Reply to determining MCID threshold for Knee Society Score to assess patient satisfaction in knee arthroplasty

Abstract

This reply addresses Wang et al.’s comments on our prior work and clarifies that the 34.5-point Knee Society Score threshold was derived from our previously validated, anchor-based approach. We agree that future research should refine predictive models by establishing consensus, standardised minimal clinically important difference cutoffs to enhance applicability and reliability in real-world clinical settings.

We thank Wang et al. for their perspective on our recently published work¹, including the discussion regarding the use of alternative minimal clinical important difference (MCID) thresholds² for the Knee Society Score (KSS) in assessing patient satisfaction. We welcome the dialogue and agree that selecting an appropriate MCID threshold is critical for evaluating clinically meaningful patient satisfaction after total knee arthroplasty (TKA).

Several prior studies have explored the MCID cutoffs for KSS^2,3,4, however, each has methodological limitations, including small sample sizes, non-standard calculation approaches, retrospective anchors susceptible to substantial response-shift bias, and collapsed or absent ‘no-change’ categories. Currently, no standardised MCID threshold exists for the KSS. In our study, we therefore adopted a threshold of 34.5 points for KSS knee score based on our previous work, which used a well-established, anchor-based linear regression method validated in Singaporean patients who underwent primary or revision TKA with follow-up at 6 months and 2 years^4,5. Furthermore, a previous study also suggested that that higher thresholds—approximately 39–40 points—may be useful in identifying patients who experience substantial clinical benefit after TKA³. In addition, the median of preoperative KSS score is 36 (interquartile range [23, 50]), with 8.3% of them had preoperative KSS score$>$65.5 that were mathematically unable to MCID. A comprehensive evaluation approach is warranted to provide thorough evaluation to those patients who may show ceiling effects when using a single metric.

In our view, the primary contribution of our paper is to provide the first multimodal, machine learning-based model to predict patient dissatisfaction across multiple scoring systems. We agree with Wang et al. that future studies are warranted to determine clinically meaningful MCID thresholds for multiple patient-reported outcome measures in large-scale, prospective cohorts. We anticipate that future work will extend and refine the models we present, incorporating richer clinical information and, ultimately, consensus-based, standardised MCID cutoffs for deployment in real-world clinical settings.