Minimal important changes in standard deviation units are highly variable and no universally applicable value can be determined

• Published in: Journal of clinical epidemiology 2022-05, Vol.145, p.92-100
• Main Authors: Tsujimoto, Yasushi, Fujii, Tomoko, Tsutsumi, Yusuke, Kataoka, Yuki, Tajika, Aran, Okada, Yohei, Carrasco-Labra, Alonso, Devji, Tahira, Wang, Yuting, Guyatt, Gordon H., Furukawa, Toshi A.
• Format: Article
• Language: English
• Subjects: anchor-based method; Credibility; Datasets; Decision theory; distribution-based method; effect size; Epidemiology; Estimates; Methods; Minimal important change; patient-reported outcome; Patients; Quality of life; Standard deviation
• ISSN: 0895-4356; 1878-5921
• DOI: 10.1016/j.jclinepi.2022.01.017

This study aims to describe the distribution of anchor-based minimal important change (MIC) estimates in standard deviation (SD) units and examine if the robustness of such estimates depends on the specific SD used or on the methodological credibility of the anchor-based estimates. We included all anchor-based MIC estimates from studies published in MEDLINE and relevant literature databases upto October 2018. Each MIC was converted to SD units using baseline, endpoint, and change from baseline SDs. We performed a descriptive analysis of MICs in SD units and checked how the distribution would change if MICs with low methodological credibility were excluded from the analysis. We included 1,009 MIC estimates from 182 studies. The medians and interquartile ranges of MICs in SD units were 0.43 (0.25 to 0.69), 0.42 (0.22 to 0.70), and 0.51 (0.28 to 0.78) for baseline, endpoint, and change SD units, respectively. Some MICs were extremely large or small. The distribution did not change significantly after excluding MICs estimated by less credible methods. The size of the universally applicable MIC in SD units could not be determined. Anchor-based MICs in SD units were widely distributed, with more than half in the range of 0.2 to 0.8.