Discordance Abounds in Minimum Clinically Important Differences in THA: A Systematic Review

Abstract

Background: The minimum clinically important difference (MCID) is intended to detect a change in a patient-reported outcome measure (PROM) large enough for a patient to appreciate. Their growing use in orthopaedic research stems from the necessity to identify a metric, other than the p value, to better assess the effect size of an outcome. Yet, given that MCIDs are population-specific and that there are multiple calculation methods, there is concern about inconsistencies. Given the increasing use of MCIDs in total hip arthroplasty (THA) research, a systematic review of calculated MCID values and their respective ranges, as well as an assessment of their applications, is important to guide and encourage their use as a critical measure of effect size in THA outcomes research.

Questions/purposes: We systematically reviewed MCID calculations and reporting in current THA research to answer the following: (1) What are the most-reported PROM MCIDs in THA, and what is their range of values? (2) What proportion of studies report anchor-based versus distribution-based MCID values? (3) What are the most common methods by which anchor-based MCID values are derived? (4) What are the most common derivation methods for distribution-based MCID values? (5) How do the reported medians and corresponding ranges compare between calculation methods for each PROM?

Methods: The EMBASE, MEDLINE, and PubMed databases were systematically reviewed from inception through March 2022 for THA studies reporting an MCID value for any PROMs. Two independent authors reviewed articles for inclusion. All articles calculating new PROM MCID scores after primary THA were included for data extraction and analysis. MCID values for each PROM, MCID calculation method, number of patients, and study demographics were extracted from each article. In total, 30 articles were included. There were 45 unique PROMs for which 242 MCIDs were reported. These studies had a total of 1,000,874 patients with a median age of 64 years and median BMI of 28.7 kg/m 2 . Women made up 55% of patients in the total study population, and the median follow-up period was 12 months (range 0 to 77 months). The overall risk of bias was assessed as moderate using the modified Methodological Index for Nonrandomized Studies criteria for comparative studies (the mean score for comparative papers in this review was 18 of 24, with higher scores representing better study quality) and noncomparative studies (for these, the mean score was 10 of a possible 16 points, with higher scores representing higher study quality). 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 the non-normal distribution of values.

Results: The Oxford Hip Score (OHS) and the Hip Injury and Osteoarthritis Score (HOOS) Pain and Quality of Life subscore MCIDs were the most frequently reported, comprising 12% (29 of 242), 8% (20 of 242), and 8% (20 of 242), respectively. The EuroQol VAS (EQ-VAS) was the next-most frequently reported (7% [17 of 242]) followed by the EuroQol 5D (EQ-5D) (7% [16 of 242]). The median anchor-based value for the OHS was 9 (IQR 8 to 11), while the median distribution-based value was 6 (IQR 5 to 6). The median anchor-based MCID values for HOOS Pain and Quality of Life were 33 (IQR 28 to 35) and 25 (14 to 27), respectively; the median distribution-based values were 10 (IQR 9 to 10) and 13 (IQR 10 to 14), respectively. Thirty percent (nine of 30) of studies used an anchor-based method to calculate a new MCID, while 40% (12 of 30) used a distribution-based technique. Thirty percent of studies (nine of 30) calculated MCID values using both methods. For studies reporting an anchor-based calculation method, a question assessing pain relief, satisfaction, or quality of life on a five-point Likert scale was the most commonly used anchor (30% [eight of 27]), followed by a receiver operating characteristic curve estimation (22% [six of 27]). For studies using distribution-based calculations, the most common method was one-half the standard deviation of the difference between preoperative and postoperative PROM scores (46% [12 of 26]). Most reported median MCID values (nine of 14) did not differ by calculation method for each unique PROM (p > 0.05). The OHS, HOOS JR, and HOOS Function, Symptoms, and Activities of Daily Living subscores all varied by calculation method, because each anchor-based value was larger than its respective distribution-based value.

Conclusion: We found that MCIDs do not vary very much by calculation method across most outcome measurement tools. Additionally, there are consistencies in MCID calculation methods, because most authors used an anchor question with a Likert scale for the anchor-based approach or used one-half the standard deviation of preoperative and postoperative PROM score differences for the distribution-based approach. For some of the most frequently reported MCIDs, however, anchor-based values tend to be larger than distribution-based values for their respective PROMs.

Clinical relevance: We recommend using a 9-point increase as the MCID for the OHS, consistent with the median reported anchor-based value derived from several high-quality studies with large patient groups that used anchor-based approaches for MCID calculations, which we believe are most appropriate for most applications in clinical research. Likewise, we recommend using the anchor-based 33-point and 25-point MCIDs for the HOOS Pain and Quality of Life subscores, respectively. We encourage using anchor-based MCID values of WOMAC Pain, Function, and Stiffness subscores, which were 29, 26, and 30, respectively.