The anchor design of anchor-based method to determine the minimal clinically important difference: a systematic review

Abstract

Background Positive results for clinical outcomes should be not only statistically significant, but also clinically significant. The minimum clinically important difference (MCID) is used to define the minimum threshold of clinical significance. The anchor-based method is a classical method for ascertaining MCID. This study aimed to summarise the design of the anchors of the anchor-based method by reviewing the existing research and providing references and suggestions. Method This study was mainly based on literature research. We performed a systematic search using Web of Science, PubMed, CNKI, Wanfang, and VIP databases. Two reviewers independently screened titles and abstracts to identify relevant articles. Data were extracted from eligible articles using a predefined data collection form. Discrepancies were resolved by discussion and the involvement of a third reviewer. Result Three hundred and forty articles were retained for final analysis. For the design of anchors, Subjective anchors (99.12%) were the most common type of anchor used, mainly the Patient’s rating of change or patient satisfaction (66.47%) and related scale health status evaluation items or scores (39.41%). Almost half of the studies (48.53%) did not assess the correlation test between the anchor and the research indicator or scale. The cut-off values and grouping were usually based on the choice of the anchor types. In addition, due to the large number of included studies, this study selected the most calculated SF-36 (28 articles) for an in-depth analysis. The results showed that the overall design of the anchor and the cut-off value were the same as above. The statistical methods used were mostly traditional (mean change, ROC). The MCID thresholds of these studies had a wide range (SF-36 PCS: 2–17.4, SF-36 MCS: 1.46–10.28), and different anchors or statistical methods lead to different results. Conclusion It is of great importance to select several types of anchors and to use more reliable statistical methods to calculate the MCID. It is suggested that the order of selection of anchors should be: objective anchors > anchors with established MCID in subjective anchors (specific scale > generic scale) > ranked anchors in subjective anchors. The selection of internal anchors should be avoided, and anchors should be evaluated by a correlation test.

Determinants of Subjective Mental and Functional Health of Critical Illness Survivors: Comparing Pre-ICU and Post-ICU Status

“…As we did not have a suitable external anchor available, we determined the relevant difference of the subjective health items using a distribution-based approach ( 34 , 35 ). We defined a subjective health change of at least ± 1 sd as a relevant decline/improvement.…”

Comparison of Chronic Sialadenitis Symptom Burden in Patients With Sialolithiasis Versus Stenosis Before and After Sialendoscopy

“…This question was then used as an anchor to calculate a minimal clinically important difference. 11 The Pearson correlation coefficient between self-rated improvement and SPIT score change was found to be −0.35, suggesting the external anchor and outcome instrument are correlated enough to calculate an MCID using anchorbased methods rather than distribution based. 12 Patients were divided into 2 groups based on data distribution of SPIT scores versus anchor question: those that reported at least 75% improvement and those who reported less than 75% improvement after intervention.…”