SEARCH

SEARCH BY CITATION

Keywords:

  • Gait;
  • Dynamic joint load;
  • Measurement error;
  • Minimal detectable change

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Objective

To estimate the test–retest reliability of the peak external knee adduction moment during walking in patients with medial compartment knee osteoarthritis (OA), and to describe the interpretation of the reported values.

Methods

A total of 31 patients diagnosed with knee OA confined primarily to the medial compartment underwent quantitative gait analyses during 2 separate test sessions at least 24 hours apart and within 1 week. The peak knee adduction moment was calculated for each patient at each session based on the mean of 5 walking trials. Reliability was estimated using the intraclass correlation coefficient (ICC2,1) and the standard error of measurement (SEM).

Results

The mean difference in peak adduction moments between test sessions was 0.1% body weight × height (BW × ht; 95% confidence interval [95% CI] −0.1, 0.3). The point estimate for the ICC was 0.86 (95% CI 0.73, 0.96). The point estimate for the SEM was 0.36% BW × ht (95% CI 0.29, 0.48).

Conclusion

The ICC suggests that the peak knee adduction moment is appropriate for use when distinguishing among patients, for example, in studies of various interventions intended to decrease dynamic load on the knee medial compartment. The SEM illustrates the importance of considering measurement error and incorporating confidence levels when interpreting an individual patient's peak knee adduction moment value.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

The impact and burden of osteoarthritis (OA) on individuals and society are substantial and are projected to increase steadily over the next 2 decades (1–3). The knee is the most commonly affected weight-bearing joint, most frequently involving the medial compartment of the tibiofemoral joint (4, 5). The need to identify modifiable risk factors for the progression of knee OA and develop appropriate interventions has been emphasized (3, 6). The potential role of quantitative gait analysis for such purposes has been specifically highlighted (3, 7, 8). The peak external adduction moment about the knee during the stance phase of walking has a strong theoretical rationale for being an important gait variable (9–11), and has received particular attention in recent arthritis research literature.

Although external joint moments must be recognized as more general measures of joint load rather than representing specific anatomic structures (12), there is evidence to suggest that the peak knee adduction moment is a valid proxy for the dynamic load on the medial compartment of the tibiofemoral joint (13–16). Consistent with what is known about bone adaptations to increased load (17, 18), previous investigators have demonstrated significant positive correlations between the peak knee adduction moment and the ratio of medial to lateral bone mineral content of the proximal tibia in healthy subjects (r = 0.56) (13) and in patients with medial compartment knee OA (r = 0.30–0.52) (15, 16). Similarly, investigators have demonstrated a significant positive correlation between the peak adduction moment and magnetic resonance imaging measures of bone size of the medial tibial plateau in healthy subjects (r = 0.63) (14).

There is also evidence to suggest that the peak knee adduction moment has important clinical implications. Although not unequivocal (19, 20), several cross-sectional studies have indicated that the peak adduction moment is significantly greater in patients with medial compartment knee OA than healthy controls, and in patients with more severe OA than those with less severe OA (21–24). Similarly, studies have indicated that the peak adduction moment is significantly correlated with radiographic measures of disease severity (r = 0.30–0.68) (22, 25). Although fewer in number, longitudinal studies have suggested that the peak adduction moment predicts radiographic disease progression in patients with medial compartment OA (odds ratio 6.46) (26) and the development of future chronic knee pain in previously asymptomatic elderly individuals (27). Additionally, although results have been mixed, the prognostic value of preoperative peak adduction moments for patients undergoing high tibial osteotomy has also been suggested (28–31).

Despite growing support for the validity of the peak knee adduction moment as a tool in the assessment of patients with knee OA, little information describing its reliability exists. Although the reliability and interpretability of self-report and imaging measures continue to be well described in patients with arthritis (32–34), investigations of the measurement properties of performance-based measures have not paralleled that work (35, 36). Kadaba et al (37) and Andrews et al (38) have previously reported test–retest reliability of gait data from healthy individuals and have provided encouraging results. Further reliability data from patients with medial compartment knee OA are required, however, to more confidently plan intervention studies and interpret an individual patient's performance. Therefore, the objective of the present study was to estimate the test–retest reliability of the peak knee adduction moment during walking in patients with medial compartment knee OA, and to describe the interpretation of reported values.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Thirty-one consecutive patients were recruited from a cohort of patients being screened for participation in a prospective study evaluating medial opening wedge high tibial osteotomy surgery. Statistical justification for sample size was based on 2 test sessions and a parameter estimation of an intraclass correlation coefficient (ICC) of 0.85 with a 95% confidence interval (95% CI) width of 0.2 (39). Inclusion criteria included pain localized to the medial tibiofemoral joint, varus alignment (mechanical axis angle) of the lower limb (40), and diagnosis of knee OA based on criteria described by Altman et al (41). Patients with prior ligamentous injury were included. Kellgren and Lawrence grades of OA severity (42) were measured from double-limb standing hip-to-ankle anteroposterior radiographs obtained during the first test session. Patients also completed the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) during the first test session (43). WOMAC scores were transformed to scores out of 100, where higher values indicated less pain, less stiffness, and greater functional status. The study was approved by the institution's Research Ethics Board for Health Sciences Research Involving Human Subjects.

Gait testing procedures.

Patients were tested during 2 separate test sessions completed at least 24 hours apart and within 1 week. During each test session, 3-dimensional gait analyses were completed using an 8-camera motion capture system (Motion Analysis Corporation, Santa Rosa, CA) synchronized with a single floor-mounted force plate (Advanced Mechanical Technology, Watertown, MA). Passive-reflective markers were placed on the patient using a modified Helen Hayes marker set (44). A single static standing trial was completed where patients were asked to stand motionless on the force plate while wearing 4 additional markers placed on the medial knee joint line and medial malleolus bilaterally. This static trial was used to determine body mass and positions of joint centers of rotation for the knee and ankle. The 4 additional markers were removed prior to the assessment of walking trials. Patients then walked barefoot across the 8-meter laboratory floor at a self-selected velocity while kinetic (1,200 Hz) and kinematic (60 Hz) data were collected. Patients were instructed to walk at their normal pace and to ignore the force plate. Walking trials were repeated until 5 clean force plate strikes (initial contact to preswing; one foot completely on the plate) from each limb were obtained.

Data analysis.

Gait data were processed using commercially available software (Motion Analysis Corporation). Kinematic and kinetic data from each trial were used to calculate moments about the knee using inverse dynamic principles. Each lower limb segment (thigh, shank, and foot) was modeled as a rigid body with a local coordinate system that coincided with anatomic axes. Inertial properties of each limb segment were approximated and translations and rotations of each segment were calculated relative to marker orientations observed during the initial static standing trial. The peak (greatest magnitude) adduction moment values were identified from each of the knee adduction/abduction moment waveforms and were averaged across the 5 trials to obtain a single peak adduction moment value for each patient on each test session. Values were normalized to body size (percentage body weight [BW] × height [ht]).

Statistical analysis was confined to the planned operative limb to be consistent with the intended clinical use of the peak adduction moment values, and to avoid potential bias and spurious precision when including both limbs of patients (nonindependent observations) in a single analysis (45). First, the validity of the observed knee adduction moments was confirmed by evaluating the correlations between the peak adduction moment (mean of the 2 test sessions), the mechanical axis angle (Pearson's correlation coefficient), and the Kellgren and Lawrence grade of severity (Spearman's rho). It was hypothesized that the peak adduction moment would be directly related to these radiographic measures and the correlation coefficient would fall within the range of previously reported values (r = 0.30–0.68) (22, 25). Test–retest data were examined graphically by plotting the difference in peak knee adduction moment between tests against the mean of the 2 tests (46). Reliability was then evaluated by calculating the point estimates for the ICC (type 2,1) and the standard error of measurement (SEM) (47–49). The ICC provided an indication of how well the peak adduction moment distinguished among patients (relative reliability), whereas the SEM provided an expression of the measurement error in the peak adduction moment in its original units (absolute reliability). The point estimate of the SEM was then used to estimate the error in an individual patient's peak adduction moment at one point in time by multiplying the SEM by the Z values for various confidence levels. Estimated error at one point in time was then multiplied by the square root of 2 (to account for measurement error on 2 test sessions) to estimate of the minimal detectable change at various confidence levels.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Thirty-one patients (21 men, 10 women) completed both test sessions. The mean ± SD time between sessions was 3.4 ± 2.0 days. Patient characteristics are summarized in Table 1. The radiographic and WOMAC scores indicated that the present sample was similar to previous studies evaluating the adduction moment in patients with medial compartment knee OA. Twenty-eight patients had the typical double-hump knee adduction moment curve. The remaining 3 patients had 1 initial peak that tapered off. The curves were consistent over the 5 trials in all patients. The sample mean ± SD percentage of stride where the peak occurred was 22.9% ± 11.2% in the first test session and 23.0% ± 11.7% in the second test session. Correlations between the peak adduction moment and mechanical axis angle (r = 0.50; 95% CI 0.18, 0.73) and Kellgren and Lawrence grade (r = 0.49; 95% CI 0.16, 0.72) were consistent with our hypothesis, similar to previous reports, and suggested that the observed values were valid. The difference between tests plotted against the mean of the tests indicated no obvious relationship or bias, although it should be acknowledged that this sample had few patients with very low adduction moments as is typical of patients with medial compartment knee OA (Figure 1). The mean difference in peak adduction moments between test sessions was 0.1% BW × ht (95% CI −0.1, 0.3). The point estimate for the ICC was 0.86 (95% CI 0.73, 0.96). The point estimate for the SEM was 0.36% BW × ht (95% CI 0.29, 0.48). Estimates of the error associated with an individual patient's peak adduction moment at one point in time and the minimal detectable change upon reassessment are reported for various confidence levels in Table 2.

Table 1. Summary statistics*
 Value
  • *

    Values are the mean ± SD (range) unless otherwise indicated. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; BW = body weight; ht = height.

  • Mechanical axis angle.

Age, years44 ± 11 (21–65)
Height, meters1.74 ± 0.09 (1.55–1.88)
Mass, kg87.22 ± 20.02 (49.90–153.60)
Body mass index, kg/m229.02 ± 6.98 (20.79–58.17)
WOMAC, total score out of 10057.95 ± 22.10 (21.23–98.91)
Varus alignment, degrees6.27 ± 4.23 (1.10–15.30)
Peak knee adduction moment: test 1, %BW × ht2.54 ± 0.96 (0.67–4.16)
Peak knee adduction moment: test 2, %BW × ht2.45 ± 0.96 (0.73–4.22)
Kellgren and Lawrence grade, no. of patients 
 16
 27
 36
 412
thumbnail image

Figure 1. Scatterplot of the differences versus the means of the test and retest peak knee adduction moments. BW = body weight; ht = height.

Download figure to PowerPoint

Table 2. Estimates of the error associated with a patient's peak knee adduction moment at one point in time and the minimal detectable change upon reassessment at various confidence levels*
Confidence level, %Measurement error (± %BW × ht)Minimal detectable change (± %BW × ht)
  • *

    Estimates are based on the point estimate for the standard error of measurement (SEM) of 0.36

  • %BW × ht observed in the present sample. See Table 1 for definitions.

  • SEM × Z value.

  • SEM × Z value × equation image.

950.701.00
900.590.83
850.510.73
800.460.65
750.410.59
500.240.34

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

The present ICC of 0.86 can be described as indicating excellent test–retest reliability of peak knee adduction moment values in patients with medial compartment knee OA awaiting high tibial osteotomy. This reliability coefficient is the ratio of between-patient variability to total variability, where total variability represents the variability attributed to between-patient variability plus within-patient variability. Within-patient variability is the variability associated with measuring each patient over time. Therefore, a high test–retest reliability coefficient implies that the between-patient variability is high relative to the within-patient variability (i.e., there is relatively little variation in the measurements within a particular patient over time). The ICC is therefore considered a relative measure of reliability that provides an indication of how well a measure is capable of differentiating among the patients in whom the measurements are taken (48). An ICC of 0.86 suggests that the peak knee adduction moment is appropriate for use when distinguishing among patients, such as might be done in clinical trials of various interventions. This finding is consistent with previous reliability reports that have evaluated healthy subjects. Kadaba et al (37) tested 40 healthy subjects on separate days and reported that the knee adduction moment (abduction/adduction moment waveform expressed as a function of the gait cycle) was highly repeatable (coefficient of multiple correlation 0.9). Andrews et al (38) tested 11 healthy subjects on separate days and also reported that the knee adduction moment was highly reliable (mean difference in peak between test days 0.1% BW × ht).

In addition to evaluating reliability in patients with medial compartment knee OA awaiting high tibial osteotomy, the present findings extend previous investigations by reporting the SEM. The SEM facilitates the interpretation of an individual's peak knee adduction moment by enabling estimates of measurement error, such as those provided in Table 2, based on the present findings and using a variety of confidence levels. For example, based on the SEM and the 95% confidence level, an individual patient's peak adduction moment obtained from a single quantitative gait analysis session in our laboratory (mean of 5 trials on the limb of interest) could vary ±0.7% BW × ht simply due to measurement error. Accordingly, we can be 95% confident that a patient with a peak adduction moment of 3.0% BW × ht could have a true value varying from 2.3% BW × ht to 3.7% BW × ht. Narrower ranges of values can be achieved with reductions in confidence levels (Table 2). If the adduction moment is being considered as an additional assessment tool to help make clinical decisions, for example, by determining cutoff values proposed to identify patients at risk for disease progression (26) or poor outcome following an intervention (29), then estimates of measurement error in an individual's score should be considered.

The SEM also enables estimates of minimal detectable change in the peak knee adduction moment upon reassessment. For example, the minimal detectable changes presented in Table 2 suggest that the peak adduction moment in almost all (95%) stable patients, with similar characteristics to those described in Table 1, would change by less than 1% BW × ht upon reassessment in our laboratory, the vast majority (75%) of stable patients would change by less than 0.59% BW × ht, and only half (50%) of stable patients would change by less than 0.34% BW × ht. Caution should be adopted and estimates of minimal detectable change should be considered if the adduction moment is to be used to assess change in a patient's dynamic knee joint load over time, or following an intervention. For example, if a patient's knee adduction moment changed 0.75% BW × ht (perhaps it increased with the passage of time and progression of malalignment), we could be very confident that this represents a true change because it exceeds the minimal detectable change at a very high confidence level (i.e., 85%). In contrast, if a patient's peak adduction moment changed 0.3% BW × ht following an intervention (perhaps it decreased with use of a knee or foot orthosis, gait retraining, exercise, or surgery), we could not be confident that this represents a true change because it does not exceed the minimal detectable change at even a very low confidence level (i.e., 50%).

The present reliability estimates are only generalizable to patients with characteristics similar to the present study's participants. These patients were referred to this tertiary care center for consultation regarding surgical treatment options, and all were deemed eligible for medial opening wedge high tibial osteotomy. Although radiographic and self-reported data in Table 1 indicate that the present sample is comparable with most of the previously cited studies evaluating the knee adduction moment (16, 21, 23, 24, 26, 28–31), it should be pointed out that these samples include patients with rather severe OA, and that measurement properties of the adduction moment (such as test–retest reliability) may vary with disease characteristics. Similarly, the present findings are only generalizable to patients tested using comparable testing equipment and procedures. Even though modern motion capture systems and data reduction techniques are based on common, well-accepted engineering and physics principles, the variability of knee adduction moments measured in different laboratories is presently unclear. In addition to continued longitudinal research evaluating the clinical importance of the knee adduction moment, we encourage further studies evaluating the reliability within and between laboratories to help determine its potential for widespread clinical use.

In summary, the present results support the use and facilitate the interpretation of peak knee adduction moment values. Specifically, the estimate of relative reliability suggests that the peak knee adduction moment is appropriate for distinguishing among patients, for example, in studies of various interventions intended to decrease dynamic load on the knee medial compartment. The estimates of absolute reliability, however, illustrate the importance of considering measurement error and incorporating confidence levels when interpreting an individual patient's peak knee adduction moment value.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Dr. Birmingham 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. Birmingham, Hunt, Jones, Jenkyn, Giffin.

Acquisition of data. Hunt, Jones.

Analysis and interpretation of data. Birmingham, Hunt, Jones, Jenkyn, Griffin.

Manuscript preparation. Birmingham, Hunt, Jones, Jenkyn, Giffin.

Statistical analysis. Birmingham.

ROLE OF THE STUDY SPONSOR

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Arthrex, Inc. contributed funds to a University- Industry research grant provided by the Canadian Institutes of Health Research. Arthrex, Inc. had no role in study design, data collection, data analysis, or writing of the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES