Knee osteoarthritis is a debilitating musculoskeletal disorder that profoundly impacts an individual’s physical health, mental health, and quality of life. Excessive compressive loading in the knee is related to increased pain and accelerated disease progression. Gait retraining is a potential non-surgical treatment for knee osteoarthritis that can reduce compressive knee loading during walking. This dissertation describes the design, evaluation, and clinical translation of two gait modifications that can reduce knee loading during walking:​ one by altering kinematics and another by altering muscle coordination.

Altering the foot progression angle (FPA) is a kinematic gait modification that can shift loading from the medial to the lateral compartment of the knee for individuals with isolated medial compartment osteoarthritis. On average, FPA modifications reduce the peak knee adduction moment (KAM), which is related to the medio-lateral distribution of knee loading and the rate of medial knee OA progression. However, the biomechanical efficacy of FPA modifications is variable, and there is limited evidence of their clinical efficacy. To teach individuals to modify their FPA, we developed a real-time biofeedback system that measures kinematics from a motion capture system and provides participants with haptic feedback while they walk on a force-instrumented treadmill. We used this system to investigate whether selecting FPA modifications in a personalized manner could improve their biomechanical efficacy. Healthy individuals reduced their KAM peak by an average of 19% with a personalized FPA modification, which was nearly two times the reduction that was achieved when the entire group was assigned the same modification. Next, we investigated the importance of personalization in a cohort of 107 individuals with medial knee osteoarthritis. A personalized approach increased the proportion of individuals who reduced their KAM peak with an FPA modification. Importantly, personalizing FPA modifications avoids prescribing modifications that increase an individual’s KAM peak, while a non-personalized approach would prescribe KAM-increasing modifications to 10- 44% of individuals. Third, we conducted a randomized controlled trial to investigate whether personalized foot angle modifications reduce knee pain compared to a sham gait retraining control. The intervention group was trained to walk with the FPA modification that maximally reduced their KAM peak while the control group was trained to walk more consistently with their natural FPA. After six weekly gait retraining sessions, the intervention group reduced their KAM peak and knee pain by more than the control group, but there were no significant differences in self-reported function. Together, these studies demonstrate that FPA modifications are a promising treatment for medial knee osteoarthritis, but they need to be personalized to be most effective.

Selecting personalized FPA modifications requires an expensive gait laboratory, which is a barrier to clinical translation. To address this, we trained a neural network model to predict the KAM peak using the positions of anatomical landmarks that can be identified from a 2D video camera. This model predicted changes in the KAM peak from an FPA modification with 92% accuracy, suggesting that it could be used in tandem with pose-recognition algorithms to prescribe personalized gait modifications in a clinical setting using only a smartphone camera.

Another way to reduce knee loading during walking is to change muscle coordination. Using musculoskeletal simulations, we found that changing the relative activation of two ankle plantarflexor muscles – the gastrocnemius and soleus – can reduce knee contact force. However, it was unclear whether humans can learn a new coordination strategy during a dynamic task like walking. To test this, we developed a biofeedback system that displays simple metrics of muscle activity after each step. Healthy individuals learned the new “gastrocnemius avoidance” coordination pattern after 15 minutes of walking with biofeedback. By walking with this new coordination pattern, individuals reduced their knee contact force during late stance by 38% of bodyweight.

This dissertation demonstrates that gait modifications have the potential to become an effective tool in the treatment of knee osteoarthritis, but they should be prescribed in a personalized manner. More generally, we show that humans can quickly learn changes to their walking kinematics and muscle coordination patterns with simple biofeedback. In the future, musculoskeletal simulations can be used to predict kinematic and coordination patterns that may be more favorable for different tasks and objectives. These predictions can motivate the design of biofeedback to optimize movement for a variety of different rehabilitation, injury prevention, and human performance applications.