Do chronic low back pain subgroups derived from dynamic quantitative sensory testing exhibit differing multidimensional profiles?

Introduction

Low back pain (LBP) is the largest contributor to global morbidity [1], [, 2]. Chronic LBP (CLBP) is a multidimensional condition influenced by many factors including pain characteristics, genetic, demographic, physical, psychological, social, health and lifestyle [3], [, 4].

When examined using quantitative sensory testing (QST), people with CLBP have been shown to have heightened pain sensitivity compared to healthy controls [5], [, 6], and differing pain sensitivity profiles have been derived in people with CLBP [7]. However, one recent systematic review suggests static QST (e.g. pain thresholds) may be less helpful for predicting future pain intensity than dynamic QST [8], while another found predominantly non-significant associations between dynamic QST findings and pain intensity and disability [9]. Dynamic QST includes assessment of temporal summation (TS): central nervous system-mediated increasing perception of pain following repetitive noxious stimulation [10]; and conditioned pain modulation (CPM): a surrogate measure of a person’s endogenous pain modulation capacity [11]. Combining TS and CPM findings to determine the presence of pro- or anti-nociceptive subgroups has been hypothesised as potentially clinically meaningful [12], however, such subgroups are yet to be thoroughly explored in an LBP cohort.

The exploration of subgroups based upon dynamic QST variables of CPM and TS has been investigated in participants with non-cancer related chronic pain (n=400, 18.8% CLBP) [13]. Four subgroups were determined; normal CPM and normal TS (30%), normal CPM and facilitated TS (11%), impaired CPM and normal TS (37%) and impaired CPM and facilitated TS (21%). Participants with impaired CPM and facilitated TS demonstrated higher average pain intensity and the presence of multiple pain sites. Additionally, between group differences were identified for static QST measures (heat pain threshold [HPT] and pressure pain threshold [PPT]). However, no differences between the subgroups for disability levels or demographic or psychological variables were identified [13]. Considering CLBP is known to be a multidimensional pain condition disorder [3] associated with, for example, differing pain characteristics (e.g. pain provocation with movement [14]) or lifestyle factors (e.g. sleep disturbance [15]), it is appropriate to explore broader multidimensional associations across such subgroups.

To date, no studies have derived subgroups in people with CLBP based upon dynamic QST variables and examined the multidimensional profiles of those subgroups. Therefore, Aim 1 was to subgroup people with CLBP based upon normal/abnormal CPM and TS findings using previously described methods [13]. Using dynamic QST variables to subgroup people with LBP would overcome a limitation reported in literature which has largely focused on testing static QST [16]. Aim 2 was to profile dynamic QST subgroups across a full array of multidimensional variables (pain characteristics, disability levels, psychological factors, lifestyle factors, static QST measures). Aim 3 was to evaluate whether subgroup membership was associated with pain intensity, disability and disability risk at 12-month follow up. Together, Aims 2 and 3 could help inform the clinical utility of subgrouping people based on dynamic QST profiles.

Methods

A longitudinal study was performed with data collected at baseline and 12-month follow-up. Data for this study were obtained from a broader study investigating multidimensional profiling and prognostic modelling of people with CLBP [17], [, 18], however, dynamic QST subgrouping has not been previously reported for this cohort. Data were collected between November 2012 and January 2014.

Ethical approval for the data collection (Approval Number 112/2012) and analysis in this current study (Approval Number HRE2019-0595) was granted by Curtin University’s Human Research Ethics Committee and complied with the Declaration of Helsinki. All participants gave written, informed consent.

Participants were recruited from private pain management, psychology and physiotherapy clinics in Perth, Western Australia and through regional and metropolitan communities in Western Australia via multimedia advertisements.

Inclusion criteria: age 18–70 years inclusive; LBP of ≥3 months duration; dominant axial LBP indicated by ≥60% of the pain in the low back and ≤40% in the legs [19]; average pain intensity in the last week of ≥2 points measured on an 11-point Numerical Rating Scale (NRS) [20]; and LBP-related disability of ≥5 points measured on the Roland-Morris Disability Questionnaire (RMDQ) [21]. Exclusion criteria were: history of extensive spinal surgery (greater than single-level fusion or discectomy); spinal surgery in the previous six months; presence of serious spinal conditions (e.g. acute fracture, cancer, inflammatory arthropathy); diagnosed neurological disease; pain bilaterally over the wrist or dorsum of the hand; pregnancy and inability to comprehend English.

The following variables were collected through questionnaires or physical examination.

Results

Participant (n=273) flow through the study is shown in Figure 1. The demographic characteristics of the total cohort (n=273) is presented in Table 1. Baseline descriptive data, stratified by dynamic QST subgroups for female and male participants reported separately, are shown as a supplemental file (Supplementary material 1).

Figure 1:

Flow of participants through the study.

The proportions of participants (n, %) in the derived dynamic QST subgroups at baseline were: normal CPM and normal TS: 34 (12.5); normal CPM and facilitated TS: 6 (2.2); impaired CPM and normal TS: 186 (68.1); impaired CPM and facilitated TS: 47 (17.2). For TS, 267 (97.8%) reported the initial monofilament stimulation as 0/10 on the NRS. CPM and TS change scores for each subgroup are shown in Table 2 and Figure 2.

Figure 2:

Conditioned pain modulation and temporal summation change scores for the four dynamic quantitative sensory testing subgroups.

Sex differed significantly across the four subgroups (χ²=9.93, p=0.013). Those with impaired CPM and facilitated TS had a higher than expected proportion of females (78.7%). Subsequently sex-adjusted comparisons are reported for all variables with a p-value <0.10 which included pain intensity, pain duration, body chart count, PPT and CPT at the lumbar spine and wrist, and HPT at the wrist (Table 2). After adjusting for sex, number of body chart areas remained different across subgroups (χ²=9.23, p=0.026), with both groups with impaired CPM having a significantly higher number of areas that those with normal CPM and TS. CPT at the lumbar spine (F [3]=5.36; p=0.001) and wrist (F [3]=4.50; p=0.004) also differed across subgroups (Table 3). Those with normal CPM and facilitated TS (n=6, 2.2%) had higher CPT at the lumbar spine and wrist compared to the other subgroups. Those with impaired CPM and facilitated TS (n=47, 17.2%) also had higher CPT at the lumbar spine than those with impaired CPM and normal TS (n=186, 68.1%).

Of the 273 participants included at baseline 250 participants (91.6%) responded at the 12-month follow-up. The proportion of the 250 participants (n, %) in the dynamic QST subgroups at follow-up were: normal CPM and normal TS: 26 (10.4); normal CPM and facilitated TS: 6 (2.4); impaired CPM and normal TS: 175 (70.0); impaired CPM and facilitated TS: 43 (17.2). There was no difference in age, sex or baseline pain intensity between those who did or did not complete the 12-month follow-up (Table 4). However, participants that did not complete the 12-month follow-up had higher baseline disability (RMDQ median score 11-points vs. 9-points respectively; z=2.02; p=0.043), higher baseline disability risk stratification (high risk allocation 60.9 vs. 29.2% respectively; X²=10.54; p=0.005) and were more likely to have normal CPM and normal TS (34.8 vs. 10.4% respectively; X²=12.13; p=0.016). Subgroup membership was not associated with follow-up pain intensity, LBP-related disability or disability risk stratification (SBT) at 12-months, either unadjusted or adjusted for sex (Table 5).

Discussion

This study utilised dynamic QST findings to characterise subgroups within a large cohort with axial CLBP and explored differences between these subgroups across multidimensional characteristics and outcomes at 12-month follow-up. Across the derived subgroups there were no sex-adjusted associations with pain intensity, disability levels, psychological or lifestyle variables at baseline. Number of pain areas was greater in the two subgroups with impaired CPM. Cold pain sensitivity was heightened in both subgroups with facilitated TS, however, the subgroup with normal CPM and facilitated TS was small (n=6) and findings should be interpreted with caution.

Strengths of the study included a large cohort exclusively with axial CLBP who were well characterised across multiple dimensions relevant to pain, and with a low attrition rate at 12-month follow-up. This provided specific advantages over the study by Vaegter and Graven-Nielsen [13] which was not LBP specific and was limited in the profiling variables it included. One limitation of this study is that the nylon monofilament used to assess TS was not re-calibrated between every participant and we cannot exclude the possibility that the monofilament may have functioned slightly differently with changes in humidity. Also, multiple comparisons in this study increase the possibility of type I error. However, with little research to date examining subgroups based upon dynamic QST this is a relatively exploratory study. Therefore, when profiling subgroups it may be more appropriate to maintain p-values such that there may be a greater chance of a type I error, but less chance of a type II error [35]. As such, no correction for multiple comparisons was deemed necessary.

Only one study has previously attempted subgrouping based on dynamic QST measures. Vaegter and Graven-Nielsen [13] formed the same dynamic QST subgroups in a cohort with chronic non-cancer pain. The subgroup with impaired CPM and facilitated TS demonstrated higher average pain intensity, more widespread pain and heightened heat pain sensitivity. The subgroup with impaired CPM and normal TS also had heightened heat pain sensitivity compared to those with normal CPM and normal TS. There may be numerous reasons for the different findings between these two studies. Firstly, the methods utilised for testing CPM and TS differed between studies. Vaegter and Graven-Nielsen [13] utilised repeated cuff pressure stimuli to test TS, and cuff pressure as both the conditioning and test stimuli for CPM testing. There is no gold standard CPM testing protocol [36]. There is also no consensus on categorising CPM or TS as normal or impaired based upon cut-off scores, and these differed in our study from the cut-off scores utilised by Vaegter and Graven-Nielsen [13]. However, the cut-off scores chosen in the current study were consistent with a previous LBP cohort study (CPM) [25] and the minimum clinically important difference on an 11-point NRS (TS) [26]. Finally, the proportion of participants per subgroup differed compared to Vaegter and Graven-Nielsen [13]. The studies were similar with the largest subgroup being those with impaired CPM and normal TS (68.1 vs. 37.0%) and our smallest subgroup with normal CPM and facilitated TS (2.2%) was also the smallest (11.2%) for Vaegter and Graven-Nielsen [13]. Our study may suggest that CLBP populations more commonly exhibit impaired CPM than other chronic non-cancer pain complaints, but the findings may also reflect methodological differences, highlighting the difficulty in comparing studies without standardised testing protocols for dynamic QST. Similar to the findings from our study, the subgroups derived by Vaegter and Graven-Nielsen [13] did not differ across disability level, pain catastrophising or fear of movement. The lack of association between the derived subgroups and pain and disability levels is probably also consistent with previous systematic reviews [9], [, 16]. Finally, the risk of type II error must be acknowledged as the sample size of each subgroup may have limited the power of the study to detect differences across profiling variables.

In our study, those participants with impaired CPM and facilitated TS demonstrated greater cold pain sensitivity at the lumbar spine than those with impaired CPM and normal TS. Taking into consideration that the small subgroup with facilitated TS (but normal CPM) also had heightened cold pain sensitivity may suggest an association between TS and cold pain sensitivity. Indeed, post-hoc analysis suggests a weak correlation between TS change score and CPT (Lumbar, Spearman’s rho=0.26; wrist, Spearman’s rho=0.22). While impaired CPM and facilitated TS may indicate altered descending nociceptive modulation, the drivers of such heightened pain sensitivity may be related to variables not investigated in this study, e.g. physical activity levels, genetics etc., and this postulate requires further examination.

Our findings suggest dynamic QST subgroups do not explain pain intensity and LBP-related disability at baseline or 12-month follow-up. This is consistent with previous multidimensional prognostic modelling of this cohort where CPM and TS were not identified as prognostic factors [18]. The clinical utility of QST measures (static, dynamic or subgroups thereof) may not be in their ability to predict future pain and disability, but rather in their modification of treatment effects [37]. Yarnitsky [38] has detailed early evidence that people with dynamic QST profiles characterised by impaired CPM may benefit from specific pharmacological interventions targeting underlying mechanisms (e.g. Tapentadol), while those with facilitated TS may benefit from other medications (e.g. ketamine). Further investigation of these subgroups as treatment effect modifiers may be warranted.

Our study found a difference in body chart count across subgroups, with the two subgroups with impaired CPM more likely to have a higher body area count compared to those with normal dynamic QST responses. In contrast, Vaegter and Graven-Nielsen [13] found that only the subgroup with impaired CPM and facilitated TS had significantly higher number of pain areas compared to the other subgroups, with the authors suggesting this combination might be a biomarker for widespread pain. This disparity between studies could be attributed to the different cohort profiles. Approximately 85% of the cohort utilised by Vaegter and Graven-Nielsen [13] presented with widespread pain or multiple spinal areas of pain, whereas our cohort presented dominantly with axial CLBP. The methodology for measuring this variable also differed between the two studies. Another study has investigated CPM in people with more localised CLBP vs. those with CLBP and more widespread pain [39]. Their findings were consistent with the present study, that impaired CPM is associated with more widespread pain. Unfortunately that study did not also test for TS, somewhat limiting comparison.

This is the first study that considered the relationship between dynamic QST subgroups and pain intensity, LBP-related disability and disability risk stratification at two points in time. At 12-month follow-up, subgroups did not differ across these variables. The 12-month follow-up analysis did not consider what interventions the cohort had received during this period, which may have influenced outcomes. It is unknown whether cold sensitivity findings persisted at 12-month follow-up as static QST measures were not collected. Furthermore, participants lost to follow-up had higher baseline disability and disability risk stratification and were more likely to have normal CPM and normal TS which may have influenced the results.

Overall, our findings suggest that within a cohort with CLBP a subgroup with impaired CPM and facilitated TS demonstrated heightened cold pain sensitivity. Additionally, those with impaired CPM are likely to have more widespread symptoms. These data suggest altered descending nociceptive modulation, however, the drivers of this heightened pain sensitivity require further examination. While sensory profile may act as a treatment effect modifier, future research examining the effectiveness of targeted interventions to address specific sensory phenotypes is needed.