Performance characteristics of the Kin-Com dynamometer.

Devices designed to measure forces generated by muscles are commonly used in physical therapy clinics and range in complexity from simple hand-held instruments to more complex dynamometers that measure torque during limb movements.[1] Electromechanical dynamometers can control the velocity of limb movement while measuring muscularly generated forces or torques. Little data, however, are available that describe the performance characteristics of these devices.

A logical place to begin evaluation of the measurements obtained with electromechanical dynamometers is to perform a bench study to examine the mechanical characteristics of the device, that is, to test the accuracy and replicability of measurements obtained with each of the transducers (eg, strain gauges, tachometer, potentiometer) without a human subject applying force to the lever arm. Farrell and Richards[2] described one possible method for performing such an evaluation. They examined the reliability of measurements obtained from the force, angle, and velocity transducers of the Kin-Com[R] dynamometer.(*) Our study represents an example of another method for the systematic analysis of the accuracy and replicability of measurements obtained from the Kin-Com[R] dynamometer. In both approaches, however, only the mechanical aspects of the devices were assessed and reliability of measurements obtained with subjects was not evaluated.

The Kin-Com[R] is a computer-controlled electromechanical dynamometer. The device provides resistance during isokinetic (constant-velocity) movement and during isometric muscle contractions. Investigators examining a variety of questions related to the measurement and improvement of muscle force have also used the device.[3-6]

The majority of clinical studies evaluating similar devices have focused on another widely used electromechanical dynamometer, the Cybex[R] II.([dagger]) More recently, investigators have begun to examine the measurement characteristics of other devices such as the Lido Digital Isokinetic System, [[double dagger]7-9] the Biodex B-2000 Isokinetic Dynamometer,[[sections]10,11] and the Ariel Computerized Exerciser.[[parallel]12] Some authors have found that measurements obtained with these devices are replicable between days; however, extensive testing of the performance characteristics of these devices has not been done.[13]

We evaluated the relationship between user-set velocity (velocity selected by the therapist using the Kin-Com[R] software [version 1.3]*) and the actual velocity that the lever arm moved without a subject applying force to the lever arm. We felt that this was a necessary first step in evaluating the performance characteristics of this device. Options were available in the Kin-Com[R] software to control acceleration and deceleration of the lever arm to and from constant velocity, as well as for the direction of lever arm movement. Therefore, we felt that testing a representative number of lever arm velocities in all possible combinations of acceleration and deceleration in both the upward and downward directions was essential to understanding the performance characteristics of the device. We believe that this information is important for both researchers and clinicians who plan to use the device for exercise and muscle performance measurement with patients.

The Kin-Com[R] consists of several components. For current versions of the Kin-Com[R], subject testing and exercise are controlled by the operator using a personal computer and a software program supplied with the device. Signals from the force, angle, and velocity transducers of the Kin-Com[R] are processed at 100 Hz(#) by the system's analog-to-digital board and displayed on the computer monitor. Subjects can be placed in a variety of positions for testing and exercise. The limb to be tested is attached to the dynamometer via a padded cuff, which is attached to a housing containing strain gauges. The housing can be moved by the operator along a metal lever arm to accommodate different limb lengths. In the isokinetic mode, the software allows the investigator to control the. velocity at which the lever arm will move. If the subject attempts to accelerate the limb beyond the preset velocity, the machine is designed to resist with a force equal in magnitude but opposite in direction, thereby resulting in a constant angular velocity of the limb. This has been termed "accommodating resistance."

The software also allows regulation of several other variables by using a setup menu. The Kin-Com[R] differs from some other dynamometers in that the subject does not actually move the lever arm. The lever arm is moved by a hydraulic motor that is controlled by the computer. The software allows the user to set a threshold value of force that the subject must generate before the lever arm will move. This is called the initial force. The user is also allowed to regulate the acceleration of the lever arm at the beginning of motion and deceleration of the lever arm at the end of motion by setting what the Kin-Com[R] manual refers to as turn points to high, medium, or low settings.

The Kin-Com[R] monitors the force, angle, and velocity signals through feedback loops, which monitor the signal transducers. Force measurements are obtained by load cells in the lever arm. Angle measurements are obtained by a potentiometer, and velocity measurements are obtained by a tachometer.

The purpose of our study was to assess the performance characteristics of the Kin-Com[R] dynamometer under controlled conditions. Measurements obtained from the Kin-Com[R], and simultaneous measurements obtained from our external recording system, were compared with known weights, angles, and velocities. These relationships were examined to furnish information not provided by the manufacturer. We consider this form of testing a prerequisite to further use of devices such as the Kin-Com[R]. Only when the mechanical properties of a device are known can we assess the error associated with measurements obtained when subjects are tested with the device. No attempt was made to assess the inferential capacity of measurements obtained with this device.

Methods

The analog signals from the three Kin-Com[R] transducers (force, angle, and velocity) were collected by an external recording system. These data were collected for comparison with known values of force, angle, and velocity and the measurements obtained from the Kin-Com[R] device's analog-to-digital processor. Our external recording system consisted of an AMM1 analog-to-digital board in a Keithley DAS Measurement and Control System (series 500).(**) Data acquisition was controlled by Dadisp I (version 1.0) software([dagger]) using an IBM XT personal computer."([double dagger]) Signal analysis was performed using Dadisp Worksheet version 1.5)."([dagger]) The specific device tested was the Kin-Com[R] dynamometer (model 500-11); testing procedures were controlled with Kin-Com[R] software (version 1.3).

Force

The Kin-Com[R] measures forces applied in either of two directions. A load cell consisting of four strain gauges is mounted in a housing on the lever arm. The housing slides along the lever arm to accommodate different limb lengths. These strain gauges are mounted in pairs on the top and bottom of a metal shaft in the load cell and can, therefore, measure force generated by a limb in two directions. In our study, each pair of strain gauges was tested separately so that force measurements in each direction could be evaluated. This was done by applying the complete range of loads to each pair independently.

All force measurements were obtained by placing known weights on a weight pan that was suspended from the Kin-Com[R] dynamometer's lever arm. The testing of the Kin-Com[R] device's force measurement system was conducted on 2 consecutive days. For this part of the study, the lever arm was maintained in a position perpendicular to the line of gravity. This position was determined with a gravity-referenced protractor and maintained by placing a hydraulic jack under the lever arm.

Static force measurements can be obtained by using the gravity-correction menu selection in the Kin-Com[R] device's software. This mode allows the tester to obtain force measurements by directly applying weights to the lever arm. Using this mode, the following measurements were obtained on each testing day. First, we obtained a baseline voltage measurement. Next, we determined a voltage calibration factor for conversion of voltage to newtons. We then assessed the nature of the relationships between the force measurements obtained with the Kin-Com[R] device's measurement system, the measurements obtained with the external recording system, and the known weights.

Determination of baseline voltage for the load cell. To measure the applied loads, we first needed to measure the load contribution of the weight pan alone. This measurement was then subtracted from subsequent load measurements. This procedure was conducted for both pairs of strain gauges on the 2 consecutive testing days. The housing containing the strain gauges also includes a removable bar. The bar is used to attach the various patient-machine interface pads for testing different muscle groups. The bar alone was used for attachment of the weight pan. The weight pan was placed on this bar in a standardized position and locked with a clamp. The analog force signal in this condition was collected for 1 second at 50 Hz with the external recording system. Because of the static nature of this measurement, we believed that a sampling frequency of 50 Hz was adequate. This signal was considered the baseline voltage in the system due solely to the weight pan and was subtracted from force signals that were subsequently obtained with the known weights.

Determination of the voltage calibration factor. Strain gauge voltage was converted to newtons using a calibration factor. The calibration factor was determined for each pair of strain gauges on each day of testing. The calibration factor was calculated by loading the lever arm with the maximum amount of weight our weight pan could accommodate (475.4 N). The force signal generated by this load was sampled at 50 Hz for 1 second. The calibration factor was determined by use of the following equation: (1) 475.4 N/mean [mu] [V.sup.475.4 N] - mean [mu] [V.sup.baseline] = N/[mu]V where mean [mu] [V.sup.475.4 N - mean [mu] [V.sup.baseline] = N/[mu]V the voltage signal obtained when the weight pan was loaded with 475.4 N and mean [mu.V.sup.baseline] = the mean of the voltage signal obtained when the weight pan was unloaded.

Relationship between known weights and measured loads. In this part of the study, we loaded the weight pan with known weights ranging from 22.17 to 453.23 N. We applied 20 different loads within this range to the weight pan. The loads were applied in increments of approximately 22 N (these increments were approximate because the weights were in kilogram units and we calculated the International System of Units equivalent). We randomized the presentation of the 20 different loads for each test, and the weight pan was unloaded after each load was applied. Each pair of strain gauges was loaded 20 times on 2 consecutive days. We collected the force signal for each load with our external recording system at 50 Hz for 1 second. In addition, we also recorded the force measured by the Kin-Com[R] dynamometer's measurement system.

Position/Angle

We conducted the second part of the study to assess the relationship between angular positions of the lever arm, as set with a gravity-referenced protractor, and measurements obtained with the Kin-Com[R] dynamometer's measurement system. The angular position of the lever arm was measured with a potentiometer. Measurements were obtained on 2 consecutive days. A voltage calibration factor for conversion of voltage to degrees was first determined on each testing day. Next, we assessed the nature of the relationships between the angle measurements obtained with the Kin-Com[R] device's measurement system, those obtained with the external recording system, and the known angles.

Determination of the voltage calibration factor for the potentiometer. A calibration factor (in degrees per microvolt) was calculated in order to convert voltage signals to degrees. A voltage signal was obtained for 1 second at 50 Hz with the lever arm at the angles of 105 degrees and 0 degrees relative to a vertical position (as determined with a gravity-referenced protractor([sub-section]). The calibration factor was determined by use of the following equation: (2) 105 [degrees]/mean [mu] [V.sup.105 [degrees]] - mean [mu] [V.sup.0 [degrees]] = degrees/[mu]V where 105 [degrees] = the total excursion of the lever arm used to determine the calibration factor, mean [mu] [V.sup.105 [degrees] = the mean of the voltage signal obtained when the lever arm was placed at 105 degrees, and mean [mu] [V.sup.0 [degrees]] = the mean of the voltage signal obtained when the lever arm was placed at 0 degrees.

Relationship between known angles and measured angles The gravity-referenced protractor was used to set the lever arm at the desired angles. The lever arm was initially placed in a horizontal position, and this position was considered 0 degrees. The voltage generated by the potentiometer at 0 degrees was recorded. The protractor was then used to set the angular position of the lever arm in 5-degree increments from 0 to 110 degrees. At each angular position, measurements were obtained from both the external recording system and the Kin-Com(R) dynamometer's measurement system. The voltage signal from the potentiometer was collected with our external recording system for 1 second at 50 Hz at each angular position. The angular measurements from the Kin-Com[R] were obtained from the test program screen, a calibration screen that can be accessed in the Kin-Com[R] software. The calibration screen displays a number representing the angle of the lever arm after processing the potentiometer signal through the Kin-Com[R] device's analog-to-digital board. The known angles were compared with the measured angles on 2 consecutive days.

Velocity

In the third part of the study, we assessed the relationship between the actual velocity that the lever arm moved and the user-set velocity velocity selected by the therapist using the Kin-Com[R] software). We felt that this was necessary because of the importance of lever arm control in this type of device. The actual velocity of the lever arm was determined from the angular displacement signal using the data analysis software (Dadisp Worksheet).

Relationships between user-set velocity and actual velocity of the lover arm. Movement of the lever arm was controlled with the software selection menus on the Kin-Com[R] device's computer. Using the setup menu, constant speed in the concentric/concentric mode was selected. Initial force, the minimum force applied to the strain gauge necessary before movement is initiated, was set at 0 N. This setting caused the lever arm to move during testing without the need for an externally applied force. When the lever arm was in the horizontal position, it was considered to be at 90 degrees; when the lever arm was in the vertical down position, it was considered to be at 0 degrees. These positions were verified by use of the gravity-referenced protractor. the lever arm was set to move through a 100-degree arc of motion from a position of -5 to 95 degrees. the lever arm was repositioned at -5 degrees prior to each upward and downward movement of the lever arm. The start and stop angles were measured with the protractor and by the Kin-Com.[R] At each set velocity, the movement in the upward direction was tested first, followed by the downward movement.

We used a sampling frequency of 500 Hz with our external recording system for this part of the study. We felt this sampling frequency was necessary because the lever arm would be moving at velocities as high as 210 [degrees]/s. The rate of 500 Hz was the maximum sampling frequency our recording system would allow when sampling three channels (angle, velocity, and force).

The velocity of the Kin-Com[R] dynamometer's lever arm was tested without any externally applied force. The velocities tested were in 30 [degrees]/s increments through a range of 30 [degrees] to 210 [degrees]/s. Each velocity was tested with all possible acceleration and deceleration settings (turn points set at high, medium, and low) and in both directions (upward and downward). The order of testing of the velocity/acceleration/deceleration combinations was randomized. Each of the combinations was tested on 2 consecutive days, and the order of testing of the 126 permutations was randomized separately for each day.

Determination of actual velocity. The velocity that the lever arm moved was determined by the rate of displacement, as calculated by use of the angle recording. The velocity and angle recordings were evaluated using the data analysis software (Dadisp Worksheet). The velocity recording was examined to determine the beginning of motion, the beginning and end of the constant velocity, and the point where the voltage returned to baseline.

Data Analysis

Data analysis was performed separately for each signal tested. For each of the following variables, the strength of the linear relationship was analyzed using a coefficient of determination ([r.sup.2]): (1) actual weights versus externally recorded measurements of the weights, (2) actual weights versus measurements of weights obtained from the Kin-Com,[R] (3) actual angles versus externally recorded measurements of angles, (4) actual angles versus measurements of angles obtained from the Kin-Com[R], and (5) user-set velocities versus externally recorded measurements of velocities.

The coefficients of determination were calculated with a software program.[parallel] In addition, intraclass correlation coefficients (ICC[2,1])(14) were calculated to assess the degree of agreement between days for the externally recorded measurements and the Kin-Com[R] measurements. This form of the ICC was used because this study was conducted to document the error associated with measurements obtained with our device, which may be generalizable to other machines.

Results

The results related to the force measurements are summarized in Tables 1 and 2. In afl conditions, the coefficient of determination for the measurements of weights was above .99 (Tab. 1). The agreement (ICC) between days for all conditions was also above .99 (Tab. 2).

[TABULAR DATA OMITTED]

Table 2. Reliability Estimates for
Measurements of Weights Between Days
Type of Measurement              ICC(a)
Measured(b)                      .99
Measured(c)                      .99
Kin-Com[R](d)                    .99
Kin-Com[R](e)                    .99
(a)Intraclass correlation coefficient (2,1).
(b)Externally recorded weight measurements
from strain gauge pair 1, day 1 versus day 2.
(c)Externally recorded weight measurements
from strain gauge pair 2, day 1 versus day 2.
(d)Kin-Com[R] weight measurements from strain
gauge pair 1, day 1 versus day 2.
(e)Kin-Com[R] weight measurements from strain
gauge pair 2, day 1 versus day 2.

The results related to the position/angle measurements are summarized in Tables 3 and 4. In all conditions, the coefficient of determination for the angle measurements was above .99 (Tab. 3). The agreement (ICC) between days for all conditions was 1.00 Tab. 4).

[TABULAR DATA OMITTED]

Table 4.  Reliability Estimates for
Angle Measurements Between Days
Type of Measurement           ICC(a)
External(b)                   1.00
Kin-Com[R](c)                 1.00
(a)Intraclass correlation coefficient (2,1).
(b)Externally recorded angle measurements,
day 1 versus day 2.
(c)Kin-Com[R] angle measurements, day 1 versus
day 2.

The results related to the velocity measurements are summarized in Tables 5 and 6. As shown in Table 5, the coefficient of determination for the velocity measurements was above .99 for all conditions. The agreement (ICC) between days for all conditions was 1.00 (Tab. 6).

Table 5. Linear Relationships Between Velocity Measurements
Condition                      [r.sup.2]   Slope   Intercept
User-set versus actual(a)
  Day 1                        .99         1.01    -1.43
  Day 2                        .99         1.01    -1.49
  Day 1 (up)(b)                .99         1.01    -1.23
  Day 2 (up)                   .99         1.01    -1.29
  Day 1 (down)(c)              .99         1.01    -1.64
  Day 2 (down)                 .99         1.01    -1.61
Actual velocity(d)
  Day 1 versus day 2           .99         1.00    -0.01
  Day 1 versus day 2 (up)      .99         1.00     0.06
  Day 1 versus day 2 (down)    .99         0.99     0.03
(a)Velocity selected by investigator using Kin-Com[R] software compared with
actual velocity of
lever
arm calculated from rate of displacement of lever arm.
(b)Upward movement of lever arm only.
(c)Downward movement of lever arm only.
(d)Actual constant velocity of lever arm calculated from rate of
displacement.
Table 6. Reliability Estimates for
Velocity Measurements Between Days
Actual Velocity(a)                   ICC(b)
Day 1 versus day 2                   1.00
Day 1 versus day 2 (up)(c)           1.00
Day 1 versus day 2 (down)(d)         1.00
(a) Actual constant velocity of lever arm calculated
from rate of displacement.
(b) Intraclass correlation coefficient (2,1).
(c) Upward direction of lever arm movement
only.
(d) Downward direction of lever arm movement
only.

During testing, we noted that the lever arm did not always travel through the full user-set arc of motion. We observed that the greatest excursion (closest to the user-set excursion) at constant velocity was achieved with slow velocities and high acceleration and deceleration settings. The loss of motion never exceeded 4 degrees in either the upward or the downward direction, with the largest amount of motion lost when the low deceleration setting was used.

Discussion

Measurements of Weights

There was a nearly perfect linear relationship between known weights loaded on a weight pan suspended from the Kin-Com[R] dynamometer's lever arm and both the measurements obtained from the external recording system and the measurements obtained from the Kin-Com[R] software (Tab. 1). A high degree 5% agreement between days was also shown to exist between the measurements of weights obtained with the external recording system and the measurements obtained by use of the Kin-Com[R] (Tab. 2).

This finding is important because static force measurements of the subject's limb are performed during the gravity-correction procedure. It has been demonstrated that there can be significant error associated with isokinetic torque measurements not corrected for the effect of gravity.[15,16] The findings of our study cannot be directly extrapolated to static force measurements of a subject's limb (and the associated error due to limb attachment and positioning). Our study does, however, provide information as to the performance characteristics of the strain gauges during static loading. More recent versions of the Kin-Com[R] software (eg, version 3.01) allow testing of a subject's isometric torque; therefore, testing the characteristics of the strain gauges during static loading is important clinically.

The intercepts of the lines describing the linear relationships between the measurements obtained with the Kin-Com[R] and the known weights were higher than for the relationship between known weights and the externally recorded measurements (Tab. 1). This discrepancy existed because we were unable to subtract the initial voltage due to the weight pan from the force values with the Kin-Com[R] software.

Our results were in agreement with those of Farrell and Richards.[2] They tested the relationship between measurements of known weights and measurements reported from the Kin-Com[R] software by loading the lever arm with weights in 22.3-N increments from 22.2 to 310.8 N. They reported an ICC of .99 after repeatedly loading and unloading the Kin-Com[R] dynamometer's lever arm.

The methods used by Farrell and Richards[2] to test the relationship between known weights and measurements of force from the Kin-Com[R] device's strain gauges were somewhat different than ours. The static loading phase of their study consisted of positioning the lever arm at 10-degree increments from a horizontal (0 [degrees]) to a vertical (90 [degrees]) position. They applied the full range of their known weights to the lever arm at each angular position. The authors reported that they compared the measurements of their known weights with the measurements obtained from the Kin-Com[R] only. Because we were unaware of the algorithm used by the Kin-Com[R] software to analyze the force signals, we independently processed the signal through our own analog-to-digital system to check the accuracy of the Kin-Com[R] device's force measurements.

Angle Measurements

The results presented in Table 3 demonstrate that there was a nearly perfect linear relationship between actual angular positions of the lever arm as determined with a gravity-referenced protractor and both the measurements obtained with the external recording system and the measurements obtained from the Kin-Com[R] software. A high degree of agreement was also shown to exist between days for the angle measurements obtained with the external recording system and those obtained with the Kin-Com[R] (Tab. 4).

Our findings indicate that when the signals generated by the Kin-Com[R] dynamometer's potentiometer are properly referenced, they represent angles determined with a gravity-referenced protractor. Farrell and Richards(2) also examined the relationship between angle measurements determined with an external recording device and angle measurements obtained with the Kin-Com[R]. They compared angle measurements obtained with a spirit level and a protractor with measurements taken by the Kin-Com[R] software. The authors reported that they positioned the lever arm at "various" angles and compared the Kin-Com[R] device's reported angles with the known angles. They did not provide a statistical analysis of this relationship but stated, "It was not possible to determine any difference in lever arm measurement made by either the Kin-Com[R] or the external system." Therefore, although it not possible to compare our results with those of Farrell and Richards because of different methodologies, it would appear that our results generally agree.

Measurements of Lever

Arm Velocity

The results presented in Table 5 demonstrate that constant-velocity measurements obtained with our external recording system have a nearly perfect linear relationship with the user-set velocities tested in our study. These measurements also show a high degree of agreement when compared on 2 different days (Tab. 6).

The loss of lever arm motion was only noted during the use of medium and low turn points. We never observed a loss of motion when using the high setting. Although the loss of motion never exceeded 4 degrees with the low and medium settings, and our study did not determine whether this effect would be different if a subject were applying a force to the lever arm, we felt that this loss of motion could pose a problem during the acquisition of average measures such as work or power.

Our study was designed to assess the mechanical characteristics of the Kin-Com.[R] Our Kin-Com[R] was a standard commercial device, and our results should therefore be generalizable to other Kin-Com[R] devices in clinical use if they are manufactured and maintained similar to the one we used. Our study was presented to provide a protocol for bench testing the performance characteristics of electromechanical dynamometers. We believe clinicians and researchers should examine the properties of any dynamometer they use. Portions of our study can be replicated with minimal equipment in any setting to determine the characteristics of a dynamometer used in clinical practice. We feel that such bench testing is an important first step in evaluating a device for clinical use. Further studies should be performed to assess the reliability of measurements obtained with the Kin-Com[R] when human subjects are applying force to the lever arm as the sources of error may be different. Validity studies should also be performed to assess the inferential capacity of these measurements for predicting functional outcomes.

Conclusion

A very strong linear relationship was shown to exist between the signals from the Kin-Com[R] dynamometer's transducers and known weights, angles, and user-selected velocities when tested under conditions not involving subject participation. There was also a very strong linear relationship between values obtained from the Kin-Com[R] device's processing system and known weights and angles. Our results indicate that the static measurements of force and angle that are necessary for use in the gravity-correction procedure are accurate and reliable between days.

References

[1] Mayhew TP, Rothstein JM. Measurement of muscle performance with instruments. In: Rothstein JM, ed. Measurement in Physical Therapy. New York, NY: Churchill Livingstone Inc; 1985:57-102.

[2] Farrell M, Richards JG. Analysis of the reliability and validity of the kinetic communicator exercise device. Med Sci Sports Exerc. 1986; 18:44-49.

[3] Hageman PA, Gillaspie DM, Hill LD. Effects of speed and limb dominance on eccentric and concentric isokinetic testing of the knee. J Orthop Sports Phys Ther. 1988;10:59-65.

[4] Hanten W, Ramberg C. Effect of stabilization on maximal isokinetic torque of the quadriceps femoris muscle during concentric and eccentric contractions. Phys Ther. 1988;68:219-222.

[5] Hart DL, Miller LC, Stauber WT. Effect of cooling on force oscillations during maximal voluntary exercise. Exp Neurol 1985;90:73-80.

[6] Jensen K, Di Fabio RP. Evaluation of eccentric exercise in treatment of patellar tendinitis. Phys is Ther. 1989;69:211-216.

[7] Aitkens S, Lord J, Bernauer E, McCrory M. Analysis of the validity of the Lido Digital Isokinetic System. Phys Ther. 1987;67:756. Abstract.

[8] Francis K, Hoobler T. Comparison of peak torque values of the knee flexor and extensor muscle groups using the Cybex Il and Lido 2.0 isokinetic dynamometers. J Orthop Sports Phys Ther. 1987;8:480-483.

[9] Lord J, Aitkens S, McCrory M, Bernauer E. Reliability of the Lido Digital Isokinetic System for the measurement of muscular strength. Phys Ther. 1987;67:757. Abstract.

[10] Thompson MC, Shingleton L, Kegerreis ST. Comparison of values generated during testing of the knee using the Cybex II Plus and Biodex Model B-2000 isokinetic dynamometers. J Orthop Sports Phys Ther. 1989; II: 108-115.

[11] Wilk KE, Johnson RD, Levine B. Comparison of knee extensor and flexor muscle group strength using the Biodex, Cybex, and Lido isokinetic dynamometers. Phys Ther. 1988;68: 792. Abstract.

[12] Warner MA, Duncan PW, Harned DJ, Garrett WE. A comparison of the Cybex 11 isokinetic dynamometer and Ariel Computerized Exerciser. Phys Ther. 1985;65:730. Abstract.

[13] Sapega A. Current concepts review: muscle performance evaluation in orthopaedic practice. J Bone Joint Surg [Am]. 1990;72:1562-1574.

[14] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428.

[15] Fillyaw M, Bevins T, Fernandez L. Importance of correcting isokinetic peak torque for the effect of gravity when calculating knee flexor to extensor muscle ratios. Phys Ther. 1986;66:23-29.

[16] Winter DA, Wells RP, Orr GW. Errors in the use of isokinetic dynamometers. Eur J Appl Physiol. 1981;46:397-408.

(*) Chattecx Corp, Part of Chattanooga Group Inc, 4717 Adams Rd, PO Box 489, Hixson, TN 34343. ([dagger]) Cybex, Div of Lumex, 2100 Smithtown Ave, Ronkonkoma, NY 11779.

([double dagger]) Loredan Biomedical Inc, 3650 Industrial Blvd, West Sacramento, CA 95691.

([sections]) Biodex Medical Systems, Brookhaven R&D Plaza, Box 702, Shirley, NY 11967.

([parallel]) Ariel Life Systems Inc, PO Box 1169, La Jolla, CA 92308.

(#) This information was obtained from the Kin-Com[R] operation manual and from discussions with the manufacturer.

(**) Keithley Instruments Inc, PO Box 391260, Cleveland, OH 44139.

([dagger]) DSP Development Corp, 1 Kendall Sq, Cambridge, MA 02139.

([double dagger]) International Business Machines Corp, Boca Raton, FL 33429.

([sub-section]) Sears Craftsman, Sears, Roebuck & Co, Sears Tower, Chicago, IL 60684.

([parallel]) QuattroPro, Version 1.0, Borland International Inc, 1800 Green Hills Rd, PO Box 660001, Scotts Valley, CA 95066-0001.