Percentile normative values of parameters of electrical pain and reflex thresholds

2.1 Participants

The study protocol was approved by the ethics committee of the University of Bern (KEK 147/04) and was in accordance with the Declaration of Helsinki. 300 pain-free subjects (150 males and 150 females, 18–80 years old) were studied. Exclusion criteria were: no knowledge of German language, any acute or chronic pain condition, intake of any pain medication for less that 24 h before the investigation, pregnancy (as ruled out by pregnancy test in women in reproductive age) and breast feeding. Participant selection was made so that 75 subjects were analyzed within each of the following age categories: 20–34, 35–49, 50–64 and 65–80 years. Furthermore, the male: female ratio within each age category had to be 1:1 (±1). All volunteers provided written informed consent before participating.

Gender, age, height, weight and body mass index (BMI) were recorded. Then the volunteers were asked to fill four questionnaires: Beck Depression Inventory (BDI), State-Trait-Anxiety-Inventory (STAI), Catastrophizing Scale of the Coping Strategies Questionnaire and Short-Form 36 (SF-36).

2.2 Pain tests

2.3 Statistical analysis

The estimates of percentile values of the distribution of the test responses have been obtained by computing quantile regressions for the each pain test. Quantile regression is a statistical method that can be used for the estimation of any percentiles of the response variable [15]. It can be regarded as an extension of multiple regression methods, where instead of estimating the mean of the dependent variables one estimates its quantile distribution. In particular, quantile regression is appropriate for the estimation of the percentile values on either tail of the distribution. These values can be regarded as the critical values of the pain tests, thereby reflecting normative values of the parameters analyzed. Importantly, the quantile regression method that we employed is particularly suited to non-normal distributions of the observations. It is a fully non-parametric method which does not rely on any distributional assumptions, and is therefore robust to non-normality or to any other irregularities in the distribution of the observations.

The role of covariates can be studied by including potential contributing factors as conditioning variables in the quantile regression estimations. In the present analysis, the covariates were gender, age, height, weight, body mass index, body side of testing, depression (Beck Depression Inventory), anxiety (State- Trait-Anxiety-Inventory), catastrophizing (Catastrophizing Scale of the Coping Strategies Questionnaire) and Short-Form 36. The variables according to which critical values have been stratified were chosen on the grounds both of their statistical significance and of their quantitative importance as conditioning factors in the previous study [11].

The quantile regressions have been carried out by using the command sqreg in Stata 12. This command generates estimates of quantile regressions with a bootstrapped variance-covariance matrix of the estimates. We computed the 5th, 10th, 25th, 75th, 90th and 95th percentiles of the distributions of the response variables (p⁵, p¹⁰, p²⁵, p⁷⁵, p⁹⁰ and p⁹⁵ respectively). For each response variable, all the percentiles have been estimated simultaneously in order to improve statistical efficiency. The number of bootstrapped replications in our calculations has been set equal to 10,000.

3.1 Results

Table 1 reports the estimated percentiles for single stimulus pain threshold, single stimulus reflex threshold, temporal summation pain threshold, temporal summation reflex threshold and RRF area.

The percentiles for single stimulus pain threshold have been stratified by age and the percentiles for single stimulus reflex threshold by body side (dominant vs. non-dominant). Gender-stratification was not done, since this parameter was not a significant covariate. This is consistent with previous investigations, showing that the influence of gender on pain sensitivity varies according to the type of stimulus applied [16,17].

Fig. 1 illustrates the critical values for one of the tests. In Fig. 2, the critical values for electrical pain threshold are stratified by four age classes for illustration purposes. The statistical analysis conducted on the four age classes was less precise that the analysis on two classes (Table 1); the latter specification (Table 1) may be therefore be preferred to the former one (Fig. 2) for clinical use.

Fig. 1

Illustration of critical values of widespread central hyper- and hyposensitivity. Pain threshold to single electrical stimulation, in patients with age <50 years, is shown. The vertical axis shows the kernel density, which can be interpreted as a continuous and smooth approximation to the histogram of the distribution as the width of the sub-intervals in the histogram becomes very small. summation reflex threshold); and (2) the subjects

Fig. 2

Illustration of critical values of widespread central hyper- and hyposensitivity stratified by four age classes for pain threshold after single electrical stimulation. Normative values of pain threshold to single electrical stimulation are stratified by four class ages. p^{5,10 and 25} are the percentile normative values for hypersensitivity, whereas p^{95,90 and 75} are the percentiles for hypersensitivity.

3.2 Discussion

We computed the critical normative values of electrical pain tests for widespread central hypersensitivity and hyposensitivity. The percentiles in Table 1 can be interpreted as the estimates of the statistical distribution of response variables in a pain-free population. For instance, when the single stimulus pain test is applied to individuals with age higher than or equal to 50 years, the pain threshold is estimated to be lower than 6.3 for 5% of the observations (p⁵).

The decision of which percentiles to use for the purposes of identifying potentially critical values should consider that the adoption of more extreme values (for instance, those corresponding to p⁵ and p⁹⁵) is more likely to lead to the correct identification of critical observations, but could leave out a number of observations that could also potentially be at risk. In other words, by selecting p⁵ and p⁹⁵ one is more likely to obtain a small number of false positives but a potentially large number of false negatives. By contrast, the adoption of critical values that correspond to more central percentiles of the distribution (for instance, p²⁵ and p⁷⁵) would correspond to a smaller number of false negatives, but a potentially larger number of false positives.

The decision on which percentiles to use will therefore depend on the particular clinical picture and the potential consequences of categorizing patients as positive or negative. For instance, if it is believed that the clinical risks of false negatives are greater than the risks of false positives, it would be prudent to adopt less extreme values for the critical percentiles, such as p²⁵ and p⁷⁵. Conversely, if one wants to minimize the risk of false positive, it would be preferable to adopt the extreme percentiles, i.e. p⁵ and p⁹⁵.

To date, identifying widespread sensory abnormalities in individual patients has mainly the aim of providing a pathophysiological model that can help patients and clinicians to better understand the mechanisms underlying pain and disability. Furthermore, some data suggest that patients with hyper- or hyposensitivity may be at risk of poor outcome, at least in the field of whiplash injury [2,18]. Identifying these patients may-be important to develop therapeutic strategies that prevent poor outcome. Finally, patients with widespread hypersensitivity may theoretically benefit from drugs that attenuate these processes, such as antidepressants, anticonvulsants or NMDA-antagonists; however, we are not aware of controlled studies on such a mechanism-based approach.

3.3 Conclusions

Estimates of critical normative values for each electrical pain test in a pain-free population were determined. These values can be used in clinical practice for the assessment of widespread central hyper- and hyposensitivity in individual patients and by future research on mechanism-based treatment of chronic pain conditions.