Differential Sensory and Clinical Phenotypes of Patients With Chronic Widespread and Regional Musculoskeletal Pain

Marina M. Finnern; Dieter Kleinböhl; Herta Flor; Justus Benrath; Rupert Hölzl

Disclosures

Pain. 2020;162(1):56-70. 

In This Article

Results

Evaluation of Clinical Pain

As indicated in Table 1, pain characteristics except for the spatial extension of pain did not significantly differ between patients with FMS and CBP. However, the number of pain sites and the WPI distributions of both groups overlapped substantially between 4 and 15 indicating a subgroup of patients previously diagnosed as "FMS" who neither fulfilled complete FMS criteria nor showed regionally distinct pain such as CBP. Moreover, 28 (26.2%) of the patients with CBP reported pain in all 4 body quadrants in addition to their primary pain area at the back, and many patients with FMS reported only 10 or less pain loci (cf. supplemental Figure S2, available at http://links.lww.com/PAIN/B131). The following sensory profiling tried to relate this variable clinical picture to distinguishable phenotypes differing in pressure pain sensitivity as suggested by the earlier findings of TPs distributed across diagnostic groups.

Sensory Pain Characteristics

Spatial Distribution of Pressure Pain Sensitivity. The assessment of pressure sensitivity by the standardized manual probe at 32 body sites confirmed a pronounced enhancement of sensitivity to pressure stimulation in patients with FMS compared to patients with CBP, both in terms of number of hypersensitive body sites as well as provoked pain intensities (U tests, P < 0.001; Table 2 and Table 3). PFC had neither sensitive tender nor CPs in the manual probe except for 1 participant with a sensitive TP at the knee.

As shown in Table 2, pressure hypersensitivity was not limited to the 18 TPs specified by the ACR 1990 protocol but applied also to several CPs at other sites. The number of hypersensitive TPs of patients with FMS ranged from 7 to 18 (median = 15.5), with 3 patients not fulfilling the criterion of ≥11 TPs at second testing. Moreover, 16 patients with CBP (15%) showed also 11 or more hypersensitive ACR TPs, indicating substantial sensory overlap between previous FMS and CBP diagnoses similar to the spatial spread of clinical pain. In addition, 63 (81%) patients with FMS and 33 (31%) patients with CBP reported also one or more hypersensitive control sites (0–14 vs 0–10, respectively). The total number of hypersensitive body sites, irrespective of ACR tender or control position, differed significantly between FMS and CBP (median = 19 vs 4), but ranges overlapped considerably (0–32 vs 0–26). At the same time, pain intensity ratings of the manual stimulation were consistently higher at TPs than at CPs in both groups (U test; FMS: P < 0.001; CBP: P = 0.003), but the difference was more pronounced in FMS. Average pain intensities of the manual probing (TPs and CPs combined) were also higher in FMS (U test, P < 0.001; Table 2) and increased systematically with the numbers of hypersensitive body sites (Figure 1A).

Figure 1.

Number of pressure-sensitive body sites and pain intensity of manual pressure probes. Average pain intensities of manual pressure probing as a function of number of pressure-sensitive body sites; pain intensity: VAS 0–10; sensitive site: VAS ≥1 (just painful or more); number corrected for missings. (A) All sensitive points; (B) sensitive ACR tender points; (C) sensitive control points. Boxplots: group median, 25th and 75th percentiles; whiskers: range windsorized at 1.5 × IQR outside interquartile range; circles: individual values. VAS, visual analogue scale.

The relation was mainly due to the patients with widespread pain dominated by the FMS group (Figure 1B, right half) confirmed by group-specific correlations (FMS: ρ = 0.363, P = 0.001; CBP: ρ = 0.116, P = 0.302, n. s.). However, a gradual transition exists from regional to widespread pain with a large overlap. The number × intensity correlation was limited to the 18 TPs (Figure 1B; ρ = 0.347, P = 0.065), whereas the pain intensity was not correlated with the number of sensitive CPs (Figure 1C; ρ = 0.048, n. s.). Moreover, the intensity increase could also be found in patients with CBP with more than 7 classical ACR TPs (right half of Figure 1B). Note that the scale clipping at the maximum of 18 TPs tested in FMS limited the maximal correlation and the correlation might have been even higher with an open scale as suggested by the accumulation of observations in category 18 in Figure 1B.

Cross-examination of the semiquantitative manual assessment by selected quantitative algometer tests at respective marker loci on the trapezius muscle ("trapezius TP") and the thenar eminence (extraspinal "thenar CP") reproduced the sensitivity differences between patients with FMS and CBP and between TP and CP (Table 3). Algometer pressure pain thresholds were generally lower at both sites in FMS than in CBP (multiple t tests, P < 0.001), but the corresponding pain intensity ratings of patients with FMS were higher only at the trapezius TP marker (P < 0.001; thenar: P = 0.512, n. s.).

Regional Clusters of Pressure Sensitivity. For the cluster-analytic identification of subgroups within the continuum from regional to widespread pain, a parsimonious indicator vector for the LCA of the spatial distribution of pressure sensitivity was extracted by PCAs of the initial set of 32 pain intensity ratings of the manual pressure stimulation at 18 TPs and 14 CPs. Nine principal components described the spatial distribution of the pressure test pain in both FMS and CBP (Kaiser criterion; 79.5% and 80.4% variance) with 5 dominant components (loadings ≥ 0.50; explained variance: FMS 64.1%; CBP 60.7%). These components related well to the functional body regions as defined in Table 2 (head-neck, thorax, lumbar region, upper and lower limbs; loadings in supplemental Tables, available at http://links.lww.com/PAIN/B131). Tender points and CPs loaded on separate components in FMS while loading on composite TP-CP components in CBP. Additional PCAs including the pressure pain thresholds and intensity ratings from the algometer test reproduced this group-specific dimensional structure (supplemental Table S5, available at http://links.lww.com/PAIN/B131).

Based on the described functional and regional dimensional structure of the pressure sensitivity, 2 reduced parameter sets were entered into cluster analyses at the parameter level to define sensory phenotypes for the final patient reclassification. The first LCA covered the pain intensities of the manual test at all TPs and CPs; the second analysis combined the algometer pain thresholds and suprathreshold intensities at the trapezius TP and the thenar CP with the corresponding manual probe measures. Because of the dimensional differences found, separate cluster analyses were calculated for patients with FMS and CBP at the parameter level. Figure 2 represents the results as body maps (A, B; dendrograms in supplemental Figure S4, available at http://links.lww.com/PAIN/B131).

Figure 2.

Body maps of spatial clusters of pressure sensitivity to manual probing in patients with FMS and CBP diagnoses. Regional clusters of pain intensity ratings to manual pressure probing; colors: sites of the same regional clusters on the front and back; sites without color: orphans. (A) FMS and (B) CBP. Hierarchical latent class analysis with multiscale bootstrap resampling.60 Significant clusters according arbitrary unbiased/bootstrap probabilities; AU ≥95% significant; dendrograms in supplemental Figure S4, available at http://links.lww.com/PAIN/B131. CBP, chronic primary back pain; FMS, fibromyalgia syndrome.

In FMS, LCA separated pain intensities of the manual probes at TPs and CPs in functionally and regionally distinct clusters (A). In CBP, regional clusters were retained also, but tender and control sites grouped together in the same cluster (B). In patients with FMS, the cluster of TPs on the back, shoulder, neck, and occiput formed a coherent region (pink and purple sites in Figure 2A) as did the TPs on the extremities (olive, orange) and one on the upper back (supraspinatus, dark orange). These regions of classical ACR TPs separated from the corresponding pressure-sensitive CPs. Pressure pain sensitivity at CPs was also arranged in separate regions on hands and feet (dark blue), on the clavicle and the forearm (green). The LCA of pressure sensitivities of patients with CBP resulted in a different cluster solution, reflecting predominantly anatomical vicinity regardless of the TP or CP status of the stimulation site (B). In particular, TPs and CPs on the head, upper back and front, and hip were classified together (olive, orange, red, and purple) as were the test sites on the upper and lower extremities (blue and green).

The group-specific spatial and functional organization of pressure sensitivities was confirmed by the separate LCA using only the algometer and manual test indicators from the selected exemplary trapezius TP and the thenar CP (dendrograms in supplemental Figure S4, available at http://links.lww.com/PAIN/B131).

Profiles of Pressure Pain Sensitivity and Clinical Widespread Pain. Markers of pressure pain sensitivity correlated differentially with clinical pain (FMS: number of sensitive TP × WPI: ρ = 0.287; P = 0.003, P < 0.01, corrected; CBP: n. s.) while heat pain parameters correlated scarcely and unsystematically. Including them in cluster analyses led to attrition samples too small for cross-modal phenotyping. Therefore, only the spatial distribution and the degree of hypersensitivity to percutaneous pressure were combined with central aspects of clinical pain in the final cluster analysis at the patient level. Differences in heat pain sensitivity between unimodal clusters were examined in dependent post facto comparisons.

Six classifiers comprising present pain intensity and the number of pain regions as clinical characteristics and the number of hypersensitive TPs and CPs, the pain intensity of the manual test (right trapezius TP), and the algometer pain threshold (right trapezius TP), sufficed to produce stable profiles of patients with the complete data sets remaining in the analysis (FMS: 69, CBP: 90). In the following, only the result of the final LCA using this minimal indicator set of the total sample of 159 patients with interpretable cluster solutions is reported (Table 4 and Table 5 and Figure 3; further information on LCA methods, fit criteria, etc.: supplemental Tables S10 and S11, available at http://links.lww.com/PAIN/B131).

Figure 3.

Discriminator profiles of 4 phenotypes of pain and pressure sensitivity of patients with FMS and CBP diagnoses. Polar chart of sensory-clinical profiles of the 4 main clusters of patients with the best fit (BIC/CAIC indices). Polar axes: indicator median/mean standardized to 0–100 (modified after65). BIC, Bayesian Information Criterion; CAIC, Consistent Akaike Information Criterion; CBP, chronic primary back pain; CP, control point; FMS, fibromyalgia syndrome; MPI Intensity, present pain intensity (MPI-D, scale 1, item #1); MPI, Multidimensional Pain Inventory; PPI man trap-r, pressure pain intensity at the right trapezius TP in the manual probe. PPT qst trap-r, pressure pain threshold at the right trapezius in the quantitative algometer test; Sens CPs, number of sensitive CPs; Sens TPs, number of sensitive TPs; TP, tender point; WPI, Widespread Pain Index.

The LC model with 4 clusters of patients with diagnoses of either FMS or CBP fitted best (Bayesian Information Criterion and Consistent Akaike Information Criterion) with very similar profiles irrespective of whether the number of pain sites was based on current or major pain or on the WPI. The sensory-clinical profiles differentiated best when the WPI was used (Figure 3, entropy R-squared = 0.8601). These pain phenotypes were reproduced with female data alone to control for the uneven sex distributions in the FMS and CBP groups.

The 4 clusters were characterized by medium strong clinical pain splitting into 2 cluster pairs with high vs moderate WPI and high vs low pain intensity in the manual sensitivity test (clusters 1 and 2 vs clusters 3 and 4 in Figure 3). The first cluster pair with widespread pain and high pressure sensitivity and the second cluster pair with narrow-spread pain and low pressure pain sensitivity were further differentiated into 4 sensitivity categories from low to high by the numbers of sensitive TPs and CPs. In addition, cluster 1 was separated from all other clusters by the very low pressure pain threshold in the quantitative sensory algometer test. Note that not only the number of sensitive ACR TPs but also the number of sensitive CPs differentiated the widespread pain clusters 1 and 2 from the clusters 3 and 4 who showed very few pressure-sensitive CPs. The best discriminating indicators according to absolute profile distances between the 4 clusters were the number of sensitive TPs followed by the pain intensity rating at the trapezius TP in the manual test and the WPI of clinical pain. The first 2 sensory discriminators differed significantly between all 4 clusters (number TPs: P = 0.012; pain intensity trapezius: P < 0.001; Kruskal–Wallis plus pairwise U tests, corrected; complete statistics in the supplemental material, available at http://links.lww.com/PAIN/B131). The WPI differentiated only between the high and low sensitivity cluster pairs but not within them (clusters 1 and 2 vs clusters 3 and 4: U test, P < 0.001; single contrasts, n. s.). The low pain threshold in the algometer test at the trapezius differentiated cluster 1 from the rest (P < 0.001, uncorrected), suggesting a unique phenotype within the widespread pain population.

However, the clinical and sensory profiles in Figure 3 are not accompanied by differences in clinical pain intensity, despite the obvious differences in spatial spread as well as pressure pain sensitivity. Whether these relate to differences between subgroups of diagnostic syndromes is clarified by the reclassification of clinicians' diagnoses and the diagnoses according to the Fibromyalgia Symptom scale (FS ≥12) into the 4 clusters according to LCA-derived sensory-clinical phenotypes (Table 6): FMS and CBP diagnoses were distributed meaningfully and highly significantly different over the 4 clusters (Table 4; P < 0.001; Ccorr = 0.881, df = 3). Clusters 1 and 2 with high WPI and many sensitive TPs comprise most patients with FMS (87%), and cluster 1 collected two-thirds of them. By contrast, clusters 3 and 4 with low to moderate WPI contain 91% of the patients with CBP, showing also low pressure sensitivity in the manual and the algometer test. Cluster 4 with the least sensitive patients consists only of patients with CBP and no patients with FMS at all. Cluster 3 with low to medium pressure sensitivity contains mainly patients with CBP (84.5%) but also 9 (15.5%) patients with FMS. Table 5 illustrates the close relations between the clinical profiles of widespread vs regional pain and the phenotypes of high and low pressure sensitivity. However, it shows also that the previous diagnosis FMS and CBP groups were inhomogeneous and may be reclassified into at least 4 subgroups with different sensory phenotypes.

This is clarified by cross-tabulating according to the Fibromyalgia Symptom Scale criterion of FS ≥12: The FS criterion misclassified 20% clinical diagnoses (FMS: 19%; CBP: 21%), and the cross-classification with the pain-pressure sensitivity phenotypes was more variable. The same was true for the cross-classification with the 2 cluster pairs (FS ≥12: 30%; FS <12: 14%). Moreover, the FS criterion misclassified particularly patients with CBP in clusters 3 and 4 (FMS: 14%, CBP: 24%, Ccorr = 0.310 and 0.331, df = 3, n. s.; Table 4).

Discriminant Analysis of Widespread vs Regional Pain. The discriminative power of the selected indicator set characterizing the 4-cluster classification was cross-validated with stepwise discriminant function analysis with individual indicators entering in the order of their absolute distance scores. Three canonical discriminant functions of the 4 sensory indicators alone sufficed to differentiate the 4 clusters (functions 1 × 2: Wilks' Λ = 0.093; 2 × 3: Λ = 0.595; function 3: Λ = 0.881; P < 0.001; overview: Supplemental Table S13, available at http://links.lww.com/PAIN/B131). Pain intensity and WPI added not substantially to cluster discrimination, although contributing to cluster building by LCA. The first discriminant function explained 89.8% of the variance (canonical R 2 = 0.84), the second 7.9% (canonical R 2 = 0.32), and the third 2.2% (canonical R 2 = 0.12). The number of sensitive ACR TPs had the highest weight on the first function followed by the pain intensity of the manual probe on the trapezius TP. The algometer pressure pain threshold obtained the highest weight on the second discriminant function, whereas the number of sensitive CPs contributed most to the third function. The 3 functions classified 82.4% of the original LCA groupings correctly (N = 131). Importantly, the clusters were systematically positioned in the discriminant function space where clusters 1 and 2 were well separated from 3 and 4 with less good discrimination within these pairs in accord with the cluster pair characterization above (Figure 4).

Figure 4.

Discriminant analysis validation of sensory-clinical phenotypes of widespread and regional pain. Discriminant analysis validation of sensory-clinical phenotypes of widespread and regional pain: location of patients in the 2-dimensional projection of the 3D orthogonal discriminant function space. Abscissa: function 1 = first step separation with highest weights for the standardized manual test; ordinate: function 2 = second step separation with highest weight for the quantitative algometer test; function 3 (least discriminating) omitted; numbered filled black circles = cluster centroids. Method note: stepwise linear orthogonal discriminant function analysis with 4 a priori groups, Wilks' lambda as variable selection criterion; program: SPSS version 25 (IBM, Armonk, NY). Standardized canonical discriminant function coefficients: function 1 = 0.868 × Sens TPs + 0.415 × PPI man trap-r – 0.035 × Sens CPs −0.149 × PPT qst trap-r; function 2 = 0.024 × Sens TPs + 0.179 × PPI man trap-r + 0.367 × Sens CPs + 0.915 × PPT qst trap-r; function 3 = −0.547 × Sens TPs + 0.244 × PPI man trap-r + 0.921 × Sens CPs − 0.393 × PPT qst trap-r. CP, control point; PPI man trap-r, pressure pain intensity at the right trapezius TP in the manual probe; PPT qst trap-r, pressure pain threshold at the right trapezius in the quantitative algometer test; Sens CPs, number of sensitive CPs; Sens TPs, number of sensitive TPs; TP, tender point.

The high-sensitive patients with widespread pain in clusters 1 and 2 were concentrated on the right according to their high positive values in the manual test (function 1) and separated along the axis of function 2 because of the differences in the algometer test. The singularly low-sensitive CBP patients of cluster 4 accumulated in the lower left according to their negative values on function 1. Although these 3 clusters exhibited more or less homogenous subgroups, the moderately sensitive patients with mixed pain in cluster 3 were more heterogeneous, distributed along both axes partially overlapping with the other clusters.

False classifications of patients with respect to the original 4 sensory-clinical phenotypes occurred only between clusters 1 and 2 and between clusters 3 and 4 by the first 2 discriminant functions leaving out intensity and spatial spread (WPI) of the clinical pain. Patients in cluster 4 having no sensitive TPs and reporting no pain in the manual test at the trapezius TP were 100% correctly classified. This relates to the high frequency of patients with CBP in this cluster (33; 37%) and the fewest patients with critical WPI (FS ≥12) of all clusters (cf. Table 6). Cluster 2 classification was second best with 90.5% correct. Cluster 1 (the highly pressure-sensitive cluster with widespread pain) and cluster 3 (the low sensitivity cluster with several pain loci) were predicted less well with 80.9% and 70.7% correct. Overall classification into 1 of the 2 cluster pairs by the sensory profiles was much better with error rates of 4.4% (1 or 2) and 6.6% (3 or 4).

Heat Pain Sensitivity and Widespread Pain. The a posteriori comparisons including also all patients with partially incomplete sensory sets (69 FMS; 90 CBP) showed heat pain sensitivity was generally higher in FMS than in CBP but varied considerably over different measures. Phasic heat pain thresholds did not differ, and tonic thresholds at the thenar of patients with FMS were only slightly lower (43.3 ± 1.7°C vs 44.6 ± 1.6°C; P = 0.026, n. s. corrected). The suprathreshold psychophysical function (Steven's coefficient estimator) did also not differ (complete statistics in the supplemental material, available at http://links.lww.com/PAIN/B131). The pressure pain differences between the trapezius TP and the thenar CP were not reproduced by heat pain. By contrast, the 2 phenotypical cluster pairs differed significantly in phasic and tonic heat pain thresholds both at the trapezius and the thenar (clusters 1 and 2 combined against clusters 3 and 4 combined: P = 0.049/0.002; P < 0.001/P < 0.001). Neither the psychophysical function nor the VAS rating markers of temporal summation (ΔT, ΔE) differentiated between clusters. In summary, although there is evidence of cross-modal lowering of pain thresholds, results were inconsistent and heat pain sensitivity did not add substantially to sensory phenotyping of the patients with chronic primary pain.

Comorbidity With Somatic and Mental Disorders and Psychosocial Factors in 4 Clusters of Pressure Pain Sensitivity. The 4 subgroups of patients identified were further explored with respect to the complete array of sensory and clinical pain measures including those not used in the LCA profiles and related to pain impact and coping, chronicity, functional level, comorbidity, as well as psychosocial cofactors often associated with fibromyalgia and/or CBP. The corresponding group comparisons in Table 6 qualify the current evidence on the diagnostic significance of these aspects. They are not attributable to age differences (n. s.) or external factors such as physical load at the work place or in the household (supplemental Table S14, available at http://links.lww.com/PAIN/B131). Sex ratios reflected those in the previous diagnostic groups with women predominating in clusters 1 and 2, which had also significantly higher CPG (Kruskal–Wallis, P = 0.004). Neither difference accounted for the specific differences nor their absence between the clusters.

Closely pain-related indicators, in particular, pain severity (MPI-D, part I, subscale 1), interference (MPI-D, part I, subscale 2), and impact (Fibromyalgia Impact Questionnaire), differed most between phenotypes. Present pain intensity and global pain severity were higher in patients of clusters 1 and 2 than in clusters 3 and 4 (t test: P = 0.005; U test: P = 0.012). Accordingly, functional capacity (FFBH) of patients in clusters 1 and 2 was significantly lower (P < 0.001). Coping with pain showed a similar pattern, with higher catastrophizing thoughts (PRSS) in clusters 1 and 2 than in clusters 3 and 4, whereas active coping and MPI-D Life Control did not significantly differ. The same applied for the subscales of the Fear-Avoidance-Beliefs Questionnaire. In general, differences reflected those between the diagnosis groups of FMS vs CBP. By contrast, psychosocial factors, ie, MPI-D Social Support and responses by significant others (MPI-D punishing, solicitous, and distracting) did not differ significantly between the 4 sensory clinical phenotypes nor did perceived stress load (Brief Stress Scale) or MPI-D General Activity Level.

Most importantly and contrary to previous work on clinical FMS criteria, comorbidity with mental disorders, ie, depression and self-reported mental health (ADS, SF-12-mental), did not differentiate the 4 clusters of pain phenotypes nor the diagnosis groups. Only trait anxiety (Trait Anxiety Scale of the State-Trait Anxiety Inventory) tended to be somewhat higher in clusters 1 and 2 compared with clusters 3 and 4 (P = 0.012; n. s. corrected). Major depression diagnoses were below 5% in all clusters as well as in the diagnosis groups. Perceived sleep quality (PSQI), a marker item for depression, was also not significantly different between clusters nor between the diagnosis groups.

By contrast, somatic comorbidity in terms of self-reported body complaints (GBB, total symptom burden) and somatic health (SF-12 physical component) differed strongly between the 4 clusters increasing progressively from clusters 4 to 1 (Kruskal–Wallis; P < 0.001). Clusters 1 and 2 were well separated from clusters 3 and 4 with respect to their somatic symptom burden (t test; P < 0.001, corrected) similar to previous diagnostic groups. Patients in cluster 1 with widespread pain and high pressure pain sensitivity had the highest somatic symptom scores and the lowest physical well-being. Cluster 2 lay between cluster 1 and the low-symptomatic groups of clusters 3 and 4. This pattern was preserved when the pain-related GBB subscale "musculoskeletal complaints" was taken out. The differences in self-reported somatic symptoms were not associated with previous medical diagnoses of gastrointestinal, endocrinological, or urogenital disease (cf. supplemental Table S15, available at http://links.lww.com/PAIN/B131).

The higher gastrointestinal score in clusters 1 and 2 is underlined by the distribution of FGID (irritable bowel syndrome and/or nonulcer dyspepsia) according to the self-report questionnaire (FGID; supplemental Table S16, available at http://links.lww.com/PAIN/B131). Again, FGID were most frequent in clusters 1 and 2 (27 of 68; 39.7%) and lowest in clusters 3 and 4 (13 of 91; 14.3%). The frequencies were comparable with those in the previous diagnosis groups (FMS: 29 of 73; 39.7%, CBP: 8 of 69; 11.6%). This general pattern of FGID across clusters (1 ≈ 2 > 3 > 4) was the same for women alone.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....