Characteristics including age, fear of movement, and body mass index (BMI) could be considered by clinicians when deciding on optimal exercise-based treatment regimens for patients with knee osteoarthritis (OA), according to study findings published in The Journal of Rheumatology.
Precision medicine focuses on assigning optimal treatment patient by patient. Random Forest Informed Tree-based Learning (a novel machine learning algorithm) identifies patient characteristics that most strongly influence the determination of optimal treatment. Using this algorithm, investigators sought to identify which patients with knee OA would benefit most from treatment with home internet-based exercise training (IBET) or standard physical therapy (PT).
Investigators conducted an exploratory secondary analysis using data from the PATH-IN trial (ClinicalTrials.gov Identifier: NCT02312713) to discover features related to differential improvement among patients with knee OA treated with IBET or PT compared with control treatment (usual care/waitlist [WT]).
During the PATH-IN trial, patients were randomly assigned 2:2:1 to IBET, standard PT, or WT. Participants in the IBET group received online program access across the full 12-month intervention. Physical therapy group participants received as many as 8 individual in-person treatment sessions within 4 months. Neither IBET nor PT was available to participants in the WT group across the 12-month intervention period.
The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) total score at 12-month follow-up was the primary endpoint.
A total of 303 patients (mean age, 65.1 years; 73.2% women; 73.2% White) were included in the analysis.
Age, Brief Fear of Movement (BF) scores, and BMI at baseline were the characteristics that divided participants into 6 subgroups. Treatment assigned according to these models vs a singular best treatment assigned to all patients resulted in greater improvements in average WOMAC scores at 12-months (P =.01).
Overall, IBET was the optimal treatment for patients with low BF and lower age. Physical therapy was the optimal treatment for patients with high BF, older age, and BMI between 26 and 37 kg/m2.
This study was limited by the computationally burdensome nature of the algorithm used, lack of updated radiographs and physician assessments to confirm knee OA diagnoses, and reduced generalizability among patients in other geographic areas.
The study authors concluded, “[T]ailoring referrals to specific exercise-based interventions, based on the patient characteristics identified, could result in greater impacts on OA symptoms.”
Kim S, Kosorok MR, Arbeeva L, et al. Precision medicine-based machine learning analyses to explore optimal exercise therapies for individuals with knee osteoarthritis: Random Forest Informed Tree-based Learning. J Rheumatol. Published online August 1, 2023. doi:10.3899/jrheum.2022-1039