J Manipulative Physiol Ther 2001 (Oct); 24 (8): 501–504
Scrimshaw SV, Maher C
School of Physiotherapy, Faculty of Health Sciences, University of Sydney, Lidcombe, NSW 2141, Australia
OBJECTIVE: To compare the responsiveness of the McGill Pain Questionnaire with the Visual Analogue Scale (VAS).
DESIGN: A repeated measures 2-group design was used, with subjects divided into "improved" and "non-improved" groups. The external criterion to identify improved and non-improved patients was a 7-point global perceived effect scale.
SUBJECTS: Seventy-five patients with low back pain who had participated in a randomized controlled trial of postsurgical rehabilitation were included in the study.
INTERVENTIONS: All patients completed both a VAS and McGill pain scale to describe their pain over the last 24 hours and a separate VAS to describe their current pain.
MAIN OUTCOME MEASURES: Responsiveness was evaluated by using receiver-operating characteristic curves, with the analysis repeated for a range of cut-off points on the global perceived effect scale. Secondary analyses of responsiveness were provided by the t value for independent change scores and Spearman's rank correlation coefficient (rho).
RESULTS: The study confirmed the results of earlier studies in finding that the VAS was less responsive to clinical change when used to rate current pain in comparison with pain over the last 24 hours. The study found that the VAS was more responsive than the McGill Pain Questionnaire when both instruments were used to rate pain over the last 24 hours.
CONCLUSION: The results of this study suggest that the VAS may be a better tool than the McGill Pain Questionnaire for measuring pain in clinical trials and clinical practice.
From the Full-Text Article:
The principal finding of this study was that the MPQ was less responsive to clinical change than the VAS-24. This finding has been reported previously [4–7]; however, because the time reference for the MPQ-PRI was often not clearly stated, the source of the reduced responsiveness was unclear in those studies. In this study, we controlled for the time reference by having subjects complete both an MPQ and VAS to rate pain over the last 24 hours. Under this condition the MPQ was less responsive than the VAS.
The number of items in each subscale of the MPQ is as follows: sensory (10), affective (5), miscellaneous (4), and evaluative (1). Based on psychometric theory, it would be expected that the MPQ scales that are formed from a larger number of items would be more responsive than those formed from a smaller number of items. However, this was not the case in this study; for example, the 10-item sensory subscale was consistently less responsive than the 1-item evaluative subscale. In fact, the evaluative subscale was more responsive than the MPQ score based on all 20 items.
The responsiveness of the various subscales of the MPQ has only been examined in 3 previous studies. [4, 5, 7] These papers analyzed responsiveness by using a single-group before-after design, which is regarded as a weaker design than the 2-group design used in this study.  Bellamy et a [4, 5] reported contradictory results in 2 studies. One study of rheumatoid patients noted that the evaluative subscale had similar responsiveness to the MPQ, whereas the study of patients with osteoarthritis noted much lower responsiveness with the evaluative subscale. Bellamy et al [4, 5] did not discuss this discrepancy between the 2 studies. Jenkinson et a  provided separate results for each of 4 treatment groups, with the MPQ-PRI less responsive for 1 group and more responsive for 2 groups. In the fourth group, the MPQ total score and evaluative subscale had similar responsiveness. Because of the conflicting results noted in past research and the fact that our analysis of the responsiveness of the subscale of the MPQ was not preplanned, we would advise readers to view this aspect of our study with caution.
The source of the reduced responsiveness of the MPQ compared with the VAS may be the presumed factor structure of pain that is reflected in the MPQ. Although Melzack originally proposed separate sensory, affective, and evaluative pain dimensions, these dimensions were not empirically derived. Subsequent factor analytic studies of the MPQ have both challenged and confirmed the original 3-factor structure.  The studies that have challenged the original dimensions have identified up to 7 factors, with some factors identified as additional dimensions or combinations of the original 3 dimensions.  The multidimensional scaling studies of Clarke et al  provide evidence that there are many more dimensions of pain than the 3 proposed for the MPQ. Importantly, they conclude that descriptors within each of the 20 MPQ descriptor groupings are heterogenous (ie, they do not belong to the one dimension) and so cannot be ordered with respect to intensity. This result challenges the current scoring method for the MPQ.
The present study also confirmed the results of previous studies [2, 3] that have reported that average pain measures are more responsive than current pain measures. This result has been reported for simple pain measures such as the VAS,  0-10 NRS, [2, 3] and verbal category rating scales.  In these studies, the periods over which average pain was judged were longer than in the current study—1 week in the study by Bolton and Wilkenson2 and 2 weeks in the study by Jensen et al.  The present study demonstrated that an average pain measure is more responsive than a current pain measure, even when the time reference is shortened to the last 24 hours. This result is important because in acute conditions or conditions from which patients recover quickly, it may be unreasonable to ask patients to judge their average pain over a period as long as 2 weeks.
Our study showed that the VAS is more responsive to clinical change than the MPQ. We would advise readers to use the VAS in preference to the MPQ when measuring pain in clinical trials and clinical practice.