Asymmetric loading and bone mineral density at the asymptomatic knees of patients with unilateral hip osteoarthritis




In patients with unilateral end-stage hip osteoarthritis (OA), the contralateral knee is known to be at greater risk for end-stage knee OA compared to the ipsilateral (i.e., same-side) knee. The contralateral knee is known to have increased dynamic joint loads compared to the ipsilateral knee. The present study was undertaken to examine patients who had unilateral hip OA but who did not have symptoms of knee OA, in order to detect early asymmetries in knee loading.


Data on 62 patients with unilateral hip OA were evaluated. Patients underwent gait analyses of dynamic knee loads as well as dual x-ray absorptiometry for determination of bone mineral density (BMD) in both knees. Differences between knees were compared.


Peak dynamic knee loads were significantly higher at the contralateral knee compared to the ipsilateral knee (mean ± SD 2.46 ± 0.71 percent of body weight × height versus 2.23 ± 0.81 percent of body weight × height; P = 0.029). Similarly, medial compartment tibial BMD was significantly higher in the contralateral knee compared to the ipsilateral knee (mean ± SD 0.897 ± 0.208 gm/cm2 versus 0.854 ± 0.206 gm/cm2; P = 0.033). Interestingly, there was a direct correlation between the contralateral:ipsilateral dynamic knee load and contralateral:ipsilateral medial compartment tibial BMD (ρ = 0.287, P = 0.036).


The risk of developing progressive symptomatic OA in contralateral knees is higher compared to the risk in ipsilateral knees in patients with unilateral hip OA. The present study demonstrates that loading and structural asymmetries appear early in the disease course, while the knees are still asymptomatic. These early biomechanical asymmetries may have corresponding long-term consequences, providing further evidence for the potential role of loading in OA onset and progression.

Osteoarthritis (OA) is the most common arthropathy worldwide, and a major cause of disability and impairment in quality of life. It is a chronic, slowly progressive arthropathy; hence, longitudinal studies evaluating its natural history are lengthy, costly, and often impractical to perform. Although several factors have been associated with incident OA and with OA disease severity, it has been difficult to establish both their sequence of onset and whether they are causally involved in OA pathophysiology.

The role of biomechanics has been an important area of investigation in OA, especially dynamic joint loading during physiologic activity (1). The peak external knee adduction moment, a validated gait parameter that reflects the load at the medial compartment of the knee (2), has been associated with pain, radiographic severity, and progression of knee OA (3–5). Although one study demonstrated that high adduction moments preceded the onset of knee pain symptoms (5), most of the other studies conducted to date have evaluated the relationship between dynamic joint loads and already established OA; it has not yet been clearly shown that elevated peak dynamic loads precede the development of symptomatic knee OA.

In addition to gait analyses and dynamic loading, complementary information is provided by assessing subchondral areal bone mineral density (BMD) (gm/cm2) at the tibial plateau, which reflects load history across the joint (6–9), and can be assessed noninvasively using dual x-ray absorptiometry (DXA).

We previously observed that in patients with end-stage unilateral OA of the hip, advanced OA more frequently develops in the contralateral knee (which is on the opposite side of the body from the affected hip) than the ipsilateral knee (10); moreover, we also reported that the contralateral knee was subjected to substantially higher peak dynamic loads than the ipsilateral knee, and that these asymmetries remained constant several years after hip replacement (11). These previous observations in unilateral hip OA, in which the contralateral knee is known to be at elevated risk for developing OA relative to the ipsilateral knee, suggest a unique model to study factors involved in the pathogenesis of early (presymptomatic) knee OA.

Using that model (Figure 1), we evaluated dynamic joint loading and proximal tibial BMD in patients who have unilateral hip OA but do not have OA symptoms in the knees, in order to test the hypothesis that they have elevated dynamic loads and increased tibial BMD at the contralateral knee compared to the ipsilateral knee. This would demonstrate that these loading asymmetries are present during an early presymptomatic state.

Figure 1.

Unilateral hip osteoarthritis (OA) study model. The knee contralateral to the affected hip was observed to have higher dynamic joint loads and higher medial tibial bone mineral density compared to the ipsilateral knee.



This study was approved by the institutional review board for studies involving human subjects, and written informed consent was obtained from all patients. Criteria for inclusion were the presence of symptomatic hip OA, as defined by the American College of Rheumatology criteria (12) and the presence of pain while walking (self-reported as ≥30 mm on a 100-mm visual analog scale [VAS]) (corresponding to question 1 of the VAS of the hip-directed adaptation of the Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC]) (13). Radiographic OA of the index hip (Kellgren/Lawrence [K/L] grade ≥2 [14]) was documented by anteroposterior (AP) radiography of the pelvis.

Patients were excluded if they demonstrated symptomatic OA in the contralateral hip or in either knee, with presence of pain while walking defined as a score of >30 mm (of 100 mm). Patients were also excluded if they had evidence of grade >3 radiographic OA in the contralateral hip or in either knee, according to the modified K/L scale. Other exclusion criteria included the inability to walk without assistance, the presence of an inflammatory arthropathy, a history of any lower-extremity joint replacement, or history of trauma to or arthroscopy of either knee within the preceding 6 months.

Clinical assessment.


All patients underwent radiography of the pelvis; radiographs were evaluated for K/L grade in the hips. All patients also underwent AP standing knee radiography and were evaluated for K/L grade in the knees. All K/L evaluations were performed by a single trained observer (NS).

Pain assessment.

Patients were assessed for pain in both knees and both hips using the WOMAC VAS. The WOMAC is the current standard in the analysis of pain and function in lower-extremity OA (15–17) and site-directed adaptation of the WOMAC has proven useful and feasible (18, 19). The WOMAC scores were normalized to a 100-mm scale.

Gait analyses.

All patients underwent gait analysis. Gait assessment included collection of 3-dimensional (3-D) kinematic and ground reaction force data using 4 Qualisys optoelectronic cameras (Innovision Systems) with passive markers and a multicomponent force plate with a sampling frequency of 120 Hz (Bertec). Passive markers were placed at the lateral most aspect of the superior iliac crest, the superior aspect of the greater trochanter, the lateral knee joint line, lateral malleolus, lateral calcaneus, and the head of the fifth metatarsal. For moment calculations, the joint centers of the hip, knee, and ankle were approximated, using previously described methods (20, 21). The joint center of the ankle was determined to be the midpoint of the distance from the medial to lateral malleolus. The joint center of the knee was determined to be the midpoint of the distance between the medial and lateral joint lines of the tibiofemoral joint. The joint center of the hip was determined to be 2.5 cm distal to the midpoint of the distance between the anterior superior iliac spine and the pubic tubercle.

Patients were instructed to walk at a self-determined normal speed with their conventional walking shoes on a 2-inch–thick wooden pressboard covered with linoleum. The kinetic components calculated (using processing software developed by Computerized Functional Testing Corporation) included frontal plane external joint moments at the knee (22). Gait data from 1 of 2 normal-walking-speed runs of the unaffected limb were chosen and speed-matched to 1 of 2 normal-walking-speed runs from the affected limb for comparison. The investigator choosing the runs was blinded with regard to the affected side.

The position and force data were then used to assess sagittal range of motion at the joints and to calculate 3-D external moments using inverse dynamics. The external moments that act on a joint during gait are, according to Newton's second law of motion, equal and opposite to the net internal moments produced primarily by the muscles, soft tissues, and joint contact forces. The external moments are normalized to patients' body weight × height × 100 (%BW*ht) (23). Peak forces on the lateral and medial compartments of the knees were calculated using gait data in conjunction with a statically determinate model (24).

BMD analyses.

DXA (Lunar Prodigy 7.53E; General Electric) was used to scan the bilateral proximal tibiae. AP valuations were performed with the patients supine on the examination table. Patients' legs were internally rotated ∼15° and held in full extension to ensure that scanning was performed perpendicular to the tibial shaft. Scanning in this way ensures that the fibula is clearly evident and tibial bone is not superimposed on fibular bone. Proximal tibial BMD was assessed using a method previously described by Clarke et al (9). The Lunar Prodigy software was used to assess the area (cm2), bone mineral content (BMC) (gm), and BMD (gm/cm2) for the proximal and lateral regions of interest (ROIs), as well as for the distal ROI in the tibial shaft. The height of each ROI was equal to 10% of the width of the tibial plateau, to standardize the measurement for differences in bone size between patients.

The software automatically determined the subperiosteal surface of the tibia to which each ROI extended superiorly. The internal border of the medial and lateral ROIs extended to intercondylar eminences. The cortical bone of the subchondral plate was excluded from the measurements as sclerosis in this region can affect BMD measurements. Care was taken to avoid including bone from the lip of the tibia, since osteophytes could affect BMD measurements. With the lateral ROI, care was taken not to include the fibula. The medial and lateral ROIs therefore include subchondral trabecular bone. The repeatability of these measurements has previously been assessed in subjects with mild to moderate knee OA (n = 10 subjects, each scanned twice). Coefficients of variation after repositioning of patients on the table were 3.8%, 2.0%, and 1.5% for the medial, lateral, and distal ROIs, respectively, and 3.0% for the medial-to-lateral ratio (6).

Study end points.

The primary end points for the study were the peak external knee adduction moment (defined as the external adduction moment of greatest magnitude during the stance phase of the gait cycle), total medial compartment load, and medial compartment BMD as measured by DXA. Secondary end points included sagittal plane (flexion–extension) range of motion at the lower-extremity joints, hip flexion, adduction, internal and external rotation moments, knee flexion moment, and lateral compartment BMD at the knees.

Statistical analysis.

Statistical analysis was performed using SPSS software. Paired-sample t-test was used to compare dynamic joint loads and BMD at the ipsilateral and contralateral knees. Relationships between loading parameters and BMD were evaluated by Pearson's and Spearman's correlations.


Of the 124 patients screened, 62 met the inclusion criteria and completed the study. Figure 1 illustrates the unilateral hip OA model, to relate the ipsilateral and contralateral associations with the results of this study. General patient demographic and baseline characteristics are summarized in Table 1. The mean ± SD age of the patients was 62 ± 11 years. There were 26 men and 36 women. Complete WOMAC data were missing for 1 patient. The mean ± SD baseline pain scores (adjusted to a 100-mm scale) were higher for the ipsilateral knee compared to the contralateral knee, but were extremely low in both cases (10 ± 18 mm and 5 ± 8 mm, respectively; P = 0.006).

Table 1. Characteristics of the 62 patients with unilateral hip osteoarthritis*
  • *

    Except where indicated otherwise, values are the mean ± SD. BMI = body mass index; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; VAS = visual analog scale; K/L = Kellgren/Lawrence.

Age, years62 ± 11
Sex, no. male/female26/36
BMI, kg/m228 ± 5
WOMAC pain VAS, 0–100 mm 
 Affected hip37 ± 24
 Unaffected hip6 ± 11
 Ipsilateral knee10 ± 18
 Contralateral knee5 ± 8
WOMAC stiffness VAS, 0–100 mm 
 Affected hip42 ± 29
 Unaffected hip9 ± 17
 Ipsilateral knee12 ± 18
 Contralateral knee7 ± 14
K/L grade (0–4), no. of patients 
 Affected hip 
 Unaffected hip 
 Ipsilateral knee 
 Contralateral knee 

Appropriate gait data for analyses were available for 58 patients. Gait data from 3 of the patients were excluded due to malfunction of the gait analysis equipment during the examination. Bone mineral density of the knees bilaterally was evaluated in 55 patients.

Gait results from the ipsilateral and contralateral limbs are summarized in Table 2. As expected, dynamic hip range of motion was significantly greater at the contralateral hip compared to the affected hip (P < 0.001). Peak hip moments, including the peak hip flexion (P = 0.003), adduction (P = 0.018), internal rotation (P < 0.001), and external rotation moments (P = 0.037) were significantly higher at the contralateral unaffected hip compared to the ipsilateral osteoarthritic hip.

Table 2. Gait characteristics and bone density in individual limbs*
CharacteristicIpsilateral (affected hip) limbContralateral (unaffected hip) limb
  • *

    Values are the mean ± SD. ROM = range of motion; BW = body weight.

  • P < 0.05 versus ipsilateral hip.

Flexion–extension ROM, degrees  
 Hip ROM21 ± 829 ± 6
 Knee ROM60 ± 662 ± 4
Peak moments, %BW × height  
 Hip flexion4.75 ± 1.495.66 ± 2.09
 Hip adduction3.06 ± 0.983.36 ± 0.94
 Hip internal rotation0.45 ± 0.250.57 ± 0.23
 Hip external rotation0.37 ± 0.260.46 ± 0.22
 Knee flexion1.43 ± 0.971.79 ± 1.26
 Knee adduction2.23 ± 0.812.46 ± 0.71
Knee compartmental loads, × BW  
 Medial compartment load1.96 ± 0.632.23 ± 0.52
 Lateral compartment load1.23 ± 0.351.41 ± 0.48
Bone mineral density, gm/cm2  
 Medial tibial plateau0.854 ± 0.2060.897 ± 0.208
 Lateral tibial plateau0.726 ± 0.1900.744 ± 0.195

In the knees, both primary gait outcome measures, the peak external knee adduction moment (P = 0.029) and the total medial compartment knee load (P = 0.003), were significantly higher at the contralateral knee relative to the ipsilateral knee, as was the lateral compartment load (P = 0.008). Additionally, the peak knee flexion moment appeared to be higher at the contralateral knee, but this did not reach statistical significance (P = 0.052).

Medial tibial plateau BMD was significantly higher at the contralateral knee relative to the ipsilateral knee (P = 0.033). However, there were no significant differences at the lateral tibial plateau (P = 0.469) (Table 2).

Bivariate correlations between contralateral:ipsilateral knee dynamic loading and BMD revealed that the ratio of contralateral:ipsilateral medial compartment knee BMD was directly correlated with contralateral:ipsilateral peak external knee adduction moment (ρ = 0.287, P = 0.036) and contralateral:ipsilateral knee medial compartment load (ρ = 0.351, P = 0.009).

Considering that radiographic changes at the contralateral hip in some patients were graded 3 on the K/L scale, despite minimal clinical symptoms of pain, a separate analysis was performed, excluding these participants. The asymmetry between the contralateral knee and ipsilateral knee medial tibial BMD remained significant (in 47 patients, mean ± SD 0.892 ± 0.199 gm/cm2 versus 0.845 ± 0.212 gm/cm2, respectively; P = 0.030). The asymmetry in the peak external knee adduction moment remained, but lost significance (in 46 patients, mean ± SD 2.39 ± 0.69 %BW*ht versus 2.18 ± 0.75 %BW*ht; P = 0.075). The correlation between contralateral:ipsilateral medial compartment knee BMD and peak external knee adduction moment (ρ = 0.406, P = 0.006) and contralateral:ipsilateral knee medial compartment load (ρ = 0.454, P = 0.002) strengthened and remained significant.

Similarly, an analysis that excluded knees with a K/L grade of 2 or 3 was performed. Once again, the asymmetry between the contralateral knee and ipsilateral knee medial tibial BMD remained significant (in 35 patients, 0.901 ± 0.217 gm/cm2 versus 0.851 ± 0.218 gm/cm2, respectively; P = 0.05). The asymmetry in the peak external knee adduction moment remained, but lost significance (in 36 patients, 2.51 ± 0.73 %BW*ht versus 2.31 ± 0.69 %BW*ht; P = 0.109). The correlation between contralateral:ipsilateral medial compartment knee BMD and the peak external knee adduction moment lost significance (ρ = 0.310, P = 0.079); however, the correlation with the contralateral:ipsilateral knee medial compartment load remained significant (ρ = 0.452, P = 0.008).


This study demonstrates that, in unilateral hip OA, the contralateral knee is subjected to significantly higher dynamic joint loading, as assessed by peak external knee adduction moment and by total medial compartment loads, relative to the ipsilateral knee. Importantly, this asymmetry of knee loading is observable even when knees are asymptomatic and have no clinical evidence of OA. Moreover, the significant asymmetries observed in the proximal tibial BMD of the contralateral versus ipsilateral knees also provide evidence of substantially altered load history in the knees. This study, along with data from our previous study demonstrating asymmetric progression of knee OA in those with unilateral hip OA (10), provides support for the notion that alterations in dynamic joint loading may precede the onset of symptomatic knee OA in humans. In addition, our findings support the idea of a relationship between dynamic loading and measures of load history in OA.

In a study of adults from the general population, Amin and colleagues demonstrated that the development of new chronic knee pain over 3–4 years was associated with significantly higher baseline knee adduction moments, compared to those who did not develop pain (5). The results of the current study provide further support to the notion that high dynamic loads may precede knee pain symptoms.

The asymmetry in knee loading in unilateral hip OA was first described in a population with end-stage hip disease (in patients awaiting total hip replacement) and interestingly, was shown to persist for up to 2 years, even after hip replacement and complete resolution of hip pain (11). The current study demonstrates that loading asymmetries at the knees begins early in the disease course of hip OA, suggesting that the relative overload of the contralateral knee that places it at higher risk of developing advanced OA is likely already established years before the progression of hip OA to end-stage disease. These results may have implications for interventional treatments that target unilateral hip OA in order to prevent or minimize these asymmetries early in the disease course.

Increased loading was observed in both the contralateral hip and the contralateral knee, which is presumably related to a pain response in the ipsilateral limb. This pain–loading relationship, which was previously demonstrated in knee OA to actually show increased loading at the affected joint after analgesic treatment for pain (3, 25), complicates the interpretation and evaluation of the relationship between loading and disease. Whereas loading and disease severity are directly related, pain (the principal symptom of OA) may actually result in decreased loading. In contrast, a unique aspect of this study, and of the unilateral hip OA model, is the ability to assess the knees without the confounding influence of localized knee pain.

The asymmetries noted in the subchondral BMD of the proximal tibiae, which reflected those observed in medial knee loads, provide evidence of the structural consequences of loading alterations, particularly since the relative loading and bone density asymmetries were directly associated. Although the relationship between dynamic knee loading and tibial BMD has previously been reported (6, 26), the observation of asymmetric BMDs reflecting asymmetric knee loading and asymmetric risk of progressive knee OA is novel. Interestingly, this is consistent with previous reports on OA of the hip, which suggested that increased hip BMD was associated with an increased risk of disease progression (27).

It should be noted that, while the knees of our study participants were explicitly asymptomatic, several were found to have radiographic evidence of OA. Because radiographic OA is almost universal in the age group studied, these subjects are likely more representative of a “normal” population than if those with radiographic OA had been completely excluded. It is for this reason that this study utilized a more clinically relevant definition of OA than one based purely on structural degeneration. In addition, it has previously been demonstrated that individuals who are asymptomatic, but have K/L grade 2 radiographic disease, are biomechanically indistinguishable from normal subjects who have no structural degeneration (28), whereas those who have K/L grade 2 radiographic knee OA and OA symptoms have significantly elevated peak adduction moments. This suggests that those with clinically evident OA are fundamentally different from those without clinical OA.

It should also be noted that increased loading at the contralateral knee and hip in this model of unilateral hip OA is in comparison to the ipsilateral side. It is not necessarily that the loading at the contralateral side is higher than that in a “normal” population. Nevertheless, the important concept is that this comparative asymmetry in loading precedes the corresponding asymmetric progression of knee OA in this group.

This study provides some evidence that in the contralateral knees of patients with unilateral hip OA, which are at high risk of developing progressive symptomatic OA, loading and structural asymmetries appear early in the disease course, while the knees are still asymptomatic. Thus, this model suggests that these early biomechanical asymmetries may have corresponding long-term consequences, providing further evidence of a role of loading in OA progression.


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 published. Dr. Shakoor 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. Shakoor, Thorp, Block.

Acquisition of data. Shakoor, Dua, Thorp, Mikolaitis, Wimmer, Foucher.

Analysis and interpretation of data. Shakoor, Dua, Thorp, Wimmer, Fogg.