Development of a new bed-side-test assessing conditioned pain modulation: a test-retest reliability study

2.1 Subjects

In the reliability study, 22 healthy subjects (15 men) with a mean age of 23.6 years (SD: 2.4) participated (Table 1). For the methodological study, 29 healthy subjects (nine men) with a mean age of 21.5 years (SD: 1.6) were included. The subjects were recruited through postings on the campus of Aalborg University, Denmark. The inclusion criteria for both studies were age between 18 and 40 years, healthy, pain-free on the test day, and able to speak either Danish or English. Furthermore, the subjects were asked not to consume alcohol or medication on the day of the test. The subjects were given detailed written information and verbal explanation and they signed an informed consent before participating in the study. All included subjects completed the studies with no loss to follow-up. The study was approved by the local Ethics Committee (N-20170088) and was conducted according to the declaration of Helsinki. The reliability study followed the GRRAS recommendations [18].

2.2 Experimental protocol

The same rater (JBL) conducted all tests in both cohorts.

2.2.1 Test-retest reliability: test protocol 1

Test protocol 1 was developed through pilot testing of the CPM method. Two assessment sessions were conducted in which the same test setup was used. The subjects were in a supine position and the dominant leg was used as the testing limb. As the painful phasic test stimulus, a 6 kg handheld, spring-based pressure algometer (SMI, Aalborg University, Fig. 1) was applied perpendicularly to the muscle belly of the tibialis anterior muscle for 10 s (measurement 1). The subjects rated the pain intensity on a 0–10 numerical rating scale (NRS, “0” represented “no pain” and “10” represented “worst pain imaginable”). Then a standard clamp, inducing a force of 1.3 kg, was applied extra-segmentally at the ipsilateral earlobe for 60 s as the painful tonic conditioning stimulus. After the 60 s, the subjects were asked to rate the conditioning pain intensity. Finally, the application of the test stimulus was re-introduced (measurement 2), while the clamp was still applied to the earlobe. Subsequently, the subjects rated the pain intensity of the test stimuli over the tibialis anterior. For the second measurement, the test stimulus was applied slightly proximally or distally to the first measurement to minimize any residual and carry over effects of the repeated stimuli. The same approach was repeated 24–48 h after the first test session to calculate reliability values.

Fig. 1:

Pressure algometer and clamp

3.1 Test-retest reliability: test protocol 1

Table 2 reports NRS scores for measurements 1, 2, and conditioning stimuli. The CPM effects in session 1 and session 2 were 0.3 (SD: 1.6) and 0.2 (SD: 1.0), respectively; illustrating no significant CPM effect (p-values: 0.35 and 0.29). This equals a relative decrease in pain intensity of 8.2% and 6.5%. CPM responder rates showed a CPM effect in nine subjects (40%) in session 1 and in 10 subjects (45%) in session 2.

Table 3 reports the test-retest reliability coefficients. The ICCs for measurement 1 and measurement 2 were 0.67 and 0.72, respectively (p-values: <0.01). Measured in NRS, the SEM values were 1.9 and 2.1 for measurement 1 and 2, respectively (Table 3).

In the Bland-Altman plots, LoA between sessions were −2.3 to 3.0 and −2.8 to 3.3 for measurement 1 and 2, respectively (Figs. 2 and 3). The LoA for the difference in CPM effect between sessions were −2.0 to 2.8 (Fig. 4).

Fig. 2:

Bland-Altman plot and limits of agreement for measurement 1 in the reliability study. Values are pain intensities measured with numerical rating scale.

Fig. 3:

Bland-Altman plot and limits of agreement for measurement 2 in the reliability study. Values are pain intensities measured with numerical rating scale.

Fig. 4:

Bland-Altman plot and limits of agreement for the between-session CPM effect in the reliability study. Values are pain intensities measured with numerical rating scale.

3.2 Methodological development: test protocols 2, 3 and 4

Table 4 reports CPM effect for test protocols 2, 3, and 4 and conditioning stimulus for each test protocol. The CPM effects for test protocols 2, 3, and 4 were 0.2 (SD: 1.0), −0.1 (SD: 1.1), and 0.0 (SD: 1.2), respectively (p-values: 0.29, 0.49, and 1.00). The CPM effect equals a relative decrease in pain intensity of 4.3% in test protocol 2, an increase in pain intensity of 3.3% in test protocol 3, and no changes in pain intensity in test protocol 4. Overall, 13 subjects (45%) in test protocol 2, 11 subjects (38%) in test protocol 3, and 12 subjects (41%) in test protocol 4 could be categorized as CPM responders.

4.1 Reliability study

Our study observed a CPM effect of 6–8% in the different sessions. This is lower than in the study by Graven-Nielsen et al., which observed a CPM effect of 13–18% for pressure pain tolerance and pressure pain threshold when using cuff-induced pain as conditioning stimulus and cuff pressure algometry as test stimulus [22]. Nahman-Averbuch et al. observed a CPM effect varying from 0 to 56% with heat as conditioning stimulus and various test stimuli (heat pain, pressure pain threshold, temporal summation) [12]. Discrepancies could be explained by the different use of test and conditioning stimuli in different intensity ranges. In the present study, pressure pain was used as both test and conditioning stimulus. The paradigm was applied as a “one size fits all” method with same magnitude of force being applied to all subjects as this is the approach most likely to be used for a simple bedside test. Recently, a study by Schliessbach et al. concluded that a decrease in pain threshold after a conditioning stimulus was seen in more than 10% of healthy subjects and that this does not necessarily indicate an abnormal finding [23]. In line with the present study, several other studies have reported large intra-individual variation in CPM effects in healthy subjects [12], [13], [23], [24], indicating that CPM might not be a particularly stable measurement.

Various CPM protocols have revealed different ICC values and have reported different statistics, such as SEM or LoA [11], [13]. A systematic review examining reliability of different CPM protocols ranged the relative reliability of the test stimulus from good-to-excellent when using pressure pain threshold as stimulus [11]. The review also reported that when using subjective pain rating for heat pain as test stimulus, the relative reliability was lower and ranged from poor-to-fair. In the present study, the ICC values were interpreted as “good relative reliability” (ICC 0.67–0.72) according to the recommendations [11], [20]. The measurements revealed similar ICC values for the test stimulus; both with and without conditioning stimulus. These values are similar to other studies examining the relative reliability of a test stimulus with and without conditioning stimulus [11], [13]. This implies that the bedside algometer test is reproducible. The absolute reliability can be measured as SEM, which provides a value for the measurement error in the same unit as the measurement itself [21]. This study observed SEM values for measurements 1 and 2 of 1.9 and 2.1, respectively. To our knowledge, no other studies have investigated SEM of subjective pain ratings during CPM testing. The Bland-Altman plots show similar LoA between measurements and for the CPM effect. The high values indicate large intra-individual variation, which again indicates that the absolute reliability was somewhat low. Thus, it was not possible to evoke a reliable CPM effect in this population of healthy subjects. The LoA observed in the study were lower than in Lie et al. [24]. These authors reported LoAs ranging from −3.4 to 3.3 and from −4.3 to 3.2 for the test stimulus and from −2.3 to 3.2 for the CPM effect, indicating a somewhat lower absolute reliability than in the present study.

4.2 Test stimulus

A pressure algometer with the force of either 6 kg or 10 kg was applied as a phasic test stimulus to induce pain. As expected, the 10 kg force resulted in higher pain ratings. However, the pain ratings only increased from 3.6 to 3.9 with the 6 kg force and from 4.3 to 4.5 with the 10 kg force, which is regarded a somewhat small increase. This could indicate that once a test stimulus exceeds the pain threshold, a quite substantial increase in the test stimulus is needed to further increase the pain rating in healthy subjects when using a localized pressure stimulus. Future studies are needed to examine whether this phenomenon is apparent in patients with chronic pain or whether more severe pain increases can be observed with higher test stimuli. The applied test stimuli were fixed to 6 or 10 kg and not individually based. Thus, large variations of pain ratings were seen. It is unknown whether this “one size fits all” approach could influence the CPM effect. However, when examining the individual data from the participants, no tendency towards higher test stimulus pain ratings leading to larger CPM effect or vice versa were observed.

As the outcome measure for subjective pain rating, both VAS and NRS were used. These methods have been shown to measure almost the exact same pain intensities when used simultaneously [25]. NRS was implemented since it is often used in clinical settings and as it is an easy method to administer for the subjects tested [26]. VAS has commonly been used in other CPM studies [1], [13], [27]. NRS was used in the reliability study. However, since the test stimulus pain ratings were repeated with short intervals, it was considered that the recalling of the first pain rating might bias the second pain rating. Therefore, a VAS was used in the methodological study since it was considered less prone to bias.

4.3 Conditioning stimulus

Previously, a conditioning pain intensity of VAS=4 has been used when evaluating CPM [16]. In the present study, the mean conditioning pain intensity varied from NRS 3.8 to 4.6. A study by Graven-Nielsen et al. [22] showed that different CPM responses could be obtained with the use of different conditioning pain intensities. Therefore, the methodological study introduced the use of two clamps and longer duration of the applied stimuli to increase the evoked conditioning pain intensity. These changes led to a lower conditioning stimulus intensity in the range of VAS 2.2 to 2.9. This was contrary to our expectation as it has previously been observed that stimulation of large areas results in decreased pain threshold, indicating spatial summation [15]. A study by Staud et al. [28] found that when separating the probes used for mechanical pressure by 8 cm, both fibromyalgia patients and healthy subjects reported higher pain intensity compared with separating the probes by 4 cm only. This could indicate that higher pain ratings are more likely to occur when stimulating across spinal segments. However, this is contrary to our test protocol using two clamps applied to the same earlobe. Furthermore, it is possible that some pain adaption occurred during the longer duration of the conditioning stimulus (i.e. more than 60 sec) resulting in lower pain intensities after a prolonged period of conditioning stimulus. In future studies, it is required to evaluate whether a higher conditioning stimulus will induce a larger CPM effect. This could not be examined in the present study despite the attempt to induce higher conditioning pain.

4.4 CPM responder

Our study defined a subject as a CPM responder when experiencing a CPM effect during the test, which ranged from 38 to 45% of the subjects. A study by Schliessbach et al. [23] attempted to establish reference values for the CPM effect by evaluating CPM effects at the 5th, 10th and 25th percentiles. The authors found a CPM effect of less than zero at both the 5th and 10th percentiles and concluded that the choice of cut-off value should depend on the particular clinical or scientific question. Fundamentally, such reference values should be statistically determined as z-scores from large normative cohort studies. The largest CPM study to date based on around 2,000 subjects could be used to define statistical ranges around the mean [29]. The complicated aspect is that the normative data and the size of the CPM responses vary from methodology to methodology. It has also been shown that within the same subjects, different CPM methods elicit different responses and that subjects may be a responder to one method but a non-responder to another method [13], [30]. However, it is evident that the CPM effect is a variable parameter and currently all the many parameters influencing the response are not known.

4.5 Limitations

Sex has been shown to have an influence on CPM as several studies have found stronger pain inhibitory capacity in males versus females [31], [32], [33]. The two populations in the present study consisted of 68% males in the reliability study and 31% in the methodological study. The reliability study revealed a better CPM effect than the methodological study and this might partly be explained by the inclusion of more males than females. The use of both VAS and NRS to assess pain intensities may have introduced a potential bias in the comparison between groups. Since the purpose was to compare the within-group pain intensity, the possible bias is unlikely to affect the CPM effect.