Precision Medicine Approach to Exercise-Based Interventions for Veterans with Knee Osteoarthritis

Background: Knee osteoarthritis (OA) is a leading cause of pain and disability, and Veterans have markedly greater rates than non-Veterans. Exercise is a core component of care for knee OA, associated with modest average improvements in pain and function. However, there is tremendous variability in the degree of

improvement individual patients experience following exercise-based interventions for knee OA. Further, there are different types of exercise-based interventions for knee OA, ranging from self-directed programs to individual physical therapy (PT), and it is likely that any given patient will not experience the same magnitude

of response to each of these different approaches. The overall objective of our research is to improve the effectiveness, efficiency and patient-centeredness of exercise-based services for Veterans with knee OA through a precision medicine approach that matches the intervention type with key patient characteristics.

Significance / Impact: There is currently no guidance or evidence regarding which patients benefit most from different exercise-based interventions for knee OA. Thus, there is no clarity regarding which patients should be directed to different types of exercise-based services. Innovation: This will be the first study to examine heterogeneity of treatment effects in the context of different

exercise-based interventions among Veterans with knee OA. Methods will involve novel, robust machine learning analyses. Specific Aims: 1. Develop a precision medicine treatment strategy that optimizes improvement in pain, stiffness and function, measured by the Western Ontario and McMaster Universities Osteoarthritis Index

(WOMAC), by tailoring exercise-based treatment to individual patients with knee OA. Aim 1.a. Use a causal inference-based machine learning (ML) approach to estimate patient-specific estimates of treatment outcomes (improvement in WOMAC scores) for each of the four treatments (Group PT, Individual PT, STEP-KOA, and

health education control). Aim 1.b. Apply uplift tree ML modeling to produce an interpretable tree-based model for optimal assignment of treatment for knee OA. Aim 2. Apply results of ML analyses, along with a robust partner-engaged development process, to design a randomized clinical trial (RCT) that will test the

effectiveness and cost-effectiveness of a precision medicine approach to delivering exercise-based interventions to Veterans with knee OA. Methodology: This project will involve analysis of two VA RCTs. One RCT compared Group vs. Individual PT and found comparable overall mean improvements in pain and function. The second RCT examined the STEP-

KOA intervention, which begins with home-based exercise and progresses to PT only if participants do not make clinically relevant improvements; STEP-KOA was also associated with mean improvements in pain and function. Individual PT, Group PT, and STEP-KOA are all evidence-based interventions for knee OA, varying

in the amount and type of support provided to patients and therefore the associated costs to the VA. We are preparing to scale Group PT and STEP-KOA in the VA. However, because Veterans in our RCTs varied substantially in their degree of improvement following these interventions, we believe these programs will

ultimately be of much higher value to the VA and Veterans if we are able to target their delivery using a precision medicine approach. In this project, we will apply robust ML approaches to uncover subgroups of patients who benefit most (and least) from Individual PT, Group PT and STEP-KOA. Next Steps / Implementation: Study results will directly inform an RCT that will test whether a precision

medicine approach is more effective than a “one size fits all” approach. Specifically, we plan for a 2-arm pragmatic trial that will compare Individual PT for all patients (current standard of care) with a precision medicine arm that assigns patients to one of three exercise-based interventions, based on key characteristics.