Predicting persistent pain after total knee arthroplasty using different machine learning algorithms

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: XGBoost model AUC = 0.67 for persistent pain prediction; similar AUC values reported for secondary outcomes (low OKS; failure to achieve MCID in OKS)

• Primary Outcome: Persistent pain 1 year after primary TKA, occurring in 850 of 11,755 patients (7.2% prevalence)

• Nnt Or Sensitivity: No sensitivity, specificity, positive predictive value, negative predictive value, or decision threshold reported. AUC 0.67 indicates poor discrimination (random chance = 0.50; acceptable clinical prediction typically ≥0.75). No NNT calculable from available data.

• Confidence Interval: Not reported in abstract

Clinical Application

The model as described is not feasible for direct clinical application given its poor discriminative performance (AUC 0.67) and absence of external validation. The identified risk factors are, however, clinically actionable for preoperative counselling and shared decision-making. Routine collection of 751 variables is impractical in most clinical settings. In Australia, TKA is one of the most commonly performed elective orthopaedic procedures, with approximately 60,000 procedures annually. The Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) captures revision rates but patient-reported outcomes including persistent pain are less systematically collected. RACGP and AOANJRR guidelines emphasise shared decision-making and appropriate patient selection for TKA. The identified risk factors (opioid use, BMI, neuropathic pain) are directly relevant to Australian pre-surgical optimisation pathways. PBS-listed opioids and neuropathic agents (pregabalin, duloxetine) are common in this population. The study's findings reinforce the importance of preoperative pain assessment and opioid stewardship in Australian TKA candidates, consistent with RACGP chronic pain guidelines. However, the model itself cannot be adopted without Australian-specific validation. Adults with knee osteoarthritis being considered for primary TKA, particularly those with preoperative risk factors for poor outcomes including high preoperative pain burden, younger age, valgus deformity, opioid use, neuropathic pain medication use, thyroid disease, and elevated BMI

Abstract

BACKGROUND AND PURPOSE: After total knee arthroplasty (TKA), 10-20% of patients remain unsatisfied. Well-performing clinical prediction models can provide individualized risk estimates and stratification in terms of poor outcomes, resulting in unnecessary surgeries being avoided and patients being counseled preoperatively. We aimed to create a precise, well-performing prediction model for clinical application using different machine learning algorithms to predict those patients who will have residual pain, a low total Oxford Knee Score (OKS) and the patient group who do not achieve minimally clinical important difference (MCID) in OKS 1 year after TKA. METHODS: We conducted a retrospective cohort study based on patients who had undergone primary TKA at our institution combined with 751 patient-related variables. The multivariable models used were based on the results of univariate analysis. We used the machine learning method Extreme Gradient Boosting (XGBoost). The discrimination capability of the models was measured with the area under the curve (AUC). RESULTS: 11,755 patients were included in this study. There were 850 (7.2%) patients who experienced persistent pain 1 year after TKA. The AUC was 0.67. For the secondary outcomes, the AUC values were similar. The most important variables in the model were lower preoperative OKS, younger age, valgus malalignment, lower preoperative pain OKS, use of mild opioid, neuropathic pain medicine and thyroxine, and higher body mass index. CONCLUSION: The prediction models achieved poor AUCs. It seems clear that the prediction of pain and functional outcome after TKA is difficult, even with a large patient cohort combined with 751 patient-related variables and sophisticated machine-learning algorithms.

References

1. 1.Rajamäki, A., Reito, A., Karsikas, M., Niemeläinen, M., & Eskelinen, A. (2026). Predicting persistent pain after total knee arthroplasty using different machine learning algorithms. Acta Orthopaedica. https://doi.org/10.2340/17453674.2026.45965