The relationship between knee pain and radiographic evidence of knee osteoarthritis (OA) is notoriously imperfect. In particular, conditions that distinguish individuals with symptoms from those with comparable radiographic involvement who remain asymptomatic are unclear. We investigated dynamic loading across the knee in individuals with mild radiographic OA who were distinguished by the presence or absence of knee pain.
Subjects were recruited into 3 groups: symptomatic with a Kellgren/Lawrence (K/L) grade of 2 (n = 52), asymptomatic with a K/L grade of 2 (n = 19), and asymptomatic with a K/L grade of 0 or 1 (n = 37), the latter representing a normal comparator group. Dynamic knee loading was assessed with gait analysis, and both the peak external knee adduction moment and the knee adduction angular impulse were determined.
Peak knee adduction moment and knee adduction angular impulse were 19% and 30% higher, respectively, in symptomatic K/L grade 2 individuals than in asymptomatic individuals with the same radiographic grade (P < 0.05). Conversely, the asymptomatic K/L grade 2 group did not differ from the K/L grade 0–1 normal comparator group (P = 1.00).
Among individuals with mild radiographic knee OA (K/L grade 2), those who are symptomatic have significantly higher medial compartment loads than those who are asymptomatic, whereas those who are asymptomatic do not differ from normal controls (asymptomatic K/L grade 0 or 1). These findings suggest a biomechanical component to the distinction between asymptomatic and symptomatic radiographic OA, which may be pathophysiologically important.
Clinically, knee osteoarthritis (OA) is defined by the presence of both radiographic evidence of degenerative joint changes and joint pain (1). The relationship between the severity of radiographic joint degeneration, as assessed by the Kellgren/Lawrence (K/L) radiographic OA scale (2), and knee pain has been extensively examined (3–6). Although in general the prevalence and severity of symptoms are increased when groups with moderate (K/L grade 3) or severe (K/L grade 4) radiographic disease are compared with those with mild radiographic OA (K/L grade 2) (4, 7), the presence of mild radiographic disease alone is a poor predictor of painful symptoms (3–6). Moreover, other than joint pain itself, the characteristics that distinguish individuals with radiographic knee OA who are symptomatic from those with comparable degrees of radiographic OA severity who nonetheless remain asymptomatic are unknown.
It has become clear that knee OA is at least partly mediated through altered loading during gait, and one well-described method to assess such dynamic loading is to determine the external knee adduction moment through gait analysis. The external knee adduction moment is a quantitative measure of the varus torque generated at the knee during the majority of the stance phase of gait, and is the primary predictor of the distribution of load across the tibial plateau (8). Thus, the external knee adduction moment is a standard surrogate marker of medial compartment knee joint loads; as such, it correlates closely with bone mineral density distribution in the proximal tibia in OA (9), radiographic disease severity as determined by K/L grade (10–12), and radiographic progression of knee OA (13).
Individuals with symptomatic knee OA have been shown to exert abnormally high loads across their knees during gait, as reflected in elevated peak external adduction moments (14). Relief of knee pain in persons with OA results in further increased loading of the medial knee (15, 16), presumably through release of pain-induced gait adaptations. Additionally, in women without definite evidence of radiographic disease, knee pain has been positively correlated with the magnitude of the knee adduction moment (17). These observations suggest that loading across the knees may be intimately associated with symptoms. The present study was designed to test the hypothesis that joint loading varies as a function of the presence or absence of knee pain in subjects with mild radiographic knee OA (K/L grade 2). To this end, external adduction moments were compared among individuals with equivalent degrees of radiographic OA (K/L grade 2) who were stratified based on the presence or absence of significant painful symptoms. An additional comparator group of asymptomatic subjects without radiographic OA was evaluated as a further control.
SUBJECTS AND METHODS
All subjects were enrolled after providing informed consent and with the approval of the Rush Institutional Review Board. Subjects were recruited into 1 of 3 groups: group 1 included symptomatic subjects with radiographically mild OA (K/L grade 2), group 2 included asymptomatic subjects with radiographically mild OA (K/L grade 2), and group 3 included asymptomatic subjects without radiographic OA (K/L grades 0 or 1).
The subjects in group 1 were recruited as part of a larger clinical trial studying potential treatments for OA. All data from the present work corresponded to baseline assessments taken prior to the initiation of any therapeutic interventions. Subjects from groups 2 and 3 were recruited separately to draw comparisons between subjects with OA and healthy asymptomatic subjects of the same age. Recruitment proceeded at the clinical practices of the Section of Rheumatology at Rush University Medical Center and by pamphlets and “word of mouth” in the community.
In each case, the presence of radiographic OA was defined based on the K/L grading scale as modified by Felson et al (2, 18). Subjects were excluded if either knee had a K/L grade greater than 2. Knee pain was assessed using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (19) to identify the presence of clinically symptomatic disease. Subjects were considered clinically symptomatic if they reported >30 mm on a maximum 100-mm visual analog scale of lower extremity pain (question 1 of the WOMAC pain subscale) and met the American College of Rheumatology criteria for symptomatic OA of the knee (1, 20). All clinically symptomatic subjects (group 1) exhibited bilateral knee pain and definite evidence of mild radiographic disease, defined as a K/L grade 2 in the medial compartment of their index knee. The index knee was defined as the more painful knee during level walking at the time of the evaluation and was utilized for comparisons in this study, as has been suggested by other investigators (13). All symptomatic subjects had a K/L grade of at least 1 but not higher than 2 in their nonindex knee. In addition, symptomatic subjects with K/L grade 2 were excluded if they had any prior knee or hip arthroplasty, or arthroscopy within 3 months or fracture of either lower extremity within 6 months.
Subjects were included in the 2 asymptomatic groups (groups 2 and 3) if they reported <10 mm of lower extremity pain bilaterally during level walking (question 1 of the WOMAC) and if they had not previously had subjective reports of pain in either knee. Asymptomatic subjects were further subdivided in the following manner: asymptomatic subjects with a K/L grade of 2 comprised one group (group 2), and asymptomatic subjects who had K/L grades of 0 or 1 bilaterally were included as a separate comparator group (group 3) because these subjects lacked any definite evidence of radiographic disease. All 19 subjects in the asymptomatic K/L grade 2 group (group 2) exhibited a K/L grade of 2 bilaterally at the knee. Subjects were excluded from groups 2 and 3 if they reported any history of hip, knee, or ankle surgery or fracture involving the lower extremity.
No subjects who exhibited the presence of clinically evident hip, ankle, or foot OA; significant foot disease; or the presence of any inflammatory arthritis were included in any of the 3 groups. Subjects who met the inclusion criteria and were enrolled in the study underwent gait analyses on the same day as their radiographic evaluation and assessment of symptoms via the WOMAC.
All radiographs were obtained at the same institution by the same technologist. Standard weight-bearing anteroposterior radiography of the knees in full extension was performed, and each film was graded by a single experienced and blinded observer (JAB) using the K/L grading scale as modified by Felson et al (2, 18). Briefly, radiographic OA grades were assigned based on the following criteria: grade 0 = no evidence of osteophytes or joint space narrowing (JSN); grade 1 = questionable, but no definite osteophytes or JSN; grade 2 = definite osteophytes, with or without possible JSN, or definite mild JSN with or without osteophytes; grade 3 = definite moderate JSN (at least 50%), cysts and sclerosis may be present, and osteophytes are usually present; grade 4 = severe JSN, definite osteophytes, deformity, cysts, or sclerosis (2, 18). The overall single-measure intraclass correlation coefficient (ICC) was 0.863 (95% confidence interval 0.792–0.911) for measurements repeated at 1 week.
Knee alignment was determined using Image J software (National Institutes of Health, Bethesda, MD; URL: http://rsb.info.nih.gov/ij/). The angle formed by the intersection of a line passing through the center of the femoral notch traversing the shaft of the femur with a line passing through the center of the tibial spine and traversing the shaft of the tibia was measured for all 108 subjects using standard weight-bearing anteroposterior radiographs of the knee (21). The overall single-measure ICC was 0.9822 (95% confidence interval 0.9567–0.9927) for measurements repeated at 1 month.
Three-dimensional kinematics and ground reaction forces were collected using 4 Qualisys optoelectronic cameras (Innovision Systems, Columbiaville, MI) with passive markers and a multicomponent force plate with a sampling frequency of 120 Hz (Bertec, Columbus, OH). Passive markers were placed at the following anatomic landmarks: the most lateral point 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. The joint centers of the hip, knee, and ankle were approximated following previously published methods for the calculation of external joint moments (22–24): 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, and 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. Subjects wore their own low-top walking shoes and were instructed to walk at a self-selected normal speed. Three walking trials were conducted unilaterally for the index knee of symptomatic subjects and bilaterally for the asymptomatic subjects. Kinematic and kinetic calculations were performed using software developed by Computerized Functional Testing Corporation (Chicago, IL) and included frontal plane external joint moments at the knee (24). Three-dimensional external moments and intersegmental joint forces were calculated through inverse dynamics with the lower extremity modeled as a linkage of 3 rigid bodies representing the thigh, shank, and foot assuming no deformation occurs within these 3 segments and no rotation occurs around the long axis of each segment (25). A lump mass approximation was utilized. The external knee adduction moment was calculated using the ground reaction force as measured by the force plate, the inertial properties of the limb, and the mass of the limb (25). One walking trial at the self-selected normal walking speed was analyzed for the index knee of the symptomatic subjects and for both the right and left lower extremities of the asymptomatic subjects. The median walking speed of the 3 trials was selected for analysis.
The overall peak knee adduction moment, which occurred for all subjects during midstance, was determined and expressed in standard normalized units (percent body weight times height [%BW*Ht]). In addition to the peak moment, the knee adduction angular impulse, the time integral of all the frontal plane joint moments, was also calculated for the midstance subdivision of the gait cycle (i.e., from 17% to 50% of stance time) as defined by Perry (26). Recently, we found that knee adduction angular impulse is a useful gait parameter to examine in addition to the peak knee adduction moment, because it is a more sensitive predictor of radiographic disease severity (12) and because it explains more variance in the medial-to-lateral proximal tibial bone mineral ratio (9) than the peak knee adduction moment. The peak knee adduction moment and the knee adduction angular impulse were normalized to body size using the body weight times height method (27).
A chi-square analysis was used to determine if there was a significant difference in the sex distribution among the 3 study groups. Data for both joint loading parameters were determined to be normally distributed. Because no side-to-side differences for any of the loading parameters or the knee alignment angle were observed in the asymptomatic subjects (paired t-tests), a limb was randomly selected using a random number generator for each asymptomatic subject and was compared with the index knee of the symptomatic subjects. A one-way analysis of variance with Bonferroni post hoc tests was used to examine differences in age, body mass index (BMI), gait speed, alignment, and joint loading parameters among the 3 groups. Individual backward elimination regression analyses with each of the loading parameters as the dependent variable and BMI, age, sex, alignment, and gait speed as independent variables were utilized to examine sources of variance in the loading parameters, other than knee pain, for the subjects with K/L grade 2.
The subjects included 28 men and 80 women with a mean ± SD age of 55.2 ± 8.47 years. There were no differences in sex distribution or age among the 3 groups (Table 1). The symptomatic K/L grade 2 group and the asymptomatic K/L grade 2 group did not differ in age, BMI, or gait speed; however, the symptomatic K/L grade 2 group had more varus alignment than the asymptomatic K/L grade 2 group (Table 1). There were no differences between the asymptomatic K/L grade 2 group and the asymptomatic K/L grade 0–1 group with regard to age, BMI, gait speed, or alignment (Table 1). The symptomatic K/L grade 2 subjects had higher BMI, walked slower, and had more varus alignment than the asymptomatic K/L grade 0–1 group (Table 1).
Values are the mean ± SD unless otherwise indicated. K/L = Kellgren/Lawrence; BMI = body mass index.
A chi-square analysis was used to determine if the sex distribution differed among the 3 groups. One-way analyses of variance (ANOVAs) were used to determine overall group effects for the other variables.
Significant differences between asymptomatic grade 0 or 1 and symptomatic grade 2 (one-way ANOVA Bonferroni post hoc test, P < 0.05).
Negative values indicate valgus knees.
Significant differences between asymptomatic grade 0 or 1 and symptomatic grade 2 (one-way ANOVA Bonferroni post hoc test, P < 0.05) and significant differences between symptomatic and asymptomatic grade 2 (one-way ANOVA Bonferroni post hoc test, P < 0.05).
Significant differences between symptomatic and asymptomatic grade 2 (one-way ANOVA Bonferroni post hoc test, P < 0.05).
The symptomatic K/L grade 2 group had a higher midstance peak knee adduction moment and a higher knee adduction angular impulse (mean ± SD 2.7 ± 0.71 %BW*Ht and 0.51 ± 0.16 %BW*Ht*seconds, respectively) than the asymptomatic K/L grade 2 group (mean ± SD 2.2 ± 0.72 %BW*Ht and 0.37 ± 0.14 %BW*Ht*seconds, respectively; P < 0.05) (Figure 1). The asymptomatic K/L grade 0–1 group also had a lower peak moment (mean ± SD 2.2 ± 0.57 %BW*Ht) and a lower knee adduction angular impulse (mean ± SD 0.35 ± 0.1 %BW*Ht*seconds) than the symptomatic K/L grade 2 group (P < 0.05). No differences in either the peak knee adduction moment or the knee adduction angular impulse were observed when the asymptomatic K/L grade 2 group was compared with the asymptomatic K/L grade 0–1 group (P = 1.00 for each of the loading parameters, corrected P value from Bonferroni post hoc test) (Figure 1). Therefore, both asymptomatic groups had peak knee adduction moments and knee adduction angular impulses that were 19% and 30% lower, respectively, than the symptomatic group.
Regression models examining variance in the knee adduction moment for the K/L grade 2 subjects.
To examine the potential factors, other than knee pain, that might contribute to variance in the medial compartment loading in all subjects with K/L grade 2 knees, 2 separate backward elimination regression models were utilized with the peak knee adduction moment and the knee adduction angular impulse each considered separately as dependent variables. Independent variables in the regression analyses were age, BMI, sex, gait speed, and alignment. For the peak knee adduction moment, only alignment (β = 0.521, P < 0.001) exhibited explanatory power in the model (adjusted r2 = 0.261, P < 0.001). For the knee adduction angular impulse in midstance, both alignment (β = 0.487, P < 0.001) and gait speed (β = −0.409, P < 0.001) exhibited explanatory power in the model (adjusted r2 = 0.404, P < 0.001).
Radiographic evidence of OA is common, yet the relationship between pain and degenerative structural changes in joints is poorly understood. In an attempt to characterize mechanical differences between clinically evident OA (i.e., both symptoms and radiographic findings) and asymptomatic radiographic OA, the present study compared subjects who had a similar degree of radiographic OA but who differed by the presence or absence of symptomatic knee pain. The peak knee adduction moment and the knee adduction angular impulse were both significantly higher in the symptomatic K/L grade 2 group than in the asymptomatic K/L grade 2 group. Analyses of lower extremity alignment revealed that symptomatic subjects also had significantly more varus alignment at the knee than the asymptomatic subjects with the same radiographic grade. These findings suggest that mechanical factors, such as loading across the knee, may be intimately associated with symptoms in individuals who have structural (radiographic) evidence of degeneration at the knee.
Interestingly, the magnitude of the knee adduction moment did not differ between the asymptomatic K/L grade 2 group and the comparator group of asymptomatic K/L grade 0–1 subjects even though the former group had definite radiographic evidence of OA. A power analysis using the uncorrected P value from an independent-sample t-test comparing these 2 groups was performed (28) and demonstrated that the power of our study was 80% to detect a 20% difference (P < 0.05) in gait parameters, suggesting that the lack of difference was not a result of a Type II statistical error. A difference of 20% is meaningful because it is approximately 2–3 times the average coefficient of variation between consecutive gait trials of the subjects in this study population. Therefore, it is reasonable to conclude that the knee adduction moments for these groups are comparable.
In addition to differences in moment patterns, the symptomatic K/L grade 2 group also had more varus alignment than the asymptomatic K/L grade 2 group. One limitation of the present work is that alignment measurements from full-limb radiographs were not available for all subjects. The method of measuring alignment in the present work has previously been observed to yield alignment measurements that are correlated with those from full-limb radiographs, but on average have a 3.4° valgus bias (29). To express alignment in terms of full-limb radiographs, thereby eliminating the valgus bias found previously (29), we correlated alignment as measured from full-limb radiographs (following the method described by Goker and Block ) with alignment measured from standard films for the subjects with both sets of radiographs (all 52 symptomatic subjects). We found that full-limb alignment = 0.73 × knee alignment from standard radiographs + 4.2° (r = 0.54, P < 0.05). Alignment explained variance in the knee adduction moment for the K/L grade 2 subjects, which is consistent with existing literature (30, 31).
The symptomatic K/L grade 2 group tended to walk slower than the asymptomatic K/L grade 2 group, although this difference was not statistically significant. These findings are consistent with other literature suggesting that individuals with knee OA ambulate at self-selected speeds that are slower than normal (32). However, gait speed did not explain variance in the peak knee adduction moment, in accordance with existing literature (25). Although gait speed has been observed to affect sagittal plane external joint moments (i.e., flexion and extension moments), gait speed has not been reported to affect the magnitude of external joint moments in the frontal plane (25). Gait speed did explain some of the variance in the knee adduction angular impulse, consistent with the fact that this parameter represents the time integral of the moment (12). Therefore, despite the trend toward a difference in gait speed between the groups, it is unlikely that gait speed affected the magnitude of the frontal plane moments.
Prior to the present work, distinctions in loading and gait patterns based on the presence or absence of pain symptoms have not been studied in mild radiographic OA. However, both obesity (6) and BMI (3), 2 factors that may be mechanically important, have been reported to be associated with the likelihood of reporting knee pain in the presence of definite radiographic disease. These findings are controversial because other studies have not observed such links (4, 33). Sex has also been associated with symptomatic knee OA, with women being more likely to report pain in the presence of radiographic findings than men (3, 4). The groups studied here did not differ in BMI or sex distribution, indicating that these potential confounding variables did not complicate the comparison of medial compartment loads in the present study. In addition, neither BMI nor sex had any explanatory power in the regression models explaining variance in the peak adduction moment or angular impulse. These latter findings are consistent with the work by Moisio et al, who demonstrated that normalization by the percent body weight times height method (as used in the present study) largely negates the effect of body size and sex on the knee adduction moment (27).
The relationship between the knee adduction moment and knee pain has previously been examined and the results of the present work support these previous findings. Two studies demonstrated that knee pain is inversely correlated with the knee adduction moment and that pharmacologic relief of knee pain in persons with OA results in increased loading of the medial knee (15, 16). However, this inverse association was found in symptomatic subjects only. The results of the present study, therefore, are not contrary to findings of the previous work, but suggest that lower medial compartment loads may delay the onset of knee pain in subjects with mild radiographic evidence of knee OA (K/L grade 2). This interpretation is consistent with a longitudinal study that observed that higher knee adduction moments increased the odds of developing chronic knee pain over time (34). Another recent work observed a significant positive correlation between knee pain and the magnitude of the knee adduction moment in women who lacked definite evidence of knee OA (17). Similarly, in the present work, we also observed a positive association between knee pain and the magnitude of the external knee adduction moment (i.e., subjects with pain had higher moments) in subjects with radiographic OA of equal severity (K/L grade 2).
Although the cross-sectional design of this study is a limitation, we believe that this work is important for future OA research for several reasons. First, our findings suggest that when examining gait patterns of individuals with radiographic OA, factors in addition to K/L grade, including symptoms and knee alignment, should be utilized to stratify the population before comparisons can be made. Second, based on previous work suggesting that the magnitude of the knee adduction moment is linked to disease progression (13) and on the present finding of higher medial compartment loading in the symptomatic K/L grade 2 group than in the asymptomatic K/L grade 2 group, we speculate that symptomatic K/L grade 2 subjects may be more likely to progress than asymptomatic K/L grade 2 subjects. Finally, the observation that the presence of pain is itself associated with altered medial compartment joint loading in subjects with mild radiographic evidence of OA provides the first clear identification of a biomechanical distinction between groups of individuals who differ in symptoms despite exhibiting comparable radiographic OA grades.
Dr. Block 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 design. Thorp, Sumner, Wimmer, Block.
Acquisition of data. Thorp, Wimmer, Block.
Analysis and interpretation of data. Thorp, Sumner, Wimmer, Block.
Manuscript preparation. Thorp, Wimmer, Block.
Statistical analysis. Thorp.
The authors would like to acknowledge Dr. Susan Shott for her assistance with statistical analyses and data interpretation.