Development and Internal Validation of a Predictive Model for Clinically Meaningful AOS Improvement After Primary...

Ankle osteoarthritis is a condition that generates a level of disability comparable to that of heart failure, chronic kidney failure, and end-stage hip and knee arthritis. To date, no predictive model has been described to determine which patients will experience a more pronounced improvement in patient-reported disability following total ankle replacement.

In this prospective cohort study, we included all patients between 2003 and 2023 who after joint decision making with their surgeon elected to undergo a total ankle replacement. Patients undergoing primary ankle fusion, revision total ankle replacement, with a history of previous ankle infection or Charcot neuroarthropathy, and those who declined to participate in the registry were excluded. A total ankle arthroplasty was considered to have a good response if patients showed a difference greater or equal than 28 points between the pre-operative and the 1-2-year AOS score. A multivariable logistic regression using a stepwise method was performed, adding or removing variables, and analyzing demographic variables (age at the time of surgery, smoking status, diabetes, inflammatory pathology, and body mass index), prosthesis type (Infinity, InBone II, InVision, Agility, Hintegra, Mobility, Zimmer, and STAR), etiology, and the AOS at the time of enrollment. As variables were added or removed, the model's Akaike information criterion was compared, ultimately keeping the most parsimonious version. After model selection, overfitting was evaluated using a bootstrapping method with 1000 replications, and finally, the goodness of fit was assessed with the Hosmer-Lemeshow test.

A total of 631 total ankle replacements were included. The average age at the time of surgery was 64.96 ± 9.7 years. The most frequently used total ankle replacement was Hintegra (n = 159), followed by Infinity (n = 151), Zimmer (n = 127), and Agility (n = 78). The variables included in the model were the AOS at the time of enrollment (OR 1.06, P < .01, 95% CI 1.05-1.08) and instability etiology (OR 5.1, P < .01, 95% CI 1.69-15.26). The AUC was 0.76, and when compared using the bootstrapping method, no differences were found, indicating that the model did not show overfitting. Finally, the Hosmer-Lemeshow test showed that calibration was satisfactory.

Higher baseline disability and instability-related etiology were associated with increased likelihood of clinically meaningful improvement after TAR. This internally validated model may assist preoperative counseling and shared decision making. A baseline AOS score above 63 was associated with a high probability of clinically meaningful improvement.

Level II, prospective cohort study.