A Gender-Bias-Mitigated, Data-Driven Precision Medicine System to Assist in the Selection of Biological Treatments of Grade 3 and 4 Knee Osteoarthritis: Development and Preliminary Validation of precisionKNEE.

Objective To identify key variables predictive of patient responses to microfragmented adipose tissue (MFAT) treatment in knee osteoarthritis (KOA) and evaluate its potential to delay or mitigate the need for total knee replacement (TKR). Methods We utilised a dataset comprising 329 patients treated with MFAT for KOA, incorporating variables such as gender, age, BMI, arthritic aetiology, radiological grade, and Oxford Knee Scores (OKS) pre- and post-treatment. We employed random forest regressors for model training and testing, with gender bias mitigation and outlier detection to enhance prediction accuracy. Model performance was assessed through root mean squared error (RMSE) and mean absolute error (MAE), with further validation in a TKR-suitable patient subset. Results The model achieved a test RMSE of 6.72 and an MAE of 5.38, reflecting moderate predictive accuracy across the patient cohort. Stratification by gender revealed no statistically significant differences between actual and predicted OKS improvements (p-values: males = 0.93, females = 0.92). For the subset of patients suitable for TKR, the model presented an increased RMSE of 9.77 and MAE of 7.81, indicating reduced accuracy in this group. The decision tree analysis identified pre-operative OKS, radiological grade, and gender as significant predictors of post-treatment outcomes, with pre-operative OKS being the most critical determinant. Patients with lower pre-operative OKS showed varying responses based on radiological severity and gender, suggesting a nuanced interaction between these factors in determining treatment efficacy. Conclusion This study highlights the potential of MFAT as a non-surgical alternative for KOA treatment, emphasising the importance of personalised patient assessments. While promising, the predictive model warrants further refinement and validation with a larger, more diverse dataset to improve its utility in clinical decision-making for KOA management.