# Knee strength variability between individuals across ranges of motion and hip angles

PAVOL, M. J., and M. D. GRABINER. Knee strength variability between individuals across ranges of motion and hip angles. *Med. Sci. Sports Exerc.,* Vol. 32, No. 5, pp. 985–992, 2000.

Purpose Isokinetic strength is normally measured for a single range of motion and body position. This study quantified the variability, between individuals, in the relationships between a single peak knee extension moment and the isokinetic extension moments measured for different hip angles and ranges of knee motion. Effects of hip angle, and of the starting knee angle of the range of motion, on isokinetic knee extension strength were also determined.

Methods The isokinetic knee extension strength of 10 subjects was measured at 30°·s^{−1} to a knee flexion angle of 10° from starting knee angles of 90, 75, 60, 45, and 30°, in both the seated and supine positions. Moments were normalized to the peak moment from a reference contraction.

Results Peak moments and moments at larger knee flexion angles were greater in the seated than in the supine position. The starting knee angle affected the peak moment, the angle of peak moment, and the moments over the initial and final portions of the range of motion. Peak moments were highly correlated between all hip angle-starting knee angle combinations. However, the normalized peak moments, the angles of peak moment, and the normalized angle-specific moments all varied considerably between subjects. The pooled standard deviation and average coefficient of variation of the normalized angle-specific moments between subjects were 10.5% of the normalizing moment and 15.7%, respectively. Excluding the reference contraction, between-subject variability was unaffected by hip angle or starting knee angle.

Conclusions Influences of hip angle, starting knee angle, and individual differences on isokinetic knee extension strength must be considered to ensure that the moments obtained from isokinetic testing adequately reflect the general strength capabilities of an individual.

Biomedical Engineering Center, The Ohio State University, Columbus, OH 43210; and Department of Biomedical Engineering, Lerner Research Institute, The Cleveland Clinic Foundation, Cleveland, OH 44195

Submitted for publication October 1998.

Accepted for publication July 1999.

Address for correspondence: Michael J. Pavol, Programs in Physical Therapy, Northwestern University, 645 North Michigan Avenue, Suite 1100, Chicago, IL 60611. E-mail: m-pavol@nwu.edu.

Peak moments measured during isokinetic contractions are the most commonly used quantitative descriptors in comparing strength between and within individuals (^{18}). In addition, angle-specific moments measured during such isokinetic contractions may be of interest in estimating the percentage of an individual’s strength capabilities that are employed during a task (e.g., 10,25). In most applications of isokinetic strength testing, strength is measured over only a single tested range of joint motion and in a single body position. It is, therefore, implicitly assumed that the isokinetic moments so measured fully characterize the strength capacity of the individual at the tested isokinetic velocity.

However, focusing on the task of knee extension, there is evidence that the moments measured during an isokinetic maximal voluntary contraction (MVC) are affected by the range of knee motion being tested and by the movement threshold moment of the dynamometer (^{1,11,12,22}). One would expect body position to affect isokinetic knee extension moments, as changes in hip angle affect the operating length, thus the strength, of the rectus femoris. Curiously, significant effects of hip angle on isokinetic knee extension strength have not been observed (^{3}), although such effects have been shown theoretically (^{15}) and observed, selectively, under isometric conditions (^{5,8}).

The practical consequences of the aforementioned factors on isokinetic knee extension strength measurement depend on the amount of variability that they introduce into the strength measures, that is, on the degree to which the effects of these factors are inconsistent and unpredictable between individuals. A few studies to date have looked at between-subject variability in the relationships between peak and angle-specific moments for isokinetic knee extension over a single range of knee motion at a single hip angle (^{19–21}). However, the inherent variability between individuals in the relationships between isokinetic knee extension moments measured over differing ranges of knee motion or at differing hip angles is not known. Such information is needed to evaluate the degree to which isokinetic moments measured over a single range of motion in a single body position characterize the general strength capabilities of an individual in a given direction of exertion at a joint. The greater the variability between individuals, the lesser the confidence that can be placed in a single measure of strength.

Thus, the primary purpose of this study was to quantify the between-subject variability in the relationships between a single measured peak knee extension moment and the knee extension moments measured isokinetically from different starting knee angles and at different hip angles. However, such measures of variability are only of interest if knee extension strength characteristics change with hip angle and starting knee angle. Therefore, we also characterized whether, and in what manner, the measured isokinetic knee extension moments were affected by the tested hip angle and starting knee angle.

## METHODS

### Subjects.

Ten subjects (five women) participated in this study after providing informed consent. The dominant lower extremity was reported by each subject as free of musculoskeletal pathology or abnormality. Subject characteristics were (mean ± SD): age 27.8 ± 3.2 yr, height 1.70 ± 0.08 m, and mass 67.8 ± 12.4 kg.

### Experimental protocol.

Subjects performed maximum voluntary isokinetic knee extensions with their dominant leg on a KIN-COM 500H isokinetic dynamometer (Chattanooga Corp., Chattanooga, TN). Contractions were performed at 30°·s^{−}1 to a final knee flexion angle of 10° from five different starting knee flexion angles: 90, 75, 60, 45, and 30° (where 0° was defined as maximal knee extension). Subjects were tested in each of two body positions: seated and supine. These positions introduced hip flexion angles of roughly 80° and 0°, respectively, from the anatomical position. The alignment between the axes of rotation of the knee and dynamometer arm did not vary between body positions, nor did the manner in which the distal leg was secured to the dynamometer load cell attachment. Restraint straps at the thigh and hips were employed throughout.

Three maximum voluntary contractions were performed for each hip angle-starting knee angle combination, with contractions separated by at least 1 min of rest. Subjects were instructed to reach their maximum force output as rapidly as possible at the start of each contraction and to maintain a maximal effort throughout the range of motion. Arms were kept folded across the chest. The dynamometer movement threshold used ranged from 48.9 to 75.6 N, as determined by the force passively exerted by the leg. The dynamometer arm acceleration was at the device’s high rate. The hip angle initially tested was balanced between subjects, and the order in which the starting knee angles were tested was randomized within each hip angle. A warm-up period preceded the tests at each hip angle.

Upon completion of the MVC at each hip angle, three passive isokinetic knee extensions at 30°·s^{−}1 were performed by the dynamometer over the range of 95° to 5° of knee flexion. Subjects were instructed to remain relaxed and to neither aid nor resist the motion. Practice trials were performed until consistent force traces with minimal evidence of active muscle force contribution were obtained. During data collection, any trial exhibiting an inconsistent force trace was repeated.

### Data analysis.

The knee angle, extension velocity, and extension force were recorded at 300 Hz, and the corresponding knee extension moment was computed according to the load cell moment arm. The angle, velocity, and moment signals were low-pass filtered at 20, 20, and 7 Hz, respectively, using a fourth-order, zero phase-lag, Butterworth filter. The velocity and moment at each integer knee angle were linearly interpolated from the filtered data. Data from the three MVC at each hip angle-starting knee angle combination were combined by selecting, at each integer angle, the maximum moment measured at a velocity between 27 and 33°·s^{−}1. To reduce dynamometer-induced oscillatory artifacts, the resulting moment-angle relationships were smoothed using an averaging filter with a window width of ± 4° about each angle.

Processing of the passive trials at each hip position was identical to the processing of the MVC trials, except that the average passive moment at each knee angle was computed in combining trials. The isokinetic MVC moment-angle relationships computed for each subject were subsequently corrected for the contribution of gravity and passive tissue forces to the measured moments by subtraction of the appropriate passive moment-angle relationship. The resultant relationships provided, for each knee angle, an estimate of the maximal knee extension moments actively generated during motion at the prescribed isokinetic velocity.

The isokinetic MVC moment-angle relationships of each subject were normalized by the peak moment generated during the 80° hip angle-90° starting knee angle trials. The peak moment and corresponding knee angle of peak moment were determined for each hip angle-starting knee angle combination. In addition, for each MVC moment-angle relationship, average normalized moments were computed over each 5° interval within the range of knee motion, excluding the initial and final 5° of the contraction.

For each hip angle-starting knee angle combination, the standard deviation of the normalized moments was computed across subjects at each integer knee angle. The corresponding coefficients of variation at each integer knee angle were also computed. Finally, within-subject moment variability was estimated. The variance of each subject’s normalized moments was computed across starting knee angles at each knee angle from 65 to 30°. Separate variances were computed for each hip angle and moments from the first 10° of an MVC were disregarded. The overall within-subject standard deviation was then determined from the pooled variance across subjects, hip angles, and knee angles. A similar procedure was employed to estimate an overall average within-subject coefficient of variation.

### Statistical analysis.

The main effects of hip angle and starting knee angle on the normalized peak knee extension moments were investigated through a paired *t*-test and a repeated-measures ANOVA, respectively. Data were averaged, by subject, across the opposing factor for each analysis. Two-way (hip angle by starting knee angle) repeated-measures ANOVA were performed on the knee angle of peak moment, on the average normalized moments over each of the 5° knee angle intervals of interest, and on the normalized moments at knee angles of 70, 55, 40, and 25°.

Pearson correlations were computed between the unnormalized peak moments from each hip angle-starting knee angle combination. The Mann-Whitney test was used to compare the correlations between peak moments from the same hip, or starting knee, angle to the correlations between peak moments from differing hip, or starting knee, angles. Effects of hip angle and starting knee angle on the variance in the normalized peak moments and the variance in the angles of peak moment were investigated by two-way repeated-measures ANOVA of the transformed Y scores (^{4}). The 90° starting knee angle was excluded from the ANOVA of peak moment variance and an *F*-test used to compare the variance in the 0° hip-90° starting knee angle peak moments to the pooled variance from all other starting knee angles. ANOVA of the standard deviations and coefficients of variation in the normalized angle-specific moments were used to compare the overall between-subject variability across hip angle-starting knee angle combinations.

*Post hoc* analysis employed Bonferroni-corrected paired *t*-tests, trend analysis, a Tukey test, or a Dunnett test as appropriate. All effects were considered significant at a *P* < 0.05.

## RESULTS

### Effects of hip angle on strength.

The mean peak MVC knee extension moments generated at the 80° hip angle were significantly greater than those generated at the 0° hip angle (Table 1, Fig. 1). However, the knee angle of peak moment was unaffected by hip angle (*P* > 0.05) (Table 2). At knee angles greater than 35°, significantly greater MVC knee extension moments were produced with the hips at 80° than with the hips at 0° (Table 3). Hip angle did not affect knee extension moments at knee angles of less than 35° (*P* > 0.05).

### Effects of starting knee angle on strength.

The peak knee extension moment and the knee angle of peak moment did not differ for MVC starting from knee angles of 90° or 75° (Tables 1 and 2). However, the peak moment and the knee angle of peak moment both decreased progressively as the starting knee angle decreased from 75 to 30°. Moments generated at an angle 5° from the start of an MVC were, in most cases, significantly smaller than the moments generated at this same angle during MVC started at larger knee flexion angles (Table 4). Beyond the initial 5° of an MVC, the average knee extension moments over almost all intervals of knee angles were not affected by the starting knee angle (*P* > 0.05) until the final 15° of isokinetic motion. At knee angles from 25 to 15°, average extension moments decreased linearly with increasing starting knee angle (Table 4). The effect of the starting knee angle on the average extension moments did not differ between hip angles for any interval of knee motion (*P* > 0.05 for all interactions).

### Variability in peak moments.

Peak unnormalized MVC knee extension moments were highly and significantly (*P* < 0.02) correlated between all hip angle-starting knee angle combinations. Correlation coefficients ranged from 0.72 to 0.97, with a root-mean-squared (RMS) value of 0.89 across the 45 such correlations. Peak moments measured at the same hip angle were more highly correlated, on average, than those measured at differing hip angles (RMS *r* of 0.91 vs 0.87, respectively;*P* < 0.05). However, peak moments measured from the same starting knee angle were no more related, on average, than those measured from differing starting knee angles (RMS *r* of 0.88 vs 0.89, respectively;*P* > 0.05).

Despite the high correlations observed, there was considerable variability between subjects in the normalized peak moments for all hip angle-starting knee angle combinations other than that used for the normalization (Table 1). Excluding the normalizing combination, the pooled standard deviation of the normalized peak moments between subjects, across hip angle-starting knee angle combinations, equaled 11.0% of the normalizing moment. When expressed as coefficients of variation, the variability of the normalized peak moments averaged 14.3% (range 10.3–17.6%) across the different hip angle-starting knee angle combinations. The variance in the normalized peak moments did not differ between hip angles or between starting knee angles (*P* > 0.05).

Considerable between-subject variability was also evident in selected of the knee angles of peak moment (Table 2). The variance in the knee angle of peak moment did not differ between hip angles (*P* > 0.05) but was significantly greater for a starting knee angle of 90° than for starting knee angles of 60° or less. This effect of the starting knee angle on the variance was independent of the hip angle (*P* > 0.05 for the interaction).

### Variability in angle-specific moments.

As observed in the peak moments, there was considerable between-subject variability in the angle-specific normalized MVC knee extension moments (Figs. 2 and 3). The pooled standard deviation of the angle-specific normalized moments between subjects, across all knee angles and hip angle-starting knee angle combinations, was 10.5% of the normalizing moment. The equivalent average coefficient of variation of the normalized moments was 15.7%.

The normalizing, 80° hip angle-90° starting knee angle combination exhibited significantly less overall between-subject variability in the normalized angle-specific moments than did other hip angle-starting knee angle combinations. Only the average standard deviations for starting knee angles of 30° were not greater (*P* > 0.05). However, although the normalizing combination exhibited very little variability in the neighborhood of the peak, normalizing, moment, the variability in the latter half of the range of motion approached that of the other combinations.

For these other hip angle-starting knee angle combinations, the between-subject standard deviations in the normalized angle-specific moments were essentially unaffected by hip angle or starting knee angle. Differences between hip angles in the pooled standard deviations at a given knee angle averaged only 1.6 ± 0.9% of the normalizing moment. Similarly, the range across starting knee angles of the pooled standard deviations at a given knee angle averaged 1.4 ± 0.8% of the normalizing moment. Finally, the standard deviations in the normalized angle-specific moments decreased linearly with decreasing knee angle (r^{2} = 0.84;*P* < 0.001) when pooled across all but the normalizing hip angle-starting knee angle combination.

Again excluding the normalizing hip angle-starting knee angle combination, the between-subject coefficients of variation of the normalized angle-specific moments were essentially unaffected by hip angle, although effects of the starting knee angle were apparent. Differences between hip angles in the average coefficient of variation at a given knee angle averaged only 1.4 ± 0.8%. However, coefficients of variation at knee angles within the initial 5° of the range of motion of an MVC were, in most cases, notably higher than the average coefficients at these angles for MVC started at larger knee angles. The range, across starting knee angles, of the average coefficients of variation also increased linearly (at 0.2% per degree) over the last 25° of motion. Excepting the initial 5° of motion and the normalizing hip angle-starting knee angle combination, the average coefficients of variation in the angle-specific moments varied quadratically with knee angle (r^{2} = 0.82;*P* < 0.001), with the smallest variability observed at knee angles near 53°.

For both measures of variability, the between-subject variability in the normalized angle-specific moments was much larger than the corresponding estimate of within-subject variability. The estimated within-subject standard deviation of the normalized angle-specific moments was 5.5% of the normalizing moment, giving an approximate *F*-ratio of between- to within-subject variance of 3.7. The average within-subject coefficient of variation in the normalized angle-specific moments was 6.0%.

## DISCUSSION

In most applications of isokinetic strength testing, strength is measured over only a single range of joint motion in a single body position. The results of the present study indicate that the peak and angle-specific moments measured in testing isokinetic knee extension strength can vary significantly with differences in both the range of knee motion and the hip angle employed. In addition, the results indicate that there is considerable variability between individuals in the manner in which these isokinetic knee extension moments vary across different ranges of knee motion and hip angles. The peak knee extension moment measured at 30°·s^{−}1 using a “standard” 80° hip angle-90° starting knee angle protocol only characterized knee extension strength for a different starting knee angle or hip angle to within a standard deviation of about 11% of the measured peak moment, with standard deviations as large as 16% observed.

Before discussing the results, several methodological issues should be addressed. First, in estimating strength variability, it was important that the moment-angle relationships analyzed be directly comparable; each should closely approximate the maximum voluntary capacity for active knee extension moment generation under the prescribed conditions. The best estimate, from the available data, of maximal moment-generating capacity at a given knee angle was the maximum moment measured at that angle. This was the rationale for selecting the maximum moment in combining corresponding trials.

To avoid inflated estimates of variability, it was also necessary to correct the moments for artifactory oscillations introduced by dynamometer accelerations and compliances (^{14,27}). An averaging filter with a window width approximating the average oscillation period was used. As the artifacts represent inertial effects and changes in muscle force that result from oscillations of the leg’s velocity about the isokinetic velocity, they may be assumed to have no mean effect over an oscillation period. Also, based on the isometric moment-angle relationship, it is reasonable that the true isokinetic moments would vary near-linearly over the filter window in the absence of artifacts. The filtered data should, therefore, have reasonably approximated the true isokinetic moments. Although deviations of the filtered moment-angle relationships from the ideal isokinetic relationships undoubtedly existed, these deviations are expected to be of little effect in comparison to the original artifacts.

Electromyography (EMG) was not measured in this study. This is mentioned because differences in measured knee extension strength can reflect differences in actual knee extensor force-generating capacity or differences in the level of neural excitation of these muscles. The magnitude of the opposing moments generated by the knee flexors will also affect the net knee extension moments measured. Without measures of EMG, the extent to which the observed results reflected changes in voluntary neural excitation of the knee extensors and flexors could not be determined. The origin of the results can, therefore, only be hypothesized. However, the lack of EMG is inconsequential to the study’s primary purpose of quantifying variability in maximum voluntary isokinetic knee extension moments.

It was expected that isokinetic knee extension strength would be affected by the angle of the hip, through a change in the strength of the rectus femoris. Because the rectus femoris crosses both the hip and knee, its length at a given knee angle varies with hip angle. As muscle strength is length-dependent, the strength of the rectus femoris at a given knee angle must then also vary with hip angle. Large effects of hip angle on knee extension strength were not expected, however. The rectus femoris contributes only about 17% of the total isokinetic MVC knee extension moment (^{24}). The remaining vast majority of the total extensor moment-generating capacity should have been unaffected by hip angle, as the lengths of the other knee extensors (i.e., the vasti) depend only on the knee angle.

Although effects of hip angle on isokinetic knee extension strength were indeed observed, the significant average decreases of 7–10% in the moments observed at the 0° versus the 80° hip angle were the opposite of what was expected. Theoretical models (^{15}) and *in vivo* measurements (^{17}) of rectus femoris force-length relationships would predict increases in the moment-generating capacity of the knee extensors at the 0° hip angle within the range of knee angles studied. In addition, studies of isometric and isokinetic strength have found knee extensor moments to either increase or remain unchanged with decreases in hip angle (^{3,5,8}), although most of these studies investigated only a single knee angle or hip angles that differed considerably from those of this study.

The most likely explanation for the unexpected influence of hip angle on isokinetic knee extension strength is that subjects activated their knee extensors to a lesser extent during MVC at the 0° hip angle, causing the measured strength to significantly under-represent the actual strength capacity of the musculature. Effects of hip angle on the maximal voluntary excitation of the knee extensors have, in fact, been observed (^{13,26}). Task familiarity may be a contributing factor, as individuals appear to be able to generate the largest rectus femoris knee moments at the lengths of its most frequent use (^{16}). Observations that the effects of hip angle appeared to vary greatly between subjects (Fig. 2) and that knee extension in the supine position was the less familiar task are consistent with the hypothesis that the effects of hip angle reflected differences in voluntary muscle activation more than differences in muscle moment-generating capacity. Differences in knee flexor activation likely did not contribute to the observed effect, however (^{7}).

Starting knee angle affected isokinetic knee extension strength in a manner that was predictable and consistent with the findings of earlier studies (^{11,12,28}). The observed monotonic decrease in the peak moment as the starting knee angle decreased from 75° reflects a similar change in isometric knee extension moment-generating capacity from its peak at a knee angle near 60° (^{30}). The submaximal moments observed over the initial 5–10° of an MVC reflect the fact that, for isokinetic contractions with a low movement threshold moment, the muscles’ neural excitation, active state (force-generating potential), and force output are all submaximal over the initial portion of the range of motion due to the finite times required for each of these processes (^{1,2}). Finally, the inverse linear relationship between starting knee angle and the moments measured at knee angles less than 25° is consistent with the increased force depression that accompanies increased distances of shortening in skeletal muscle (^{6}).

As the hip angle and range of knee motion employed during isokinetic testing have been shown to affect measured knee extension strength, the question arises as to whether the normal practice of characterizing strength at a given joint velocity by the peak moment measured over a single range of motion in a single body position is appropriate. The answer to this question depends on the accuracy with which the single measured peak moment characterizes the strength of the individual over the full spectrum of ranges of motion and body positions, as well as on the degree of uncertainty in the strength characteristics of the individual that is acceptable.

Variability in strength characteristics between individuals must be expected. Anatomical differences in musculotendon architecture and musculoskeletal geometry will cause variations in moment-angle behaviors between subjects. Sources of these variations include differing joint angle-muscle length relationships, differing muscle length-force capacity relationships, differing muscle force-joint moment relationships, and differences in synergist force capacity. Differing muscle fiber-type compositions will also introduce between-subject variations in the maximum rates of initial moment generation and in the muscles’ capacity for force generation at the shortening velocities associated with the isokinetic motion.

However, the potentially greatest contributors to between-subject differences in measured strength characteristics are differences in the voluntary neural excitation of muscle. Variability will result, not only from differences in excitation in performing a given task, but also from differing changes in excitation between tasks. These differences may be in the initial rate of increase of excitation, in the ability to fully activate the agonist musculature, or in the ability to sustain a voluntary maximum level of excitation. Differences in cocontraction could also be a factor.

In acknowledgment of this between-subject variability, the present study provides quantitative estimates of the errors to expect in employing a single measured peak moment to predict the strength of an individual over different ranges of motion and body positions. The reported measures of variability correspond directly to the error associated with using the group-average moment-angle behavior to predict the subjects’ strength based on their 80° hip angle-90° starting knee angle peak moments. This moment was employed as the reference value because of its common use as a measure of isokinetic knee extension strength.

The between-subject variability in the present study tended to be smaller than the comparable variability reported by Kannus and colleagues (^{19–21}) in the relationships between angle-specific isokinetic knee extension moments and the peak moment measured for the same velocity, range of motion, and hip angle. Their standard errors of estimate averaged approximately 27% (range 12–39%) of the mean angle-specific moment, as compared with the average coefficient of variation of 16% (range 2–25%) for the angle-specific moments in the present study. The larger values in the previous studies are probably attributable to the higher velocities tested (60°·s^{−}1 and 180°·s^{−}1), as well as their restricting their analysis to single angles near the ends of the tested range of motion. It is possible that the differing dynamometers and signal processing methods used could also have contributed to the between-study differences.

In the present study, an isokinetic velocity of 30°·s^{−}1 was studied. This velocity is one that is employed clinically and is among the velocities at which the largest voluntary isokinetic knee extension moments may be generated (^{18,29}), which has the desirable effect of reducing the influence of any small random factors on the results obtained. The slow velocity also provided time for the knee extensors to attain their maximum active state during the MVC, allowing both transient and steady-state behaviors to be observed. Because between-subject variations in the rate of moment generation should significantly affect only the transient behavior, greater coefficients of variation were expected during the initial, transient portion of the MVC than during the latter, steady-state portions. This was indeed the case.

An encouraging finding of the present study was that, qualitatively, the peak knee extension moment measured over a single range of knee motion and at a single hip angle appears to be a reasonable indicator of overall knee extension strength at the tested velocity. Despite significant effects of both hip angle and starting knee angle on knee extension strength, peak extension moments were highly correlated across hip angle-starting knee angle combinations. In addition, a single “standard” peak moment characterized the angle-specific moments over different hip angle-starting knee angle combinations to within an average standard deviation of 11% of the peak moment. Although not negligible, the size of this variability indicates that the effects of hip angle and starting knee angle on strength were qualitatively similar across most subjects. These results would support the continued clinical measurement of knee extension strength over only a single range of motion in a single body position.

On the other hand, where accurate quantitative approximation of an individual’s overall knee extension strength characteristics is needed, the present results are less encouraging. The results indicated that between-subject variability in the normalized peak and angle-specific moments was essentially the same for all hip angle-starting knee angle combinations except the normalizing combination, which exhibited significantly less variability. This would imply that a single, measured, isokinetic peak moment is sufficient to characterize an individual’s overall knee extension strength characteristics at the tested velocity to within an overall, baseline level of uncertainty. However, beyond this baseline, a given measure only allows a more accurate characterization of strength for conditions near-similar to those of the measurement. As a result, improving on the between-subject variability of this study would likely require individual-specific models of knee extension strength based on moment-angle relationships measured over a number of hip angle-starting knee angle combinations. Where practical considerations limit such an approach, a general model of knee extension strength will likely exhibit average errors of approximation of at least the magnitude of the variability observed in this study.

Finally, the dependence of knee extension strength on hip angle and starting knee angle cautions against using moment-angle-velocity relationships determined from isokinetic strength testing over a single range of motion in a single body position (^{9,23}) to estimate the percentage of an individual’s strength capabilities that are employed during another motor task. Isokinetic moments measured under steady-state voluntary maximal activation will significantly overestimate functional moment-generating capacity during transient changes in muscle activation or in the latter portion of sustained contractions. The converse is also true. In addition, strength relationships determined for one body position may significantly misrepresent the strength capacity in a different position.

In conclusion, the results of the present study indicate that the common practice of characterizing knee extension strength by the peak moment measured over a single range of isokinetic motion in a single body position is justified if one is willing to accept a considerable degree of uncertainty in inferring strength capabilities over ranges of motion and hip angles other than those tested. Significant influences of hip angle, starting knee angle, and individual differences on isokinetic knee extension strength must be considered to ensure that the moments obtained from isokinetic testing adequately reflect the general strength capabilities of an individual.

The authors thank A. Al-Johar for assisting in data collection. This study was partially funded by NIH R01AG10557 (to MDG).

## REFERENCES

**Keywords:**

KNEE EXTENSION; ISOKINETIC TESTING; MOMENT-ANGLE RELATIONSHIPS