There are Considerable Inconsistencies Among Minimum Clinically Important Differences in TKA: A Systematic Review

BackgroundPatient-reported outcome measures (PROMs) are frequently used to assess the impact of total knee arthroplasty (TKA) on patients. However, mere statistical comparison of PROMs is not sufficient to assess the value of TKA to the patient, especially given the risk profile of arthroplasty. Evaluation of treatment effect sizes is important to support the use of an intervention; this is often quantified with the minimum clinically important difference (MCID). MCIDs are unique to specific PROMs, as they vary by calculation methodology and study population. Therefore, a systematic review of calculated MCID values, their respective ranges, and assessment of their applications is important to guide and encourage their use as a critical measure of effect size in TKA outcomes research.Questions/purposesIn this systematic review of MCID calculations and reporting in primary TKA, we asked: (1) What are the most frequently reported PROM MCIDs and their reported ranges in TKA? (2) What proportion of studies report distribution- versus anchor-based MCID values? (3) What are the most common methods by which these MCID values are derived for anchor-based values? (4) What are the most common derivation methods for distribution-based values? (5) How do the reported medians and corresponding interquartile ranges (IQR) compare between calculation methods for each PROM?MethodsFollowing Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, a systematic review was conducted using the PubMed, EMBASE, and MEDLINE databases from inception through March 2022 for TKA articles reporting an MCID value for any PROMs. Two independent reviewers screened articles for eligibility, including any article that calculated new MCID values for PROMs after primary TKA, and extracted these data for analysis. Overall, 576 articles were identified, 38 of which were included in the final analysis. These studies had a total of 710,128 patients with a median age of 67.7 years and median BMI of 30.9 kg/m(2). Women made up more than 50% of patients in most studies, and the median follow-up period was 17 months (range 0.25 to 72 months). The overall risk of bias was assessed as moderate using the Jadad criteria for one randomized controlled trial (3 of 5 ideal global score) and the modified Methodological Index for Non-randomized Studies criteria for comparative studies (mean 17.2 +/- 1.8) and noncomparative studies (mean 9.6 +/- 1.3). There were 49 unique PROMs for which 233 MCIDs were reported. Calculated values were classified as anchor-based, distribution-based, or not reported. MCID values for each PROM, MCID calculation method, number of patients, and study demographics were extracted from each study. Anchor-based and distribution-based MCIDs were compared for each unique PROM using a Wilcoxon rank sum test given non-normal distribution of values.ResultsThe WOMAC Function and Pain subscores were the most frequently reported MCID value, comprising 9% (22 of 233) and 9% (22 of 233), respectively. The composite Oxford Knee Score (OKS) was the next most frequently reported (9% [21 of 233]), followed by the WOMAC composite score (6% [13 of 233]). The median anchor-based values for WOMAC Function and Pain subscores were 23 (IQR 16 to 33) and 25 (IQR 14 to 31), while the median distribution-based values were 11 (IQR 10.8 to 11) and 22 (IQR 17 to 23), respectively. The median anchor-based MCID value for the OKS was 6 (IQR 4 to 7), while the distribution-based value was 7 (IQR 5 to 10). Thirty-nine percent (15 of 38) used an anchor-based method to calculate a new MCID, while 32% (12 of 38) used a distribution-based technique. Twenty-nine percent of studies (11 of 38) calculated MCID values using both methods. For studies reporting an anchor-based calculation method, a question assessing patient satisfaction, pain relief, or quality of life along a five-point Likert scale was the most commonly used anchor (40% [16 of 40]), followed by a receiver operating characteristic curve estimation (25% [10 of 40]). For studies using distribution-based calculations, all articles used a measure of study population variance in their derivation of the MCID, with the most common method reported as one-half the standard deviation of the difference between preoperative and postoperative PROM scores (45% [14 of 31]). Most reported median MCID values (15 of 19) did not differ by calculation method for each unique PROM (p > 0.05) apart from the WOMAC Function component score and the Knee Injury and Osteoarthritis Outcome Score Pain and Activities of Daily Living subscores.ConclusionDespite variability of MCIDs for each PROM, there is consistency in the methodology by which MCID values have been derived in published studies. Additionally, there is a consensus about MCID values regardless of calculation method across most of the PROMs we evaluated.