Purpose: Pressure-based auditory feedback was employed to facilitate a gait modification in the gait lab and at home in individuals with medial knee osteoarthritis (OA). We evaluated subjects’ responses immediately and three weeks post training, as well as their responses after a three-week washout from feedback. Our primary variable of interest was the knee adduction moment (KAM), a surrogate measure of compartmental knee loads. We hypothesized that after training subjects will walk with a lower KAM. A reference group was included to ensure any observed changes in KAM were due to gait retraining and not simply due to footwear and/or interaction with the feedback system.
Methods: The participant-blind, parallel group study with an allocation ratio of 2 feedback: 1 recording (reference) was IRB-approved and registered on clinicaltrials.gov (NCT02955225). All participants provided informed consent. Eligible participants were: ≥40 years old, BMI≤38 kg/m2, capable of walking without assistance, knee pain ≥6 months, KL grade II or III in the medial tibiofemoral joint, knee pain >30 mm on the Visual Analog Scale (VAS), no pain in adjacent joints >20 mm, no presence of systematic inflammatory arthritides, no history of joint replacement, no surgical arthroscopy or trauma within three months, and no intra-articular injection within six weeks for cortisone or four months for hyaluronan. An a priori power calculation, adjusted for potential dropouts, suggested a recruitment plan of 25 and 15 subjects for the feedback and recording group, respectively. Each subject received a pair of standardized shoes (Flex OA, Dr. Comfort) and a smartphone installed with a customized app. An active wireless, pressure-detecting insole (OpenGo, Moticon GmbH) replaced the original standardized shoe insole on the index side and an identical but inactive insole on the contralateral side. The active insole of the feedback group transmitted lateral pressure data in real-time to the smartphone app. Whenever lateral pressures at the time occurrence of peak KAMs (KAM1 & KAM2 in early and late stance, respectively) exceeded the preset thresholds, auditory cues were generated. The thresholds were customized within the range of 75-95% depending on individual tolerance before any gait retraining. Subjects were instructed to avoid the auditory cues by shifting their plantar pressure medially. The recording group used the active insole and smartphone app to record pressure during walking instead of receiving feedback. Both groups practiced their tasks during a training session as a part of the first lab visit. They continued practicing their group-specific tasks for 15 min/day and wore the standardized shoe for 6 hr/day for 6 day/week for three weeks at home until the Week 3 assessment. Between the Week 3 and 6 assessment, the feedback group underwent a washout of feedback, by joining the recording group to record pressure following the same schedule. Each of the three lab visits included a gait test and surveys of pain (VAS), the Lower Extremity Functional Score (LEFS), and the Knee Injury and Osteoarthritis Outcome Score (KOOS). Using traditional marker-based motion analysis (Qualysis & Bertcec), five walking trials were acquired at baseline, immediately, three weeks, and six weeks post-intervention. Baseline trials were collected in subjects’ own shoes and at their comfortable walking pace. Baseline and immediate assessments occurred during the first lab visit separated by the training session. For the feedback group, insole-generated auditory cues were provided during the immediate assessment but not the Week 3 and 6 assessments. KAM1 and KAM2 were analyzed using mixed linear models (SAS University Edition, SAS Institute Inc.). The KAM waveforms were analyzed using principal component (PC) analysis, and the PC scores were compared also using the described approach. Any significant findings were examined post hoc using pairwise comparisons adjusted for multiplicity using Tukey’s method.
Results: Between Oct 2016 and Dec 2019, 109 out of 654 individuals passed the phone screening and came to the clinic for informed consent and further screening. 45 met all eligibility requirements. 30 were randomized into the feedback group and 15 into the recording group. Six individuals decided not to participate after randomization and one was determined ineligible after further review, resulting in 38 subjects participating in the first lab visit. Baseline characteristics, gait variables, knee moments, and self-reported outcome measures were not different between study groups, except a higher knee flexion moment of the recording group (p = 0.014). 21 out of 26 subjects in the feedback group and all 12 subjects in the recording group completed all visits. Immediately after feedback training, KAM1 was 7.3% lower than baseline (-0.23 %Bw*Ht, p = 0.021), irrespective to the significantly reduced cadence (Fig. 1A). Lower scores of PC1 and PC2 (both p < 0.001) suggested a lower overall KAM magnitude during stance and a lower ratio between early and late stance of KAM (Fig. 1B & C), respectively. A significant difference in KAM1 was not retained at Week 3 (p = 0.169) and W6 (p = 0.323), as between-subject variability increased over time. To review the varied individual responses, the feedback group was divided into two subgroups (substantial & partial responders). We verified the threshold of 6% in reference to the literature and using a receiving operating characteristic curve generated by the study data. Changes in subgroup membership were examined (Table 1). For example, after three weeks of feedback training, 10 out of 16 subjects remained in the substantial responder subgroup, and 3 out of 10 subjects improved and changed their subgroup membership from the partial responder subgroup. Over the study period, the feedback group reported less pain on VAS (p < 0.001), and improved symptoms, physical function relating to daily life, sports and recreation, and knee-related quality of life (KOOS: p = 0.041, < 0.001, 0.029, 0.040, respectively). The recording group did not show a significant difference in KAM1 (p = 0.0350; Fig. 1A). One subject during all assessments and five subjects during W3 were considered as substantial responders. While the recording group reported less pain (VAS, p = 0.040), they did not report significant changes in other outcome measures.
Conclusions: Individuals with medial knee OA can use pressure-based auditory feedback to improve knee loads. Lower KAM1 was evident immediately post training in the lab. Most subjects maintained or improved from their immediate performance after continuing feedback training at home for three weeks. For some subjects who had less desirable outcome, future work is warranted to identify biomechanical, psychological, and behavioral factors that led to the varied responses. Pressure-based auditory feedback has potential to facilitate gait modification outside of a lab environment to enhance and reinforce a gait modification in order to improve knee loads in those with medial knee OA.
