Effects of increasing walking cadence on gait biomechanics in adults with knee osteoarthritis

Abstract

Gait retraining is a strategy to manage altered loading patterns and pain characteristic of knee osteoarthritis. Lower walking cadence is associated with higher knee joint loading, vertical ground reaction forces, and risk for cartilage worsening. Therefore, we determined the acute effects of increasing walking cadence on measures of lower extremity loading and knee pain in knee osteoarthritis. Twenty-five participants with knee osteoarthritis (age = 62.5 ± 7.2; 76.0 % female) walked at fixed speed on an instrumented treadmill from which baseline cadence was measured. Five, randomized experimental cadence conditions (2 %, 4 %, 6 %, 8 %, or 10 % over baseline cadence) were completed. Real-time auditory and visual feedback on cadence was provided while kinematics and ground reaction forces were sampled. Linear mixed effects models evaluated the effect of cadence on knee adduction and flexion moment peaks and impulses, impact loading metrics (vertical ground reaction force impact peak, vertical average and instantaneous loading rates), and knee pain. Increasing cadence by 2–10 % did not significantly change knee adduction moment peaks or impulse. Peak knee flexion moment increased by 3–32 % and knee flexion moment impulse reduced by 2–9 % with increases in cadence, but these results were not significant (peak knee flexion moment, p = 0.070; knee flexion moment impulse, p = 0.085). Increasing cadence significantly increased the vertical impact peak (p < 0.001), and the vertical average (p = 0.010), and instantaneous (p = 0.007) loading rates. Small increases in cadence at a fixed gait speed does not significantly change surrogate measures of knee joint loading or pain, but does increase measures of impact loading.

1. Introduction

Knee osteoarthritis (OA) is a highly burdensome disease that affects more than 19 % of American adults aged 45 years and older (Lawrence et al., 2008). Chronic pain and disability are consequences of knee OA and contribute to low quality of life and morbidity. Yet, no curative treatments are available. Altered walking mechanics, particularly lower extremity joint loading, play a significant role in the etiology of knee OA (Felson, 2013; Vincent et al., 2012). The external knee adduction and flexion moments (KAM; KFM), surrogates of knee joint forces, are associated with the presence, severity, progression, and symptoms of OA (Chehab et al., 2014). The vertical ground reaction force (GRF), another indicator of the load applied to the lower extremity during gait, contributes significantly to KAM and KFM and is related to injury. Therefore, rehabilitation treatments that aim to optimize loading patterns may be a promising approach for OA management.

Among load-modifying treatments for knee OA, gait retraining has grown in popularity because it is conservative, noninvasive, and does not necessitate the long-term patient adherence that is required when using assistive devices or diet and exercise interventions. During gait retraining, patients are taught a new motor behavior to modify a gait parameter associated with pain and/or disease progression. Recent evidence suggests that cadence (i.e., number of steps per minute) may be a promising parameter to target in gait modification treatments for knee OA. While cadence modifications have been widely utilized as a mechanism to reduce loading and pain in healthy and injured runners (Allen et al., 2016; Bowersock et al., 2017a; Bramah et al., 2019; Heiderscheit et al., 2011; Lenhart et al., 2014; Miller et al., 2020; Willy et al., 2016), only recently has its application to knee OA been considered.

Compared to those without knee OA, individuals with knee OA walk with a lower cadence (Mills et al., 2013). Additionally, a study of 691 patients with medial tibiofemoral OA found that walking with a low cadence, controlling for gait speed, is associated with increased peak KAM, peak KFM, and KFM impulse (Hart et al., 2021). It is proposed that these alterations in loading also explain the increased risk for cartilage worsening observed among individuals who walk with a low compared to high cadence (Hart et al., 2020). Together these data further support the modification of preferred cadence to improve gait mechanics in knee OA.

There is a dearth of evidence regarding the effects of cadence modification on both loading and knee pain in knee OA. In a study of patients with patellofemoral OA, increasing preferred walking cadence by 10 % at a fixed gait speed resulted in immediate reductions in peak KFM and KFM impulse (Hart et al., 2023). The opposite occurred when preferred cadence was decreased by 10 %. While these results are promising, it is unclear whether (1) the results are generalizable to community-dwelling adults with knee OA defined through clinical rather than radiographic criteria, and (2) whether increasing cadence acutely affects knee pain. Radiographs are costly and inaccessible yet required to diagnose and identify those with patellofemoral OA who only represent a subset of individuals with knee OA. Moreover, there is discordance between radiographic and symptomatic knee OA: many individuals with radiographic knee OA do not experience pain, while those with knee pain may not have radiographic evidence of disease (Bedson and Croft, 2008; Finan et al., 2013). Therefore, studying individuals with clinically defined knee OA and confirmed knee pain during walking – a treatment target of gait retraining – can provide critical insights that radiographic studies may overlook.

Finally, a 10 % increase in cadence may not be achievable or sustainable for all individuals with knee OA, especially if their preferred cadence is already high or during periods of increased pain. Examining incremental increases in cadence can contribute to our understanding of a more achievable cadence modification “dose” that is applicable to real-world scenarios where patients may only be able to implement smaller changes. Understanding the amplitude of change in lower extremity loading for incremental increases in cadence (i.e., dose–response relation) is relevant for developing tailored gait retraining interventions (Favre et al., 2016) for individuals with varying levels of mobility and pain severity. Even modest reductions in knee joint moments observed in the lab can result in clinically meaningful improvements over time and numerous steps if the gait modification is adopted and retained. Understanding the impact of smaller cadence increases is therefore vital for developing realistic and effective gait retraining strategies that can benefit a broader range of patients with knee OA who experience pain with walking.

The objective of the current study was to investigate the acute effects of increasing walking cadence on measures of lower extremity loading in adults with knee OA. Our primary aim was to test the hypothesis that increasing walking cadence by 2–10 % at a fixed gait speed will decrease peak external KAM and peak external KFM. Our secondary aim was to test the hypothesis that increasing cadence would decrease other measures of impact loading that contribute to KAM and KFM, including the vertical GRF impact peak (VIP), vertical average loading rate (VALR), and vertical instantaneous loading rate (VILR). Finally, we determined the effect of increasing cadence on acute changes in knee pain.

2. Methods

2.1. Participants

Participants were recruited in the Boston metropolitan area through flyers, word-of-mouth, and social media advertisements. To be included in the study, participants had to be ≥45 years of age and report a history of knee pain for ≥3 months, and knee stiffness after periods of inactivity (e.g., sleeping, sitting) that lasts <30 min. These criteria are based on the National Institute for Health and Care Excellence (NICE) guidelines (Conaghan et al., 2008) which have 90 % specificity for ruling in knee OA and are commonly used in clinical trials for knee OA (Dziedzic et al., 2014; Hinman et al., 2017). In addition, participants had to report knee pain ≥3 out of 10 while walking (0 = no pain; 10 = worst pain imaginable). Exclusion criteria included the following: (1) walk with an assistive device; (2) history of knee replacement or knee surgery within the last year; (3) intra-articular knee injection within the last 3 months; (4) unable to walk on a treadmill for 25 min; (5) have a disease or condition that affects lower extremity function; and (6) skin allergy to adhesives.

An a priori sample size calculation was conducted with G*Power software. We used data from the running literature as studies in walking and knee OA were unavailable at the time. In runners, moderate effect sizes were reported for decreases in loading with increased cadence (Cohen’s D = 0.42–0.54). We estimated that 25 participants, with 6 measurements per participant, and an assumed correlation among repeated measures of 0.75, would yield > 80 % power to detect small effect sizes (≥0.15) with a two-tailed alpha of 0.05.

Ethical approval for the study was obtained from the Institutional Review Board at Northeastern University (# 20-04-10) and all participants provided written consent prior to participation.

2.2. Questionnaires

Participants completed questionnaires about demographics and clinical characteristics related to knee pain. Participants reported their comorbidities from a list of medical conditions, including cardiovascular disease, hypertension, diabetes, and rheumatological disease, and indicated any other chronic health conditions that were not listed. Knee pain and physical function were evaluated using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). The index (i.e., more symptomatic) knee was determined with a Visual Analog Scale (VAS). The VAS asked, “How much knee pain have you had in the past week?” with integer scores ranging from 0 (no pain) to 100 (worst pain imaginable). The knee with the higher VAS score was defined as the index knee. If VAS scores were equal, the index knee was defined with the question “Which knee causes you more difficulty when walking?”.

2.3. Experimental protocol

Participants completed a single-session gait analysis where they walked on an instrumented split-belt treadmill (Bertec Corp., Columbus, OH, USA) wearing standardized walking shoes (Gel-Excite 6, ASICS Corp., Kobe, Japan). The treadmill was set to a comfortable walking speed, selected by the participant during a 5-minute familiarization period. The speed selected during this period was fixed for the duration of the experimental protocol. Baseline cadence was first measured. Participants then completed 5 experimental cadence conditions: +2 %, +4 %, +6 %, +8 %, and +10 % over their preferred cadence (Fig. 1). Conditions lasted 2 min and were performed in a randomized order to minimize learning effects. After each condition, the treadmill was stopped while participants rested for 1 min and verbally reported their knee pain on an 11-point Numeric Rating Scale (NRS; 0 = no pain; 10 = worst pain imaginable).

Fig. 1.

Treadmill protocol for experimental cadence modification.

A custom-written code in LabVIEW (National Instruments Corporation, Austin, TX, USA) was used to monitor and provide the participant with visual cues on cadence in real-time using the GRF signal from the treadmill. Numbers portrayed in green indicated that participants were within ±1 step/minute of the target cadence. Numbers portrayed in red indicated that participants were outside of the ±1 step/minute of the target. Visual cues were supplemented with auditory cues from a metronome set to the target cadence. Feedback was provided 100 % of the time for all experimental conditions, with visual cues updating every 10 steps. Participants were instructed to walk to the beat of the metronome and to keep their cadence within ±1 step/minute of the target. The allowable variation of ±1 step/min was to (1) account for the natural variability in cadence as someone attempts to modify their natural gait, and (2) to provide a reasonable amount of feedback that promoted motor learning.

2.4. Data acquisition

Prior to walking, participants were equipped with 52 retroreflective markers affixed to their pelvis and lower extremities. After static and dynamic hip trials were collected for kinematic model building purposes, 22 anatomical markers were removed. The remaining tracking markers were located on the sternum (1), C7 vertebrae (1), each acromion process (2), the low back (1), and each of the 2nd and 5th meta-tarsal heads (2). In addition, 25 markers attached to rigid clusters were located on the pelvis (3), and the right and left thighs (8), shanks (8), and feet (6). For the duration of each walking condition, marker trajectories were sampled at 100 Hz with a 10-camera motion capture system (Qualisys AB, Göteborg, Sweden) after starting the auditory and visual feedback. The GRF data were simultaneously sampled at 1200 Hz using the two force plates embedded side-by-side in the treadmill.

Kinematic and GRF data were processed with Visual3D software (C-Motion, Inc., Germantown, MD, USA). Data were low-pass filtered using a bidirectional, 4th order Butterworth filter with a cutoff frequency of 12 Hz. Initial contact and terminal stance events were identified using a vertical GRF threshold of 20 N. External frontal and sagittal plane knee joint moments were computed using a routine inverse dynamics approach. Impulses were calculated by integrating the adduction and flexion components of the frontal and sagittal plane moment waveforms, respectively, throughout stance with respect to time. Knee joint moments were calculated in newton-meters (Nm) and normalized to mass times height (kg•m). The peak knee adduction angle was calculated during stance phase. The peak knee flexion angle was calculated during early stance, between initial contact and midstance. The VIP, VALR, and VILR were extracted from the vertical GRF during early stance phase based on previously described methods (Futrell et al., 2018; Hunt et al., 2010). In brief, the VIP was defined as the first local maxima of the vertical GRF waveform. The VALR and VILR were calculated as the average and peak slope between 20–50 % of the duration from heel-strike to VIP. All outcomes were calculated as the average of ten successive steps for each walking condition during any period where 1) the exact target cadence was met; and 2) the foot made clean contact with the force plate without crossing over to the other treadmill belt.

2.5. Statistical analysis

Data from the index knee were analyzed for the current study. Continuous variables were described as mean ± standard deviation (SD). Categorical data were described using frequencies (n) and percentages (%). Separate linear mixed effects models for repeated measures were used to evaluate the effect of cadence on each outcome (i.e., 1st and 2nd KAM peaks and impulses, KFM peak and impulse, impact loading variables, and NRS). This approach was used to account for missing data and biases caused by multiple observations per participant. Cadence condition was modeled as a fixed effect and participant was modeled as a random effect to capture individual-specific variability in responses. By modeling participants as a random effect, we account for individual differences in response patterns to increasing cadence, ensuring that the within-subject changes in loading are accurately reflected in the model. Bayesian Information Criterion (smallest value preferred) was used to select the optimal model covariance structure (either compound symmetry or unstructured) to best capture the correlation between repeated measures. Model assumptions were evaluated by examining the normality of the residual histograms and the plots of the residuals against the predicted values. When significant main effects were found, Tukey’s post-hoc test was used for multiple comparisons to evaluate where the outcome variables differed between the preferred and increased cadence conditions (i.e., 95 % CIs did not include 0). Using the same approach described above, changes in step length and stance time were analyzed as additional verification that cadence was increasing. Results were considered statistically significant at p < 0.05. Effect sizes (Cohen’s D) were calculated by dividing the estimate for the fixed effect by the √sum of variances of random effects to capture the correlated nature of the within-subjects data. Effect sizes were interpreted as very small (<0.20), small (0.20 to <0.50), medium (0.50 to <0.80), or large (≥0.80). All statistical analyses were performed with SAS software (version 9.3, SAS Institute, Cary, NC, USA).

3. Results

Demographics are summarized in Table 1. Twenty-five adults with clinically defined knee OA participated in the study. The participants had a mean ± SD age 62.5 ± 7.2 years, a mean BMI of 28.9 ± 6.3 kg/m², and 19 (76.0 %) were women. One participant could not reach the 8 % condition; another could not reach the 10 % condition. Missing data occurred for one additional participant during an 8 % condition due to equipment malfunction.

Table 1.

Summary of participant characteristics (n = 25).

Demographics and clinical characteristics
Age, years	62.5 ± 7.2
Sex, % female	76
Body mass, kg	81.4 ± 17.2
Height, m	1.7 ± 0.1
Body mass index, kg/m2	29.4 ± 6.7
Number of comorbidities	0.9 ± 1.2
History of knee injury, %	20
Pain characteristics
WOMAC Pain, 0–20	5.7 ± 2.6
WOMAC Stiffness, 0–8	3.2 ± 1.6
WOMAC Function, 0–68	19.6 ± 7.9
VAS Pain, 0–100	41 ± 18
Knee pain while walking, 0–10	4.4 ± 1.4
Duration of knee pain, years	14.5 ± 13.0
Presence of pain in non-index knee, %	80
Gait characteristics
Cadence, steps/minute	100.4 ± 11.1
Gait speed, m/s	0.9 ± 0.2

WOMAC, Western Ontario McMaster Universities Osteoarthritis Index

VAS, Visual Analog Scale

Overall, increasing cadence by 2–10 % at a fixed gait speed did not significantly change peak knee joint moments, angles, or impulses (Fig. 2). The adjusted means in Fig. 3 show that there was no significant main effect of cadence on 1st peak KAM (F = 2.31, p = 0.075), KAM impulse (F = 1.53, p = 0.188), or peak knee adduction angle (F = 0.96, p = 0.463). There was a significant main effect of cadence on 2nd peak KAM (F = 2.52, p = 0.033); however, post-hoc comparisons revealed no significant differences between the preferred cadence and any of the higher cadence conditions (Table 2). On average, increasing cadence by 2–10 % increased peak KFM by 3–32 %, reduced KFM impulse by 2–9 %, and increased peak knee flexion angle during early stance by 8–19 %, but these results were not statistically significant (peak KFM, F = 2.37, p = 0.070; KFM impulse, F = 1.99, p = 0.085; peak knee flexion angle, F = 2.51, p = 0.058; Table 2).

Fig. 2.

Group ensemble data for the (A) knee adduction moment, (B) knee flexion moment, and the (C) vertical ground reaction force across all cadence conditions.

Fig. 3.

Adjusted means [95 % CI] of knee joint loading and kinematics, impact loading, and spatiotemporal variables within participants, across cadence conditions.

Table 2.

Model-estimated changes in gait biomechanics between preferred and increased cadence.

	Conditions	Mean Difference [95 % CI]*	P	d
Knee Moments and Kinematics
KAM 1st peak, Nm/kg•m	100 % vs. 102 %	0.009 [−0.002, 0.021]	0.201	0.460
	100 % vs. 104 %	0.007 [−0.005, 0.018]	0.555	0.456
	100 % vs. 106 %	−0.001 [−0.012, 0.011]	0.999	0.448
	100 % vs. 108 %	0.007 [−0.005, 0.018]	0.583	0.456
	100 % vs. 110 %	0.002 [−0.009, 0.014]	0.991	0.452
KAM 2nd peak, Nm/kg•m	100 % vs. 102 %	−0.001 [−0.013, 0.011]	0.999	0.394
	100 % vs. 104 %	−0.003 [−0.015, 0.008]	0.956	0.388
	100 % vs. 106 %	−0.011 [−0.022, 0.001]	0.098	0.380
	100 % vs. 108 %	−0.009 [−0.021, 0.003]	0.278	0.384
	100 % vs. 110 %	−0.009 [−0.021, 0.003]	0.236	0.382
KAM impulse, Nm•s/kg•m	100 % vs. 102 %	0.001 [−0.001, 0.002]	0.863	0.104
	100 % vs. 104 %	−0.0001 [−0.002, 0.002]	0.999	0.099
	100 % vs. 106 %	−0.0004 [−0.002, 0.001]	0.956	0.098
	100 % vs. 108 %	−0.001 [−0.002, 0.001]	0.853	0.098
	100 % vs. 110 %	−0.001 [−0.002, 0.001]	0.882	0.097
Peak Knee Adduction Angle, deg	100 % vs. 102 %	0.324 [−0.733, 1.380]	0.930	0.041
	100 % vs. 104 %	0.028 [−0.734, 0.791]	1.000	0.031
	100 % vs. 106 %	0.451 [−0.582, 1.483]	0.755	0.046
	100 % vs. 108 %	0.592 [−0.335, 1.519]	0.385	0.051
	100 % vs. 110 %	0.446 [−0.520, 1.411]	0.711	0.046
KFM peak, Nm/kg•m	100 % vs. 102 %	0.004 [−0.018, 0.026]	0.992	0.291
	100 % vs. 104 %	0.017 [−0.014, 0.049]	0.547	0.348
	100 % vs. 106 %	0.015 [−0.014, 0.043]	0.608	0.337
	100 % vs. 108 %	0.025 [−0.004, 0.054]	0.127	0.381
	100 % vs. 110 %	0.019 [−0.007, 0.045]	0.256	0.356
KFM impulse, Nm•s/kg•m	100 % vs. 102 %	−0.002 [−0.009, 0.005]	0.966	0.267
	100 % vs. 104 %	−0.001 [−0.009, 0.006]	0.995	0.270
	100 % vs. 106 %	−0.007 [−0.014, 0.0004]	0.078	0.258
	100 % vs. 108 %	−0.005 [−0.012, 0.003]	0.408	0.262
	100 % vs. 110 %	−0.004 [−0.001, 0.004]	0.654	0.264
Peak Knee Flexion Angle, deg	100 % vs. 102 %	0.891 [−0.759, 2.541]	0.564	0.419
	100 % vs. 104 %	1.359 [−1.115, 3.833]	0.546	0.437
	100 % vs. 106 %	2.038 [−1.261, 5.337]	0.420	0.462
	100 % vs. 108 %	1.822 [−0.114, 3.758]	0.074	0.454
	100 % vs. 110 %	2.194 [−0.595, 4.983]	0.185	0.468
Impact Loading
VIP, BW	100 % vs. 102 %	0.008 [−0.010, 0.026]	0.800	1.015
	100 % vs. 104 %	0.020 [−0.002, −0.038]	0.021	1.020
	100 % vs. 106 %	0.023 [−0.004, −0.041]	0.006	1.022
	100 % vs. 108 %	0.025 [−0.008, −0.045]	0.001	1.024
	100 % vs. 110 %	0.025 [−0.006, −0.043]	0.003	1.023
VALR, BW/s	100 % vs. 102 %	0.192 [−0.655, 0.271]	0.791	0.649
	100 % vs. 104 %	0.420 [−0.884, 0.043]	0.091	0.672
	100 % vs. 106 %	0.438 [−1.102, 0.227]	0.352	0.673
	100 % vs. 108 %	0.805 [−1.522, −0.087]	0.022	0.711
	100 % vs. 110 %	0.660 [−1.393, 0.074]	0.095	0.696
VILR, BW/s	100 % vs. 102 %	0.650 [0.128, −1.428]	0.140	0.774
	100 % vs. 104 %	0.873 [−0.036, −1.710]	0.038	0.791
	100 % vs. 106 %	0.895 [−0.115, −1.675]	0.002	0.792
	100 % vs. 108 %	1.345 [−0.206, −2.484]	0.001	0.827
	100 % vs. 110 %	1.620 [−0.475, −2.766]	<0.001	0.848
Spatiotemporal
Step length, m	100 % vs. 102 %	−0.005 [−0.021, 0.010]	0.930	0.713
	100 % vs. 104 %	−0.017 [−0.032, −0.001]	0.023	0.705
	100 % vs. 106 %	−0.029 [−0.045, −0.014]	<0.001	0.696
	100 % vs. 108 %	−0.039 [−0.055, −0.024]	<0.001	0.688
	100 % vs. 110 %	−0.048 [−0.064, −0.032]	<0.001	0.683
Stance time, s
	100 % vs. 102 %	−0.017 [−0.028, 0.006]	<0.001	2.093
	100 % vs. 104 %	−0.031 [−0.043, 0.020]	<0.001	2.051
	100 % vs. 106 %	−0.048 [−0.060, 0.037]	<0.001	2.010
	100 % vs. 108 %	−0.062 [−0.073, 0.050]	<0.001	1.972
	100 % vs. 110 %	−0.068 [−0.080, 0.057]	<0.001	1.954

^*Mean differences are calculated as Increased Cadence – Preferred Cadence.

KAM, Knee adduction moment.

KFM, Knee flexion moment.

VIP, Vertical impact peak.

VALR, Vertical average loading rate.

VILR, Vertical instantaneous loading rate.

There was a significant main effect of cadence on all impact loading outcomes. Increasing cadence by 2–10 % resulted in linear increases in the VIP (F = 5.64, p = 0.0001), VALR (F = 3.89, p = 0.010), and VILR (F = 4.18, p = 0.007). Post-hoc comparisons (Table 2) showed significant differences in VIP between baseline and the 4 % condition (p = 0.021), 6 % condition (p = 0.006), 8 % condition (p = 0.001), and 10 % condition (p = 0.003). Though the model-adjusted means for VALR increased as cadence increased, significant differences were only observed between baseline and the 8 % condition (p = 0.022). Significant differences in VILR were identified between baseline and the 4 % condition (p = 0.038), 6 % condition (p = 0.002), 8 % condition (p = 0.001) and 10 % condition (p < 0.001). Reductions in step length occurred when cadence was increased by at least 4 %; stance time was significantly reduced across all cadence conditions (Table 2). Finally, there was no significant main effect of cadence on pain (F = 0.36, p = 0.876). Mean NRS at baseline was 2.3 ± 1.9 at baseline, and generally remained unchanged with increases in cadence (Fig. 3).

4. Discussion

The current study is the first to examine the immediate effects of small increases (<10 %) over preferred walking cadence on gait biomechanics and pain in adults with clinically defined knee OA. Our hypotheses were generally not supported. We found that modest increases in cadence at a fixed gait speed did not significantly alter surrogate measures of knee joint loading or knee pain but did increase measures of impact loading. These findings suggest that increasing cadence as a strategy to reduce loading in knee OA may require further research prior to clinical utility.

One other study has investigated the effects of a cadence modification on knee joint loading in knee OA (Hart et al., 2023). In contrast to our results, Hart et al found that a 10 % higher cadence is associated with significantly lower peak KFM and KFM impulse in adults with patellofemoral OA. Conversely, reducing cadence by 10 % was associated with significantly higher peak KFM and KFM impulse. The authors also reported comparable findings for the effects of cadence modification on 2nd peak KAM and KAM impulse. In contrast, we did not observe an inverse relation between cadence and loading. While our results did show a similar trend toward decreasing 2nd peak KAM as cadence increased, post-hoc tests revealed that these results were not statistically significant. Further, increasing cadence did not change the 1st peak KAM or KAM impulse. The 2nd peak KAM often corresponds to increases in loading as the foot pushes off in preparation for the next step, and is associated with greater OA severity (Mündermann et al., 2005) and varus thrust (Mahmoudian et al., 2016). Thus, further investigation into the trend towards reductions in 2nd peak KAM is warranted.

Knee adduction angle accounts for more than half of the variance in 1st peak KAM during walking (Schmitz and Noehren, 2014). However, we found no significant changes in peak knee adduction angles with increasing cadence, which may help explain why small increases in cadence did not substantially affect frontal plane knee loading. Interestingly, we observed a trend toward small increases in peak KFM and knee flexion angle during early stance, though these results were not statistically significant. While these findings contrast previous research in running, where increasing cadence reduces sagittal plane knee loading and kinematics (Bowersock et al., 2017b; Heiderscheit et al., 2011), a recent study in patients following anterior cruciate ligament reconstruction similarly found minimal effect of increasing cadence on both sagittal plane knee loading and kinematics (Garcia et al., 2024).

There are several possible explanations for the differing results between the study by Hart et al and the current one. We anticipated that increasing cadence at a fixed gait speed would decrease impact loading because vertical displacement of the center of mass reduces as step length decreases (Derrick et al., 1998). In contrast, we observed increases in impact loading with increased cadence. One possible explanation is that the auditory feedback from the metronome, along with the verbal instruction to “walk to the beat of the metronome” inadvertently encouraged participants to punctuate their steps. This would also explain the observed increases in impact loading, as well as the trend toward increasing peak KFM, despite significant reductions in step length. In a study of runners, metronome-based cues to increase cadence unexpectedly increased tibial acceleration while verbal cues to reduce step length had the opposite effect (Anderson et al., 2023). These findings suggest that the effects of gait modifications can vary depending on cueing methods and intensity, likely due to differences in how participants interpret and apply the cues. Similarly, Ulrich et al. found that using an augmented reality system to modify step length during overground walking increased cadence and reduced peak KAM and KFM in some, but not all, participants with medial knee OA (Ulrich et al., 2023). These studies highlight the need to consider the specific feedback method when designing gait retraining interventions, as different types of cues can influence loading patterns in distinct ways. Our results indicate that metronome feedback may inadvertently affect impact loading, highlighting the need for further research comparing different methods of feedback delivery and their potential confounding effects to optimize gait modification and retraining strategies for knee OA.

The short practice time for each cadence condition may have also influenced participant biomechanics. This study was designed to investigate the immediate effects of increasing cadence as opposed to the long-term or learned effects of increasing cadence. The duration of each condition was selected based on prior literature suggesting that it takes less than 20 strides to collect stable estimates of walking cadence (Hollman et al., 2010). However, motor learning principles may suggest that increased practice time may increase automaticity (Fitts and Posner, 1967), thus optimizing task performance. Participants were experienced treadmill users walking at a self-selected speed and were generally able to achieve the targeted increase in cadence. However, it was beyond the scope of this study to determine the effects of practice structure on changes in loading. As a result, it cannot be ruled out that the effects of increasing cadence on impact loading may have been different had an alternative practice structure been utilized.

Furthermore, we recruited adults with a clinical definition of knee OA based on NICE guidelines as opposed to patients from a sports medicine clinic with radiographically confirmed patellofemoral OA. Because interindividual differences in social-cognitive factors, pain, cognition, and age are all known to influence motor performance and skill acquisition (Leech et al., 2022; Wulf and Lewthwaite, 2016), incongruous findings between study populations is not completely unexpected. We consider using a representative sample of community-dwelling adults a strength of the study, as it increases generalizability and avoids costs associated with radiographs. However, our results also suggest that using a clinical definition of knee OA or knee pain with walking, independent of other biomechanical criteria, may not be specific enough for identifying the individuals who may benefit most from a cadence modification.

Increasing preferred cadence did not alter knee pain while walking in our sample. An effective gait modification would ideally reduce both knee joint loading and pain. Although the current study was not designed to assess the long-term effects of increased cadence on knee pain, the short-term results suggest that, despite increased impact loading, increasing cadence did not immediately exacerbate knee pain. While these findings indicate that short-term cadence modifications may be tolerated, further research is needed to determine whether the unchanged knee pain could also reflect the lack of change in knee joint loading and kinematics, the short duration at which cadence was increased, or both.

A strength of the study includes the real-time monitoring and validation of changes in cadence during the experimental conditions. Previous studies in running and walking typically report using visual confirmation by the researcher to ensure that the target cadence is achieved. Although this method is simple, it is subject to error. Using the GRF signal to monitor cadence in real-time instead provided a more rigorous yet efficient method to increase fidelity. Nonetheless, findings from this study should be considered in light of the following limitations. Symptomatic knee OA can manifest as different phenotypes (Deveza et al., 2019), and this heterogeneity could have contributed to the significant variability observed in the results. A further limitation was that we did not conduct any further screening to exclude those with high preferred cadence. Additional screening for potential “responders” is an efficient method that should be utilized in future studies that facilitates the development of more targeted interventions (Charlton et al., 2023). Finally, the experimental protocol was completed in standardized shoes on a treadmill. This could have impacted comfort level and gait mechanics, further limiting the generalizability of the results to overground or free-living walking.

5. Conclusion

The study findings indicate that small increases in cadence at a fixed gait speed does not significantly change surrogate measures of knee joint loading or knee pain but does increase measures of impact loading. Continuation of research into the effects of increasing preferred walking cadence in those with knee OA will facilitate efforts toward developing clinically accessible interventions for disease management.