Meaningful within‐person change estimates do not identify responders to treatment

Intervention effects are routinely evaluated by assessing the statistical significance of mean differences between groups—for example, a treatment group may have significantly better patient-reported outcomes than the control group. However, a significant improvement in the treatment group may not be substantial enough to be important. The minimally important change (MIC) concept was proposed to estimate the minimum amount of change that is important or meaningful to patients. 1 Similarly, meaningful within-patient change (MWPC) is “a range of score changes on a clinical outcome assessment that reflects a clinically meaningful change”.2 Lansdall et al.2 used care partner retrospective reports of “somewhat worse” as the basis of estimating the MWPC for the Clinical Dementia Rating-Sum of Boxes (CDR-SB) in a study of 422 adults with Alzheimer's disease. Landsdale, Cummings, and Andrews3 and Cummings4 suggested that MWPC estimates cannot be used to evaluate between-group differences. These authors argued that MWPC thresholds should be used to identify individual patients who have changed by a meaningful amount.3, 4 MIC and MWPC are estimates based on averages of individuals’ perceptions. But these group-level averages do not indicate if individuals have changed significantly.5 In other words, MIC and MWPC do not indicate whether an individual is a responder to treatment. Assessing whether a person has changed requires an evaluation of individual statistical significance, such as the minimal detectable change (MDC), which represents the “smallest amount of change in a measurement that is statistically greater than measurement error”.6 At a significance threshold of p = 0.05, the MDC = 1.96 2 $\sqrt 2 $ SEM, where the SEM (standard error of measurement) = standard deviation at baseline (SDbaseline) 1 − r e l i a b i l i t y $\sqrt {1 - reliability} $ 7or 1.96 times the standard error of the difference: ( S D b a s e l i n e 2 + S D f o l l o w u p 2 ) ( 1 − r e l i a b i l i t y ) $\sqrt {(SD_{baseline\ }^2 + SD_{followup}^2)( {1 - reliability} )} $ . The MDC estimates for the data in the current study,2 based on p < 0.05 (2.77 * SEM), range from 1.30 to 1.77 CDR-SB points (ES = 1.24). The MDC threshold estimate can be used to assess whether an individual has experienced a statistically significant change in CDR-SB over time. MWPC estimates are averages of individual perceptions and do not represent all individuals. The MWPC reported by Landsdale et al ranged from 1.07 (ES = 1.02) to 2.25 (ES = 1.57). MDC estimates can be larger than MWPC or MIC estimates because they account for measurement error. To ensure both meaningfulness and statistical significance are accounted for when categorizing individuals as changed or unchanged, we recommend reporting, in addition to MWPC estimates, the statistical significance of individual change (e.g., MDC) and the individual's retrospective perceptions of their change. The authors have nothing to report. The authors declare no conflicts of interest. The authors have no applicable funding to report. The authors have no disclosures. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.