To estimate the inter- and intrarater reliability and agreement of instrumented knee joint proprioception measurement in subjects with knee osteoarthritis (OA) and healthy subjects; to assess the effect of variations in the measurement procedure on agreement parameters.
Proprioception was measured by a computer-controlled knee angular motion-detecting device in a movement-detecting task. The angular displacement between the starting position and the position at the instant of movement detection by the subject was recorded. Two raters independently assessed knee joint proprioception. After 14 days the assessment was repeated. Complete data were obtained from 24 subjects with knee OA and 26 healthy subjects. The inter- and intrarater reliability coefficients (intraclass correlation coefficients [ICC]) and inter- and intrarater agreement measures (standard error of measurement [SEM] and minimal detectable difference [MDD]) were calculated. Additionally, the effect of changing the velocity of angular displacement and applying headphone music during the measurement on the absolute error (i.e., SEM and MDD) was estimated at the second occasion.
Interrater reliability was good in subjects with knee OA and healthy subjects (ICC 0.91 and 0.89, respectively). Interrater agreement was higher in subjects with knee OA than in healthy subjects (SEM 2.13° versus 0.43°, MDD 5.90° versus 1.19°). Intrarater reliability was good in subjects with knee OA and healthy subjects (ICC 0.91 and 0.86, respectively). Intrarater agreement (SEM and MDD) was 2.26° and 6.26° in subjects with knee OA and 0.39° and 1.08° in healthy subjects. The original measurement and the 2 variations in measurement showed comparable measurement errors for subjects with knee OA and healthy subjects.
In knee OA subjects and healthy subjects, knee proprioception measurement shows adequate intra- and interreliability. However, the absolute measurement error is rather high. Therefore, this measurement has limited value in the assessment of individual subjects, but can be recommended for scientific research in groups of individuals.
Knee osteoarthritis (OA) is a leading cause of limitations in daily functioning in the elderly (1). Inaccurate proprioception has been suggested to be a risk factor for the development of limitations in function in patients with knee OA (1–3). Proprioception can be defined as the conscious and unconscious perception of joint movement and joint position (4–6). Proprioception is decreased in patients with knee OA compared with elderly controls (7–12). Although many studies have measured proprioception in patients with knee OA (1, 7, 8, 10–27), information on the reproducibility of the methods used to assess proprioception is rarely provided.
Reproducibility concerns the degree to which repeated measurements of a stable characteristic provide similar results. For the quantification of reproducibility, 2 types of measures can be distinguished: reliability and agreement (28–30). Reliability parameters assess whether persons in a group can be distinguished from each other, despite measurement errors (28). Reliability is expressed as the intraclass correlation coefficient (ICC). Agreement parameters assess how close the results of the measurements are within individuals by estimating the absolute measurement error in repeated measurements (29, 30).
Adequate reliability and agreement indicate that a measurement is appropriate for use both in scientific research to describe characteristics in groups of patients and in clinical practice to adequately assess individual patients. However, when agreement is lower (i.e., considerable measurement error is present), the assessment can still be sufficient for use in groups of patients, but may be too imprecise to adequately define the individual patient's level of proprioceptive accuracy. Therefore, knowledge of the reproducibility of proprioception measures is needed to establish the utility of these measures in scientific research and clinical practice. Although information has been presented concerning the reliability parameters of joint proprioception measurement (13, 15), information concerning the agreement parameters is as yet unavailable.
The goal of this study was to estimate the inter- and intrarater reliability and the inter- and intrarater agreement of instrumented knee joint proprioception measurement in subjects with knee OA and in healthy subjects. An additional goal was to assess the effect of variations in the measurement procedure on agreement parameters.
SUBJECTS AND METHODS
Two measurement sessions were carried out within a time frame of 2 weeks. Two raters (both physical therapists and trained to perform the proprioception measurement) independently performed the measurements. Both raters were blinded for the outcome of all other measurements. Rater 1 and rater 2 measured proprioception at day 1. At day 14, rater 1 repeated the measurement. Additionally, at day 14 rater 1 performed the measurement using 2 different protocols to assess the impact of protocol variations on the measurement agreement. Measurements were performed both in subjects with knee OA and in healthy subjects.
Subjects with OA of the knee were recruited in an outpatient rheumatology and rehabilitation clinic in the Netherlands. The inclusion criterion was OA diagnosed according to the clinical American College of Rheumatology criteria (31). These criteria include pain and a minimum of 3 of the following criteria: age >50 years, morning stiffness ≤30 minutes, crepitus on active movement of the knee joint, palpable or visible bony enlargement, bony tenderness at the knee joint margins, and no palpable warmth of synovium. The exclusion criteria were as follows: presence of prosthesis at the lower extremity, steroid injection within 2 months prior to inclusion, presence of neurologic disorders (e.g., stroke, Parkinson's disease, or poliomyelitis), presence of other rheumatoid or orthopedic disorders, recent (<1 year) history of a lower extremity fracture, history of ligament deficiency, insufficient control of the Dutch language, and hearing problems.
Healthy subjects were recruited from a student population of an allied health faculty. The exclusion criteria were presence, or history, of a severe injury of the lower extremity; a history of knee surgical procedure, or waiting list for knee operation; the presence of any neurologic, rheumatoid, or orthopedic disorders; and insufficient control of the Dutch language. Ethical review board approval from the Slotervaart Hospital in Amsterdam was obtained, and all participants provided written informed consent.
To assess proprioception (i.e., the threshold of detection of passive motion) of the knee, a device was designed following the recommendations of Sharma (5) and Pai et al (12). The device consisted of a chair with a computer-controlled motor and transmission system and 2 attached free-moving arms (Figure 1). Each arm supported the subject's shank and foot and moved in the sagittal plane. The joint of each arm was moved by a computer-controlled stepper motor and transmission system for angular displacement. The foot/ankle was attached with an air splint to the footrest, which was a moving component of the apparatus (32). Angular motion was detected by angular displacement and force transducers. Attached to the chair was an upward-bending tray to prevent visual input of the moving knee. Two handheld buttons were attached to the tray. The seat of the chair consisted of a gel pad to prevent any vibrating sensation and movement of the skin. This device provides a measurement of angular displacement, while eliminating or minimizing visual and auditory stimuli, vibrations, cutaneous tension, and pressure cues to limb motion.
Subjects were seated in a semi-reclining position with the back supported and the knee hanging over the edge of the apparatus, which is 5 cm proximal to the popliteal fossa. The knees were placed in 90° flexion and the hips in 70° flexion.
Measurement of knee joint proprioception.
The measurement procedure consisted of a knee joint movement detection task. Standard instructions were given to each participant. Each time, the leg was moved to a starting position of 30° knee flexion. Upon reaching this position, movement stopped. Following a random delay, the knee was then extended further with an angular velocity of 0.3°/second. Participants were instructed to push a handheld button at the moment of definite detection of knee joint position change. The angular displacement between the starting position at 30° flexion and the position in the extension direction at the instance when the button was pushed was recorded as the measure of knee joint proprioception (32). This means that a low value (i.e., a small difference between the knee joint angle at onset of movement and the knee joint angle at the moment of detection of knee joint position change) indicates good proprioception.
The participants were asked to put on short pants and take off shoes and socks. After attaching the foot/ankle to the footrest with an air splint, participants were told that the test could begin. A training session was started after standardized instruction was given. The participants were told: “Both legs will be moved to the start position, when both legs stop the test begins. At that exact time the rater will say ‘yes’ and point a finger in the air. After a random time period, movement of one of the legs will occur. When you feel movement in the knee joint please push the handheld button, corresponding to the side of knee movement.” The participants were asked to concentrate during the entire measurement. Then the measurement was started. When a detection mistake was made, the test was repeated. Measurements were performed 3 times per leg (i.e., 6 times per participant) by both rater 1 and rater 2 on day 1. The order of the 6 repetitions was randomized to ensure that participants would not know beforehand which of the 2 legs would be tested in a specific measurement. At day 14, rater 1 performed the same measurement. The average of the 3 measurements per leg per occasion (day 1 or day 14) of the standard procedure (in degrees) was used to estimate the reproducibility (comprising both reliability and agreement) of this measurement of knee joint proprioception between raters (interrater reproducibility) and occasions (intrarater reproducibility). A detailed description of these analyses is provided below.
Additionally, at day 14, 2 variations in the measurement of knee joint proprioception were performed by the same rater. In the first variation, the angular velocity was reduced from 0.3°/second to 0.1°/second. In the second variation, participants listened to music through headphones during the measurement to eliminate any remaining auditory input related to the onset of knee joint movement (i.e., the sound of the device's stepper motors starting). Both variations were also performed 3 times per leg, and the order in which participants' left and right legs were tested was randomized. The 3 separate measurements per leg of the standard procedure at day 14 and the 2 variations in measurement were used to calculate the within-session agreement parameters at day 14 (see statistical analyses below).
For all analyses, the following sources of variance were used: participant, rater, time of measurement, knee, and interaction between these variables. To express reproducibility between raters, interrater reliability and interrater agreement were estimated (30). To express reproducibility between 2 occasions (day 1 and 14) of the proprioception measurement by rater 1, intrarater reliability and intrarater agreement were estimated (30).
Interrater reliability and agreement.
The ICC(2,1) was calculated as the ratio of variance between participants and between the 2 raters and total variance. The standard error of measurement (SEM) was calculated by taking the square root of the error variance consisting of the following sources of variance: participants; rater; knee; interaction between participant and knee; interaction between participant and rater; and interaction between participant, rater, and knee. The SEM was used to calculate the minimal detectable difference (MDD). To compute the MDD as the 95% confidence interval limits of the SEM, the SEM has to be multiplied by 1.96 (for the 95% interval) and by the square root of 2 for the difference scores (1.96 × √2 × SEM) (33, 34).
Intrarater reliability and agreement.
The ICC(2,1) was calculated as the ratio of variance between participants within one rater and total variance. The SEM was calculated by taking the square root of the error variance of the following sources of variance: participant; time of measurement; knee; interaction between participant and knee; interaction between participant and time of measurement; and interaction between participant, knee, and time of measurement. The SEM was calculated across both occasions (35).
Impact of variations in measurement on intrarater agreement.
In addition, at day 14 the within-session SEM and MDD of the 3 repeated measurements were calculated for the original measurement and the 2 variations in measurement performed by rater 1, taking into account the following sources of variance: participant, knee, and the interaction between participant and knee. For reliability, an ICC >0.70 was regarded as adequate (36). To calculate the ICC, SEM, and MDD, a two-way random effects model of analysis of variance was performed, using SPSS software for Windows, version 12.0.1 (SPSS, Chicago, IL).
A total of 24 (8 men, 16 women) subjects with knee OA participated in the study. Mean ± SD age was 61.3 ± 9.8 years, weight was 84.5 ± 17.9 kg, height was 1.68 ± 0.09 meters, and body mass index (BMI) was 30.2 ± 7.1 kg/m2. A total of 26 (10 men, 16 women) healthy subjects participated in this study. Mean ± SD age was 20.6 ± 3.1 years, weight was 69.4 ± 12.3 kg, height was 1.75 ± 0.08 meters, and BMI was 22.5 ± 2.9 kg/m2.
Mean ± SD values for the proprioception measurement, generalized over the 2 raters and the 2 occasions, were 8.88° ± 6.82° for subjects with knee OA and 1.87° ± 1.24° for healthy subjects. To assess reproducibility parameters, the mean of the 3 repeated measurements per leg per session was used. The within-session correlations at day 1 for rater 1 and rater 2 were 0.821 and 0.876, respectively.
The inter- and intrarater reliability (as expressed by the ICC) and agreement (as expressed by the SEM and MDD) are presented in Table 1. Reliability was high in both subjects with knee OA and healthy subjects. Intra- and interrater reliability were comparable with each other. Likewise, intra- and interrater agreement were comparable.
Table 1. Inter- and intrarater reliability and agreement of proprioception measurement in subjects with knee OA and healthy subjects*
Subjects with knee OA
ICC (95% CI)
ICC (95% CI)
OA = osteoarthritis; ICC = intraclass correlation coefficient; 95% CI = 95% confidence interval; SEM = standard error of measurement; MDD = minimal detectable difference.
The within-session intrarater agreement as expressed by the SEM and MDD at the second session are presented in Table 2. The differences in SEM and MDD between the 2 measurement variations compared with the original measurement were minimal.
Table 2. Within-session intrarater agreement of the original proprioception measurement and the 2 variations of proprioception measurements in subjects with knee OA and healthy subjects*
The goal of this study was to estimate the inter- and intrarater reliability and inter- and intrarater agreement of the instrumented knee joint proprioception measurement in subjects with knee OA and in healthy subjects. An additional goal was to explore the effects of a change in angular velocity and the addition of headphone music on agreement coefficients.
Reliability was found to be adequate both within and between raters, for both subjects with OA and healthy subjects. Reliability estimates were almost equal in subjects with OA and healthy subjects. However, the slightly higher reliability observed in OA subjects can be explained by a larger variance in measurement results and therefore higher ICCs (5, 10).
In healthy subjects, inter- and intrarater agreement parameters were better than in subjects with OA, indicating a lower measurement error for the procedure in healthy subjects than in OA subjects. Measurement error for healthy subjects was 0.4°, whereas it was 2.2° in subjects with OA. This finding suggests that in subjects with OA, within-person variability has a considerable impact on the level of agreement in the assessment of proprioception. Due to a decrease in proprioceptive accuracy, OA subjects may be less likely to detect repeated knee joint position change at the same degree of angular deviation. In addition to OA subject variance, the level of agreement in the assessment of proprioception is also influenced by intrarater variance. Therefore, both within- and between-subject differences in proprioception must be interpreted with caution in subjects with OA. Even a considerable difference in result between 2 measurements may not be indicative of a genuine difference in proprioceptive accuracy, but instead is likely to be an expression of general proprioceptive inaccuracy. This is also reflected by the rather large MDDs (>4°) found for the population of OA subjects.
In addition to subject and rater variance, other sources of error could have been responsible for variation in outcome. One source of error could have been the fixation of the foot/ankle of subjects. Small differences in the positioning and fixation of the leg between the 2 raters and between the 2 sessions (day 1 and day 14) could have been a reason for variation in measurement outcome. A second source of error could have been the environmental circumstances at the time of the measurement. The subjects' attention can be influenced by surrounding noises. A third source of error could have been the alertness of the subjects during the measurement. Changes in alertness might influence the timing of detection of knee motion. Therefore, to minimize the impact of these potential sources of error, the protocol was standardized to a high degree, the 2 raters were specifically trained to be mindful of subject positioning and instructing, and subjects were measured during the same time of day on both occasions. We therefore believe that these potential sources of error did not have a major impact on the SEM and MDD.
Variations in measurement procedure had no impact on intrarater agreement. In the first variation, the angular velocity during the measurement was reduced from 0.3°/second to 0.1°/second. However, the measurement error between the original and the variation in measurement did not change substantially. This is not in accordance with previous studies, in which proprioceptive acuity was found to improve with increasing velocities of joint movement (37–39). An explanation for this difference in results could be the calculation of the SEM and MDD. In our analyses, the SEM was measured with the variables subject, knee, and the interaction between subject and knee as random variables, resulting in an absolute measurement error. This absolute measurement error represents more precisely the within-subject differences.
The second variation in the measurement procedure, music by headphones, did not substantially affect the agreement of the measurement. This means that the auditory cue of the starting of the stepper motor indicating the start of knee joint movement did not lead to substantially different results, compared with a condition where this cue was absent.
Many studies have measured proprioception in patients with knee OA (1, 7, 8, 10–27); however, information on the reproducibility of the methods used to assess proprioception is rarely provided. The studies providing information on reproducibility all used a different method for the measurement of knee joint proprioception: weight bearing or non–weight bearing, start position flexion or extension, and a velocity of angular displacement of 0.1°/second to 5°/second. All these factors could have influenced the reproducibility. In general, the measurement of proprioception can be divided into 2 categories: by the detection of joint movement (i.e., joint movement sense) and by the detection of joint position (i.e., joint position sense). Our findings concern joint movement reliability and do not apply to joint position sense measurements. It can be expected that joint movement sense and joint position sense are related to each other, i.e., that both are expressions of proprioception. Future research could examine the relationship between these 2 joint senses.
The results of our study are in agreement with the study by Sharma et al (21). A similar device and measurement procedure for the detection of joint movement were used. The measurement was found to have high intrarater reliability, which is in line with our results. However, no information was presented concerning the interrater reliability, agreement, and variations in measurement. In the study by Marks (15), the joint position sense was measured, whereas in our study the sense of joint movement was measured. Although Marks (15) previously reported high reliability for the measurement of knee joint proprioception, these results are difficult to compare due to the considerable differences in the proprioception measurements used in the study by Marks and our own study.
It is believed that the knee joint position during proprioception measurement influences the accuracy of the measurement. It has been demonstrated that proprioception is more accurate in the middle range than at the end range (40). In our study, subjects were measured from a starting position of 30° flexion, which is a position commonly present in daily life (e.g., during walking and other transfers). Measurements were made while the knee moved toward extension, i.e., toward the end of the range of motion. It is possible that this has resulted in an underestimation of the degree of proprioceptive accuracy in some patients with OA.
In conclusion, in persons with knee OA and healthy subjects, the measurement of knee proprioception shows adequate intra- and interreliability. The absolute measurement error is rather high. Therefore, this measurement has limited value in the assessment of individual patients, but can be recommended for scientific research in groups of patients.
Mr. van der Esch 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. Hurkmans, van der Esch, Ostelo, Dekker, Steultjens.
Acquisition of data. Hurkmans, van der Esch.
Analysis and interpretation of data. Hurkmans, van der Esch, Ostelo, Knol, Dekker, Steultjens.
Manuscript preparation. Hurkmans, van der Esch, Ostelo, Knol, Dekker, Steultjens.
Statistical analysis. Hurkmans, van der Esch, Ostelo, Knol, Dekker, Steultjens.
We gratefully acknowledge Mr. M. Paalman and colleagues of the VU University Medical Center, Amsterdam, The Netherlands, for manufacturing the instrumented knee joint proprioception measurement.