Choice of external criteria in back pain research: Does it matter? Recommendations based on analysis of...

Understanding a change score is indispensable for interpretation of results from clinical studies. One way of determining the relevance of change scores is through the use of transition questions that assesses patients' retrospective perception of treatment effect. Unfortunately, results from studies using transition questions are difficult to compare since wording of questions and definitions of important improvements vary between studies. The objectives of this study were to determine the consequence of using different transition questions on pain and disability measures and make proposals for a standardised use of such questions. Two hundred and thirty-three patients with low back pain and/or leg pain were recruited. Participants were followed over an 8-week period and randomised to two groups receiving a 7- (TQ1) and 15-point (TQ2) transition question, respectively, in addition to a numeric rating scale evaluating the importance of the perceived change. Four external criteria were generated using both stringent and less stringent standards to dichotomise patients. Discrimination was determined using area under the receiver operating characteristic curve (ROC(auc)) and responsiveness using standardised response mean (SRM). Results demonstrated small variations in ROC(auc) across the external criteria for all outcome measures. 7% more patients were classified as improved in the group receiving the 15-point TQ compared to the 7-point TQ (stringent standard). SRMs were higher for the retrospective TQs in primary sector patients compared to the serial measures with no difference between TQ1 and TQ2. On the basis of our findings we have outlined a proposal for a standardised use of transition questions.