Individuals with severe knee osteoarthritis (OA) exhibit altered proximal walking mechanics compared with individuals with less severe OA and those without knee pain




To examine hip, pelvis, and trunk walking biomechanics in individuals with medial compartment knee osteoarthritis (OA) of varying radiographic disease severities and others without knee pain.


Hip, pelvis, and trunk kinematics and hip kinetics were assessed in 75 individuals with radiographically confirmed OA and 20 asymptomatic individuals. Differences in peak hip adduction and abduction angles, the amount of contralateral pelvic drop, and peak lateral trunk lean measured by 3-dimensional gait analysis were examined using analyses of variance. Peak external hip abduction and adduction moments were compared using analyses of covariance, with gait speed as a covariate.


Those with severe OA disease exhibited significantly less peak hip adduction (5.0°), but more ipsilateral trunk lean toward the study extremity (5.0°) than all other groups (P < 0.001). Those in the severe (1.1°) and asymptomatic (0.9°) groups also exhibited significantly less lateral trunk lean toward the non-study extremity (P = 0.004). No significant differences in hip joint moments or contralateral pelvic drop were observed between the groups (P > 0.02).


Gait changes at the hip and trunk were evident across the groups based on radiographic disease severity and compared with those without knee pain. Although not conclusive from this cross-sectional study design, the results provide initial evidence to support increased lateral trunk lean as being a compensatory response to the disease.


Deviations from normal gait often accompany lower extremity musculoskeletal pathologies; therefore, an understanding of gait abnormalities is essential for optimal rehabilitation of such conditions. It is well known that individuals with osteoarthritis (OA) of the knee, particularly those with medial joint compartment involvement, exhibit altered gait biomechanics compared with those without knee OA (1). Given the role of knee joint loading in the pathogenesis of OA, many studies have investigated the external knee adduction moment (2–4), a validated proxy of medial compartment knee joint load during walking (5, 6). In addition to being higher in individuals with knee OA than in those without knee OA when walking at the same speed (1, 7), the knee adduction moment is known to be a predictor of disease severity (2), pain severity (8), and the risk of disease progression (9). Accordingly, much research has been conducted in studying the knee adduction moment and factors that influence its magnitude.

Previous studies have shown that, in addition to clinical measures such as static lower extremity alignment (3, 10) and disease severity (2), gait biomechanics at other joints may be related to knee adduction moment magnitude. In particular, biomechanics at the hip, pelvis, and trunk have all been shown to affect dynamic joint loading at the knee (11–14). The importance of these proximal joints is highlighted by findings from Chang et al (12), which showed that increased internal hip abduction moments (equivalent to external hip adduction moments) attributed to higher hip abduction strength were protective against OA progression in the knee over an 18-month period. Accordingly, the focus has shifted from solely examining changes at the knee during walking in those with knee OA to understanding factors that influence proximal gait mechanics.

Radiographic disease severity is one factor that has been suggested to be associated with gait mechanics proximal to the knee. Recent studies have reported differences in hip moments between those with and without knee OA (11) and between those who are and are not on a surgical waitlist (15), and support the earlier findings of Chang et al (12). In particular, those with greater disease severity exhibited lower external hip adduction moments and increased external hip abduction moments. Between-extremity differences in hip kinematics in those with knee OA scheduled for high tibial osteotomy surgery have also been reported (16). However, despite previous studies showing a protective effect of lateral trunk lean against increased knee adduction moment magnitudes (13, 14), it is unknown if those with differing levels of disease severity exhibit differences in the amount of self-selected lateral trunk lean during gait.

Importantly, it is believed by many authors that these gait differences observed at the hip and trunk are compensatory mechanisms that occur secondary to the onset of OA, and are influenced by structural deterioration, pain, or other symptoms (11, 16, 17). That is, individuals with knee OA adopt these adaptations and may modify them as the disease progresses. Although longitudinal evidence to support this claim is lacking and difficult to obtain, a cross-sectional examination of any differences in these hip and trunk biomechanics based on disease severity would provide some initial support. Therefore, the purpose of this study was to compare hip and trunk biomechanics in those with and without knee OA, with particular importance placed on differences based on radiographic disease severity. It was hypothesized that individuals with knee OA would demonstrate differences in frontal plane hip and trunk biomechanics during walking compared with those without knee OA, and that within OA participants, differences would be evident across varying grades of radiographic disease severity. In particular, we hypothesized that increasing disease severity would be associated with lower amounts of hip adduction angulation and external moments, and more lateral trunk lean and external hip abduction moments.


This cross-sectional observation study involved 95 participants. Twenty asymptomatic controls (15 women, 5 men) and 75 individuals (38 women, 37 men) with radiographically confirmed medial compartment knee OA participated. Those with knee OA were recruited via local print media and from our database of previous knee OA study participants for enrollment in a randomized controlled trial of hip strengthening (18). All of the data described in the current study (gait analysis and Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC] [19]) were acquired at baseline prior to any intervention. Participants with knee OA were included if they were ages >50 years and had knee pain on most days of the previous month (average level >3 on an 11-point numerical rating scale). Other inclusion criteria consisted of: 1) varus malalignment as determined by short-film radiographs, 2) predominance of pain/tenderness over the medial region of the knee, and 3) osteoarthritic changes in the medial compartment of the tibiofemoral joint (osteophytes and/or joint space narrowing). Exclusion criteria included: 1) questionable radiographic knee OA (Kellgren/Lawrence [K/L] scale grade 1 [20]); 2) lateral tibiofemoral joint space width less than medial; 3) symptoms originating predominantly from the patellofemoral joint as determined by clinical examination; 4) knee surgery or intraarticular corticosteroid injection within 6 months; 5) current or past (within 4 weeks) oral corticosteroid use; 6) systemic arthritic conditions; 7) history of hip or tibiofemoral/patellofemoral joint replacement or tibial osteotomy; 8) any other muscular, joint, or neurologic condition affecting lower extremity function; and 9) unable to ambulate without a gait aid.

The asymptomatic control participants were recruited from the community via local newspapers and other print media. All of the participants were ages >50 years and had to have been completely free of any knee or hip pain for the previous 6 months. Consistent with the OA group, the following exclusion criteria applied to the asymptomatic control group: 1) current or past (within 4 weeks) oral corticosteroid use; 2) systemic arthritic conditions; 3) history of hip or tibiofemoral/patellofemoral joint replacement or tibial osteotomy; 4) any other muscular, joint, or neurologic condition affecting lower extremity function; and 5) unable to ambulate without a gait aid. On presentation to the laboratory for testing, all of the control participants underwent a clinical examination by a trained musculoskeletal physiotherapist (MAH) to ensure normal lower extremity joint function and the absence of knee, hip, or back pain. They were also required to rate knee pain using a numerical rating scale (11-point scale, where 0 = no pain and 10 = the worst pain ever) and complete the pain subscale of the WOMAC. Any control participant reporting any pain using either of these 2 methods was excluded from the study. The research was approved by the University of Melbourne Human Research Ethics Committee and the Department of Human Services Radiation Committee. All of the participants provided written informed consent.

Radiographic analysis.

The OA group underwent a single, semiflexed, posteroanterior weight-bearing short-film radiograph with the feet externally rotated 10° to confirm the presence of OA, to assess radiographic severity, and to measure lower extremity alignment. Disease severity of the study knee (the one reported as more symptomatic in cases of bilaterally eligible knees) was assessed by the same 2 examiners using the K/L scale grading system (20), where higher grades indicate greater OA severity (our interrater reliability on a subset of 30 radiographs: κ = 0.87). To confirm the presence of varus malalignment, mechanical lower extremity alignment was measured on the study extremity using techniques and an equation described by Kraus et al (21). Using the equation, values less than 182° corresponded to varus alignment (180° on a full-length radiograph), whereas values greater than 182° corresponded to valgus alignment. To reduce a subject bias based on disease severity, the number of individuals with mild (K/L scale grade 2), moderate (K/L scale grade 3), and severe (K/L scale grade 4) disease recruited in the larger randomized controlled trial study was capped (n = <30) and data from the first 25 participants in each severity group were included in the present study.

Gait analysis.

A Vicon motion analysis system with 8 cameras operating at 120 Hz (Vicon) and Workstation software, version 5.2 (Vicon), was used to measure gait biomechanics. The standard Vicon Plug-in-Gait lower extremity marker set (Vicon) was used, and additional markers were attached to the medial knee and ankle during a single static standing trial to determine the relative positioning of joint centers. Ground reaction force data were measured by 2 force platforms (Advanced Mechanical Technology) embedded in the floor at the midpoint of a 10-meter walkway, sampled at 1,080 Hz in synchrony with the cameras. The participants walked barefoot at a self-selected pace. Several practice trials ensured that the participants walked naturally and that the entire foot of the test extremity landed on the force platform. Kinematic and kinetic data were filtered using the Woltring filtering function within Workstation software, version 5.2 (22). Joint moments were calculated via inverse dynamics (Vicon Plug-in-Gait, version 2; Vicon) and normalized for body weight and height (Nm/%BW × Ht) (23). Peak external hip adduction (both early and late stance peaks) and abduction moments during stance were identified for each trial. The pelvic coordinate system was constructed from markers on the anterior and posterior superior iliac spines, from which peak drop of the pelvis during ipsilateral stance was identified from the frontal plane pelvic obliquity angle in relation to the laboratory coordinate system (pelvic Cardan angles in tilt-obliquity-rotation sequence).

The trunk (thorax) coordinate system was constructed from markers at the sternal notch, T2, and T10, as suggested by Baker (24). Peak lateral trunk lean over the study (ipsilateral) and non-study (contralateral) extremities when each were in stance was identified for each trial (first angle in the Cardan sequence representing trunk orientation in relation to the laboratory coordinate system). Lastly, peak hip adduction angle during stance and frontal plane hip angle at initial contact were identified.

Statistical analysis.

Each participant completed 5 walking trials with successful force platform strikes, and mean values for the variables listed above were obtained by averaging discrete values across the trials. With the exception of sex, differences between the groups (mild, moderate, severe, and asymptomatic controls) for demographic and kinematic data were examined using one-way analyses of variance. Sex differences were examined using a chi-square test. However, given the potential influence of walking speed on gait kinetics (25), differences between the groups for joint moments were examined using one-way analyses of covariance, with walking speed used as a covariate. When significant differences existed, Tukey's honest significant difference post hoc analyses were conducted to assess the nature of these differences. All of the data were analyzed using SPSS, version 15 (SPSS), and a conservative alpha level of 0.01 due to multiple comparisons. An a priori power analysis indicated that our sample size allowed us to detect moderate effect sizes in all variables (d = 0.35) with 80% power.


Demographic, clinical, and gait data for all of the participants are summarized in Table 1. Those in the severe group had significantly more varus malalignment than those with mild disease (P = 0.001). No other demographic or clinical variables were significantly different between the groups. Of note, however, is that WOMAC pain was only compared among those with knee OA (P = 0.12) because those in the control group reported no knee pain (therefore, a WOMAC score of 0).

Table 1. Demographic, clinical, and gait data for those with mild, moderate, and severe radiographic knee OA as well as control participants*
 Control (no knee pain)Mild (K/L scale grade 2)Moderate (K/L scale grade 3)Severe (K/L scale grade 4)P
  • *

    Values are the mean ± SD unless otherwise indicated. OA = osteoarthritis; K/L = Kellgren/Lawrence; N/A = not applicable; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

  • Indicates significantly different (P < 0.01) from the severe OA group.

  • Indicates significantly different (P < 0.01) from the control (no knee pain) group.

  • §

    Indicates significantly different (P < 0.01) from the control (no knee pain) group and the severe OA group.

  • P values are based on adjustments for walking speed via analyses of covariance.

 Sex, male:female5:1510:1514:1113:120.16
 Age, years63.2 ± 12.461.2 ± 7.763.6 ± 8.468.0 ± 6.60.06
 Height, meters1.65 ± 0.061.65 ± 0.081.67 ± 0.081.68 ± 0.090.34
 Mass, kg69.3 ± 12.173.7 ± 14.078.4 ± 14.881.7 ± 20.20.05
 Alignment, degreesN/A178.2 ± 1.5176.6 ± 2.7175.4 ± 2.60.001
 WOMAC painN/A6.4 ± 3.26.8 ± 2.68.1 ± 3.10.12
 Walking speed, meters/second1.38 ± 0.171.28 ± 0.161.21 ± 0.171.17 ± 0.16< 0.001
Kinematics, degrees     
 Frontal plane angle at initial contact (positive = adduction)1.7 ± 3.01.4 ± 3.80.5 ± 4.3−0.5 ± 4.30.24
 Peak hip adduction9.4 ± 3.18.6 ± 3.76.4 ± 4.05.0 ± 3.9< 0.001
 Peak pelvic drop4.0 ± 1.74.3 ± 2.33.0 ± 1.82.8 ± 2.30.02
 Peak trunk lean to study extremity1.6 ± 2.42.3 ± 2.03.1 ± 2.05.0 ± 2.2< 0.001
 Peak trunk lean to non-study extremity0.9 ± 2.02.5 ± 2.02.9 ± 2.9§1.1 ± 1.50.004
Joint moments (% body weight × height)     
 First hip adduction5.98 ± 0.875.55 ± 0.985.27 ± 1.104.86 ± 0.940.07
 Second hip adduction5.55 ± 0.984.91 ± 0.964.59 ± 1.024.55 ± 1.430.82
 Hip abduction0.37 ± 1.210.65 ± 1.190.74 ± 1.351.27 ± 1.840.04
 Peak knee adduction3.12 ± 0.882.94 ± 0.882.98 ± 1.153.12 ± 0.860.69

Many significant differences existed in the gait variables across the groups. A significant difference was observed in self-selected gait speed (F[3,91] = 7.31, P < 0.001) and post hoc tests indicated that those with mild disease walked significantly faster than those with moderate or severe disease, whereas all 3 OA groups were significantly slower than those in the asymptomatic control group. A significant difference was also observed in the maximum hip adduction angle (F[3,91] = 6.76, P < 0.001) during stance (Figure 1B), with the severe group exhibiting less adduction compared with the 3 other groups. No statistical differences in frontal plane hip angle at initial contact were observed (F[3,91] = 1.41, P = 0.24), although the severe group only was in abduction at this point in the gait cycle (Figure 1B).

Figure 1.

Ensemble averages across the stance phase of gait for A, ipsilateral lateral trunk lean angles and B, pelvic obliquity angles for each study group. Traces were created by normalizing each participant's data to 100% of stance (0% = initial contact, 100% = foot off) and averaging across the individuals within a group at each data point. Note that the data are only shown for the stance phase of the study extremity and do not include values during the entire contralateral stance phase. Lines correspond to control (gray line), mild (dotted line), moderate (solid line), and severe (broken line).

Ensemble averages across the stance phase for trunk and pelvis kinematics are shown in Figure 1. Ensemble average curve patterns for frontal plane hip angles and external hip joint moments were similar to those reported in previous studies (11, 15, 16) and are not reported here. Peak trunk lean values were significantly different between the groups. Not only were differences observed in the peak trunk lean over the study extremity (F[3,91] = 10.94, P < 0.001), but also over the contralateral extremity when it was in stance (F[3,91] = 4.71, P = 0.004). Those in the severe group exhibited significantly more ipsilateral trunk lean than all other groups, whereas severe and asymptomatic control group participants exhibited significantly less trunk lean to the contralateral extremity compared with the moderate group. Lastly, no significant differences in pelvic drop during stance (Figure 1B) were present (F[3,91] = 3.50, P = 0.02). Post hoc analyses on all of the kinematic variables were conducted while co-varying for walking speed, and showed that the results were not different than those reported above.

When examining hip joint kinetics (Table 1), although increases in disease severity corresponded to decreases in early and later stance external hip adduction moments, these differences were not statistically significant before or after co-varying for walking speed (F[3,91] = 2.38, P = 0.07 for early stance; F[3,91] = 0.30, P = 0.82 for late stance). Finally, increasing disease severity was associated with larger external hip abduction moments in early stance, although differences between the groups were not statistically significant before or after controlling for walking speed (F[3,91] = 2.88, P = 0.04).


Results from this study indicate that those with severe radiographic knee OA demonstrate specific differences in frontal plane hip and trunk walking biomechanics compared with those with less severe radiographic changes or those without knee OA. Although not conclusive from this cross-sectional study design, these findings support earlier suggestions that people with knee OA modify their walking patterns at some point after disease onset and potentially increase the amounts of compensation, most notably in the later stages of the disease (11, 16, 17).

Few studies have investigated proximal frontal plane gait biomechanics in individuals with knee OA, and our results support earlier findings of kinematic differences. Patients with severe knee OA walked with less hip adduction throughout stance and were the only group to exhibit hip abduction at initial contact. This gait difference was also found by Briem and Snyder-Mackler in their study of 32 individuals with moderate medial compartment knee OA who adopted a more abducted hip pattern in their involved extremity compared with their uninvolved extremity throughout much of stance (16). Huang et al (17) also showed that individuals with severe knee OA walk with more hip abduction throughout stance than healthy individuals.

Our findings of no differences in peak external hip adduction or abduction moments are inconsistent with previous reports (11, 15, 16). We did observe trends in our data that were consistent with these previous studies (lower external hip adduction moments and higher external hip abduction moments with increasing disease severity), but these between-group differences did not reach statistical significance. One key difference in our study was that we statistically controlled for walking speed, whereas others did not. It is known that, all else being equal, higher walking speeds are associated with larger ground reaction forces (26) and associated joint moments (27) and that those with less severe OA, and especially asymptomatic individuals, walk faster (28). Therefore, the higher hip adduction moments in the asymptomatic participants and those with less severe OA in our cohort may be explained by greater self-selected walking speeds. However, this does not explain why higher abduction moments have been observed in those with increasing disease severity in the present as well as previous studies (11).

Our study builds on previous studies in some important ways. First, we report and compare frontal plane trunk lean angles. Lateral trunk lean is receiving increased attention in the knee OA literature due to its common observation clinically, as well as its beneficial effect on medial compartment knee joint loading (13, 14). Our findings suggest that the amount of lateral trunk lean over the affected knee increases with increasing disease severity. Increased trunk lean lateralizes the ground reaction force vector, which results in concomitant decreases in external knee and hip adduction moments and increases in initial stance external abduction moments at these 2 joints. We have shown in a previous study (14) that the relationship between the knee adduction moment and lateral trunk lean is independent of disease severity, varus alignment, or walking speed, suggesting that trunk lean provides a consistent mechanical benefit based on magnitude of lean. Therefore, the patients with severe disease in our cohort would have received the greatest mechanical benefit since they leaned the most. Importantly, by offloading the painful knee compartment, this compensation has the potential to provide symptomatic benefits. Indeed, although not reported, a small but significant correlation between WOMAC pain scores and stance extremity trunk lean was observed in our study (Pearson's r = 0.25, P = 0.03). Our findings of increased lean with disease severity may suggest that leaning is a learned response that is developed over the course of the disease and becomes more prevalent in the advanced stages when factors such as pain, malalignment, and ultimately knee loading typically increase. However, although our results provide support for a potential relationship between trunk lean and disease severity, given the nature of our cross-sectional design, we cannot conclude that any given individual will increase their amount of trunk lean as their disease progresses.

Although participants with milder and moderate OA severity and those without knee pain exhibited near similar amounts of trunk lean between extremities, those with severe disease in our study walked with significantly more ipsilateral trunk lean when the affected extremity was in contact with the ground compared with the contralateral extremity (5.0° versus 1.1°). In addition to not being similar, given that those with mild and moderate OA exhibited 2.5° and 2.9° of trunk lean over the contralateral extremity, respectively, these data suggest that participants with severe radiographic disease favor their most painful extremity throughout not only ipsilateral, but contralateral, stance (exhibited as reduced lateral lean over this extremity). Although the benefits for the most painful extremity with this strategy have been outlined above, the potential detriment to the other extremity cannot be ignored. It is possible that by maintaining an upper body posture away from the stance extremity, medial compartment loading of the unaffected/less affected extremity during gait may be increased, which may place that knee at an increased risk of disease development and/or progression. This is particularly relevant, given the prevalence of bilateral knee involvement. However, future research is needed to investigate this possibility.

We also measured pelvic obliquity but found no significant differences between the groups. It is possible, however, that since the trunk and pelvis likely do not operate independently of each other, any true discrepancies in pelvic drop data may have been masked by between-group differences in lateral trunk lean. Specifically, visual observation and re-inspection of 3-dimensional coordinate data showed that increasing ipsilateral lateral trunk lean was associated with raising the contralateral anterior superior iliac spine marker (i.e., a “hip hike” gait pattern).

Our study has identified differences in the amount of a suggested gait compensation, lateral trunk lean, employed by individuals with severe radiographic knee OA severity. We have also shown that differences in kinematics at the hip exist between groups with differing severity. A limitation of our study is that we cannot rule out the possibility of concomitant articular cartilage degeneration in the asymptomatic group, because although this group was completely free of knee pain, and therefore does not have symptomatic knee OA according to accepted clinical classification criteria (29), we did not acquire radiographs for these individuals. Despite this limitation, the asymptomatic nature of the control group still provides a useful comparison with our knee OA cohort. Although not a primary objective of this study, we also did not find a significant difference in the knee adduction moment between our control group and the OA groups, which contradicts some previous studies (1). That said, many recent studies (15, 30–33) have also failed to show significant differences in peak knee adduction moment magnitudes between those with and without knee OA, which may possibly be explained by the absence of a protective effect that trunk lean provides (i.e., those without knee pain do not lean as much and thus do not experience the load-reducing benefits of increased lateral trunk lean).

Another important limitation of this study that must be addressed is the demographic status of the asymptomatic control group. Recognizing that we could not feasibly control for all of the potential confounders, our primary aim was to match the control group based on age as closely as possible. We were successful in this regard, but less so for other potentially important variables such as sex and body mass, although statistical analysis failed to show significant differences regarding these parameters between the groups (P > 0.05). Although future researchers may wish to control for variables such as muscle strength that may potentially impact hip and trunk biomechanics (34), it could be argued that this may be inappropriate, given that muscle weakness is a normal finding associated with the disease process (thus resulting in a nonrepresentative sample of asymptomatic controls) and may, in fact, explain some of the differences in gait patterns that we have observed.

A final limitation of our study was that we categorized our OA participants based on the K/L grading scale, which although widely used and reported in the knee OA literature, has questionable clinical appropriateness. Participants were categorized as having mild (K/L scale grade 2), moderate (K/L scale grade 3), or severe (K/L scale grade 4) OA, as per the guidelines. However, this scale is highly dependent on the number and size of joint line osteophytes as well as joint space narrowing, which although well-accepted signs of the disease process and consistent with its pathogenesis, are commonly discordant with symptoms (35, 36). Therefore, in lieu of an accepted categorization scheme that accounts for both radiographic and symptomatic severity, interpretation of our findings should be cautious.

Our findings have clinical importance and show that individuals with severe radiographic knee OA demonstrate proximal gait changes compared with asymptomatic individuals and those with lesser grades of radiographic disease severity. The cross-sectional nature of our study means that it is not possible to determine when and why these gait changes occur; however, it is feasible that excessive lateral trunk lean may be a natural compensatory response to the disease. Future studies employing a longitudinal study design are needed to confirm this hypothesis.


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. Hunt 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. Hunt, Wrigley, Hinman, Bennell.

Acquisition of data. Hunt.

Analysis and interpretation of data. Hunt, Wrigley, Bennell.


The authors wish to thank Ms Fiona McManus and Dr. Mark Creaby for assistance with data collection.