Substantial Inconsistency and Variability Exists Among Minimum Clinically Important Differences for Shoulder Arthroplasty Outcomes: A Systematic Review.

Abstract

As the value of patient-reported outcomes becomes increasingly recognized, minimum clinically important difference (MCID) thresholds have seen greater use in shoulder arthroplasty. However, MCIDs are unique to certain populations, and variation in the modes of calculation in this field may be of concern. With the growing utilization of MCIDs within the field and value-based care models, a detailed appraisal of the appropriateness of MCID use in the literature is necessary and has not been systematically reviewed. We performed a systematic review of MCID quantification in existing studies on shoulder arthroplasty to answer the following questions: (1) What is the range of values reported for the MCID in commonly used shoulder arthroplasty patient-reported outcome measures (PROMs)? (2) What percentage of studies use previously existing MCIDs versus calculating a new MCID? (3) What techniques for calculating the MCID were used in studies where a new MCID was calculated? The Embase, PubMed, and Ovid/MEDLINE databases were queried from December 2008 through December 2020 for total shoulder arthroplasty and reverse total shoulder arthroplasty articles reporting an MCID value for various PROMs. Two reviewers (DAK, MAM) independently screened articles for eligibility, specifically identifying articles that reported MCID values for PROMs after shoulder arthroplasty, and extracted data for analysis. Each study was classified into two categories: those referencing a previously defined MCID and those using a newly calculated MCID. Methods for determining the MCID for each study and the variability of reported MCIDs for each PROM were recorded. The number of patients, age, gender, BMI, length of follow-up, surgical indications, and surgical type were extracted for each article. Forty-three articles (16,408 patients) with a mean (range) follow-up of 20 months (0.75 to 68) met the inclusion criteria. The median (range) BMI of patients was 29.3 kg/m2 (28.0 to 32.2 kg/m2), and the median (range) age was 68 years (53 to 84). There were 17 unique PROMs with MCID values. Of the 112 MCIDs reported, the most common PROMs with MCIDs were the American Shoulder and Elbow Surgeons (ASES) (23% [26 of 112]), the Simple Shoulder Test (SST) (17% [19 of 112]), and the Constant (15% [17 of 112]). The ranges of MCID values for each PROM varied widely (ASES: 6.3 to 29.5; SST: 1.4 to 4.0; Constant: -0.3 to 12.8). Fifty-six percent (24 of 43) of studies used previously established MCIDs, with 46% (11 of 24) citing one study. Forty-four percent (19 of 43) of studies established new MCIDs, and the most common technique was anchor-based (37% [7 of 19]), followed by distribution (21% [4 of 19]). There is substantial inconsistency and variability in the quantification and reporting of MCID values in shoulder arthroplasty studies. Many shoulder arthroplasty studies apply previously published MCID values with variable ranges of follow-up rather than calculating population-specific thresholds. The use of previously calculated MCIDs may be acceptable in specific situations; however, investigators should select an anchor-based MCID calculated from a patient population as similar as possible to their own. This practice is preferable to the use of distribution-approach MCID methods. Alternatively, authors may consider using substantial clinical benefit or patient-acceptable symptom state to assess outcomes after shoulder arthroplasty. Although MCIDs may provide a useful effect-size based alternative to the traditional p value, care must be taken to use an MCID that is appropriate for the particular patient population being studied.