How scoring limits the usability of minimal important differences (MIDs) as responder definition (RD): an exemplary demonstration using EORTC QLQ-C30 subscales.

Abstract

The recommended method for establishing a meaningful threshold for individual changes in patient-reported outcome (PRO) scores over time uses an anchor-based method. The patients assess their perceived level of change and this is used to define a threshold on the PRO score which may be considered meaningful to the patient. In practice, such an anchor may not be available. In the absence of alternative information often the meaningful change threshold for assessing between-group differences, the minimally important difference, is used to define meaningful change at the individual level too. This paper will highlight the issues with this, especially where the underlying measurement scale is not continuous. Using the EORTC QLQ-C30 as an example, plausible score increments ("state changes") are calculated for each subscale highlighting why commonly used thresholds may be misleading, including leading to sensitivity analyses that are inadvertently testing the same underlying threshold. The minimal possible individual score change varies across subscales; 6.7 for Physical Functioning, 8.3 for Global Health Scale and Emotional Functioning, 11.1 for fatigue, 16.7 for role functioning, cognitive functioning, social functioning, nausea and vomiting, pain and 33.3 for single items. The determination of meaningful change for an individual patient requires input from the patients but being mindful of the underlying scale ensures that these thresholds are also guided by what is a plausible change for patients to achieve on the scale.