pain area, soreness area and knee pain duration showed the same feature (except for a few relations, Table 3). The KL grade correlated weakly with the TS, but no other correlations to the experimental pain measures were found. The KL correlated with other clinical pain parameters (maximal pain during the last 24 h, duration of OA pain, area of clinical pain and area of soreness). The pain intensity was significantly correlated with the knee PPT, TS and CPM (Fig. 2). Further, the pain duration was significantly correlated (Spearman’s correlation, nonlinear) with the knee PPT, TS and CPM.
3.4 The pain sensitivity index for OA patient classification
Figure 2 The correlation between maximal pain intensity during the last 24 h [visual analogue scale (VAS) 24 h] and averaged pressure pain threshold (PPT) from the knee points (1–8), averaged increase in PPT from the knee points (1–8) during CPM and averaged increase in evoked pain intensity during repeated mechanical stimulation (mean VAS sum) elicited from the eight knee points. The linear correlation coefficients (r) and p-values are given (**p < 0.01). © 2014 European Pain Federation - EFICâ
Based on the analysis (see Methods) the pain sensitivity index was based on (1) PPT from the most sensitive assessment site on the knee (minimum PPT from point 1–8); (2) VAS sum as a measure of TS to repeated pressure stimulation of the tibialis anterior muscle (VASsumTA); and (3) the CPM ratio assessed at the TA (CPMTA). Experimental pain measures for both the most sensitive point on the knee (local sensitization) and lower leg point (spreading sensitization) were performed to provide information about different sensitization processes. To obtain unidirectionality towards high pain sensitivity [low PPT, low CPM ratio, high TS (VAS sum)], the equation was established as follows:
Index =
PPTTA − duringCPM VASsum TA ; CPM TA = PPTKneeMostSensitive ⋅ CPM TA PPTTA − beforeCPM Eur J Pain 19 (2015) 1406--1417
1413
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 4 Based on the z-score of the pain sensitivity index, the table provides the number of persons identified as sensitized (for each VAS pain group) by z-scores larger than 1.96 (± 2 SD). 24 h maximal VAS
Identified as sensitized (n)
Identified as sensitized (%)
Mean z-score (± SD) for the sensitized
VAS 0–9 VAS 10–39 VAS 40–69 VAS 70–100
2/64 22/81 26/70 25/66
3 27 37 38
4.35 ± 1.28 4.63 ± 2.68 5.54 ± 4.20 6.87 ± 5.87
For those identified as sensitized, the averaged z-scores were calculated. SD, standard deviation; VAS, visual analogue scale.
The z-score was calculated by subtracting the index by the mean index from control subjects (meancontrols = 5.80) and divide by the SD of the index from the control subjects (SDcontrols = 7.07). The 95% confidence interval of the zero-mean and unit variance z-scores (± 1.96) was then used to define the pain sensitivity when the z-score was above 1.96, and desensitization when below −1.96. This procedure, independent of the unit of measure, can be applied to any measurement and has recently been used for separating OA patients into different groups (Schiphof et al., 2013). For the PCA analysis (see Methods), the three principal components with the highest loadings were PPTKneeMostSensitive (loading = 0.986), VASsumTA (loading = 0.863) and CPMKneeMostSensitive (loading = 0.799). Despite the limitations of the PCA analysis, the three variables with the highest loadings were identical to the more simple analysis using Pearson correlations, with the exception of CPMTA (low Pearson correlations so providing additional information) and CPMKneeMostSensitive (high loading in PCA). As CPM assessed from the knee will be confounded by the local knee sensitization process, the best measure of the general CPM effects was considered the measure from a muscle distal to the knee (CPMTA). The z-scores of the pain sensitivity index were used to identify the proportion of sensitized subjects (Table 4). Ninety-seven per cent of the control subjects were within the normal range (
A mechanism-based pain sensitivity index to characterize knee osteoarthritis patients with different disease stages and pain levels L. Arendt-Nielsen1, L.L. Egsgaard1,2, K.K. Petersen1, T.N. Eskehave1,2, T. Graven- Nielsen1, H.C. Hoeck2,3, O. Simonsen4 1 2 3 4
Center for Sensory-Motor Interaction, Aalborg University, Denmark CCBR, Aalborg, Denmark C4Pain, Aalborg, Denmark Sygehus Vendsyssel, Frederikshavn, Denmark
Correspondence Lars Arendt-Nielsen E-mail: [email protected] Funding sources The study was supported by The Danish National Advanced Technology Foundation, Aase and Ejnar Danielsens Foundation, Lions Club in Denmark, The Danish Council for Technology and Innovation (09-052174), The Bevica Foundation and The Danish Rheumatic Association. Conflicts of interest None declared. Accepted for publication 25 November 2014 doi:10.1002/ejp.651
Abstract Background: In a cohort of well-characterized patients with different degrees of knee osteoarthritis (OA) and pain, the aims were to utilize mechanism-based quantitative sensory testing (QST) to (1) characterize subgroups of patients; (2) analyse the associations between clinical characteristics and QST; and (3) develop and apply a QST-based knee OA composite pain sensitivity index for patient classification. Methods: Two hundred seventeen OA pain patients and 64 controls were included. Kellgren and Lawrence (KL) grading scores were obtained, and pressure pain thresholds (PPTs), temporal summation of pain to repeated painful pressure stimulation and conditioning pain modulation (CPM) were assessed. Associations between pain score/area/duration, radiological findings and QST-related parameters were analysed. A pain sensitivity index was developed and applied based on PPT, temporal summation and CPM. z-Score, as statistical tool, was calculated for statistically comparing the pain index of a single patient with a healthy control group. Results: High knee pain associated with low KL grade showed particular signs of pain sensitization. Patients showed significant associations between clinical knee pain intensity/duration and lowering of knee PPTs (p < 0.01), facilitation of temporal summation (p < 0.01), reduction of CPM function (p < 0.01) and high pain sensitivity index (p < 0.01). The index classified 27–38% of the OA patients and 3% of the controls as highly sensitive with no association to KL. The index increased for high knee pain intensities and long pain duration. Conclusions: Radiological scores, contrary to clinical pain intensity/ duration, were poorly associated with QST parameters. The pain sensitivity index could classify OA patients with different degrees of OA and pain.
1. Introduction Osteoarthritis (OA) is a common condition with increasing prevalence (Woolf and Pfleger, 2003; Jinks et al., 2007; Neogi, 2013) due to demographic and lifestyle changes. Pain and impaired function are the clinical representations of OA (Breivik et al., 2008; 1406 Eur J Pain 19 (2015) 1406--1417
Read and Dray, 2008; Neogi, 2013). Furthermore, psychosocial variables influence the pain and contribute to the variability in pain (Sluka et al., 2012). For the individual patient, severe knee joint damage may cause little pain and vice versa (Davis et al., 1992; Hannan et al., 2000; Neogi et al., 2009), but at a population level this discordance is less evident (Bedson © 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
What’s already known about this topic? • Pain is most often not associated with clinical features such as joint damage in osteoarthritis. • Pain sensitization seems to play an important role in osteoarthritis. What does this study add? • Use of quantitative sensory testing to characterize patient groups with knee osteoarthritis (OA). • Associate clinical features of OA with quantitative sensory testing. • Developing and applying a composite pain sensitivity index for profiling OA patients.
and Croft, 2008). To some degree this diversity in the clinical presentation of OA pain (Lanyon et al., 1998; Dieppe and Lohmander, 2005; Felson, 2005) has been associated with different degrees of pain sensitization in individual patients (Hannan et al., 2000; Dieppe and Lohmander, 2005; Finan et al., 2013; Schiphof et al., 2013). Pain intensity and OA duration are found to be associated with this development of sensitization (Fernández-de-las-Peñas et al., 2007, 2009; Arendt-Nielsen et al., 2010). If the peripheral nociceptive drive from the joint is reduced/terminated (e.g., knee replacement), the pain sensitization will normalize (Graven-Nielsen et al., 2012; Aranda-Villalobos et al., 2013), but unfortunately 10–34% of the patients develop chronic postoperative pain (Beswick et al., 2012) with continued pain sensitization (Petersen et al., 2015). Therefore, it would have major impact on management strategies and drug development if quantitative biomarkers could be developed for phenotyping OA pain patients to design optimal individualized management strategies. Quantitative sensory testing (QST) has the potential to be used for phenotyping of pain patients (Rolke et al., 2006a,b). In OA, pressure pain algometry has mainly been used to quantify localized muscle/joint pain sensitivity (Kosek and Ordeberg, 2000; Bajaj et al., 2001; Imamura et al., 2008; Arendt-Nielsen et al., 2010; Graven-Nielsen et al., 2012; Suokas et al., 2012; Skou et al., 2014). Recently, more advanced QST technologies have been developed and applied in OA to assess changes in the central pain mechanism such as temporal summation (TS) and descending pain inhibition (Arendt-Nielsen et al., 2010; Graven-Nielsen et al., 2012). Most QST studies in OA (1) are based on relatively small cohorts of patients with insufficient power to © 2014 European Pain Federation - EFICâ
grade the responses according to the pain status; (2) use only few QST parameters; and (3) do not utilize the complementary information that can be gained from combining a variety of mechanism-based QST tools. The aims of the present cohort study in patients with different degrees of knee OA and pain were to investigate how to utilize mechanism-based QST to (1) characterize subgroups of patients; (2) analyse the associations between clinical characteristics (pain intensity, pain area, pain duration, radiological scores) and mechanism-based QST [pressure pain, pain sensitivity maps, TS, conditioning pain modulation (CPM)]; and (3) develop and apply a QST-based knee OA composite pain sensitivity index for patient classification.
2. Methods 2.1 Participants A total of 217 patients (115 women) and 64 control subjects (32 women) were enrolled (Table 1). The patients and control subjects were recruited from the patient database at CCBR (https://dk.ccbr.com/ccbr-aalborg) and the clinical trial unit C4Pain (www.C4Pain.com), Aalborg, Denmark. The database consists of volunteers who are interested in participating in studies and clinical trials focusing on knee OA. All the patients in the database have verified knee OA and have been referred from GPs, physiotherapy clinics or recruited through newspaper advertisements. Patients from the database were invited to participate in the present study. Those responding were clinically evaluated and scored according to the American College of Rheumatology (ACR) OA criteria (Altman et al., 1986). If the diagnosis was confirmed and they fulfilled the ACR OA criteria, they were included in the study. If X-ray images were more than 6 months old, new X-ray images were taken and radiologically scored (see below). None of the participants included in the study had participated in other clinical trials for the last 12 months. The control subjects were mainly recruited among patients’ family members, and they were to fulfil the same exclusion criteria. Control subjects had no symptomatic OA with maximal pain less than 10 mm for the last 24 h on the visual analogue scale (VAS). The control subjects were invited to have X-ray images taken. The exclusion criteria comprised secondary OA, psychiatric conditions hindering the ability of the patient to participate in the study, pregnancy, previous/current drug(s) or alcohol abuse, previous neurological or concomitant musculoskeletal disorders, lack of collaboration ability and subjects unable to abstain from analgesic medication for at least 72 h prior to the knee pain assessment. The data collected from the cohort of patients participating in this study may provide the basis for additional studies. Eur J Pain 19 (2015) 1406--1417
1407
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 1 Demographic data from the low-pain control group [VAS (0–9), n = 64] and the three pain groups (n = 217). Groups
Controls (VAS 0–9)
VAS 10–39
VAS 40–69
VAS 70–100
Men (n) Women (n) Total (n) Age ± SD (years) BMI ± SD (kg/m2)
32 32 64 67.5 ± 4.4 27.1 ± 4.1
35 46 81 62.8 ± 8.3 27.5 ± 3.7
34 36 70 64.9 ± 7.6 28.0 ± 3.3
33 33 66 63.8 ± 8.1 29.6 ± 5.1
KL = 0 KL = 1 KL = 2 KL = 3 KL = 4
3 25 28 0 1
4 17 46 11 3
3 11 33 12 11
2 6 38 13 7
Men (n)
Women (n)
Age ± SD (years)
BMI ± SD (kg/m2)
8 28 66 19 10
4 31 79 17 12
60.1 ± 8.7 63.7 ± 8.0 64.7 ± 7.3 64.3 ± 7.1 67.8 ± 7.7
26.3 ± 3.3 27.0 ± 4.0 28.2 ± 3.8 29.3 ± 5.6 29.5 ± 3.9
KL = 0 KL = 1 KL = 2 KL = 3 KL = 4
Total 12 59 145 36 22
X-rays were not taken from seven control subjects with VAS = 0. The four groups are separated based on the clinical pain ratings (maximal pain the last 24 h). BMI, body mass index; KL, Kellgren and Lawrence; SD, standard deviation; VAS, visual analogue scale.
2.2 Patient characterization 2.2.1 Radiologic assessment of OA All participants had a bilateral weight bearing, 10° fixed flexion posteroanterior X-ray taken of the knees. The medial and lateral tibiofemoral joint space and tibial plateau were measured using a posteroanterior view. The superior and inferior osteophytes were evaluated. The X-ray beam was centred on the joint line of the knee in the popliteal space of a 10° caudal angulation. The degree of radiologic OA was evaluated using Kellgren and Lawrence grading scale (KL0-4) (Kellgren and Lawrence, 1957). The KL grading did not include the patellofemoral joint.
2.2.2 Clinical assessment of pain The maximal pain intensity during the past 24 h prior to the visit was assessed on a 100 mm VAS where 0 mm defined ‘no pain’ and 100 mm defined ‘maximal pain’. The participants were informed about the pain scale, instructed in detail how to use it and informed that the maximal knee intensity pain over the last 24 h could have occurred during rest, movement or during the night. The participants were grouped according to the maximal knee pain intensity as: VAS (0–9) (low-pain control group), VAS (10–39), VAS (40–69) and VAS (70–100). The knee in which OA was diagnosed and which showed the highest pain intensity was identified as the index knee. Clinical pain areas (individual conditions evoking pain, e.g., walking) and soreness areas (evoked by palpation) were mapped by the subjects on the anterior and posterior aspects of the index knee and later quantified in arbitrary units using a digitizer software (ACECAD, model 1408 Eur J Pain 19 (2015) 1406--1417
D9000, Taipei, Taiwan). The subjects were asked to indicate the duration (in years) during which they had had knee pain (in case they had any pain). This knee pain duration is often longer than the one of clinically verified OA. As all participants recruited were cognitively and adequately functioning, the instruction procedure could be highly standardized.
2.3 QST for characterizing subgroups of OA patients 2.3.1 Pressure pain threshold (PPT) assessment A computer-controlled pressure algometer (Center for Sensory-Motor Interaction, Aalborg University, Aalborg, Denmark) was used to assess the PPTs. The pressure algometer consisted of a round aluminium footplate with a padded contact surface of 1 cm2 fixed to the top of a piston. The pressure stimulation was feedback controlled via recordings of the actual force. The computer-controlled pressure algometer was placed perpendicularly to the test site and applied an increasing pressure of 60 kPa/s until the subjects defined the pressure as pain and pressed a stop button. For all experimental procedures, the subjects were asked to rest in a supine position with their target knee mildly raised and placed on a vacuum pillow in order to fixate the knee when assessing the PPTs. Ten test sites were located and marked: eight test-specific sites in the peripatellar region and two control points according to our previous recommendations (Arendt-Nielsen et al., 2010). The 10 points were defined as (1) 2 cm distal to the inferomedial edge of patella; (2) 2 cm distal to the inferolateral edge of patella; (3) 3 cm lateral to the centre of the lateral edge of patella; (4) 2 cm © 2014 European Pain Federation - EFICâ
L. Arendt-Nielsen et al.
proximal to the superolateral edge of patella; (5) 2 cm proximal to the superior edge of patella; (6) 2 cm proximal to the superomedial edge of patella; (7) 3 cm medial to the centre of the medial edge of patella; (8) on the centre of patella; (9) the belly of the tibialis anterior muscle; and (10) the belly of the extensor carpi radialis muscle. The PPTs were measured three times with an interval of minimum 25 s at each site, and the average of the three determinations was calculated for further analysis. The sequence for assessing the 10 individual locations was randomized.
2.3.2 Pressure pain sensitivity maps Pressure pain sensitivity maps (PPT maps) were constructed by interpolation of the mean PPT values across all subjects from the peripatellar sites using an inverse distanceweighted interpolation to compute PPT values of unknown locations by applying the known mean PPT values and locations (Arendt-Nielsen et al., 2010; Binderup et al., 2010). Those PPT maps were projected onto three-dimensional MRI-based contour models of a knee to obtain a visual impression of the pressure pain sensitivity distribution. The PPT maps were developed in relation to the clinical groupings based on the knee VAS pain scores and KL grading.
2.3.3 TS of pressure pain The computer-controlled pressure algometer (Center for Sensory-Motor Interaction, Aalborg University) was used to assess TS to repeated pressure-induced pain. The mean PPT values from the different locations were used to examine the extent of TS by applying 10 successive stimuli with an intensity at mean PPT and delivered at a frequency of 0.5 Hz (1 s stimulation and 1 s interval without stimulation) (Nie et al., 2005; Arendt-Nielsen et al., 2010). A constant contact pressure of 60 kPa was maintained between each stimulus, which was not painful. During repeated stimuli, the pain intensity was assessed by the patient on an electronic VAS ranging from 0 to 100 mm (Center for Sensory-Motor Interaction, Aalborg University) with end points defined as no pain and maximal pain, respectively. The electronic VAS ratings were interfaced to a computer using the LabViewbased Mr. Kick software (Center for Sensory-Motor Interaction, Aalborg University, Denmark). The VAS following each stimulus was extracted and the sum of the 10 increasing VAS scores (VAS sum) was used for analysis. This VAS sum was used as this measure was associated with other measures of summation such as the slope of increase in VAS or the ratio between the VAS from the 10th and the first stimulus. The mean VAS sum elicited from stimulation of sites 1–8 was used for statistical analysis.
2.3.4 CPM A computer-controlled cuff algometer (NociTech, Aalborg, Denmark) was used to evoke CPM (Yarnitsky et al., 2010). A © 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
24 in. tourniquet cuff was placed approximately 3 cm proximal to the cubital fossa on the contralateral arm to the target knee. The tourniquet applied a constant pressure of 36 kPa (approx. 270 mmHg) during which the subjects reported their pain on the electronic VAS. Two minutes after the subjects had reached a VAS score of 60 mm, the PPT recordings of the 10 points were commenced in randomized order. The PPTs were measured once at each of the 10 points. If needed, the volunteers could increase the pain intensity by performing grip movements around a soft rubber ball. For the knee, the mean effect assessed from points 1–8 was calculated. The mean ratio between PPT during cuff stimulation and PPT at baseline was defined as the CPM ratio.
2.4 Pain sensitivity index and z-score analysis Among all the relevant parameters to be included in a composite measure (the pain sensitivity index), an analysis was made to find the parameters that provided the most complementary information associated with pain sensitivity, i.e., the parameter correlating (Pearson’s correlation) least with the most sensitive PPT value measured from the knee. The PPT value from the knee was chosen as the reference for the analysis as this location is the main origin of the clinical pain problem. The rationale for this analysis was to identify which parameters in combination provided most entropy (obtaining maximal information from the data). The two parameters least correlated with the PPT value from the most sensitive location on the knee were the TS provoked for tibialis anterior (VASsumTA, Pearson correlation −0.059; two-tailed p-value 0.327) and the CPM ratio based on the test stimulus applied to tibialis anterior (CPMTA Pearson correlation −0.053; two-tailed p-value 0.379). The two parameters strongest correlated with the knee PPT value were PPT from the tibialis anterior (Pearson correlation 0.970; two-tailed p-value < 0.001) and the PPT assessed from the arm (Pearson correlation 0.967; two-tailed p-value < 0.001). Based on this information the pain sensitivity index was constructed. The index was used to calculate a z-score for all participants. Calculating the z-score from the index transforms the data into a standard normal distribution with zero mean and unit variance, i.e., z-values above 0 indicate a gain in index (sensitized), and z-values below 0 indicate a reduction in index (non-sensitized). The use of z-scores facilitates the comparison of the pain sensitivity of a single patient with a group of healthy control subjects (Rolke et al., 2006a,b). A principal component analysis (PCA) was also used as a tool to reduce the number of parameters to be included in the pain sensitivity index. The PCA is based on linear relationships like Pearson’s correlation. The PCA algorithm cannot take the a priori clinical considerations into account (e.g., we would like to use the PPTKneeMostSensitive as one component in the pain sensitivity index). The data from the Pearson’s correlation and the PCA were compared to ensure that the optimal set of data was selected for the calculation of the index. Eur J Pain 19 (2015) 1406--1417
1409
Pain sensitization in osteoarthritis
2.5 Statistical analysis Data are presented as mean and standard deviation (SD). The statistical analysis was conducted using SPSS 19.0 software (SPSS Inc., Chicago, IL, USA). In case of significant factors detected by the analysis of variance (ANOVA), Bonferroni post hoc test was performed to identify the pattern of significant differences. The level p < 0.05 was considered significant. The statistical analyses for the different parameters were performed according to the following plan.
L. Arendt-Nielsen et al.
2.8.1.1 Correlation between maximal pain intensity during the past 24 h and QST parameters Due to a linear distribution of the parameter, Pearson correlations were used for linear estimates of the correlation between maximal pain intensity during the past 24 h and the other QST parameters (PPT, TS, CPM).
2.8.1.2 Correlation between PPT measured from the most sensitive part of the knee and the other QST parameters
2.6 Patient characterization
A one-way ANOVA was performed for the demographic parameters (gender, age and BMI) to test for differences between groups [control subjects, VAS (0–9), VAS (10–39), VAS (40–69), VAS (70–100)].
To identify which QST parameters correlated least with the selected reference (PPT from the most sensitive point on the knee), a correlation matrix composed of Pearson correlations was used. This served as an analysis to identify which parameters should be included in the sensitization index. Pearson correlations were performed due to a linear distribution of the parameters.
2.6.2 Comparing OA groups with different pain ratings and KL scores
2.8.1.3 Correlation between clinical and QST parameters
The combination of different pain ratings and KL scores resulted in four groups (‘low pain’/‘low KL’; ‘low pain’/‘high KL’; ‘high pain’/‘low KL’; ‘high pain’/‘high KL’) where low pain was defined as VAS (0–51), and high pain was defined as VAS (52–100); low KL was defined as KL 0, 1 and 2, and high KL was defined as KL 3 and 4. A one-way ANOVA, adjusted for age, BMI and gender was performed for each assessment site (knee, arm, TA) on PPT, TS, CPM measurements and z-scores to test for differences between these pain/KL groups. The pressure pain sensitivity maps for the four groups were not analysed statistically.
Spearman’s correlations of clinical parameters (pain intensity, pain area, pain duration, radiological scores) and the mechanism-based QST responses were performed. Spearman’s correlations were performed due to non-linear distribution of some parameters.
2.6.1 Demographic data
2.7 QST for characterizing subgroups of OA patients
2.9 Pain sensitivity index The z-statistics used for classification of patients based on the pain sensitivity index as described in the sections above.
3. Results 3.1 Patient characterization
2.7.1 Group comparison of PPTs, TS and CPM
3.1.1 Demographic data
A one-way ANOVA was performed for each assessment site (knee, arm, TA) on PPT, TS and CPM measurements to test for differences between groups [control subjects; VAS (10– 39); VAS (40–69); VAS (70–100)]. The pressure pain sensitivity maps were not statistically analysed.
The demographic data from the study population are summarized in Table 1.
2.8 Association between clinical features and QST
The experimental pain measures for the four combinations (high pain, low pain, high KL, low KL) are given in Table 2. The high pain/low KL group showed significantly higher pain sensitivity and a higher proportion of sensitized subjects when compared with the low pain/low KL group as also reflected in the pressure pain sensitivity maps (Fig. 1A). The pain sensitivity index at group level (z-score; Table 2) was
2.8.1 Correlation analyses A variety of correlation analyses were used and the individual tests are described below. For all analyses the correlation coefficient (r), slopes and significance levels (p) were calculated. 1410 Eur J Pain 19 (2015) 1406--1417
3.1.2 Comparing OA groups with different pain ratings and KL scores
© 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 2 The experimental pain responses in four different OA patient groups (n = 217) based on high (VAS > 51), low (VAS < 51) last 24 h maximal VAS and on high KL (3 and 4) and low KL (1 and 2). Group/pain response
Mean PPT (1–8) (kPa)
Mean PPT (9–10) (kPa)
Mean CPM ratio (1–8)
Mean VAS sum (1–8)
z-score
% of sensitized
Low pain/low KL (n = 85) Low pain/high KL (n = 24) High pain/low KL (n = 76) High pain/high KL (n = 32)
448 ± 230 471 ± 222 379 ± 212 385 ± 147
373 ± 173 398 ± 179 320 ± 167 341 ± 134
1.277 ± 0.752 1.087 ± 0.603 0.927 ± 0.679 0.866 ± 0.448
25.203 ± 16.527 29.415 ± 20.921 43.283 ± 22.622 34.533 ± 18.923
1.6 ± 3.2 1.3 ± 2.2 2.9 ± 4.9 2.0 ± 2.9
30 25 39 34
Knee test points are 1–8 and the leg/arm test points 9–10. The z-score of the pain sensitivity index is calculated for the group and the proportions (% of patients) of sensitized (z-score >1.96) are indicated. The connecting lines indicate significant differences between groups (p < 0.02). The statistical analysis was adjusted for age, gender and body mass index. Mean ± standard deviation values are given. CPM, conditioning pain modulation; KL, Kellgren and Lawrence; OA, osteoarthritis; PPT, pressure pain threshold; VAS, visual analogue scale.
Figure 1 (A) The mean pressure pain threshold knee maps for four different patient group combinations based on high [visual analogue scale (VAS) ≥ 51], low (VAS < 51) maximal VAS during the last 24 h and on high Kellgren and Lawrence (KL) (3 and 4) and low KL (0, 1 and 2). (B) The top row shows the mean pressure pain threshold knee maps for the control subjects (VAS 0–9) and the three knee pain groups with different maximal VAS during the last 24 h. The bottom panel shows the mean pressure pain threshold knee maps for the different patient groups based on KL. KL 0 and KL 1 were combined.
© 2014 European Pain Federation - EFICâ
Eur J Pain 19 (2015) 1406--1417
1411
Pain sensitization in osteoarthritis
1412 Eur J Pain 19 (2015) 1406--1417
– r = −0.429** p = 0.000 r = −0.533** p = 0.000 –
–
–
r = 0.311** p = 0.000
r = 0.674** p = 0.000 r = 0.523** p = 0.000 r = 0.338** p = 0.000 r = −0.197** p = 0.001 r = −0.193** p = 0.001 r = 0.285** p = 0.000 –
r = 0.321** p = 0.000
r = 0.227** p = 0.000 r = −0.157** p = 0.008 r = 0.311** p = 0.000
r = 0.119* p = 0.046 –
r = 0.889** p = 0.000 – –
–
r = 0.437** p = 0.000 r = 0.274** p = 0.000 r = −0.160** p = 0.007 r = −0.124* p = 0.038 r = 0.243** p = 0.000 –
r = 0.285** p = 0.000 r = −0.131* p = 0.029 –
–
r = −0.301** p = 0.000 r = 0.142** p = 0.017
Mean VAS sum (1–8) Mean PPT (9–10) Mean PPT (1–8) Kellgren and Lawrence OA duration (years)
z-Score pain sensitivity index
Mean CPM ratio (1–8)
Mean VAS sum (1–8)
Mean PPT (9–10)
Mean PPT (1–8)
Kellgren and Lawrence
OA duration (years)
Soreness area
r = 0.544** p = 0.000 r = 0.520** p = 0.000 r = 0.562** p = 0.000 r = 0.303** p = 0.000 r = −0.342** p = 0.000 r = −0.302** p = 0.000 r = 0.482** p = 0.000 r = −0.345** p = 0.000 r = 0.412** p = 0.000
For the TS effect (mean of all 10 test points), the ANOVA showed a significant difference between groups (p < 0.001). The TS effect was significantly lower in low knee pain control subjects with VAS (0–9) compared with knee pain patients with VAS (40–69) (p < 0.001, 18.3 ± 14.2 vs. 38.7 ± 22.3) and knee pain patients with VAS (70–100) (p < 0.001, 18.3 ± 14.2 vs. 40.2 ± 20.5), respectively. Further, the TS effect was significantly lower for knee pain patients with VAS (10–39) compared with knee pain patients with VAS (40–69) (p < 0.001, 26.2 ± 18.2 vs. 38.7 ± 22.3) and knee pain patients with VAS (70– 100) (p < 0.001, 26.2 ± 18.2 vs. 40.2 ± 20.5), respectively. The TS correlated as the only experimental pain parameter with the KL score (r = 0.119, p = 0.046) (Table 3).
Pain area
3.2.1.2 TS to repeated pressure stimulation
Soreness area
For PPTs at knee and control points (arm, leg), the ANOVA showed a significant difference between groups (p < 0.001). The control group had significantly higher PPTs when comparing the average PPT from sites 1–8 and the averaged PPTs from control points [leg (9), arm (10)] with VAS (40–69) group (PPPT 1–8 = 0.004; PPPT 9–10 = 0.02) and VAS (70–100) group (PPPT 1–8 < 0.001; PPPT 9–10 < 0.001), respectively. When grouping subjects based on the KL grades, no differences in any of the PPT measures were found across groups with the following PPT (average points 1–8) values KL0: 440 ± 276 kPa, KL1: 477 ± 266 kPa, KL2: 427 ± 201 kPa, KL3: 427 ± 188 kPa and KL4: 364 ± 163 kPa. The mean pressure pain sensitivity maps were calculated as function of the pain intensity (control subjects and three pain groups) and KL grading (Fig. 1B). For KL grading 2 and 3, similar maps were seen with no correlation between KL and PPT (Table 3; Fig. 2). As few subjects had KL 0 (Table 1), the two groups 0 and 1 were combined.
Pain area
3.2.1.1 PPT
Maximal VAS least 24 h
3.2.1 Group comparison of PPTs, TS and CPM
Table 3 The table shows Spearman’s correlation coefficient (r) and p-value for mechanism-based pain parameters and clinical pain manifestations for the three knee pain groups.
3.2 QST for characterizing subgroups of OA patients
Mean CPM ratio (1–8)
significantly different between groups (ANOVA; p < 0.0143) with the high pain/low KL group significantly higher compared with the low pain/low KL group (p < 0.008).
Knee test points are 1–8 and the leg/arm test points 9–10. Non-significant findings are not listed (−). */** Significant correlation at p < 0.05/p < 0.01 level (two tailed). CPM, conditioning pain modulation; OA, osteoarthritis; PPT, pressure pain threshold; VAS, visual analogue scale.
L. Arendt-Nielsen et al.
© 2014 European Pain Federation - EFICâ
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
3.2.1.3 CPM
For the CPM effect (mean of all 10 test points), the ANOVA showed a significant difference between the groups (p < 0.001). The CPM effect was significantly larger in low knee pain control subjects with VAS (0–9) compared with the knee pain patients with VAS (40– 69) (p = 0.001, 576.4 ± 170.4 vs. 445.5 ± 198.7 kPa) and the knee pain patients with VAS (70–100) (p < 0.001, 576.4 ± 170.4 vs. 366.8 ± 181.0 kPa), respectively. Further, the CPM effect was significantly larger in knee pain patients with VAS (10–39) compared with knee pain patients with VAS (70–100) (p < 0.001, 501.0 ± 227.7 vs. 366.8 ± 181.0 kPa).
3.3 Association between clinical features and QST The correlations of clinical characteristics and QST measures showed significant associations (Table 3). The maximal VAS during the last 24 h correlated significantly with all clinical and experimental measures. The pain area, soreness area and knee pain duration showed the same feature (except for a few relations, Table 3). The KL grade correlated weakly with the TS, but no other correlations to the experimental pain measures were found. The KL correlated with other clinical pain parameters (maximal pain during the last 24 h, duration of OA pain, area of clinical pain and area of soreness). The pain intensity was significantly correlated with the knee PPT, TS and CPM (Fig. 2). Further, the pain duration was significantly correlated (Spearman’s correlation, nonlinear) with the knee PPT, TS and CPM.
3.4 The pain sensitivity index for OA patient classification
Figure 2 The correlation between maximal pain intensity during the last 24 h [visual analogue scale (VAS) 24 h] and averaged pressure pain threshold (PPT) from the knee points (1–8), averaged increase in PPT from the knee points (1–8) during CPM and averaged increase in evoked pain intensity during repeated mechanical stimulation (mean VAS sum) elicited from the eight knee points. The linear correlation coefficients (r) and p-values are given (**p < 0.01). © 2014 European Pain Federation - EFICâ
Based on the analysis (see Methods) the pain sensitivity index was based on (1) PPT from the most sensitive assessment site on the knee (minimum PPT from point 1–8); (2) VAS sum as a measure of TS to repeated pressure stimulation of the tibialis anterior muscle (VASsumTA); and (3) the CPM ratio assessed at the TA (CPMTA). Experimental pain measures for both the most sensitive point on the knee (local sensitization) and lower leg point (spreading sensitization) were performed to provide information about different sensitization processes. To obtain unidirectionality towards high pain sensitivity [low PPT, low CPM ratio, high TS (VAS sum)], the equation was established as follows:
Index =
PPTTA − duringCPM VASsum TA ; CPM TA = PPTKneeMostSensitive ⋅ CPM TA PPTTA − beforeCPM Eur J Pain 19 (2015) 1406--1417
1413
Pain sensitization in osteoarthritis
L. Arendt-Nielsen et al.
Table 4 Based on the z-score of the pain sensitivity index, the table provides the number of persons identified as sensitized (for each VAS pain group) by z-scores larger than 1.96 (± 2 SD). 24 h maximal VAS
Identified as sensitized (n)
Identified as sensitized (%)
Mean z-score (± SD) for the sensitized
VAS 0–9 VAS 10–39 VAS 40–69 VAS 70–100
2/64 22/81 26/70 25/66
3 27 37 38
4.35 ± 1.28 4.63 ± 2.68 5.54 ± 4.20 6.87 ± 5.87
For those identified as sensitized, the averaged z-scores were calculated. SD, standard deviation; VAS, visual analogue scale.
The z-score was calculated by subtracting the index by the mean index from control subjects (meancontrols = 5.80) and divide by the SD of the index from the control subjects (SDcontrols = 7.07). The 95% confidence interval of the zero-mean and unit variance z-scores (± 1.96) was then used to define the pain sensitivity when the z-score was above 1.96, and desensitization when below −1.96. This procedure, independent of the unit of measure, can be applied to any measurement and has recently been used for separating OA patients into different groups (Schiphof et al., 2013). For the PCA analysis (see Methods), the three principal components with the highest loadings were PPTKneeMostSensitive (loading = 0.986), VASsumTA (loading = 0.863) and CPMKneeMostSensitive (loading = 0.799). Despite the limitations of the PCA analysis, the three variables with the highest loadings were identical to the more simple analysis using Pearson correlations, with the exception of CPMTA (low Pearson correlations so providing additional information) and CPMKneeMostSensitive (high loading in PCA). As CPM assessed from the knee will be confounded by the local knee sensitization process, the best measure of the general CPM effects was considered the measure from a muscle distal to the knee (CPMTA). The z-scores of the pain sensitivity index were used to identify the proportion of sensitized subjects (Table 4). Ninety-seven per cent of the control subjects were within the normal range (
