Comparison of five definitions of meaningful change on patient-reported outcomes in four SWOG clinical trials

Purpose The minimally important difference (MID) represents how much change on a patient-reported outcome (PRO) is considered meaningful for defining a patient as improving or worsening. MIDs can be based on a statistical distribution or anchored to another PRO. This study used four SWOG clinical trials with treatments of known toxicities and positive quality of life (QOL) results as a silver standard to compare different MID definitions. Methods Two breast cancer trials compared true acupuncture (S1200) to sham acupuncture and waitlist control or duloxetine (S1202) to placebo to improve pain. S9346 compared continuous versus intermittent androgen deprivation therapy in metastatic prostate cancer. S1404 compared pembrolizumab with physician’s choice of interferon or ipilimumab in melanoma. Four MID distribution definitions (30% change, 50% change, half a standard deviation (0.5SD), reliable change index (RCI)) and one MID prospective anchor-based definition were compared. Chi-square statistics for intervention effect compared different definitions of MID. Results In S1200, the 0.5SD and RCI performed the best with the anchor-based method also performing well. In S1202, the anchor method, 0.5SD and 50% change performed well. For S9346, the anchor method performed well in physical function comparisons. For S1404, RCI and 30% change performed well for worsening QOL. Across all comparisons, 0.5SD had the highest mean chi-square values for worsening symptoms and RCI had the highest mean chi-square values for improvement. Conclusion Results suggest that some distribution-based MID methods may be sensitive to change for symptom and QOL PROs. Anchor-based methods may be the most sensitive for function PROs.