Subjects
Twenty-four individuals (14 women and 10 men) with no history of neurological disease or major orthopedic lesions were included in the study. The sample size was based on the recommendations of Fleiss, i.e., that 15 to 20 subjects would be required for estimating the reliability of a quantitative variable [25]. The subjects' mean age was 41 years (SD 7.9 years), mean height 174 cm (SD 8.4 cm), mean weight 74 kg (SD 12.6 kg), and median activity level 4 (quartiles 4 to 5, range 2 to 9) according to the Tegner activity level scale, equal to moderately heavy work or recreational sports such as jogging, bicycling, or cross-country skiing [26]. The Research Ethics Committee at Lund University approved the study. All subjects gave their written informed consent to participate in the study.
Kinesthesia test
Kinesthesia was measured in a specifically designed apparatus, which has been described and used previously on patients with ACL injury and uninjured subjects [16, 17, 27–29]. The apparatus consists of a large rectangular platform. A new platform, mounted inside a steel frame has been constructed in order to make the device easier to use for older subjects (the platform was previously placed on the floor) (Figure 1). Mounted at one end is an electric motor with a wire. The wire is connected to a movable T-shaped sled to which a plastic splint is attached for fixation and positioning of the lower limb and foot. A metal bar is attached to the center of the sled, and pulling the wire in either direction causes the sled to rotate like the hand of a clock along the natural arc of extension or flexion of the knee. The arrow-shaped tip of the sled points to an analog scale on the platform (i.e., goniometer) to record movements in increments of 0.25°. The use of ball-bearings allows movements with little friction.
The subject lies in a lateral decubitus position, with the lower leg in the plastic splint. The splint supports the posterolateral part of the leg, but also has a slight anterior curve (to avoid valgus stress at the knee). The oversized construction allows for differences in the girth of the lower leg. Two bars mounted on the platform serve as guides for placing the thigh and trunk in a standard position, with the hip joint semiflexed. The knee joint was carefully positioned at the center of rotation. Markings on the platform allow accurate positioning of the knee in the different starting positions of knee joint flexion: 20° and 40°. Zero degrees is defined as full extension. The upper thigh and hip rest on a foam pillow (which can be adjusted to different heights, due to more extreme varus/valgus angulations), and pillows were also placed under the back to help the subject relax during the test. Care was taken to reduce any external stimuli of limb movement except those from the knee joint and surrounding structures. To minimize cutaneous sensations during the tests, all subjects wore short pants and a thick woolen sock, and the knee had no contact with the underlying surface. Visual cue of the leg was reduced by the subject's position, and closed eyes during the test, and auditory impulses were reduced during the threshold trial by earmuffs and a tape recorder playing a sound imitating the motor.
Measurements of the TDPM were performed towards extension (TE) and flexion (TF) from the two starting positions, 20° and 40° knee joint flexion, giving the variables TE20, TE40, TF20, and TF40. The subjects were asked to close their eyes, concentrate on their knee and respond (by raising their hand) when they felt any sensation of movement in their knee. The tape recorder was then turned on and, after a delay of 5 to 15 seconds (this information was not given to the subjects), the motor started to move the leg at a calibrated angular velocity of 0.5°·s-1. When the subject responded, the assessor stopped the motor and the movement was registered in degrees. The median values of three consecutive measurements of TE20, TE40, TF20, and TF40 were determined [16, 17, 27–29]. Higher values indicate poorer proprioceptive acuity [17]. The subjects were tested twice (test 1 and test 2), at about the same time of day with an interval of approximately one week, median value 7 days (quartiles 6–7, range 2–12 days).
The different starting positions were chosen so as to be within the working range of the knee during ordinary weight-bearing activities/exercise. Since the range of motion may differ between individuals (e.g., some individuals may have an extension deficiency), the most extreme joint positions were excluded. Thus, the tension in the muscles, capsule and ligaments was kept below high levels to avoid more variable tissue tensions between individuals, and to allow the subjects to relax without having their leg forced to maximum extension.
A slow speed was chosen to ensure that the subjects could not detect a sudden onset of motion and to maximally stimulate the joint receptors and minimize the contribution from muscle receptors. The tests were performed on both legs; the right leg being tested first, by shifting the apparatus arrangement from one side of the platform to the other.
Statistical analysis
Since no differences were found between the men and the women, the results were analyzed together. No statistically significant difference was found between the variables in the right and left legs. To avoid the subjectivity in choosing one of the legs, the average of the right and left leg, i.e., (right+left)/2, for each variable was used for statistical analyses [30]. However, the results were confirmed using the results from the right and left legs separately in the analyses.
A number of statistical methods of assessing test-retest reliability were used: 1) mean difference and 95% CI, 2) the two-way random effect model (absolute agreement definition), single measure ICC and 95% CI (ICC2,1 according to Shrout & Fleiss [31]), and 3) the Bland and Altman method of assessing agreement for individual subjects, which includes a scatter plot of the differences between test 1 and test 2 (test 2 minus test 1) against their mean with 95% limits of agreement (LOA) (i.e., mean difference ± 1.96 SDdiff) [23]. Systematic bias can easily be estimated from these "Bland & Altman plots", e.g., if the values from the second test are greater than the values from the first test, the mean difference between the tests will be positive, and if the values from the second test are smaller than the values from the first test, the mean difference between the tests will be negative. If zero is included in the 95% CI, no significant systematic change in the mean is present. The plots also show indications of heteroscedasticity, i.e., larger variability for higher test values. In such cases, spreading out of data for larger values will be observed in the plots. Heteroscedasticity can be revealed by calculating a correlation coefficient between the absolute difference and the average of the test sessions. Performing a logarithmic transformation decreases this relationship. Without this log-transformation of heteroscedastic data, the LOA will be wider apart than necessary for low values and narrower than they should be for larger values [24]. Since heteroscedasticity was found in the present data (i.e., spreading out of data for larger values with a significant relationship between the absolute difference and the average of the test sessions), which is exemplified in Figure 2, a log-transformation (log
e
) was applied prior to calculation of the LOA (exemplified in Figure 3). The log-transformed LOA were then back-transformed (antilogged), giving values that can be interpreted in relation to the original scale. Using this transformation, the limits of the ratio of the two tests (LOAratio) were obtained [24]. For example, a LOAratio ranging between 0.80 and 1.20 times means that one test may differ from another by 20% below, i.e., a 20% decrease in TDPM, to 20% above, i.e., a 20% increase in TDPM in an individual.