The inhibitory effect of conditioned pain modulation on temporal summation in low-back pain patients

Participants

All patients referred between February and October (2019) to the Spine Center of Southern Denmark, Lillebaelt Hospital with low back pain as the primary complaint were invited to participate in the study. The Spine Center of Southern Denmark is a large regional spine-care unit which attends to patients on referral from other hospital departments and general medical practitioners, chiropractors and medical consultants in private practice.

Inclusion criteria were: 18+ years of age, ability to speak and understand Danish, referred with a primary complaint of low back pain (defined as dorsal pain, muscle tension, or stiffness localized below the lower costal margin and above the inferior gluteal folds, with or without sciatica), a completed SpineData questionnaire [13], written consent for the use of SpineData for research, and no severe psychiatric disorders.

Patients were excluded from the study if they later withdrew their consent for participation, if the experimental procedure was not tolerated, or in the event of unexpected technical difficulties.

Descriptive and clinical data

Descriptive and clinical data was collected using an electronic, online questionnaire (SpineData). Descriptive data included age, sex, height and weight. Clinical data included current LBP intensity, worst and average LBP intensity over the last 14 days (0–10 Visual analog scale), pain area (quantified from digital pain drawings as convex hull polygons [14]) and LBP duration [13].

QST protocol

Participants were assessed by a research assistant trained in performing a standardized QST test protocol. The current study represents a pre-planned limited analysis of data collected as part of a larger QST test battery. An overview of the complete QST test protocol will be published elsewhere.

For the current analyses, the following QST data were included (see Figure 1):

Figure 1:

Study flow diagram.

QST instructions

The participant information about the QST procedures followed a fixed, pre-authored manuscript, which is available (in Danish) on request.

Data management

Study data were collected and managed using REDCap electronic data capture tools hosted by ‘Open,’ a research support initiative in the Region of Southern Denmark. REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for data integration and interoperability with external sources [15], [16].

Statistical analyses

Comprehensive summary statistics of QST data and descriptive variables are presented as both parametric and non-parametric data. Distribution of QST data are presented as density plots.

The assumptions of homoscedasticity and normality of residuals for a two-way, repeated-measures ANOVA test were not fulfilled by the data. Most likely this is a result of the nature of the visual analogue pain scale, which is bounded at both ends, resulting in left-skewed data.

Instead data is presented visually and analyzed using non-parametric tests, which do not permit for analysis of variance or interactions and have less power.

Distributions are presented graphically as density plots. Density plots can be thought of as histograms which have been smoothed out and have the benefit of not being affected by the arbitrary number of bins used in histogram plotting. Also, density plots allow for simultaneous, comparative presentation of several distribution plots without obscuring overlaying plots.

CPM responder/non-responder group differences in clinical variables were analyzed non-parametrically (Wilcoxon test). Comprehensive summary statistics thereof are provided in appendix.

Descriptive summary statistics

A total of 149 patients participated in the study (60 women, and 89 men).

Descriptive data is summarized in Table 1. A frequency table of clinical ICD10 diagnoses (primary and secondary) are presented in Appendix. The most common diagnoses were Low back pain, Muscle strain, Lumbago with sciatica, Lumbar and other intervertebral disc disorders with radiculopathy, Other spondylosis, Spinal stenosis, Other spondylosis with radiculopathy and Other biomechanical lesions.

Summary statistics of QST

The median pressure probe employed for CPM test stimuli was the 8 kg probe (individual pressure pain threshold + 1 kg) (n=149, median=8, q1=6, q3=10, min=3, max=10).

Distributions of pain intensity (VAS) with single and repeated pressure stimulation before and after CPM are illustrated in Figure 2 as density plots.

Figure 2:

Density plot (smoothed histogram) of pain scores (VAS) for mechanical pressure pain, before/after cold pressor test and with single/repeat stimulation. Vertical lines represent mean values.

Summary statistics of quantitative sensory testing is presented in Tables 2 and 3.

CPM and TS effects

Two-way repeated measures ANOVA (Pillai statistic) indicated statistically highly significant effects of both TS, CPM and their interaction. However, as described in the Method section, the underlying assumptions of ANOVA were not fulfilled.

Unpaired, single-sample Wilcoxon rank-sum test (mu=0) support significant effects of both Temporal Summation and Conditioned Pain Modulation. See Table 4.

Unpaired, single-sample Wilcoxon rank-sum test (mu=0) support significant effects of both Temporal Summation and Conditioned Pain Modulation. See Table 4. Significant differences in the CPM effect was observed when comparing single and repeated stimuli (Paired Wilcoxon test, p<0.0001) and when comparing repeated stimuli to temporal summation (Paired Wilcoxon test, p<0.0001), but not when comparing single stimulus to temporal summation (Paired Wilcoxon test, p≥0.05). Similarly, a significant difference in the effect of temporal summation was observed before/after CPM (Paired Wilcoxon test, p<0.0001).

Figure 3 illustrates the distribution of the CPM effect for both single and repeated (TS) stimulation. As evident from Table 4, a significant CPM effect was observed for both test stimuli.

Figure 3:

Distribution density plot (smoothed histogram) of the Conditioned Pain Modulation effect on single/repeat mechanical pressure stimulation, i.e. the individual change in pain before/after CPM. Vertical lines represent mean values.

Similarly, Figure 4 illustrates the distribution for the TS effect on pressure pain before and after CPM. As evident from Table 4, a significant TS effect was observed at both points in time.

Figure 4:

Distribution density plot (smoothed histogram) of the Temporal Summation effect before/after cold-pressor test, i.e. the individual difference in pain with single versus repeated pressure. Vertical lines represent mean values.

The effect of Temporal Summation was found to be significantly attenuated after CPT (paired Wilcoxon test, p<0.0001), with the mean TS effect dropping from 17.42 δVAS to 11.67 δVAS.

Effect size

The mean CPM effect of cold-pressor pain on single pressure pain was −5.24 (VAS) (Table 4), corresponding to −29% of mean single pressure pain intensity before CPT (Table 2).

Similarly, the mean CPM effect of cold-pressor pain on repeated pressure pain was −10.99 (VAS) (Table 4), corresponding to −31% of mean repeated pressure pain intensity before CPT (Table 2).

Lastly, the mean CPM effect of cold-pressor pain on TS was −5.75 (VAS) (Table 4), corresponding to −33% of mean TS effect before CPT (Table 4).

CPM responders/non-responders

Using the simple CPM responder criteria, the frequency (number) of CPM responders/non-responders was observed to be 75/74 for Single stimuli, 103/46 for Repeat stimuli, and 84/65 for Temporal summation respectively. For the strict CPM responder criteria the frequencies were 137/12, 142/7, and 124/25 respectively, and for the reverse CPM responder criteria 40/109, 64/85, and 55/94.

Group (CPM responder vs. non-responder) differences in clinical pain are presented in Table 5, subdivided into each of the three non-responder criteria (simple, strict, reversed) and each test stimulus (single, repeated and temporal summation). Significant differences were observed in only six of the 45 tests.

Conditioned pain modulation of temporal summation of pain

Two previous publications about the CPM effect on TS [11], [12] were not in agreement. Holden et al. [11] reported no significant differences in TS, whereas Sirucek et al. [12] found a significant CPM induced inhibition of TS. The current data align with the findings of Sirucek et al.

Both Holden et al. and Sirucek et al. applied the conditioning and testing stimuli at the same segmental levels of spinal innervation, albeit on contra-lateral sides. By contrast, the current study applied the test and conditioning stimuli at different segments of spinal innervation. Applying the test and conditioning stimuli at the same segmental level could obfuscate any diffuse descending inhibitory effect of the conditioning stimulus: There is evidence that nociceptive input can lead to contralateral sensitization on the same segment [17] and it is thus conceivable that the intense pain of the conditioning stimulus could interfere with the integration of nociceptive input from the test stimulus and the descending diffuse noxious inhibitory control from higher centers.

The two previous and the current study are different in other ways, including the QST methods employed. As Kennedy et al. [6] observed CPM is sensitive to differences in test method and this could also account for the differences observed between the three studies.

Earlier publications have examined the effects of CPM on second pain elicited by electrical or heat stimulation on healthy individuals [18] and fibromyalgia patients [19]. These authors reported greater inhibitory CPM effect on second pain than first pain, dependent on clinical status and sex. There may be an element of slow onset secondary pain to temporal pain summation and vice versa, but the test methods are not directly comparable and the findings should be compared with care.

All-in-all, there is some evidence to suggest that perturbations in CPM is a feature of chronic low-back, but CPM methodology is important and the literature is not in agreement [2].

Clinical considerations

Several other publications have used TS and CPM as indicators of perturbed central pain modulation in chronic clinical pain states – see e.g. O’Brien et al. [20] and den Bandt et al. [2] for reviews on fibromyalgia and low-back pain patients respectively. To our knowledge the effect of CPM on TS of pressure pain in a clinical population has not been reported before.

The current data is based exclusively on low-back pain patients seen in a specialized hospital spine center. It is therefore not possible to say whether the effects of CPM and TS differ from healthy individuals or other clinical populations.

In any clinical context however, the relevance of a CPM effect is likely to be in the distinction between individuals as CPM responders or CPM non-responders [21]. A CPM non-response could reflect an attenuated diffuse, descending noxious inhibitory control. This in turn is probably important for development of widespread hyperalgesia and chronification of pain [22].

On the face of it, the results presented in Table 5 suggested that significant differences in clinical presentation were observed between CPM responders and non-responders. However, on closer examination this must be questioned: The table included forty-five p-values, which suggests that it would be appropriate to correct for multiple statistical comparisons. Such correction methods (e.g. Bonferroni correction) generally rely on stricter significance criteria, and as the six significant p-values in Table 5 ranged from 0.01 to 0.05, they would likely be concluded insignificant using any correction method. Also, the three clinical measures in question were current pain, typical pain and worst pain which are not independent, but correlated (post-hoc Pearson correlation coefficient=[0.75; 0.9]).

Regarding the best way to dichotomize CPM responses into responders and non-responders, the three criteria resulted in very different frequencies. With the strict criteria, very few patients (n=7–25 – see Results subsection CPM responders/non-responders) were categorized as non-responders. Conversely, with the reverse criteria more than half (n=85–109) were categorized as non-responders. As the current study did not include a healthy control group, we can not determine the CPM non-responder criteria with the best discriminatory ability. However, a previous study of healthy controls can be cautiously contrasted with the present study. Locke et al. [23] also used CPT to induce inhibitory CPM responses, but used pressure pain thresholds as the test stimuli (+5% higher thresholds corresponding to an inhibitory CPM effect) and was focused on identifying CPM responders. In other words, Locke et al. considered CPM non-responders to be those participants with either a small or no inhibitory CPM response, which would correspond to our reverse criteria. In contrast to our LBP population, Locke et al. found that only 7% of healthy controls were CPM non-responders by such reverse criteria. Even with the laxer simple criteria a considerably higher proportion of LBP patients (n=46–74) were categorized as CPM non-responders in our population.

No differences in reported LBP duration were found between CPM responders and non-responders, irrespective of the criteria. Additionally, Gerhardt et al. [5] suggested that altered CPM might drive pain extent (area), but we found no difference in pain extent between CPM responders and non-responders in the current data. At first look, this challenges the impact of pain duration on development of perturbed central pain modulation, but it should be recalled that the current study population consisted exclusively of chronic low-back pain patients in whom such perturbed modulation may be common.

Thus, no difference in clinical pain intensity, duration or extent between CPM responders and non-responders should be inferred from these data. We reiterate however, that the present data are from a select subgroup of chronic LBP patients. We can not rule out, that a comparison with healthy controls or acute LBP patients would demonstrate significant differences.

Restrictions imposed by a VAS

The CPM effect was most pronounced in relation to the pain induced by repeated pressure stimuli. This could possibly reflect an undesired characteristic of the Visual Analogue Scale or methodological caveat: As the VAS is fixed at both ends, there is a risk of flooring and ceiling effects which may limit the usefulness of standardized QST tests [24]. If the test stimulus pain intensity prior to CPM is low, there is little remaining room on the VAS to indicate even lower scores following CPM. As the pain induced by repeated pressure was about double that of a single stimulus, there was more room available for a drop in pain scores following CPM. The similarity in relative effect size of CPM on single and repeated pressure stimuli could be interpreted as reflecting such a flooring effect. Thus, in CPM testing care should be taken to ensure that the intensity of the test stimuli is sufficiently high to avoid a flooring effect and ensuing type B error.

Strengths and weaknesses

Clinical pain status was recorded at a single point in time prior to QST testing and we can say nothing about the temporal development of clinical pain in relation to CPM responder/non-responder status.

As discussed above, only chronic LBP patients were included and whilst significant differences in CPM effect have been reported across different chronic pain diagnoses [25], important differences within pain patient groups have also been reported [5, 26, 27]. The current results should not be uncritically extrapolated to other populations and we can not draw conclusions regarding any potential differences between our study population and healthy individuals.

It is known that attention and expectations [28], [29] may influence the CPM response. No attempt was made to control or measure attention in the current study, but conversely a strict and uniform testing protocol was followed throughout.