Assessment of CPM reliability: quantification of the within-subject reliability of 10 different protocols

Abstract

Abstract Background and aims Conditioned Pain Modulation (CPM) is a well-established phenomenon and several protocols have shown acceptable between-subject reliability [based on intraclass correlation coefficient (ICC) values] in pain-free controls. Recently, it was recommended that future CPM test-retest reliability studies should explicitly report CPM reliability based on CPM responders and non-responders (within-subject reliability) based on measurement error of the test stimulus. Identification of reliable CPM paradigms based on responders and non-responders may be a step towards using CPM as a mechanistic marker in diagnosis and individualized pain management regimes. The primary aim of this paper is to investigate the frequency of CPM responders/non-responders, and to quantify the agreements in the classification of responders/non-responders between 2 different days for 10 different CPM protocols. Methods Data from a previous study investigating reliability of CPM protocols in healthy subjects was used. In 26 healthy men, the test-stimuli used on both days were: Pain thresholds to electrical stimulation, heat stimulation, manual algometry, and computer-controlled cuff algometry as well as pain tolerance to cuff algometry. Two different conditioning stimuli (CS; cold water immersion and a computer-controlled tourniquet) were used in a randomized and counterbalanced order in both sessions. CPM responders were defined as a larger increase in the test stimulus response during the CS than the standard error of measurement (SEM) for the test-stimuli between repeated baseline tests without CS. Results Frequency of responders and non-responders showed large variations across protocols. Across the studied CPM protocols, a large proportion (from 11.5 to 73.1%) of subjects was classified as CPM non-responders when the test stimuli standard error of measurements (SEM) was considered as classifier. The combination of manual pressure algometry and cold water immersion induced a CPM effect in most participants on both days (n=16). However, agreement in the classification of CPM responders versus non-responders between days was only significant when assessed with computer-controlled pressure pain threshold as test-stimulus and tourniquet cuff as CS (κ=0.36 [95% CI, 0.04–0.68], p=0.037). Conclusions and implications Agreements in classification of CPM responders/non-responders using SEM as classifier between days were generally poor suggesting considerable intra-individual variation in CPM. The most reliable paradigm was computer-controlled pressure pain threshold as test-stimulus and tourniquet cuff as conditioning stimulus. However while this CPM protocol had the greatest degree of agreement of classification of CPM responders and non-responders across days, this protocol also failed to induce a CPM response in more than half of the sample. In contrast, the commonly used combination of manual pressure algometry and cold water immersion induced a CPM effect in most participants however it was inconsistent in doing so. Further exploration of the two paradigms and classification of responders and non-responders in a larger heterogeneous sample also including women would further inform the clinical usefulness of these CPM protocols. Future research in this area may be an important step towards using CPM as a mechanistic marker in diagnosis and in developing individualized pain management regimes.