Title: Discriminative ability of two different external anchors: improvement appears better than importance. Authors and affiliation: Henrik H. Lauridsen, DC, MSc, PhD*, Claus Manniche, MD, DMSc‡, Werner Vach, PhD §, Niels Grunnet-Nilsson DC, MD, PhD*, Jan Hartvigsen, DC, PhD*† * Clinical Locomotion Science, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark. † Nordic Institute of Chiropractic and Clinical Biomechanics, part of Clinical Locomotion Science. ‡ Backcenter Funen, Ringe, part of Clinical Locomotion Science. § Department of Statistics, University of Southern Denmark, Odense. Abstract Background The minimal important change (MIC) has gained acceptance as an important questionnaire attribute to interpret change scores. It allows patients to be reliably classified as either importantly improved or unchanged in clinical trials, and easily interpretable statistics such as proportions and numbers needed to treat can be reported based on the MIC. Estimating the MIC using the anchor-based method is, however, not without difficulties. For instance issues relating to calculation methods, measurement error, rater perspective, population specificity and baseline dependence have been raised, and little attention has been paid to how improvement and importance of a treatment outcome should be interpreted. The purpose of this study is to explore the adequateness of two different external anchors using the anchor-based MIC distribution method and probability of improvement/importance curves. Methods Two hundred and twenty-four patients with chronic low back pain and/or leg pain were recruited from an out-patient hospital back pain clinic and followed over an 8-week period. Participants received the Danish Oswestry Disability Index (ODI), the numeric rating scale for pain (NRSpain) and two different external anchors. These were patient-rated and measured the dimensions of improvement and importance of the obtained improvement. All patients were dichotomised as having a positive treatment outcome or a neutral/unchanged treatment outcome using each of the external anchors. The anchor-based MIC distribution method was applied and logistic regression analyses generated probability curves for each anchor. Results The correlation between patient reported outcomes (PRO) and external anchors ranged from 0.53 to 0.73. The anchor-based MIC distribution method showed similar MIC ranges for pain and disability for both external anchors (ODI: [4.4-16.3] for improvement and [6.9-17.3] for importance and NRSpain: [2.0-3.3] and [2.0-3.9]). However, the probability curves generally showed smaller false positive/negative fractions using improvement compared to importance as an external anchor. Applying the MIC ranges to the probability curves revealed probability ranges of [0.24-0.55] (improvement) and [0.37-0.57] (importance) for the ODI and [0.53-0.64] (improvement) [0.55-0.62] (importance) for the NRSpain. Conclusion This study combines the anchor-based MIC distribution method and probability curves to evaluate the adequateness of two different external anchors when choosing specific MIC values. The MIC ranges for the included pain and disability PRO’s have relatively large proportions of misclassified patients using both anchors. However, the improvement criterion had better discriminative abilities compared to the importance criterion. Consequently, we recommend using improvement and not importance as an anchor to determine the MIC.