Expressing treatment-associated changes

The Osteoarthritis Research Society International is attempting to establish a consensus on outcome measures, so as to facilitate comparability between different clinical studies in this area of research. There are no general recommendations on how changes in outcome measure (effect size) should be expressed. We therefore used data from a recently published study to express change as: a) the mean of the change from baseline divided by the individual baseline, b) the median of the change from baseline divided by the individual baseline, c) the mean of the change from baseline divided by the SD of baseline and d) the median of change from baseline divided by the SD of baseline. The results show that the correlations between different ways of expressing effect sizes were poor and the perceived relative magnitudes of various effects depended on how they were expressed. Organisations aiming at consensus ought to recommend the way in which change ought to be expressed. Until they do, authors should justify their choice of expression, particularly if it critically influences their conclusion, and/or present their data as fully as possible on a website.