Reference values of conditioned pain modulation

2.1 Design and setting

The study was carried out at the University Department of Anesthesiology and Pain Medicine, Inselspital Bern, Switzerland, according to good clinical practice guidelines and in accordance with the Declaration of Helsinki. Approval by the local Ethics Committee (KEK 066/13) was obtained. The study was registered with clinicaltrials.gov (NCT02377180). All participants gave written informed consent prior to enrolment.

2.2 Subjects

Subjects were recruited by advertisement in local newspapers, among members of local clubs and associations and by word of mouth. In- and exclusion criteria were mentioned in the advertisement. Exclusion criteria were chronic pain, pain at the time of testing, intake of any analgesic medication within 24 h before the test, intake of antidepressants, opioids, benzodiazepines or anticonvulsants within 1 week before the test, any neurological disease and sensory deficits.

We aimed at recruiting 150 subjects. A sample size calculation was not performed. One-hundred and eighteen subjects responded to an advertisement in local newspapers. Eighty-seven of them were recruited and successfully tested. Of the remaining 31 subjects, 27 could not be called back or reached by phone for an appointment, one was no longer willing to participate and three did not show up for the experiment. Furthermore, we distributed the advertisement by e-mail among local clubs and associations (sport clubs, choir singers, music societies, etc.) that were identified by internet search in the region around Bern. Another 59 subjects could be recruited by this effort. After reaching a total sample size of 146 subjects, we stopped the recruitment efforts, as we felt they would not be worth only four additional subjects. Participants received a compensation of 50 Swiss Francs for the inconvenience, time and travel expenses.

2.3 Variables of interest

The following variables were recorded: gender, age, weight, height, body mass index, Beck Depression Inventory fast screen (BDI) [16], state trait anxiety inventory (STAI Trait) [17] and Pain Catastrophizing Scale [18]. The average sleep quality during the week before the experiment was rated by the participants on a 0–10 numeric rating scale (NRS), with 0=poorest and 10=best imaginable sleep quality.

2.4 Conditioned pain modulation

Figure 1 illustrates the time course of the experiment. Pressure and electrical stimulation were the test stimuli. The cold pressor test was the conditioning stimulus. The experiment was carried out in a quiet room, with the participants placed comfortably in supine position. The upper body was elevated by 30–45 degrees. A series of training measurements was performed until the subjects were familiar with the procedure. After multiple training measurements, baseline values were recorded using one single measurement at each test-site. Single measurements, as compared to three, do not lead to relevant measurement error in the tests that we applied [19].

Fig. 1:

Time course of experiment. EPT=electrical pain threshold; NWRT=nociceptive withdrawal reflex threshold; PPDT=pressure pain detection threshold; NRS=numeric rating scale (0=no pain, 10=worst pain imaginable).

2.5 Statistical analyses

Baseline values of QST are presented as means±SD. The CPM effect is reported both as absolute difference between the measurement during and before the cold pressor test, as well as the percent change from baseline. This is in accordance with published recommendations on CPM testing [20].

Quantile regression analyses were performed to calculate reference values for CPM effect. This analysis can be used to estimate quantiles within a given sample, without requiring normal distribution of measurements [21]. Quantile regression is similar to multiple regression. However, instead of estimating the mean of the dependent variables, it estimates their quantile distribution. The percentile values that are calculated can be regarded as the critical values of the tests, thereby reflecting their reference values. Two sets of regressions were performed using electrical and pressure CPM as dependent variables, respectively. Age, sex, body-mass index (BMI), sleep quality, BDI, STAI trait and catastrophizing scale were independent variables. This is in accordance with previous studies that found associations of quantitative sensory tests and CPM with demographic variables, psychosocial factors, and sleep [22], [23], [24], [25]. The 5th, 10th and 25th percentile were estimated, analogous to previous studies [22], [26], using bootstrapped standard errors with 1,000 replications and the “sqreg”-command in STATA (STATA SE 13, College Station, TX, USA).

3.1 Baseline measurements

Baseline values of EPT, PPT and NWR before the cold pressor test are displayed in Table 2. There were 2 missing values for EPT and 1 for PPT because of device malfunction. The baseline NWR could be recorded in only 73 of the 146 subjects (50%), as pain become intolerable before a reflex could be elicited. Among those 73 subjects, 64 showed a measurable reflex during the cold pressor test.

3.2 CPM with pressure pain thresholds

Pressure pain thresholds rose by 98 kPa from baseline (SD 103), which corresponds to a 27.5% increase (SD 26.6) (Table 3). Multiple linear regression with pressure CPM as dependent variable showed none of the explanatory variables to have a significant influence on the CPM effect [F(7, 134)=0.5, p=0.83, r²=0.02]. The percentiles of reference values are presented in Table 4 for both absolute values in kPa and percent change. They are also illustrated in Fig. 2A.

Fig. 2:

Illustration of the distribution of conditioned pain modulation (CPM) for the pressure pain paradigm (A) and for the electrical pain paradigm in females (B) and in males (C). The short-dashed line indicates the 5th percentile, the long-dashed line indicates the 10th percentile and the dashed-dotted line indicates the 25th percentile of reference values (10th and 25th coincide in fig. B). To illustrate the difference between males and females, both graphs (B) and (C) are combined in (D). The y-axis indicates the Kernel density of the distribution, which can be interpreted as a smooth approximation to the histogram, as the width of the histogram bars becomes very small. PPT=pressure pain thresholds; EPT=electrical pain thresholds.

3.3 CPM with electrical pain thresholds

Electrical pain thresholds rose by 1.5 mA from baseline (SD 1.6), corresponding to a 17.7% increase (SD 20.3). Multiple linear regression with electrical CPM as dependent variable showed that sex had a significant influence on the magnitude of CPM effect, with a regression coefficient of −0.078 (p=0.036). This indicates that female sex was associated with a 7.8% lower electrical CPM effect than male sex. The remaining independent variables (age, BMI, sleep quality, BDI, STAI and catastrophizing scale) did not influence CPM [F(7, 134)=1.32, p=0.24, r²=0.06]. CPM effects and percentiles of reference values for EPT are therefore displayed separately for males and females (Tables 3 and 4, Fig. 2B–D).

3.4 CPM with nociceptive withdrawal reflex

NWR thresholds rose by 1.2 mA (SD 2.5) from baseline, which corresponds to a 16.8% increase (SD 28.6) (Table 3). Multiple linear regression with reflex CPM as dependent variable showed none of the explanatory variables to have a significant influence on the CPM effect [F(7, 55)=0.29, p=0.95, r²=0.04]. The percentiles of reference values are presented in Table 4 for both absolute values and percent changes. Because of the many missing values, we omit the presentation in a figure.

4.1 CPM effect

The conditioning cold pressor test induced an average increase in thresholds for all three test modalities. However, over 10% of pain-free subjects had negative CPM effects for PPT and NWR, and one out of four subjects did not show an increase in electrical pain thresholds during the cold pressor test. The lack of positive CPM effect in pain-free subjects has been observed in previous studies [14], and the reasons remain unclear. It has to be stressed that the construct validity of the CPM paradigm in measuring endogenous pain modulation in humans is unknown. This uncertainty is due to lack of a reference standard to assess sensitivity, specificity and likelihood ratios of CPM paradigms [27]. It is possible that CPM has limited sensitivity in measuring endogenous pain modulation. In this case, the negative CPM effect in part of pain-free participants would represent false negative results in detecting functional endogenous pain modulation. An alternative explanation is that endogenous pain modulation is truly impaired in individuals who display a negative CPM effect, and this impairment does not translate into the presence of a pain syndrome. If this is true, an intriguing question would be whether these individuals are more likely to develop a pain syndrome in the future, compared to those who display a positive CPM effect. Large-scale longitudinal studies over a time course of several years would be required to answer this question.

Previous studies on CPM in large samples are sparse and have been mostly descriptive in nature. Locke et al. [14] examined 125 pain-free volunteers and found that 116 of them (92%) exhibited a CPM effect that was greater than the inherent measurement error of their test stimulus (i.e. <5.3%). Nine subjects (8%) did not show a positive CPM effect, which is roughly consistent with our findings. Differences in baseline pain thresholds between males and females were detected, but the CPM effect did not differ between genders. Skovbjerg et al. investigated CPM in a sample of 2,199 Danish people, whereby having pain was not an exclusion criterion [15]. Females had lower CPM than males, and no association of CPM with age or BMI was detected. The average CPM effect expressed in percent change from baseline was 32–39%, similar to the effect observed in the present study.

4.2 Potential applications

The percentiles that we determined can be used as reference to assess CPM in individual patients (Table 4). The choice of the 5th, 10th or 25th percentile as cut off for normal values depends on the particular clinical or scientific question. Only few patients would be identified as having dysfunctional pain modulation when the 5th percentile is chosen as cut-off. The number of cases categorized as having dysfunctional CPM increases progressively by choosing the 10th and 25th percentile. Therefore, the choice of the cut-off depends on the specific consequences of categorizing individual patients as having functional or dysfunctional CPM.

Noticeable in this regard is the finding that the majority of percentile reference values is negative, which is explained by the non-negligible proportion of subjects who had a CPM-effect less than zero. It seems illogical that decreases in pain thresholds after application of conditioning stimulus, even if modest, may be considered as normal. A “normal” decrease in pain threshold is against the common understanding of the CPM concept, whereby functional endogenous modulation should always be associated with increases in pain thresholds. However, it has to be considered that this assumption is derived from animal models. In humans, a simple test paradigm consisting of a test and a conditioning stimulus is unlikely to reflect the complex neuronal, cognitive and emotional processes involved in endogenous pain modulation. The main implication of this finding is that a slightly negative CPM effect does not necessarily reflect abnormal CPM. As mentioned above, a negative CPM effect in part of pain-free subjects has been observed in previous investigations. Therefore, we assume that if those investigations had had a sufficient sample size or had calculated reference values, they would have likely estimated negative percentiles.

4.3 Strengths and limitations

To our knowledge, the present study is the first one that specifically aimed at determining reference values of CPM. We applied a statistical approach that is designed to estimate reference values, in particular the percentiles of CPM, taking into account the potential influence of covariates [21]. A larger sample size may have allowed a more precise estimation of the percentiles. This is particularly true for the NWR, which could not be elicited in a large proposition of subjects. This phenomenon has been observed previously [28]. While the substantial number of missing values of NWR may have compromised the estimation of the reference values, we decided to present the analysis in the light of the lack of available data. Clearly, the estimates pertaining to the NWR have to be taken with caution. A final limitation is that the results apply to three of the several available CPM paradigms.