Total hip arthroplasty (THA) in the elderly population has been linked to osteoporosis-related problems. In osteoporotic individuals, changing bone density is a risk factor for revision surgery. The purpose of the study was to create and test machine learning (ML) models to predict revision surgery in individuals with osteoporosis following primary non-cemented THA.

Investigators looked back at a series of 350 patients with osteoporosis (T-score less than or equal to 2.5) who had primary non-cemented THA at a tertiary referral facility. All patients were followed for at least two years (range: 2.1 to 5.6). To estimate the likelihood of revision surgery, four ML algorithms were created and tested using discrimination, calibration, and decision curve analysis.

At a mean follow-up of 3.7 years following primary non-cemented THA in osteoporotic patients, the overall incidence of revision surgery was 5.2%. Revision THA was performed in nine patients (50%) due to periprosthetic fracture, aseptic loosening/subsidence in five patients (28%), and periprosthetic joint infection in two patients (11%), and dislocation in two individuals (11%). Female sex, BMI (>35 kg/m2), age (>70 years), American Society of Anesthesiology score (3), and T-score were the best predictors of revision surgery in patients following original non-cemented THA. All four ML models performed well in discrimination (AUC range: 0.78 to 0.81), calibration, and decision curve analysis.

The ML models described in the study were highly accurate in predicting revision surgery in osteoporotic patients following primary non-cemented THA. The reported ML models have the potential to be employed by orthopedic surgeons for preoperative patient counseling and optimization in osteoporotic patients to enhance the results of primary non-cemented THA.

Reference:journals.lww.com/jaaos/Abstract/2022/05150/Artificial_Neural_Networks_Can_Predict_Early.7.aspx