Are Ordinal Data Produced by Pain Threshold Assessments Reliable and Do they Demonstrate Responsiveness to Treatment?

Abstract

Abstracts 431 I Lund, J Kowalski, T Lundeberg Karolinska Institutet, Stockholm, Sweden The use of visual analogue scale data as numerical has beencriticised, as the size of the distance between the differentsteps is unknown. Its reliability and use at a group level havealso been discussed (Svensson, 2000). An alternative pain assessment is pain thresholdmeasuring. Pain thresholds have been shown to begenerally lower within but also outside the primary painlocation in patients with osteo-arthritic knee pain. Themeasured thresholds were lower compared to healthycontrols and were shown to increase following successfultreatment (Wilder-Smith et al , 2001). This may indicate thatpain threshold changes could be a measure of painalleviation and be evaluated anywhere, eg in the hand. Thedata created from threshold measurements could beregarded as comprising an ordered structure but withoutmetric properties. In this study we wanted to examine whether the ordinaldata produced by such a technique are reliable and ifresponsiveness to treatment would be demonstrated.To measure pain thresholds a PainMatcher® (CefarMatcher AB, Sweden) constant low current (15 mA)stimulator delivered electrical stimulation to the skin of athumb and index finger at random velocity and with afrequency of 10 Hz and raised intensity by increased pulsewidth. Electrical thresholds for sensory detection and pain weremeasured in 93 healthy volunteers (64 women, 29 men; 19-42 years) and in nine women with hip osteo-arthritits(48-70 years). Test-retest reliability was assessed repeatedlyfor two days. Responsiveness was measured before and after treatment with transcutaneous electrical stimulation in the healthy volunteers and before and six months afterhip replacement surgery in the patients with osteo-arthritis.Statistical hypothesis testing was done by Friedmananalysis of variance and calculated by STATISTICA 5.5.Statistical analyses of the rank-invariant methods bySvensson were evaluated by SYSRAN 1.0 for Matlab 6.0. The rank-invariant analyses showed no statistical evidencefor systematic disagreement in the test of reliability.Responsiveness to treatment following intervention wasdetected as systematically increased thresholds both inhealthy volunteers and patients. The measuring techniquewith PainMatcher could be a valuable tool for clinical painevaluation.