Knee joint stiffness during walking in knee osteoarthritis




To investigate the construct validity of walking knee stiffness as a measure to differentiate between individuals with and without knee osteoarthritis (OA) and the construct validity of walking knee stiffness as related to self-reported knee stiffness. The contributors to walking stiffness and its relationship with loading rate and adduction moment are also investigated.


Thirty-seven individuals with knee OA and 11 asymptomatic controls participated. Knee stiffness was calculated during walking as the change in knee flexion-extension moment divided by the change in knee flexion angle. Forward-stepwise regression models and Pearson's correlation coefficients were used to evaluate the relationships between variables.


Knee stiffness in walking was significantly greater in the OA group (mean ± SD 10.1 ± 4.4 Nm/°/kg × 100) compared with the controls (mean ± SD 5.6 ± 1.5 Nm/°/kg × 100) (P < 0.001). Knee excursion range explained 39% of the variance in walking knee stiffness (B = −0.736, P < 0.001) and knee extensor moment a further 7% (B = 6.974, P = 0.045). In the OA group, walking knee stiffness was not associated with self-reported stiffness (r = 0.029; P = 0.863). For the OA group, greater self-reported stiffness was associated with lower peak knee adduction moment (B = −0.354, P < 0.001).


The construct validity of walking knee stiffness is supported. The poor correlation between walking stiffness and self-reported stiffness suggests the 2 measures evaluate different aspects of knee stiffness. Since a measure of walking stiffness is likely to provide valuable information, future research evaluating its clinical significance is merited.


Knee osteoarthritis (OA) is a common disease of the lower extremity, with radiologic signs of OA evident in 30–40% of people age >65 years (1). Knee OA is generally viewed as a disorder of cartilage and subchondral bone that results in a clinical syndrome of symptoms evolving from pathophysiologic changes within the joint (2). Knee stiffness is commonly reported by individuals experiencing knee OA, and as such is one of 6 criteria used in the clinical diagnosis of knee OA (3). Knee stiffness has important clinical implications for patients with knee OA. Increasing self-reported knee stiffness is associated with a significantly greater risk of developing incident osteophytes (4) and for demonstrating progressive osteophyte growth over time (5). Furthermore, self-efficacy for physical tasks in knee OA is related to the sensation of knee stiffness (6), and stiffness also displays a modest association with physiologic predictors of the risk of falls in older adults (7). Therefore, knee stiffness is an important symptom associated with knee OA and warrants evaluation.

Knee OA-related stiffness is commonly quantified with the Western Ontario and McMasters Universities Osteoarthritis (WOMAC) Index (8). Of the 24 questions included in this self-report tool, only 2 relate to stiffness, which is defined as restriction or slowness in the ease with which the joint can be moved. Although widely used in knee OA studies, test–retest reliability of the WOMAC stiffness subscale is low (reliability coefficients 0.48–0.61) (8). This low reliability may be related to the fact that stiffness is difficult to both define and quantify with knee OA patients who demonstrate substantial variability in their interpretation of what constitutes stiffness (3). Furthermore, research has shown that the WOMAC subscales of pain, stiffness, and physical function are influenced by factors other than the parameters they purport to assess (9), including presence of fatigue, depression, and low back pain. Subjective self-reported knee stiffness could therefore be quite different from an objective mechanical measure of stiffness. Quantification of knee stiffness during walking (referred to as walking knee stiffness throughout this study) may provide additional insight into knee stiffness in individuals with OA. Although such a measure may also be influenced by factors such as fatigue, it would be performance-based and objective.

From a biomechanic perspective, walking knee joint stiffness in the sagittal plane can be estimated directly using 3-dimensional gait analysis. Joint stiffness during the stance phase of gait is defined as the change in joint moment divided by the change in joint angle (10). Theoretically, greater knee stiffness results from reduced knee excursion and/or higher knee moments. Importantly, reduced knee excursion is one of the most frequently observed characteristics of OA gait (11, 12), which suggests it will influence walking knee joint stiffness. The influence of sagittal plane knee moments on knee stiffness in walking is also unknown. Therefore, the relative influence of knee excursion and knee moments upon walking knee stiffness in OA requires investigation.

Owing to the influence on knee stiffness, it has been postulated that reduced knee excursion and higher knee moments can result in increased loading rates and a greater risk of developing knee OA (10). However, the relationship between walking knee joint stiffness and knee joint loading rates is unknown. Furthermore, given the link between self-reported stiffness and OA progression (5), it is possible that knee joint stiffness is associated with the knee adduction moment, an established biomechanic marker of OA progression (13).

The main aims of this study were to investigate the construct validity of walking knee stiffness as a measure that can differentiate between individuals with and without knee OA, and to investigate the construct validity of walking knee stiffness as being related to the recognized measure of self-reported knee stiffness in people with knee OA. Secondary aims were to investigate in patients with knee OA the extent to which knee flexion excursion and knee extensor moment contribute to knee joint stiffness in walking, and to examine the relationship between stiffness (self-reported and walking stiffness) and markers of joint loading (rate of loading and peak knee adduction moment). It was hypothesized that individuals with knee OA would demonstrate greater walking knee joint stiffness than the asymptomatic controls, that self-reported knee stiffness would be correlated with walking joint stiffness (r ≥ 0.43 at 80% statistical power), that greater walking knee joint stiffness would be associated with decreased knee excursion, and that stiffness (self-reported and walking) would be positively associated with markers of loading.



Thirty-seven patients with knee OA and 11 asymptomatic controls were recruited from the community in Melbourne, Australia, via newspaper advertisements and local clubs. Participants with knee OA fulfilled clinical and radiographic criteria as described by the American College of Rheumatology (formerly the American Rheumatism Association; age >50 years, knee pain, and osteophytes on radiographs) (14). All participants with knee OA demonstrated medial tibiofemoral osteophytes (although concomitant lateral tibiofemoral OA or patellofemoral OA were not excluded) and had experienced knee pain, which averaged >3 on an 11-point numerical pain rating scale on most days of the previous month. Exclusion criteria included a history of hip or knee joint replacement, knee surgery or injection in the previous 6 months, current use of a gait aid and any condition affecting gait, or the ability to complete testing. Control participants were age >50 years and reported no history of knee pain, injury or pathology. Due to ethical constraints, control participants did not undergo radiographic evaluation to exclude signs of OA. The study was approved by the University of Melbourne Human Research Ethics Committee. All participants provided written informed consent.

Symptom and disease severity assessment.

Symptomatic and radiographic severity of knee OA were evaluated in the OA cohort. Symptoms were evaluated using the WOMAC Index (8) with regard to pain (range 0–20, where higher scores indicate greater pain), stiffness (range 0–8, where higher scores indicate greater stiffness), and physical function (range 0–68, where higher scores indicate poorer function). Radiographic severity of tibiofemoral OA was assessed from an anteroposterior weight-bearing film using the Kellgren/Lawrence scale, where 0 = normal, 1 = possible osteophytes, 2 = minimal osteophytes and possible joint space narrowing, 3 = moderate osteophytes, some narrowing and possible sclerosis, and 4 = large osteophytes, definite narrowing, and severe sclerosis (15).

Gait analysis.

Knee OA and control participants performed shod walking trials at 1 meter/second (±10%) and 1.2 meters/second (±10%), respectively, with walking speed monitored using the forward velocity of the pelvis. Force plate data were collected at 1080 Hz using 2 AMTI force plates (Advanced Mechanical Technology, Watertown, MA). Synchronized 3-dimensional kinematic data were collected at 120 Hz using a 6-camera VICON 612 motion analysis system (Vicon, Oxford, UK). Reflective markers were placed on the pelvis and lower extremities to define the segments of the pelvis, thigh, lower leg, and foot, as well as the joint centers of the hip, knee, and ankle. Each participant performed 5 successful walking trials (where the required walking speed was attained and contact was made with the force plate), and mean data were used for analyses. Joint angles were calculated using a joint coordinate system approach, and moments were calculated using inverse dynamics techniques (version 1.9, Vicon Plug-In-Gait, Oxford, UK). The most symptomatic knee was utilized for data analyses for OA participants, and a randomly selected side was analyzed for control participants.

In running and hopping, dynamic knee joint stiffness is typically defined as the change in sagittal plane joint angle in response to the applied joint moment from initial ground contact to peak knee joint flexion (16). In contrast with running and hopping, the stance phase of walking is typically associated with an initial increase in the peak knee flexion moment, followed by a knee extensor moment (Figure 1). Therefore, we calculated walking knee stiffness as the change in sagittal plane joint angle in response to the applied joint moment over the period from peak flexion moment to peak weight-acceptance phase knee joint flexion angle (or peak extension moment, whichever occurred first).

Figure 1.

Sample knee moment (in Newtons × meters); knee angle time history (positive moment indicates internal extensor moment).

Knee joint flexion moment and dynamic walking stiffness were normalized for body mass (Nm/kg and Nm/°/kg, respectively). The knee flexion angle immediately prior to ground contact (initial knee angle), peak knee flexion angle during the weight-acceptance phase of stance, knee excursion (difference between peak knee flexion and initial knee angle), and peak knee extensor moment were also determined. Test–retest reliability of walking knee stiffness was determined in a population of 11 adults with medial knee OA tested 7 days apart (intraclass correlation coefficient [3, 5] 0.95) (17).

The rate of loading of ground reaction force and the peak knee adduction moment were used as markers of knee joint loading. Rate of loading was normalized for body weight (N/second/body weight), and the peak knee adduction moment was normalized for body weight and height (Nm/body weight × height %).

Statistical analyses.

Analyses were performed using the Statistical Package for the Social Sciences (version 15, Norusis/SPSS, Chicago, IL) using an alpha level of 0.05. Data were checked for normality prior to analyses. Demographic characteristics of the OA and control groups were compared at baseline via independent t-tests and chi-square tests. Since walking speed was different between the 2 groups, data were examined to determine whether walking speed was linearly related to the dependent variables measured for each group. There was no linear relationship between walking speed and walking knee joint stiffness for the OA group (r2 < 0.001, P = 0.872), or for the control group (r2 = 0.186, P = 0.185). The participants' ages also differed between the groups, but there was no linear relationship between age and knee joint stiffness in walking for the OA group (r2 = 0.045, P = 0.209) or for the control group (r2 < 0.001, P = 0.786). Therefore, neither walking speed nor age were included as covariates, and an independent t-test was used to compare dependent variables between the groups. A forward stepwise regression model was used to determine the contribution of knee excursion range and peak knee extension moment to the walking knee joint stiffness in the OA group. Pearson's correlation coefficient (r) was used to determine the degree of correlation between self-reported knee stiffness and walking knee stiffness in the OA participants. Finally, forward stepwise regression models were used to determine whether stiffness could predict markers of joint loading in the OA group. Separate analyses were performed for the dependent variables of rate of loading and peak knee adduction moment using the independent variables of walking speed, age, disease severity, self-reported stiffness, and walking knee stiffness.


Participant characteristics are presented in Table 1. The OA group was significantly older (P = 0.012) and heavier (P = 0.015) than the control group, and therefore demonstrated a significantly higher body mass index (P = 0.003). There were similar proportions of women (62–64%) within the OA and control groups. Within the OA cohort, a spread of radiographic disease severity was evident, but the majority of individuals demonstrated moderate (27%, grade 3) or severe (41%, grade 4) disease. The OA group walked at a slower mean ± SD walking speed (1.01 ± 0.04 meters/second) than the control group (1.23 ± 0.05 meters/second) (P < 0.001) during the gait analysis.

Table 1. Participant characteristics*
 Knee OA group (n = 37)Control group (n = 11)Mean difference (95% CI)P
  • *

    Values are mean ± SD unless otherwise indicated. OA = osteoarthritis; 95% CI = 95% confidence interval; NA = not assessed.

  • Measured by the WOMAC Index, higher scores indicate worse symptoms (pain range 0–20, stiffness range 0–8, and function range 0–68).

  • Assessed by Kellgren/Lawrence disease severity system; higher scores indicate more severe radiographic change.

Age, years65.6 ± 9.257.3 ± 9.58.3 (1.9, 14.7)0.012
Height, meters1.63 ± 0.091.65 ± 0.070.02 (−0.07, 0.04)0.551
Body mass, kg78.2 ± 11.168.6 ± 10.69.6 (2.0, 17.2)0.015
Body mass index, kg/m229.5 ± 4.125.3 ± 3.14.2 (1.6, 7.0)0.003
Sex, no. (%)   0.61
 Male14 (38)4 (36)  
 Female23 (62)7 (64)  
Self-reported symptom severity NA  
 Pain9 ± 3   
 Stiffness4 ± 2   
 Physical function30 ± 11   
Disease severity, no. (%) NA  
 Grade 13 (8)   
 Grade 29 (24)   
 Grade 310 (27)   
 Grade 415 (41)   

Between-group comparisons revealed that walking knee joint stiffness was significantly greater in the OA group (mean ± SD 10.1 ± 4.4 Nm/°/kg × 100) compared with the control group (mean ± SD 5.6 ± 1.5 Nm/°/kg × 100; P < 0.001). The 95% confidence interval (95% CI) for the difference between means was 1.8, 7.3 Nm/°/kg × 100. The mean ± SD knee excursion range during the weight-acceptance phase of gait was 9.4° ± 4.2° in the OA group and 15.5° ± 3.4° in the control group (95% CI −8.8°, 3.3°). The OA group appeared to contact the ground with a greater mean ± SD initial knee flexion (9.7° ± 5.2°) than the control group (3.7° ± 5.0°) (95% CI 2.5°, 9.5°). Similar mean ± SD peak knee flexion was observed for the OA group (19.1° ± 5.4°) and for the control group (19.2° ± 5.0°) (95% CI −3.7°, 3.6°). The mean ± SD peak knee extensor moment was 0.385 ± 0.175 Nm/kg and 0.524 ± 0.160 Nm/kg in the OA and control groups, respectively (95% CI −0.257, 0.198 Nm/kg).

In the OA cohort (n = 37), the forward stepwise multiple regression revealed that the knee excursion range and the peak knee extensor moment remained in the final model as significant predictors of walking knee stiffness, explaining 46% of the variance in this parameter (P < 0.001). The knee excursion range accounted for 39% (adjusted r2 = 0.369, B = −0.736 [95% CI −1.02, −0.45]; P < 0.001), and the peak knee extensor moment accounted for a further 7% (adjusted r2 = 0.425, B = 6.974 [95% CI 0.27, 13.68]; P = 0.045). The nature of this relationship was such that a smaller knee excursion range and greater knee extensor moment were associated with greater walking knee stiffness.

In the OA group, walking knee joint stiffness was not associated with the self-reported stiffness measured by the WOMAC Index (r = 0.029 [95% CI −0.32, 0.39]; P = 0.863). The relationship between walking and self-reported stiffness is illustrated in Figure 2. In the OA group, the mean ± SD rate of loading was 16.6 ± 4.1 N/body weight/second, while the mean ± SD peak knee adduction moment was 3.80 ± 0.93 Nm/body weight × height %. Results of regression modeling revealed none of the independent variables to be significant predictors of rate of loading. Only self-reported stiffness remained in the final model as a significant predictor of the peak knee adduction moment, explaining 31% of its variance (B = −0.354 [95% CI −0.53, −0.18]; P < 0.001). The nature of this relationship was such that greater self-reported stiffness was associated with a lower peak knee adduction moment.

Figure 2.

Correlation of walking knee stiffness (in Newtons × meters/°/kilograms × 100) against self-reported stiffness (Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC], range 0–12, where higher scores indicate greater stiffness), showing 95% confidence interval bands.


Knee OA-related self-reported stiffness is a substantial problem in those with knee OA, influencing self-efficacy for physical tasks (6) and being associated with a risk of falls in older adults (7). Utilizing 3-dimensional gait analysis, the current study has described a functional measure of knee joint stiffness during walking. As hypothesized, this study demonstrated that walking knee stiffness is significantly greater in people with knee OA compared with healthy controls. Contrary to our hypothesis, self-reported knee stiffness was not associated with walking knee stiffness. Consistent with the study hypothesis, reduced knee excursion was found to be a major contributor to the increased stiffness observed in OA. There was no association between walking knee stiffness and rate of loading; however, self-reported stiffness was associated with the peak knee adduction moment. Interestingly, the nature of this relationship was contrary to that hypothesized, with greater self-reported stiffness being associated with a lower peak knee adductor moment.

Across all study participants, non-normalized values for walking knee stiffness ranged from 1.8 to 19.4 Nm/°, with the mean value for control participants being ∼4 Nm/° and for OA participants ∼8 Nm/°. Published data for dynamic torsional knee stiffness in healthy participants are in the region of 6–11 Nm/° for hopping (16), and 7–24 Nm/° for running (18, 19). Values for walking have not been found in the published literature; however, it would be anticipated that they would be less than those observed for hopping and running. Knee stiffness values obtained for walking in the current study are therefore at the lower end of the range observed for more dynamic forms of locomotion. The relatively large range in walking knee stiffness values is likely reflective of the inclusion of both healthy and OA participants. The range for healthy study participants is 1.8–5.7 Nm/°, indicating all values are below those reported for running. In contrast, the range for OA participants (2.3–19.4 Nm/°) extends into the range reported for running, with 20 of the 37 OA participants having values above 7 Nm/°.

The hypothesis that individuals with knee OA would demonstrate greater walking knee stiffness than that of asymptomatic controls was supported by the study results, with OA participants exhibiting on average twice the stiffness of the controls. This finding supports the construct validity of the measure of walking knee stiffness. While the lower sample size of the control group is acknowledged, the 95% CIs for the 2 groups indicate no overlap of stiffness values. The 95% CI for the difference between the means does highlight, however, that there is uncertainty regarding the absolute magnitude of the difference between the population means. Although we report the magnitude of walking knee stiffness to be approximately double that of matched controls, it is yet to be determined how much walking knee stiffness must increase in order to influence physical function and other symptoms. Further work evaluating the clinical use of this measure is therefore suggested.

Despite joint stiffness typically being reported by OA patients (8, 20), and a consistently high walking knee stiffness being measured for the OA participants of the current study, walking knee stiffness was not associated with self-reported stiffness measured by WOMAC. This is contrary to the study hypothesis, and indicates that walking knee stiffness and self-reported knee stiffness measure different constructs. Our study was powered at 80% to detect a correlation greater than r = 0.43, with a correlation of less than this not considered to be clinically important. Figure 2 highlights that all OA participants had relatively high walking stiffness values compared with the control group mean of 5.6 Nm/°/kg × 100. In contrast, the WOMAC scores ranged widely on the scale of 1–8. Therefore, regardless of the level of self-reported stiffness, the majority of knee OA participants in the current study walked with a stiff knee and were impaired during locomotion. It is likely that the sole use of the WOMAC to evaluate knee stiffness will not detect the degree to which stiffness influences function. Although 3-dimensional gait analysis is not generally a feasible measure of stiffness for the clinical setting, it is useful in research studies incorporating gait analysis. Future research should be directed towards the development of simple clinical tools for evaluating walking knee stiffness to augment self-reported symptoms.

For individuals with knee OA, joint stiffness will be influenced by passive elements such as structural changes in articular cartilage, peripheral osteophyte formation, stiffening of the supporting ligaments (21), and increased joint capsular pressure (22), as well as active elements including muscle strength, activation and co-contraction, joint angle, range of motion, and angular velocity (18). The walking knee stiffness quantified in the current study incorporates both the passive and active elements of stiffness, while the WOMAC score is a subjective perception score. It is unclear whether the WOMAC score is related to passive and/or active elements of stiffness. The pendulum oscillation testing described by Oatis and colleagues (23) that quantifies passive elements of stiffness found no correlation with the WOMAC score, suggesting that WOMAC scores represent something other than pure passive stiffness. Walking knee stiffness encompasses features of pathology and is therefore suggested to provide a distinct and functional measure of stiffness. Since active elements contributing to knee stiffness, such as quadriceps strength/activation and joint range of motion, have been found to differ between those with medial knee OA and asymptomatic individuals, these factors likely contribute to the differences in walking knee stiffness observed between the study groups.

The hypothesis that walking knee stiffness, for the OA group, would be associated with a decrease in knee excursion is supported by the study results, with knee excursion range accounting for 39% of the variance in knee stiffness. Our data suggest, as confirmed by others, that greater knee flexion at ground strike contributes to the resulting reduction in knee excursion range in knee OA (11, 12). Further analysis of the sagittal plane knee joint moment in the current study revealed that the initial knee flexion moment had lower magnitude for the OA group than the controls (mean ± SD −4.3 ± 1.4 Nm/kg compared with −6.5 ± 2.0 Nm/kg). This is likely a result of a reduced requirement for hamstring activity to control knee hyperextension because of the relatively flexed knee at initial contact for the OA group. The lower initial flexion moment and lower peak extension moment result in a lower value for the change in knee moment during weight-acceptance. This will contribute to a lower knee joint stiffness, indicating that the higher walking knee stiffness is contributed to primarily by the lower knee excursion observed for the OA group. This supports the suggestion of Farley and colleagues that joint stiffness may be influenced by initial joint angle (24). From a clinical perspective, an observation of reduced knee excursion during the stance phase of gait, indicated by greater knee flexion at ground impact, may suggest elevated joint stiffness.

A stiffer knee joint was expected to result in increased joint loading, as indicated by peak knee adduction moment and rate of loading of ground reaction force. Neither self-reported nor walking stiffness contributed to rate of loading. The study was powered at 80% to detect a correlation greater than r = 0.55; a correlation of less than this was not considered clinically important. Contrary to the study hypothesis, self-reported stiffness negatively correlated with the peak knee adduction moment, while walking stiffness demonstrated no relationship. Hurwitz et al (25) also noted a significant negative correlation between the WOMAC stiffness subscale score and the peak knee adduction moment (r = −0.34, P < 0.05). It is possible that a perception of increased knee stiffness reflects reduced dynamic frontal plane instability during gait, which could partially explain the higher peak knee adduction moment observed. Future longitudinal studies should assess whether patients with higher levels of self-reported knee stiffness demonstrate slower disease progression over time, as a function of the reduced peak knee adduction moment, which possibly indicates a clinical implication of this relationship.

A possible limitation of the study is the different walking speeds utilized by the OA patients and the controls (1.0 meters/second compared with 1.2 meters/second). This difference resulted from OA participants in general walking most comfortably at the slower speed and the controls walking at the slightly faster speed. In terms of the influence on stiffness, we would expect greater stiffness with faster locomotion, as indicated by the comparison of walking stiffness with running values in the literature (18, 19). Therefore, the finding of greater stiffness for the OA participants when walking slower is the opposite of what would be expected if walking speed was the only difference between groups. The differences in age and body mass between the OA individuals and controls are unlikely to have affected the study results since age was not related to stiffness, and stiffness was corrected for body mass. It is also acknowledged that the relatively small sample size limits the generalizability of the study results, particularly in light of the presented 95% CI for the comparison of means.

In conclusion, this study describes an approach to objectively measure knee joint stiffness during walking. Walking knee stiffness was found to be greater with the presence of knee OA, and reduced knee excursion was associated with this increase. This observation may be indicative of pathophysiologic changes observed at the joint level and in neuromuscular recruitment patterns. Given the functional nature of the knee stiffness measure reported in this study, it is likely to provide additional information to the previously described methodologies, such as self-report tools. Therefore, measurement of knee stiffness during walking may complement existing stiffness measures used for evaluating dysfunction and treatment outcomes. Further research to develop clinical tools to evaluate knee stiffness during walking is merited.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Dixon had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Dixon, Hinman, Crossley.

Acquisition of data. Creaby, Kemp.

Analysis and interpretation of data. Dixon, Hinman, Creaby, Crossley.


The authors wish to acknowledge the contributions of Tim Wrigley and Ben Metcalf in providing support with data collection and technical assistance in the laboratory.